[go: up one dir, main page]

CN120278406A - Sponge city high-topography rainwater management method and system - Google Patents

Sponge city high-topography rainwater management method and system Download PDF

Info

Publication number
CN120278406A
CN120278406A CN202510772801.8A CN202510772801A CN120278406A CN 120278406 A CN120278406 A CN 120278406A CN 202510772801 A CN202510772801 A CN 202510772801A CN 120278406 A CN120278406 A CN 120278406A
Authority
CN
China
Prior art keywords
prediction
model
rainfall
predictor
fine
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202510772801.8A
Other languages
Chinese (zh)
Other versions
CN120278406B (en
Inventor
王宸
熊洁
郭勤
魏伟
周予进
余豪
林彦珉
刘扬
杜文青
裴华君
胡晓晨
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Communications Design Institute Co Ltd Of Jiangxi Prov
Original Assignee
Communications Design Institute Co Ltd Of Jiangxi Prov
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Communications Design Institute Co Ltd Of Jiangxi Prov filed Critical Communications Design Institute Co Ltd Of Jiangxi Prov
Priority to CN202510772801.8A priority Critical patent/CN120278406B/en
Publication of CN120278406A publication Critical patent/CN120278406A/en
Application granted granted Critical
Publication of CN120278406B publication Critical patent/CN120278406B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • GPHYSICS
    • G06COMPUTING OR CALCULATING; COUNTING
    • G06QINFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00Administration; Management
    • G06Q10/06Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063Operations research, analysis or management
    • GPHYSICS
    • G06COMPUTING OR CALCULATING; COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/04Architecture, e.g. interconnection topology
    • G06N3/044Recurrent networks, e.g. Hopfield networks
    • G06N3/0442Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
    • GPHYSICS
    • G06COMPUTING OR CALCULATING; COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/08Learning methods
    • G06N3/084Backpropagation, e.g. using gradient descent
    • GPHYSICS
    • G06COMPUTING OR CALCULATING; COUNTING
    • G06QINFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00Administration; Management
    • G06Q10/04Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • GPHYSICS
    • G06COMPUTING OR CALCULATING; COUNTING
    • G06QINFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/06Energy or water supply
    • GPHYSICS
    • G06COMPUTING OR CALCULATING; COUNTING
    • G06QINFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/10Services
    • G06Q50/26Government or public services

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Economics (AREA)
  • Human Resources & Organizations (AREA)
  • Strategic Management (AREA)
  • General Physics & Mathematics (AREA)
  • Tourism & Hospitality (AREA)
  • Health & Medical Sciences (AREA)
  • Marketing (AREA)
  • General Business, Economics & Management (AREA)
  • General Health & Medical Sciences (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Development Economics (AREA)
  • Data Mining & Analysis (AREA)
  • Biophysics (AREA)
  • Mathematical Physics (AREA)
  • General Engineering & Computer Science (AREA)
  • Computing Systems (AREA)
  • Educational Administration (AREA)
  • Molecular Biology (AREA)
  • Evolutionary Computation (AREA)
  • Computational Linguistics (AREA)
  • Software Systems (AREA)
  • Primary Health Care (AREA)
  • Biomedical Technology (AREA)
  • Artificial Intelligence (AREA)
  • Quality & Reliability (AREA)
  • Operations Research (AREA)
  • Game Theory and Decision Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Water Supply & Treatment (AREA)
  • Public Health (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses a sponge city high-topography rainwater management method and system, in particular relates to the technical field of rainwater prediction management, and aims to solve the problem of poor high-topography rainwater prediction management scheduling; the method comprises the steps of firstly collecting and preprocessing data such as elevation, permeability coefficient, soil layer thickness and vegetation coverage, generating a continuous physical field, mapping a facility pipe network, dividing grid units, constructing a multi-element predictor based on the central attribute of the grid units, issuing edge node increment fine adjustment after cloud training, automatically extracting observation and prediction residual increment training neural networks after heavy rain to generate confidence intervals, dividing risks according to the confidence intervals and flood control water depth, minimum infiltration threshold values, outputting gate opening and pump set power suggestions, and accordingly remarkably improving extreme storm peak prediction precision and response timeliness of the high-altitude small river basin, and achieving fine risk perception scheduling and rainwater resource utilization.

Description

Sponge city high-topography rainwater management method and system
Technical Field
The invention relates to the technical field of rainwater prediction management, in particular to a method and a system for managing high-topography rainwater in a sponge city.
Background
A Sponge City (Sponge City) is an ecological City development mode for improving the absorption, storage, purification and reuse capacities of the City to rainwater by simulating natural hydrologic cycle, and the core aim is to enable the City to permeate, retain, store and purify the rainwater like a Sponge during rainfall and release and utilize stored water resources during drought, so that the problems of waterlogging, drought, water pollution and the like are relieved, and sustainable management of the urban water resources is realized.
In the extreme storm management of a small basin in a high topography of a sponge city, a single statistical rainfall and runoff model or a pure data driving method is relied on, the amplification effect of the sudden fluctuation of the high topography on the peak time sequence and the intensity of the runoff of the storm is ignored, and the hydrodynamic conservation relation is not brought into the model constraint, so that systematic prediction errors of the peak time delay or the lower intensity usually occur under the condition of short-time heavy rainfall, for example, the rainfall of the small basin in a certain mountain region drops by 90 millimeters in 1 hour, the traditional model predicts the peak value by 60 millimeters and delays for 20 minutes, the regulation and storage tank fails to open floods in time, the precise effect of the waterlogging prevention and control and groundwater supply in the high topography region is seriously restricted, and the water circulation scheduling management of the city is affected.
Disclosure of Invention
In order to overcome the defects in the prior art, the invention provides a method and a system for managing high-topography rainwater in a sponge city, which are used for solving the problem of poor prediction management and scheduling of the high-topography rainwater in the background art.
In order to achieve the above purpose, the present invention provides the following technical solutions:
a sponge city high topography rainwater management method comprises the following steps:
Carrying out data acquisition and preprocessing on a high topography area, combining drilling permeability coefficient, soil layer thickness and remote sensing vegetation coverage interpolation, generating a continuous physical field after removing anomalies, mapping facility and pipeline information to an elevation coordinate system, outputting a facility information layer and a facility attribute list after topology repair, and dividing grid units;
Constructing a predictor based on the attribute of the grid unit center, issuing edge nodes to locally perform incremental fine adjustment after cloud training, dynamically matching the master predictor according to a scene performance matrix, generating a global update package through federal learning aggregation, and synchronously updating local model prediction;
After each storm event, the edge node automatically extracts the latest observation and prediction residual error, and uses the latest observation and prediction residual error as a training sample to carry out incremental iteration on the neural network, and after fine adjustment is finished, a random shielding mechanism is used for forward reasoning, so that the concentrated trend and fluctuation range of the predicted value are counted, and a confidence interval of the model is generated;
and dividing the high risk, the low risk and the safety unit by the edge node according to the prediction confidence interval, the flood control water depth threshold value and the minimum infiltration threshold value, and generating gate opening and pump group power suggestions.
In a preferred embodiment, the data acquisition and pretreatment are performed on the high topography area, and the drilling permeability coefficient, the soil layer thickness and the remote sensing vegetation coverage are combined to interpolate, so that a continuous physical field is generated after the abnormality is removed, and the specific process is as follows:
The method comprises the steps of obtaining original elevation data of a high topography area, carrying out projection transformation, unifying the original elevation data to the same coordinate system, applying Gaussian smoothing filtering to the projected elevation data, removing measurement noise to form smooth elevation information, carrying out polynomial error correction on the smooth elevation information according to ground control points, and outputting corrected elevation information;
performing soil and vegetation parameter interpolation on the high-relief area, obtaining soil permeability coefficient data and soil layer thickness data from drilling sampling, and obtaining vegetation coverage data from remote sensing classification;
Respectively applying a kriging method interpolation to the soil permeability coefficient, the soil layer thickness and the vegetation coverage data to generate a continuous permeability coefficient field, a soil layer thickness field and a vegetation coverage field;
And removing abnormal values of the interpolated permeability coefficient field, the soil layer thickness field and the vegetation coverage field, and interpolating again to fill the hole.
In a preferred embodiment, the facility and pipeline information is mapped to an elevation coordinate system, a facility information layer and a facility attribute list are output after topology repair, and grid cells are partitioned, wherein the specific process is as follows:
extracting facility and pipeline information from the regional design drawing, wherein the facility and pipeline information comprises spatial positions and construction parameters of a permeation tank, a regulating reservoir and a main drainage pipeline;
Mapping the extracted facility and pipeline information into a coordinate system of corrected elevation information, generating a facility information layer, performing topology repair on broken or overlapped pipe sections in the facility information layer, and outputting a complete facility attribute list;
Calculating local gradient and curvature for each position in a GIS environment or a special script based on the corrected elevation information, calling a Delaunay triangulation algorithm, and automatically generating a series of triangular grid units by taking a side length function as a constraint;
taking the central elevation, the permeability coefficient, the soil layer thickness and the vegetation coverage of each grid unit as initial conditions, taking the real-time rainfall intensity and the boundary water level observation value as boundary input, and calculating the water depth change and the flow velocity distribution of each unit in each time step through a finite volume iteration method.
In a preferred embodiment, a predictor is constructed based on the attribute of the grid cell center, and the edge node is issued to locally perform incremental fine tuning after cloud training, which specifically comprises the following steps:
Constructing four types of complementary predictors based on the central elevation, the permeability coefficient, the soil layer thickness, the vegetation coverage, the historical rainfall time sequence and the corresponding water depth observation value attribute of each grid unit;
the predictor comprises a simulation error correction predictor which is trained by using a decision tree regression algorithm with a physical simulation result and a sensor observation difference as targets;
Physical equation consistency constraint and water depth observation are taken as double losses, and a physical information enhancement neural network is trained, so that output fitting observation meets hydrodynamic conservation;
taking the multidimensional attribute and the historical yield of the grid cells as inputs, training an attribute association predictor based on an integrated tree algorithm, and mining nonlinear interaction among soil, vegetation and topography;
Taking the time dependence characteristics of rainfall and runoff production as input, training a long-period and short-period memory network, and capturing the fluctuation trend after rainfall mutation;
after the cloud end completes batch training for the first time, each predictor is respectively issued to each drainage basin edge node to carry out local fine adjustment, and the fine adjustment process adopts residual errors between current rainwater observation and batch training results as new samples to carry out iterative updating.
In a preferred embodiment, the active predictor is dynamically matched according to the scene performance matrix, and the global update package is generated through federal learning aggregation, and then the local model prediction is synchronously updated, which comprises the following specific processes:
Dividing typical heavy rain events of short-time heavy rainfall, long-time weak rainfall and uneven spatial distribution into different scene categories according to a historical back-detection library, and carrying out statistics and analysis on peak arrival time, peak height and fluctuation trend performance indexes of each predictor under each category to form a scene performance matrix;
according to the current rainfall time sequence and the terrain section division, automatically matching the most-fit scene category, extracting a predictor which shows the optimal performance under the scene category from the matrix to serve as a main model, and triggering a standby model calling mechanism if the deviation between a water depth curve output by the main model and real-time observation data exceeds a preset threshold value, and sequentially switching among suboptimal models until the prediction precision is restored;
the rainfall time sequence comprises a curve inflection point of accumulated rainfall and rainfall intensity abrupt change time;
After the local fine adjustment of the edge nodes of each drainage basin is completed, uploading the difference between the local model parameters and the initial model parameters to a cloud terminal, and collecting the parameter difference from each node by the cloud terminal;
for the same network layer or decision tree branch structure, adopting a majority of improvement choosing strategies, namely keeping updating for the parameter directions of which most nodes are improved, and regarding the parameters of which part of nodes are changed as noise suppression not to be adopted;
Forming a unified global update package after screening, and sending the global update package back to each edge node, wherein each edge node is applied to a local model for prediction after receiving the update package;
After the global updating deployment is completed, each edge node immediately carries out cross-domain back measurement, a new generation model is respectively operated in the original typical short-time heavy rainfall, long-time weak rainfall and space distribution uneven scene, root mean square error and maximum absolute error index are calculated, and compared with the historical version result,
If the error index of the drainage basin is raised or not expected, an online performance feedback reporting mechanism is automatically triggered, and the error distribution of the drainage basin and the real-time observation residual are uploaded to the cloud.
In a preferred embodiment, after each heavy rain event, the edge node automatically extracts the latest observation and prediction residuals and uses them as training samples to perform incremental iterations on the neural network, as follows:
after each round of storm event is finished, automatically summarizing the difference value between the real water depth sequence recorded by the local sensor and the predicted water depth sequence output by the integrated predictor, constructing a residual error data set containing a time stamp and corresponding residual errors, and selecting the latest 50 residual errors as fine adjustment samples according to descending order for the latest time period data in the residual error data set;
After sample preparation is completed, loading a physical information enhancement neural network model on an edge node, setting an iterative training process taking least square residual error sum as a loss function, and carrying out 10 rounds of back propagation optimization by adopting parameters with a learning rate of 0.005 and a batch size of 10;
After fine tuning is completed, the updated model parameter snapshot and the latest residual statistics are stored together, and the old model is replaced, so that the model parameter snapshot is immediately used for water depth prediction in the next period.
In a preferred embodiment, after the fine tuning is completed, forward reasoning is performed by using a random masking mechanism, and the central trend and the fluctuation range of the predicted value are counted to generate a confidence interval of the model, wherein the specific process is as follows:
After the fine tuning is finished, the system carries out self-checking on the new model at the edge node, namely, the latest 50 observation inputs are used, the predicted output is recalculated and compared with the true value, if the root mean square error is more than ten percent lower than the root mean square error before the fine tuning, the online iteration is confirmed to be effective, otherwise, the online iteration is returned to the last effective parameter state and the cloud end is reported;
after the online fine tuning is completed, an uncertainty quantization link is entered, forward reasoning is executed on the connection nodes of the random shielding part in the network for the input data at the same moment, and the output fluctuation of the model under different internal states is simulated;
generating a group of water depth predicted values through each reasoning, integrating the predicted results, calculating an arithmetic mean value and a standard deviation, and deducing upper and lower limits of a confidence interval based on the mean value and the standard deviation;
after obtaining the confidence interval, comparing the confidence interval with a preset flood control and infiltration threshold value:
When the confidence upper limit is close to or exceeds the flood control threshold, more regulation spaces are automatically reserved and emergency diversion strategies are started in advance;
When the confidence lower limit is far lower than the lower osmosis requirement, the operation power of the pump set is reduced, and natural osmosis is carried out;
if the confidence interval falls entirely within the safety interval, then regular scheduling is performed.
In a preferred embodiment, the edge node divides the high risk, low risk and safety units according to the prediction confidence interval and the flood depth threshold value and the minimum infiltration threshold value, and generates gate opening and pump group power suggestions as follows:
Comparing the upper limit and the lower limit of the prediction confidence interval of each triangular grid unit with a preset flood control depth threshold value and a preset minimum infiltration threshold value one by one, and dividing the units into three types of high risk, low risk and safety;
When the upper limit of the confidence interval exceeds the flood control water depth threshold, determining as a high-risk grid unit, and preferentially expanding the use proportion of the regulating reservoir;
when the lower limit of the confidence interval is lower than the minimum infiltration threshold, determining as a grid unit with low risk, preferentially reserving the infiltration pool capacity suggestion, and performing natural infiltration treatment;
And for the grid cells in the safety interval, generating a default opening degree and a power interval of the gate and the pump set according to conventional scheduling logic.
A sponge city high topography rainwater management system for realize above-mentioned sponge city high topography rainwater management method, include:
The high-topography network dividing module is used for carrying out data acquisition and preprocessing on a high-topography area, combining drilling permeability coefficient, soil layer thickness and remote sensing vegetation coverage interpolation, generating a continuous physical field after abnormality removal, mapping facility and pipeline information to an elevation coordinate system, outputting a facility information layer and a facility attribute list after topology repair, and dividing grid units;
The grid fine adjustment prediction module is used for constructing a predictor based on the attribute of the grid unit center, issuing edge nodes to locally perform incremental fine adjustment after cloud training, dynamically matching the master predictor according to the scene performance matrix, generating a global update package through federal learning aggregation, and synchronously updating local model prediction;
The prediction model adjustment module is used for automatically extracting the latest observation and prediction residual errors by the edge nodes after each storm event, performing incremental iteration on the neural network by using the latest observation and prediction residual errors as training samples, performing forward reasoning by using a random shielding mechanism after fine tuning is finished, and counting the concentrated trend and the fluctuation range of the predicted values to generate a confidence interval of the model;
The management advice generation module is used for dividing the high risk, the low risk and the safety unit by the edge node according to the prediction confidence interval, the flood control water depth threshold value and the minimum infiltration threshold value, and generating the gate opening and pump group power advice.
The invention has the technical effects and advantages that:
The method comprises the steps of carrying out systematic data acquisition and preprocessing on a high-relief area, generating a continuous physical field by fusing drilling penetration coefficient, soil layer thickness and remote sensing vegetation coverage interpolation, mapping penetration pool, regulation pool and pipeline information to form a facility information layer, dividing grid units, constructing a multi-element predictor based on grid unit center elevation and physical properties, issuing edge node local increment fine tuning after cloud training, relying on a scene performance matrix and a federal learning mechanism to dynamically match and synchronize an optimization model, automatically extracting an observation prediction residual error after each storm event for online increment iteration of a neural network, generating a confidence interval by random shielding forward reasoning, dividing risk grades by the edge node according to the confidence interval and a flood control water depth and a minimum infiltration threshold value, and outputting gate opening and pump power advice, thereby remarkably improving the prediction precision and response speed of extreme storm peaks of the high-relief small-flow areas, realizing refined risk perception scheduling, and enhancing storm prevention and control and rainwater resource utilization effects.
Drawings
Fig. 1 is a flow chart of a method for managing high topography rainwater in a sponge city.
Fig. 2 is a schematic structural diagram of a sponge city high profile rain water management system according to the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Embodiment 1 As shown in fig. 1, the sponge city high topography rainwater management method comprises the following steps:
Carrying out data acquisition and preprocessing on a high topography area, combining drilling permeability coefficient, soil layer thickness and remote sensing vegetation coverage interpolation, generating a continuous physical field after removing anomalies, mapping facility and pipeline information to an elevation coordinate system, outputting a facility information layer and a facility attribute list after topology repair, and dividing grid units;
Constructing a predictor based on the attribute of the grid unit center, issuing edge nodes to locally perform incremental fine adjustment after cloud training, dynamically matching the master predictor according to a scene performance matrix, generating a global update package through federal learning aggregation, and synchronously updating local model prediction;
After each storm event, the edge node automatically extracts the latest observation and prediction residual error, and uses the latest observation and prediction residual error as a training sample to carry out incremental iteration on the neural network, and after fine adjustment is finished, a random shielding mechanism is used for forward reasoning, so that the concentrated trend and fluctuation range of the predicted value are counted, and a confidence interval of the model is generated;
and dividing the high risk, the low risk and the safety unit by the edge node according to the prediction confidence interval, the flood control water depth threshold value and the minimum infiltration threshold value, and generating gate opening and pump group power suggestions.
Step one, constructing digital twin data enhanced by physical information under high topography, which comprises the following specific steps:
carrying out data acquisition and preprocessing, obtaining original elevation data of a high topography area, carrying out projection transformation, unifying the original elevation data to the same coordinate system, applying Gaussian smoothing filtering to the projected elevation data, removing measurement noise to form smooth elevation information, carrying out polynomial error correction on the smooth elevation information according to ground control points, and outputting corrected elevation information;
performing soil and vegetation parameter interpolation, obtaining soil permeability coefficient data and soil layer thickness data from drilling sampling, and obtaining vegetation coverage data from remote sensing classification;
Respectively applying a kriging method interpolation to the soil permeability coefficient, the soil layer thickness and the vegetation coverage data to generate a continuous permeability coefficient field, a soil layer thickness field and a vegetation coverage field;
Removing abnormal values of the interpolated permeability coefficient field, the soil layer thickness field and the vegetation coverage field, interpolating again to fill the hole, and ensuring that each data field is smooth and has no isolated abnormality;
And (3) fusing the facility and pipe network information, extracting facility and pipe information from the regional design diagram, including spatial positions and construction parameters of a permeation pond, a regulating reservoir and a main drainage pipe, mapping the extracted facility and pipe information into a coordinate system of corrected elevation information, generating a facility information layer, topologically repairing broken or overlapped pipe sections in the facility information layer, and outputting a complete facility attribute list.
Taking a small river basin test point area of a certain mountain as an example, the corrected elevation information refers to digital elevation Data (DEM) corrected by projection transformation, filtering smoothing and ground control points, and the elevation value of each position of the ground surface is recorded, wherein the elevation value is usually measured by a laser radar or a total station;
For example, in the design drawing, the plane coordinate of a certain infiltration tank is (x=512345.67 m, y= 341234.56 m), the design elevation of the bottom plate is 365.00m, the coordinate is aligned to a grid unit in the DEM after correction, the surface elevation value (such as 367.25 m) of the unit can be read, and then the burial depth of the bottom plate of the infiltration tank relative to the terrain is calculated (367.25 m-365.00 m=2.25m).
For another example, a section of main drainage pipeline from node a (x=512500.00 m, y= 341200.00m, and height 370.10 m) to node B (x=512600.00 m, y= 341350.00m, and height 368.50 m), pipe diameter 500mm, and slope 0.16, and after mapping the two spatial positions and elevations with DEM one by one, a facility information layer including three-dimensional position, pipe diameter, and slope can be generated in GIS.
Through the mapping, each facility or pipe section obtains accurate three-dimensional coordinates and physical parameters, and reliable basic data based on high topography fluctuation is provided for subsequent hydraulic simulation, network topology verification and intelligent scheduling.
It should be noted that, the raw elevation data refers to a digital high Cheng Dian cloud or grid that is not subjected to any processing, and generally comes from the following paths:
airborne LiDAR (laser radar) aerial survey, namely, carrying a laser radar sensor on an unmanned plane or a fixed-wing aircraft above a small high-topography drainage basin of a sponge city to be built, and carrying out contour scanning on the whole drainage basin (such as a mountain area with an average gradient of more than or equal to 15 degrees in a range of 10km 2) to obtain three-dimensional coordinates (X, Y and Z) of each point;
satellite remote sensing (such as SRTM and ALOS) downsampling DEM, namely supplementing boundary areas in peripheral areas with insufficient coverage of airborne LiDAR by adopting satellite DEM data with 30m or 12.5m resolution;
The ground GNSS static measurement comprises the steps of arranging a double-frequency GNSS reference station at a key control point (GCP) position, obtaining an absolute elevation, and correcting an navigation point cloud with millimeter-level precision;
after the acquisition is completed, the airborne LiDAR point cloud is fused with the satellite DEM and the ground GNSS control points, all data are unified to the same projection coordinate system, and the original elevation data covering the whole high-topography small drainage basin are output.
Based on the corrected elevation information, calculating local gradient and curvature for each position in a GIS environment or a special script, and constructing a nonlinear grid side length function according to the local gradient and curvature, so that the high-topography steep slope region is divided, a gentle region can be properly coarsened, a Delaunay triangulation algorithm is called, and a series of triangular grid units are automatically generated by taking the side length function as a constraint. Each unit is a basic calculation unit, and the geometric center, the vertex and the adjacent relation of the basic calculation unit are determined by the subdivision process;
And after grid generation is completed, continuous field data constructed in pretreatment stages such as elevation, permeability coefficient, soil layer thickness, vegetation coverage and the like can be extracted on the central coordinate of each triangle unit immediately, and spatial overlapping is carried out on the continuous field data and the facility information layer.
And carrying out equation configuration and numerical solution, namely applying a mass conservation equation and a momentum conservation equation to each triangular unit based on multidimensional attributes of the grid units so as to describe rainfall input, surface runoff and permeation confluence processes, firstly, taking the central elevation, the permeation coefficient, the soil layer thickness and vegetation coverage of each grid unit as initial conditions, taking real-time rainfall intensity and boundary water level observation values as boundary input, and calculating the water depth change and flow velocity distribution of each unit in each time step through a finite volume iteration method. In the solving process, introducing a permeability loss term in the form of Darcy's law aiming at the permeation process, so that the lower permeation quantity is related to the permeability coefficient and the soil layer thickness;
the numerical solution is executed on the cloud cluster in parallel, and the water depth time sequence and the flow velocity time sequence of each unit are output in real time, so that physical simulation results are provided for neural network training and local prediction.
Taking the space coordinates and the time stamp as inputs, constructing a depth feedforward neural network, setting nodes and layers according to the nonlinear characteristics of the watershed, and generating a water depth predicted value at a corresponding moment at the output end of the network;
In the network training, two parts of losses are introduced simultaneously, namely, firstly, data fitting loss, a network predicted value and a water depth value observed by each sensor under a rainfall event are compared, and secondly, physical residual loss is realized, the degree of unsatisfied equation is calculated by substituting predicted output into a mass conservation equation and a momentum conservation equation, and the residual is used as a punishment item;
by adjusting the network weight, the trained network can accurately restore the historical observation water depth and can naturally follow the hydrodynamics rule in the prediction process;
it should be noted that, in the initial training stage, global historical data is adopted for batch training, and then small batch fine adjustment is performed on the cloud end aiming at an extreme storm event, so that the universality and the local precision are both considered.
Comparing the physical numerical solution result with the neural network prediction result, evaluating the accuracy of water depth prediction by taking root mean square error as an index, and adjusting the permeability loss coefficient in numerical solution and the physical residual error weight in network training aiming at the error trend until the calibration error is converged within a preset range;
After calibration is completed, a physical solver and a neural network model are packaged and deployed to an edge computing node, the node receives rainfall intensity, water level and soil humidity data in real time in a storm process, short-time water depth prediction is rapidly given by the neural network, then local numerical solution is combined for correction, and finally a comprehensive prediction result is pushed to a control unit through a message middleware for intelligent scheduling of a gate and a pump group, so that high-precision prediction and low-delay response are ensured in a high-topography sponge city rainwater management scene.
Step two, after space discrete grid division and attribute mapping are completed, generalized adaptation to the diversity of terrains, soil and vegetation is realized through cooperation of a multi-element predictor and a federal learning mechanism, and the specific steps are as follows:
Four types of complementary predictors are constructed based on the central elevation, permeability coefficient, soil layer thickness, vegetation coverage, historical rainfall time sequence, corresponding water depth observation values and other attributes of each grid unit;
Firstly, training a simulation error correction predictor by using a decision tree regression algorithm with a physical simulation result and a sensor observation difference as targets, and capturing systematic deviation of physical solution in a local extreme process;
Physical equation consistency constraint and water depth observation are taken as double losses, and a physical information enhancement neural network is trained, so that output fitting observation meets hydrodynamic conservation;
taking the multidimensional attribute and the historical yield of the grid cells as inputs, training an attribute association predictor based on an integrated tree algorithm, and mining nonlinear interaction among soil, vegetation and topography;
Finally, taking the time dependence characteristics of rainfall and runoff production as input, training a long-period memory network, and capturing the fluctuation trend after rainfall mutation;
After the cloud end completes batch training for the first time, each predictor respectively sends the result to each drainage basin edge node for local fine adjustment, and the fine adjustment process adopts residual errors between current rainwater observation and batch training results as new samples for iterative updating, so that the model can take account of global universality and local scene characteristics;
a specific example is that in the construction stage of a multi-element predictor, each model type is trained in a targeted manner based on the elevation of a unit center, the permeability coefficient, the soil layer thickness, the vegetation coverage and the actual measured water depth of historical rainfall;
For example, for a certain grid unit (center elevation 350 m, permeability coefficient 0.000012 m/s, soil layer thickness 2.5 m, vegetation coverage 65%), the small time sequence rainfall in the process of 15 days of heavy rain in 2024, 6 months is [20, 30, 40, 10] mm/h, the corresponding observed water depth is [0.10, 0.30, 0.40, 0.20] m, the physical simulation solution outputs [0.08, 0.25, 0.35, 0.15] m, the residual errors of the two are [0.02, 0.05] m, the residual error sequence and the grid unit characteristics are taken as samples, and a decision tree regression algorithm (maximum tree depth 5 layers, minimum leaf node samples 10) is adopted to train a simulation error correction predictor, so that the simulation error correction predictor can automatically correct the underestimation deviation of the physical solution at the rainfall peak under the condition similar heavy rain.
On the basis, the physical simulation result and the observed water depth of the same unit are taken as dual targets, and a physical information enhancement network is designed, wherein a network structure 5 layer, 64 neurons in each layer and an activation function are selected from ReLU, and training is performed by simultaneously minimizing a water depth fitting error and an equation residual error (introducing a mass conservation equation residual error and a momentum conservation equation residual error into a loss function), so that prediction is ensured to be in accordance with measured data and not to violate a hydrodynamic rule. Meanwhile, the multidimensional attribute and the accumulated yield (such as the maximum hour yield corresponding to the heavy rain of 0.15 m) of the unit are used as input, an integrated tree algorithm (100 subtrees and the learning rate of 0.1) is used for training an attribute association predictor, nonlinear interaction of terrains, soil and vegetation on the yield is deeply excavated, finally, the rainfall time sequence and the yield sequence at the corresponding moment are input into a long-short-term memory network (50 hidden nodes and 4 hours in time step) to capture the dynamic characteristics of water depth fluctuation after rainfall mutation, and a time sequence prediction model for the water depth trend in the future 1-2 hours is constructed.
Dividing typical heavy rain events such as short-time heavy rainfall, long-time weak rainfall, uneven spatial distribution and the like into a plurality of scene categories according to a historical back-detection library, and carrying out statistics and analysis on key performance indexes such as peak arrival time, peak height, fluctuation trend and the like of each predictor under each category to form a scene performance matrix;
In actual operation, firstly, according to the current rainfall time sequence (comprising the accumulated rainfall curve inflection point and rainfall intensity mutation time) and terrain section division, the most fit scene category is automatically matched, a predictor which is optimal in performance under the category is extracted from a matrix to be used as a main model, if the deviation between the water depth curve output by the main model and real-time observation data exceeds a preset threshold value, a standby model calling mechanism is triggered, and the sub-optimal models are sequentially switched until the prediction precision is recovered.
After each basin edge node completes local fine tuning, uploading the difference between local model parameters and initial model parameters (marked as parameter difference) to a cloud, collecting the parameter difference from each node by the cloud, aiming at the same network layer or decision tree branch structure, adopting a majority of improvement and replacement strategies, wherein the improved parameter direction appears for most nodes, updating is reserved;
After global updating deployment is completed, each edge node immediately carries out cross-domain loop test, a new generation model is respectively operated in the scenes of original typical short-time heavy rainfall, long-time weak rainfall, space distribution inequality and the like, root mean square error and maximum absolute error indexes are calculated and are compared with historical version results, if the error index of a certain drainage basin is raised or is not expected, an online performance feedback reporting mechanism is automatically triggered, error distribution of the drainage basin and real-time observation residual errors are uploaded to a cloud, the cloud preferentially adjusts the depth of a corresponding network layer or decision tree according to the fed-back error mode in the subsequent federal learning round, the adaptive evolution of the model to cross-domain changes such as extreme climate, vegetation evolution and construction disturbance is realized, and the generalization capability and the predictive robustness under various working conditions of the high-topography small drainage basin are continuously improved through federal fine tuning and cross-domain verification closed loops of continuous iteration.
The method is characterized in that the storm time is defined by a continuous period of time when the real-time rainfall intensity is continuously higher than a preset storm intensity threshold value, specifically, the time when the rainfall intensity is first broken through is determined as the beginning of the storm by threshold value filtration, then the time when the rainfall intensity falls below the threshold value and keeps stable and does not rise again is monitored as the end of the storm, the interval between the two is the storm time, and in order to avoid misjudgment caused by short pulsation, the time when the rainfall intensity is continuously lower than the threshold value after falling back reaches the minimum duration set by an expert is required, and the end of the storm can be determined.
Thirdly, performing edge online fine tuning and uncertainty quantification, namely after multi-model integration and federal updating are completed, performing local incremental fine tuning on edge nodes and generating prediction uncertainty so as to improve the risk protection capability of the model on quick response and scheduling decision of environment time-varying characteristics, wherein the method comprises the following specific steps of:
Performing online incremental fine adjustment, automatically summarizing the difference value between a real water depth sequence recorded by a local sensor and an estimated water depth sequence output by an integrated predictor after each round of storm event is finished, constructing a residual error data set containing a timestamp and a corresponding residual error, and selecting the latest 50 residual errors as fine adjustment samples according to descending order aiming at the latest time period data in the residual error data set so as to ensure that the used samples can fully reflect the current soil moisture content, vegetation coverage change and model offset caused by burst drainage strategy adjustment;
after sample preparation is completed, loading a physical information enhancement neural network model on an edge node, setting an iterative training flow taking least square residual error and as a loss function, and carrying out 10 rounds of back propagation optimization by adopting parameters with a learning rate of 0.005 and a batch size of 10, wherein in the process, only a feature extraction layer and an output layer which are related to time-dependent feature and space attribute mapping in a network are subjected to parameter updating, so that the fine tuning speed is ensured, and the calculation cost caused by full-quantity retraining is avoided;
After fine tuning is completed, the updated model parameter snapshot and the latest residual statistics are stored together, an old model is replaced immediately, and the model is used for water depth prediction in the next period immediately so as to support high-precision water depth prediction in the subsequent period.
After the fine tuning is finished, the system carries out self-checking on the new model at the edge node, namely, the latest 50 observation inputs are used, the predicted output is recalculated and compared with the true value, if the root mean square error is reduced by at least 10% compared with that before the fine tuning, the online iteration is confirmed to be effective, otherwise, the online iteration is returned to the last effective parameter state and the cloud end is reported.
Performing uncertainty quantization, namely entering an uncertainty quantization link immediately after online fine adjustment is completed, and performing forward reasoning on input data at the same moment for a plurality of times (30 times, for example) at random shielding part connecting nodes in a network so as to simulate output fluctuation of a model in different internal states;
generating a group of water depth predicted values by each reasoning, integrating the 30 groups of predicted results, calculating an arithmetic mean value and a standard deviation of the predicted results, and describing the concentration trend and the discrete degree of the model under the current environmental condition, and deducing the upper limit and the lower limit of a 95% confidence interval based on the mean value and the standard deviation so as to reflect the uncertainty range of the prediction;
After the confidence interval is obtained, comparing the confidence interval with a preset flood control and infiltration threshold value:
When the confidence upper limit is close to or exceeds the flood control threshold, more regulation spaces are automatically reserved and emergency diversion strategies are started in advance;
When the confidence lower limit is far lower than the lower osmosis requirement, the pump set operating power can be properly reduced so as to promote natural osmosis;
If the confidence interval wholly falls in the safety interval, conventional scheduling is executed according to the central trend value;
all uncertainty quantization results and corresponding input and output data are reported to the cloud in the form of a structured message for performance feedback and model improvement of subsequent federal learning, and closed loop optimization and risk controllability between end-edge clouds are realized.
The flood control threshold, the infiltration requirement and the safety interval are determined according to actual measurement or expert opinion, and can be adjusted according to actual conditions.
And fourthly, performing risk perception scheduling and Yun Bianduan cooperation, namely entering a scheduling decision stage after completing online fine adjustment and uncertainty quantification, and combining a prediction confidence interval with a flood control and infiltration demand threshold value to realize risk perception intelligent scheduling, wherein the method comprises the following specific steps of:
After on-line fine tuning and uncertainty quantification are completed, the edge node firstly performs one-by-one comparison on the upper limit and the lower limit of the prediction confidence interval of each triangular grid unit, a preset flood depth threshold value and a minimum infiltration threshold value, and the units are divided into three types of high risk, low risk and safety.
For the grid unit judged to be high in risk, the upper limit of the confidence interval exceeds the flood control water depth threshold value, the using proportion suggestion of the regulating reservoir is preferentially enlarged, and a preliminary control instruction of the gate opening degree range (such as 60% -80%) and the available power interval (such as 30% -50%) of the pump set of the unit is generated;
For the units judged to be low in risk, if the lower limit of the confidence interval is lower than the minimum infiltration threshold, making a proposal of preferentially reserving the capacity of the infiltration pond, setting the lower limit of the pump set power to 0% and the upper limit to the acceptable minimum running power (such as 10%), so as to promote natural infiltration;
For the units in the safety interval, generating a default opening degree and a power interval of the gate and the pump unit according to conventional scheduling logic, forming an edge scheduling table by taking the units as units according to the preliminary scheduling result, wherein the edge scheduling table comprises risk levels, gate opening degree suggestion intervals, pump unit power suggestion intervals and corresponding triggering conditions of each unit, and completing second-level generation and local storage at an edge node for global optimization call of a cloud;
After receiving the edge schedule uploaded by all edge nodes and the corresponding confidence interval report, the cloud firstly gathers the scheduling suggestions of each grid unit, builds a global water balance equation set to ensure that the total inflow rate, the total infiltration rate and the total discharge rate are within the allowable range, simultaneously adopts a mode based on multi-scene simulation to simulate the water level response and the pump energy consumption performance after the combined execution of each edge suggestion, screens out the complete scheduling scheme meeting the optimal control and the optimal energy consumption of waterlogging through parallel calculation, defines the use sequence of each regulating and accumulating tank, the opening curve of each gate in the future period (such as every 5 minutes) and the time-period power setting of a pump group, and accompanies the risk evolution prediction and emergency switching logic of each unit.
The operation and maintenance platform superimposes a risk level region spectrum color block diagram in a GIS interface, marks each triangular grid unit with warning red, warning yellow and safety green according to high risk, medium risk and safety three levels respectively, simultaneously displays a prediction confidence interval, a recommended gate opening interval and a pump group power interval of each grid unit in a floating panel mode above each grid unit, and pushes control instructions containing an accurate opening curve and a power setting time sequence of the corresponding grid unit in the global scheduling packet to respective edge nodes through a safety channel;
After the execution of each edge node is finished, the actual water depth and energy consumption data are continuously monitored and recorded, the actual water depth residual error and the energy consumption deviation are calculated and are compared with recommended values in a global scheduling packet, an execution effect report is generated, the report comprises a residual error distribution map, a deviation statistical value and a unit list exceeding a tolerance threshold (such as +/-5%), the report is uploaded to a cloud in real time, after the cloud aggregates the report of each node, the historical performance and the feedback of the current round are compared to form a new performance feedback report, a typical mode aiming at the water depth error, the higher energy consumption or the lower energy consumption in the report is fed back to a federal learning system to serve as an important improvement object in a next round of distributed federal learning collaborative updating and cross-domain verification stage, the modeling capacity and the scheduling accuracy are continuously improved in a plurality of rounds of iterations through a terminal cloud closed-loop mechanism, the dynamic self-adaption to sudden climate and field disturbance is realized in actual operation, and the long-term stable operation of the sponge city high-altitude rain management system is ensured.
The method comprises the steps of carrying out systematic data acquisition and preprocessing on a high-relief area, generating a continuous physical field by fusing drilling penetration coefficient, soil layer thickness and remote sensing vegetation coverage interpolation, mapping penetration pool, regulation pool and pipeline information to form a facility information layer, dividing grid units, constructing a multi-element predictor based on grid unit center elevation and physical properties, issuing edge node local increment fine tuning after cloud training, relying on a scene performance matrix and a federal learning mechanism to dynamically match and synchronize an optimization model, automatically extracting an observation prediction residual error after each storm event for online increment iteration of a neural network, generating a confidence interval by random shielding forward reasoning, dividing risk grades by the edge node according to the confidence interval and a flood control water depth and a minimum infiltration threshold value, and outputting gate opening and pump power advice, thereby remarkably improving the prediction precision and response speed of extreme storm peaks of the high-relief small-flow areas, realizing refined risk perception scheduling, and enhancing storm prevention and control and rainwater resource utilization effects.
Embodiment 2A sponge city high topography rainwater management system, as shown in figure 2, specifically comprises:
The high-topography network dividing module is used for carrying out data acquisition and preprocessing on a high-topography area, combining drilling permeability coefficient, soil layer thickness and remote sensing vegetation coverage interpolation, generating a continuous physical field after abnormality removal, mapping facility and pipeline information to an elevation coordinate system, outputting a facility information layer and a facility attribute list after topology repair, and dividing grid units;
The grid fine adjustment prediction module is used for constructing a predictor based on the attribute of the grid unit center, issuing edge nodes to locally perform incremental fine adjustment after cloud training, dynamically matching the master predictor according to the scene performance matrix, generating a global update package through federal learning aggregation, and synchronously updating local model prediction;
The prediction model adjustment module is used for automatically extracting the latest observation and prediction residual errors by the edge nodes after each storm event, performing incremental iteration on the neural network by using the latest observation and prediction residual errors as training samples, performing forward reasoning by using a random shielding mechanism after fine tuning is finished, and counting the concentrated trend and the fluctuation range of the predicted values to generate a confidence interval of the model;
The management advice generation module is used for dividing the high risk, the low risk and the safety unit by the edge node according to the prediction confidence interval, the flood control water depth threshold value and the minimum infiltration threshold value, and generating the gate opening and pump group power advice.
The above formulas are all formulas for removing dimensions and taking numerical calculation, and specific dimensions can be removed by adopting various means such as standardization, and the like, which are not described in detail herein, wherein the formulas are formulas for acquiring a large amount of data and performing software simulation to obtain the latest real situation, and preset parameters in the formulas are set by those skilled in the art according to the actual situation.
The above embodiments may be implemented in whole or in part by software, hardware, firmware, or any other combination. When implemented in software, the above-described embodiments may be implemented in whole or in part in the form of a computer program product. The computer program product comprises one or more computer instructions or computer programs. When the computer instructions or computer program are loaded or executed on a computer, the processes or functions described in accordance with embodiments of the present application are produced in whole or in part. The computer may be a general purpose computer, a special purpose computer, a network of computers, or other programmable devices. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions may be transmitted from one website site, computer, server, or data center to another website site, computer, server, or data center by wired or wireless means (e.g., infrared, wireless, microwave, etc.). The computer readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. that contains one or more sets of available media. The usable medium may be a magnetic medium (e.g., floppy disk, ATA hard disk, tape), an optical medium (e.g., DVD), or a semiconductor medium. The semiconductor medium may be a solid state ATA hard disk.
It should be understood that, in various embodiments of the present application, the sequence numbers of the foregoing processes do not mean the order of execution, and the order of execution of the processes should be determined by the functions and internal logic thereof, and should not constitute any limitation on the implementation process of the embodiments of the present application.
Those of ordinary skill in the art will appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.
In the several embodiments provided in the present application, it should be understood that the disclosed system, apparatus and method may be implemented in other manners. For example, the apparatus embodiments described above are merely illustrative, e.g., the division of the units is merely a logical function division, and there may be additional divisions when actually implemented, e.g., multiple units or components may be combined or integrated into another system, or some features may be omitted or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be an indirect coupling or communication connection via some interfaces, devices or units, which may be in electrical, mechanical or other forms.
The units described as separate units may or may not be physically separate, and units shown as units may or may not be physical units, may be located in one place, or may be distributed over a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
In addition, each functional unit in the embodiments of the present application may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit.
The foregoing is merely illustrative of the present application, and the present application is not limited thereto, and any person skilled in the art will readily recognize that variations or substitutions are within the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims (9)

1.一种海绵城市高地势雨水管理方法,其特征在于,包括如下步骤:1. A method for managing rainwater in high terrain of a sponge city, characterized by comprising the following steps: 对高地势区域的进行数据采集与预处理,并结合钻孔渗透系数、土层厚度与遥感植被覆盖度插值,剔除异常后生成连续物理场,并将设施与管道信息映射到高程坐标系,拓扑修补后输出设施信息图层及设施属性清单,划分出网格单元;Data collection and preprocessing are performed on high-relief areas. The data are interpolated based on the borehole permeability coefficient, soil thickness and remote sensing vegetation coverage. After removing anomalies, a continuous physical field is generated. The facility and pipeline information is mapped to the elevation coordinate system. After topological repair, the facility information layer and facility attribute list are output and the grid units are divided. 基于网格单元中心的属性构建预测器,并在云端训练后下发边缘节点本地进行增量微调,根据场景性能矩阵动态匹配主用预测器,并通过联邦学习聚合生成全局更新包后同步更新本地模型预测;The predictor is built based on the attributes of the grid unit center, and after training in the cloud, it is sent to the edge node for local incremental fine-tuning. The main predictor is dynamically matched according to the scenario performance matrix, and the local model prediction is synchronously updated after a global update package is generated through federated learning aggregation. 在每次暴雨事件后,边缘节点自动提取最新观测与预测残差,并用作训练样本对神经网络开展增量迭代,微调完成后,使用随机屏蔽机制进行前向推理,统计预测值的集中趋势与波动范围,生成模型的置信区间;After each rainstorm event, the edge nodes automatically extract the latest observations and forecast residuals and use them as training samples to perform incremental iterations on the neural network. After fine-tuning, the random masking mechanism is used for forward reasoning to statistically calculate the central trend and fluctuation range of the forecast values and generate the confidence interval of the model. 边缘节点根据预测置信区间与防洪水深阈值、最小下渗阈值划分高风险、低风险与安全单元,生成闸门开度和泵组功率建议。The edge node divides high-risk, low-risk and safe units according to the prediction confidence interval, flood control depth threshold and minimum infiltration threshold, and generates gate opening and pump group power recommendations. 2.根据权利要求1所述的一种海绵城市高地势雨水管理方法,其特征在于:对高地势区域的进行数据采集与预处理,并结合钻孔渗透系数、土层厚度与遥感植被覆盖度插值,剔除异常后生成连续物理场,具体过程如下:2. A method for managing rainwater in high terrain in a sponge city according to claim 1, characterized in that data is collected and preprocessed in high terrain areas, and interpolated by combining borehole permeability coefficient, soil thickness and remote sensing vegetation coverage, and a continuous physical field is generated after eliminating anomalies. The specific process is as follows: 获取高地势区域的原始高程数据,进行投影变换,统一至同一坐标系,对投影后高程数据施加高斯平滑滤波,去除测量噪声,形成平滑高程信息,再根据地面控制点对平滑后高程信息进行多项式误差校正,输出校正后高程信息;Obtain the original elevation data of the high-relief area, perform projection transformation, unify to the same coordinate system, apply Gaussian smoothing filtering to the projected elevation data, remove measurement noise, form smoothed elevation information, and then perform polynomial error correction on the smoothed elevation information according to the ground control points, and output the corrected elevation information; 对高地势区域进行土壤与植被参数插值,从钻孔采样获得土壤渗透系数数据与土层厚度数据,从遥感分类获得植被覆盖度数据;Soil and vegetation parameters are interpolated for high-relief areas, soil permeability data and soil thickness data are obtained from borehole sampling, and vegetation coverage data are obtained from remote sensing classification; 分别对土壤渗透系数、土层厚度和植被覆盖度数据应用克里金法插值,生成连续的渗透系数场、土层厚度场和植被覆盖度场;Kriging interpolation is applied to soil permeability coefficient, soil thickness and vegetation coverage data to generate continuous permeability coefficient field, soil thickness field and vegetation coverage field. 对插值后的渗透系数场、土层厚度场及植被覆盖度场进行异常值剔除,再次插值填补空洞。The outliers in the interpolated permeability field, soil thickness field and vegetation coverage field are removed, and the gaps are filled by interpolation again. 3.根据权利要求2所述的一种海绵城市高地势雨水管理方法,其特征在于:并将设施与管道信息映射到高程坐标系,拓扑修补后输出设施信息图层及设施属性清单,划分出网格单元,具体过程如下:3. A method for managing rainwater in high terrain in a sponge city according to claim 2, characterized in that: the facility and pipeline information is mapped to the elevation coordinate system, the facility information layer and the facility attribute list are output after topological repair, and the grid units are divided. The specific process is as follows: 从区域设计图中提取设施与管道信息,包括渗透池、调蓄池及主排水管道的空间位置与构造参数;Extract facility and pipeline information from the regional design map, including the spatial location and structural parameters of the infiltration tank, storage tank and main drainage pipeline; 将提取出的设施与管道信息映射到校正后高程信息的坐标系中,生成设施信息图层,对设施信息图层中的断裂或重叠管段进行拓扑修补,输出完整的设施属性清单;Map the extracted facility and pipeline information to the coordinate system of the corrected elevation information, generate a facility information layer, perform topological repair on the broken or overlapping pipe segments in the facility information layer, and output a complete facility attribute list; 以校正后高程信息为基础,在GIS环境中对每个位置计算局部坡度和曲率,调用Delaunay三角剖分算法,以边长函数为约束自动生成一系列三角形网格单元;Based on the corrected elevation information, the local slope and curvature are calculated for each location in the GIS environment, and the Delaunay triangulation algorithm is called to automatically generate a series of triangular mesh units with the edge length function as a constraint; 将每个网格单元的中心高程、渗透系数、土层厚度与植被覆盖度作为初始条件,将实时降雨强度和边界水位观测值作为边界输入,通过有限体积迭代方法,计算各单元在每个时间步内的水深变化和流速分布。The central elevation, permeability, soil thickness and vegetation coverage of each grid cell are taken as initial conditions, and the real-time rainfall intensity and boundary water level observations are taken as boundary inputs. The finite volume iteration method is used to calculate the water depth change and flow velocity distribution of each cell in each time step. 4.根据权利要求3所述的一种海绵城市高地势雨水管理方法,其特征在于:基于网格单元中心的属性构建预测器,并在云端训练后下发边缘节点本地进行增量微调,具体步骤如下:4. A method for managing rainwater in high terrain in a sponge city according to claim 3, characterized in that: a predictor is constructed based on the attributes of the grid unit center, and after training in the cloud, it is sent to the edge node for local incremental fine-tuning, and the specific steps are as follows: 基于每一网格单元的中心高程、渗透系数、土层厚度、植被覆盖度以及历史降雨时序与对应水深观测值属性,构建四类互补的预测器;Four complementary predictors are constructed based on the central elevation, permeability, soil thickness, vegetation coverage, historical rainfall time series and corresponding water depth observation attributes of each grid cell. 预测器包括以物理仿真结果与传感器观测差值为目标,使用决策树回归算法训练仿真误差修正预测器;The predictor includes a simulation error correction predictor trained using a decision tree regression algorithm with the difference between the physical simulation result and the sensor observation as the target; 以物理方程一致性约束和水深观测为双重损失,训练物理信息增强神经网络,使输出拟合观测又满足水动力学守恒;Taking the physical equation consistency constraint and water depth observation as double losses, the physical information enhanced neural network is trained so that the output fits the observation and satisfies the conservation of hydrodynamics. 将网格单元多维属性与历史产流量作为输入,训练基于集成树算法的属性关联预测器,挖掘土壤、植被、地形之间的非线性交互;Taking the multi-dimensional attributes of grid cells and historical runoff as input, the attribute association predictor based on the ensemble tree algorithm is trained to mine the nonlinear interactions between soil, vegetation and terrain. 以降雨与产流的时间依赖特征为输入,训练长短期记忆网络,捕捉雨量突变后的波动趋势;Taking the time-dependent characteristics of rainfall and runoff as input, the long short-term memory network is trained to capture the fluctuation trend after the sudden change of rainfall; 各预测器在云端首次完成批量训练后,分别下发至各流域边缘节点进行本地微调,微调过程采用当次雨水观测与批量训练结果之间的残差作为新样本进行迭代更新。After each predictor completes batch training for the first time in the cloud, it is sent to the edge nodes of each basin for local fine-tuning. The fine-tuning process uses the residual between the current rainfall observation and the batch training results as a new sample for iterative update. 5.根据权利要求4所述的一种海绵城市高地势雨水管理方法,其特征在于:根据场景性能矩阵动态匹配主用预测器,并通过联邦学习聚合生成全局更新包后同步更新本地模型预测,具体过程如下:5. A method for managing rainwater in high terrain in a sponge city according to claim 4, characterized in that: the main predictor is dynamically matched according to the scene performance matrix, and the local model prediction is synchronously updated after a global update package is generated through federated learning aggregation. The specific process is as follows: 根据历史回测库,将短时强降雨、长时弱降雨和空间分布不均典型暴雨事件划分为不同场景类别,并对每一类别下各预测器在峰值到达时间、峰值高度和波动趋势性能指标进行统计与分析,形成场景性能矩阵;According to the historical backtesting database, short-term heavy rainfall, long-term weak rainfall and typical rainstorm events with uneven spatial distribution are divided into different scenario categories, and the peak arrival time, peak height and fluctuation trend performance indicators of each predictor under each category are statistically analyzed to form a scenario performance matrix; 根据当前降雨时序和地形区段划分,自动匹配最契合的场景类别,从矩阵中提取场景类别下表现最优的预测器作为主用模型,若主用模型输出的水深曲线与实时观测数据偏差超出预设阈值,则触发备用模型调用机制,在次优模型之间依次切换,直至恢复预测精度;According to the current rainfall time series and terrain segment division, the most suitable scenario category is automatically matched, and the best predictor under the scenario category is extracted from the matrix as the main model. If the water depth curve output by the main model deviates from the real-time observation data beyond the preset threshold, the backup model calling mechanism is triggered, and the suboptimal models are switched in sequence until the prediction accuracy is restored; 降雨时序包括累积雨量曲线拐点、降雨强度突变时间;The rainfall time series includes the inflection point of the cumulative rainfall curve and the time of sudden change of rainfall intensity; 在各流域边缘节点完成本地微调后,将本地模型参数与初始模型参数之差上传至云端,云端汇总来自各节点的参数差分;After completing local fine-tuning at each watershed edge node, the difference between the local model parameters and the initial model parameters is uploaded to the cloud, which aggregates the parameter differences from each node; 对同一网络层或决策树分支结构,采用多数改进取舍策略;For the same network layer or decision tree branch structure, the majority improvement selection strategy is adopted; 在经过筛选后形成统一的全局更新包,并将全局更新包下发回各边缘节点,各边缘节点在接收更新包后,应用于本地模型进行预测;After screening, a unified global update package is formed and sent back to each edge node. After receiving the update package, each edge node applies the local model for prediction; 全局更新部署完成后,各边缘节点立即开展跨域回测,将新一代模型在原有的典型短时强降雨、长时弱降雨及空间分布不均场景中分别运行,计算均方根误差和最大绝对误差指标,并与历史版本结果进行对比;After the global update deployment is completed, each edge node immediately conducts cross-domain backtesting, running the new generation model in the original typical short-term heavy rainfall, long-term weak rainfall and spatially uneven distribution scenarios, calculating the root mean square error and maximum absolute error indicators, and comparing them with the historical version results; 若存在流域的误差指标出现回升或未达预期,自动触发在线性能反馈报告机制,将流域的误差分布与实时观测残差一并上传至云端。If the error index of a watershed rises or falls short of expectations, the online performance feedback reporting mechanism will be automatically triggered, and the error distribution of the watershed and the real-time observation residuals will be uploaded to the cloud. 6.根据权利要求5所述的一种海绵城市高地势雨水管理方法,其特征在于:在每次暴雨事件后,边缘节点自动提取最新观测与预测残差,并用作训练样本对神经网络开展增量迭代,具体过程如下:6. A method for managing rainwater in high terrain in a sponge city according to claim 5, characterized in that: after each rainstorm event, the edge node automatically extracts the latest observation and prediction residuals and uses them as training samples to perform incremental iterations on the neural network, and the specific process is as follows: 在每轮暴雨事件结束后,自动汇总本地传感器记录的真实水深序列与集成预测器输出的预计水深序列之间的差值,构建出包含时间戳与对应残差的残差数据集,对残差数据集中的最新时段数据,按降序选择最近50条残差作为微调样本;After each round of heavy rain events, the difference between the actual water depth sequence recorded by the local sensor and the expected water depth sequence output by the integrated predictor is automatically summarized to construct a residual data set containing timestamps and corresponding residuals. For the latest period data in the residual data set, the most recent 50 residuals are selected in descending order as fine-tuning samples; 在样本准备完成后,于边缘节点加载物理信息增强神经网络模型,并设置以最小二乘残差和为损失函数的迭代训练流程,采用学习率0.005、批量大小为10的参数进行10轮反向传播优化;After the sample preparation is completed, the physical information enhanced neural network model is loaded on the edge node, and an iterative training process with the least squares residual sum as the loss function is set, and 10 rounds of back propagation optimization are performed with parameters of a learning rate of 0.005 and a batch size of 10; 完成微调后,将更新后的模型参数快照与最新残差统计一并保存,并替换旧版模型,立即用于下一时段的水深预测。After fine-tuning is completed, the updated model parameter snapshot is saved together with the latest residual statistics, and the old model is replaced and immediately used for water depth prediction in the next period. 7.根据权利要求6所述的一种海绵城市高地势雨水管理方法,其特征在于:微调完成后,使用随机屏蔽机制进行前向推理,统计预测值的集中趋势与波动范围,生成模型的置信区间,具体过程如下:7. A method for managing rainwater in high terrain in a sponge city according to claim 6, characterized in that: after fine-tuning is completed, a random shielding mechanism is used for forward reasoning, the central tendency and fluctuation range of the predicted value are statistically calculated, and the confidence interval of the model is generated. The specific process is as follows: 微调结束后,系统在边缘节点对新模型进行自检:使用最新50条观测输入,重新计算预测输出并与真实值对比,若均方根误差较微调前下降大于等于百分之十,则确认本次在线迭代有效,否则退回至上一次有效参数状态并上报云端;After fine-tuning, the system performs a self-check on the new model at the edge node: using the latest 50 observation inputs, recalculates the predicted output and compares it with the true value. If the root mean square error decreases by more than 10% compared to before fine-tuning, the online iteration is confirmed to be valid. Otherwise, it returns to the last valid parameter state and reports to the cloud. 在完成在线微调后,进入不确定性量化环节,对同一时刻的输入数据,在网络中随机屏蔽部分连接节点执行前向推理,模拟模型在不同内部状态下的输出波动;After completing the online fine-tuning, the uncertainty quantification phase begins. For the input data at the same time, some connected nodes are randomly blocked in the network to perform forward reasoning, simulating the output fluctuations of the model under different internal states. 每次推理均生成一组水深预测值,将预测结果汇总后,计算算术平均值与标准差,基于平均值与标准差,推导出置信区间的上下限;Each inference generates a set of water depth prediction values. After summarizing the prediction results, the arithmetic mean and standard deviation are calculated. Based on the mean and standard deviation, the upper and lower limits of the confidence interval are derived. 在得到置信区间后,将置信区间与预设的防洪与下渗阈值进行比对:After obtaining the confidence interval, compare the confidence interval with the preset flood control and infiltration thresholds: 当置信上限接近或超出防洪阈值时,自动预留更多调蓄空间并提前启用应急分流策略;When the confidence upper limit approaches or exceeds the flood control threshold, more storage space is automatically reserved and the emergency diversion strategy is activated in advance; 当置信下限远低于下渗要求时,则降低泵组运行功率,进行自然渗透;When the confidence lower limit is far lower than the infiltration requirement, the operating power of the pump set is reduced to allow natural infiltration; 若置信区间整体落在安全区间内,则执行常规调度。If the confidence interval as a whole falls within the safe interval, regular scheduling is performed. 8.根据权利要求7所述的一种海绵城市高地势雨水管理方法,其特征在于:边缘节点根据预测置信区间与防洪水深阈值、最小下渗阈值划分高风险、低风险与安全单元,生成闸门开度和泵组功率建议,具体过程如下:8. A method for managing rainwater in high terrain of a sponge city according to claim 7, characterized in that: the edge node divides high-risk, low-risk and safe units according to the prediction confidence interval, flood control depth threshold and minimum infiltration threshold, and generates gate opening and pump group power recommendations, and the specific process is as follows: 对各三角网格单元的预测置信区间上下限与预设防洪水深阈值和最小下渗阈值进行逐一比对,将单元分为高风险、低风险和安全三类;The upper and lower limits of the predicted confidence interval of each triangular grid unit are compared one by one with the preset flood control depth threshold and minimum infiltration threshold, and the units are divided into three categories: high risk, low risk and safe; 当置信区间上限超过防洪水深阈值,则判定为高风险的网格单元,优先扩大调蓄池使用比例;When the upper limit of the confidence interval exceeds the flood control depth threshold, the grid cell is judged as high-risk and priority is given to expanding the proportion of storage ponds; 当置信区间下限低于最小下渗阈值,则判定为低风险的网格单元,优先保留渗透池容量建议,进行自然下渗处理;When the lower limit of the confidence interval is lower than the minimum infiltration threshold, the grid cell is judged as low-risk, and the infiltration pool capacity recommendation is prioritized for natural infiltration treatment; 对处于安全区间的网格单元,则按照常规调度逻辑生成闸门与泵组的默认开启度和功率区间。For grid cells in the safety zone, the default opening degree and power range of the gates and pump groups are generated according to the conventional scheduling logic. 9.一种海绵城市高地势雨水管理系统,用于实现权利要求1-8中任一项所述的一种海绵城市高地势雨水管理方法,其特征在于,包括:9. A high-altitude rainwater management system for a sponge city, used to implement a high-altitude rainwater management method for a sponge city as claimed in any one of claims 1 to 8, characterized in that it comprises: 高地势网络划分模块,用于对高地势区域的进行数据采集与预处理,并结合钻孔渗透系数、土层厚度与遥感植被覆盖度插值,剔除异常后生成连续物理场,并将设施与管道信息映射到高程坐标系,拓扑修补后输出设施信息图层及设施属性清单,划分出网格单元;The high-relief network division module is used to collect and preprocess data in high-relief areas, and interpolate the borehole permeability coefficient, soil thickness and remote sensing vegetation coverage, remove anomalies, generate a continuous physical field, and map the facility and pipeline information to the elevation coordinate system. After topological repair, the facility information layer and facility attribute list are output to divide the grid units; 网格微调预测模块,用于基于网格单元中心的属性构建预测器,并在云端训练后下发边缘节点本地进行增量微调,根据场景性能矩阵动态匹配主用预测器,并通过联邦学习聚合生成全局更新包后同步更新本地模型预测;The grid fine-tuning prediction module is used to build a predictor based on the attributes of the grid unit center, and send it to the edge node for local incremental fine-tuning after training in the cloud. It dynamically matches the main predictor according to the scenario performance matrix, and generates a global update package through federated learning aggregation and synchronously updates the local model prediction; 预测模型调整模块,用于在每次暴雨事件后,边缘节点自动提取最新观测与预测残差,并用作训练样本对神经网络开展增量迭代,微调完成后,使用随机屏蔽机制进行前向推理,统计预测值的集中趋势与波动范围,生成模型的置信区间;The prediction model adjustment module is used to automatically extract the latest observations and prediction residuals from the edge nodes after each rainstorm event, and use them as training samples to perform incremental iterations on the neural network. After fine-tuning, the random masking mechanism is used for forward reasoning to statistically calculate the central trend and fluctuation range of the predicted values and generate the confidence interval of the model. 管理建议生成模块,用于边缘节点根据预测置信区间与防洪水深阈值、最小下渗阈值划分高风险、低风险与安全单元,生成闸门开度和泵组功率建议。The management suggestion generation module is used for edge nodes to divide high-risk, low-risk and safe units according to the prediction confidence interval, flood control depth threshold and minimum infiltration threshold, and generate gate opening and pump group power recommendations.
CN202510772801.8A 2025-06-11 2025-06-11 A method and system for managing rainwater in high-altitude areas of sponge cities Active CN120278406B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202510772801.8A CN120278406B (en) 2025-06-11 2025-06-11 A method and system for managing rainwater in high-altitude areas of sponge cities

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202510772801.8A CN120278406B (en) 2025-06-11 2025-06-11 A method and system for managing rainwater in high-altitude areas of sponge cities

Publications (2)

Publication Number Publication Date
CN120278406A true CN120278406A (en) 2025-07-08
CN120278406B CN120278406B (en) 2025-08-12

Family

ID=96241346

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202510772801.8A Active CN120278406B (en) 2025-06-11 2025-06-11 A method and system for managing rainwater in high-altitude areas of sponge cities

Country Status (1)

Country Link
CN (1) CN120278406B (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN120742700A (en) * 2025-09-08 2025-10-03 宁波九荣智控有限公司 Urban pipe network multi-pump station intelligent joint control system based on SWMM dynamic coupling model and self-adaptive optimization method
CN121257865A (en) * 2025-12-03 2026-01-02 陕西省水利电力勘测设计研究院(集团)有限公司 Trans-regional flood risk prediction management system and method based on federal learning
CN121365853A (en) * 2025-12-19 2026-01-20 中国电建集团华东勘测设计研究院有限公司 Methods, devices, and equipment for dynamic water allocation in wetland communities based on multimodal sensing

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
AR109623A1 (en) * 2018-02-16 2019-01-09 Pescarmona Enrique Menotti PROCESS AND SYSTEM OF ANALYSIS AND HYDROLOGICAL MANAGEMENT FOR BASINS
CN111651885A (en) * 2020-06-03 2020-09-11 南昌工程学院 A smart sponge city flood forecasting method
CN114970315A (en) * 2022-04-19 2022-08-30 河海大学 An urban water accumulation simulation and rapid prediction method based on deep learning of spatial dynamic features
CN117273193A (en) * 2023-08-07 2023-12-22 广西电网有限责任公司电力科学研究院 A neural network-based waterlogging risk prediction method and system for distribution equipment
CN117332542A (en) * 2023-11-30 2024-01-02 南京师范大学 A multi-scale adaptive selection urban flood modeling and simulation method
CN118094105A (en) * 2024-03-25 2024-05-28 广州探域科技有限公司 Model increment fine tuning method, system, equipment and medium based on dynamic information
CN119358186A (en) * 2024-12-27 2025-01-24 深圳市旗扬特种装备技术工程有限公司 Sponge city design optimization method and system based on rainfall calculation
CN119648506A (en) * 2024-12-23 2025-03-18 浙江大学建筑设计研究院有限公司 A method and system for managing rainwater in high terrain of a sponge city
CN119669702A (en) * 2025-02-18 2025-03-21 广州赋安数字科技有限公司 Method and device for constructing a multi-source ocean data deep learning analysis and prediction platform

Patent Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
AR109623A1 (en) * 2018-02-16 2019-01-09 Pescarmona Enrique Menotti PROCESS AND SYSTEM OF ANALYSIS AND HYDROLOGICAL MANAGEMENT FOR BASINS
US20190354873A1 (en) * 2018-02-16 2019-11-21 Lucas Pescarmona Analysis system and hydrology management for basin rivers
CN111651885A (en) * 2020-06-03 2020-09-11 南昌工程学院 A smart sponge city flood forecasting method
CN114970315A (en) * 2022-04-19 2022-08-30 河海大学 An urban water accumulation simulation and rapid prediction method based on deep learning of spatial dynamic features
CN117273193A (en) * 2023-08-07 2023-12-22 广西电网有限责任公司电力科学研究院 A neural network-based waterlogging risk prediction method and system for distribution equipment
CN117332542A (en) * 2023-11-30 2024-01-02 南京师范大学 A multi-scale adaptive selection urban flood modeling and simulation method
CN118094105A (en) * 2024-03-25 2024-05-28 广州探域科技有限公司 Model increment fine tuning method, system, equipment and medium based on dynamic information
CN119648506A (en) * 2024-12-23 2025-03-18 浙江大学建筑设计研究院有限公司 A method and system for managing rainwater in high terrain of a sponge city
CN119358186A (en) * 2024-12-27 2025-01-24 深圳市旗扬特种装备技术工程有限公司 Sponge city design optimization method and system based on rainfall calculation
CN119669702A (en) * 2025-02-18 2025-03-21 广州赋安数字科技有限公司 Method and device for constructing a multi-source ocean data deep learning analysis and prediction platform

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
ASIF, M 等: "Geospatial identification of possible rainwater harvesting locations within a high-altitude Gilgit River basin, Pakistan", THEORETICAL AND APPLIED CLIMATOLOGY, 26 July 2024 (2024-07-26) *
徐美;刘舒;孙杨;姚永慧;: "利用洪涝模型进行城市内涝风险快速识别与预警", 武汉大学学报(信息科学版), no. 08, 3 August 2020 (2020-08-03) *

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN120742700A (en) * 2025-09-08 2025-10-03 宁波九荣智控有限公司 Urban pipe network multi-pump station intelligent joint control system based on SWMM dynamic coupling model and self-adaptive optimization method
CN121257865A (en) * 2025-12-03 2026-01-02 陕西省水利电力勘测设计研究院(集团)有限公司 Trans-regional flood risk prediction management system and method based on federal learning
CN121365853A (en) * 2025-12-19 2026-01-20 中国电建集团华东勘测设计研究院有限公司 Methods, devices, and equipment for dynamic water allocation in wetland communities based on multimodal sensing

Also Published As

Publication number Publication date
CN120278406B (en) 2025-08-12

Similar Documents

Publication Publication Date Title
CN120278406B (en) A method and system for managing rainwater in high-altitude areas of sponge cities
CN109978235B (en) Waterlogging water level prediction method based on sample learning
CN103279671B (en) Municipal water disaster Risk Forecast Method based on RBF neural-cloud model
CN120026669A (en) Real-time monitoring method, device and system for foundation pit deformation based on BIM technology
CN120450214B (en) Urban waterlogging risk management method and system based on multi-model coupling
CN113570132A (en) Wind power prediction method for space-time meteorological feature extraction and deep learning
CN120218183A (en) Dynamic optimization and adaptive learning method and storage medium for large water conservancy system models
CN120410451A (en) A natural resource engineering data intelligent management method
CN120409192A (en) Numerical simulation method and system for urban waterlogging retreat process based on physics-data fusion
CN120526540B (en) Intelligent risk perception and multi-stage progressive early warning system for flood storage and detention areas
CN120952318A (en) Artificial Intelligence-Based Water Resource Predictive Analysis Method
CN118780499B (en) A multi-scale and multi-scenario method for predicting and evaluating ecosystem service trade-offs and synergies
CN119989893A (en) A karst watershed runoff prediction method based on reinforcement learning and deep learning
CN120634160B (en) Intelligent rainwater regulation and storage method and system
CN119179998B (en) BIM dynamic model construction method for pumped storage engineering
CN119692740B (en) Comprehensive intelligent harbour water resource management and control method
CN120450452B (en) Hydrological weather coupling forecast early warning system
CN121032334A (en) A Smart Prediction System for Groundwater Storage Variables Based on Multi-Source Data Fusion and Parameter Optimization
CN121352499A (en) A Method for Cost Risk Prediction and Control of Municipal Drainage Projects Based on Big Data and Artificial Intelligence
CN121279075A (en) Method and system for constructing and correcting equivalent rain and flood model of cell in real time
CN121189792A (en) Water supply and drainage monitoring management method based on intelligent early warning
CN121365853A (en) Methods, devices, and equipment for dynamic water allocation in wetland communities based on multimodal sensing
CN120671452A (en) Drainage basin hydrologic prediction method based on digital twin
CN121352101A (en) Dynamic forecasting method for ice water surface elevation along water return area of reservoir in flood season
CN120448398A (en) A dynamic surface data updating method based on numerical weather forecast model

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant