CN112084229A

CN112084229A - Method and device for identifying abnormal gas consumption behaviors of town gas users

Info

Publication number: CN112084229A
Application number: CN202010730983.XA
Authority: CN
Inventors: 王勋; 陈婷婷; 王倩微; 乔佳; 孙俊芳; 丁斌; 柴家凤; 石书强
Original assignee: Beijing Gas Group Co Ltd
Current assignee: Beijing Gas Group Co Ltd
Priority date: 2020-07-27
Filing date: 2020-07-27
Publication date: 2020-12-15

Abstract

The invention provides a method and a device for identifying abnormal gas consumption behaviors of town gas users, wherein the method comprises the following steps: acquiring gas consumption data of urban gas public service users, and performing data cleaning on the gas consumption data of the users to obtain data to be analyzed; carrying out feature extraction on data to be analyzed to obtain feature data; analyzing the characteristic data by a single user to determine abnormal users; and carrying out same-class user outlier detection on the characteristic data to determine abnormal users. The characteristics of the gas consumption data of the gas users are extracted, on one hand, a single user is analyzed to obtain the variation of the characteristics, on the other hand, a clustering method is adopted to cluster the similar users to find abnormal users, and the gas consumption condition of the abnormal users is further analyzed through comparison of the two aspects. In addition, relevant characteristics of the users are abstracted through the gas consumption data of the users, and key abnormal users are found out for follow-up supervision by combining curve fitting of single users and cluster analysis of similar users.

Description

Method and device for identifying abnormal gas consumption behaviors of town gas users

Technical Field

The invention relates to the field of computers, in particular to a method and a device for identifying abnormal gas consumption behaviors of town gas users.

Background

Some users with abnormal gas consumption exist in town gas users, the reasons may be decoration, shutdown, equipment replacement, operation type change, gas theft and the like, the abnormal users have management difficulty and safety risks, particularly the gas theft users cause enterprise asset loss, belong to illegal behaviors, and the users are identified and checked at the same time.

The users have inevitable gas consumption data abnormity, the abnormal data mining mainly uses deviation detection, and in a mathematical sense, the deviation refers to abnormal examples in classification, special cases which do not meet rules, or values of observation results which are inconsistent with model predicted values and change along with time, and the like. The basic target of deviation detection is to find meaningful differences between observed results and reference values, and the main deviation technologies include clustering, sequence abnormality, nearest neighbor method, multidimensional data analysis and the like. In addition to identifying abnormal data, abnormal data mining also aims to find implicit models among the abnormal data for intelligent analysis and prediction.

The problem of anomaly detection is widely studied in the field of statistics, and users usually model data points with a certain statistical distribution and then determine whether anomalies occur according to the distribution of the points by using an assumed model. A large number of anomaly testing methods for different distributions have been developed, which are respectively suitable for different situations: firstly, data distribution condition; whether the data distribution parameters are known; number of abnormal data; and fourthly, abnormal data types (higher or lower than the general sampling values). Representative of this aspect are the single point diagnostics based on the "mean shift" model proposed by Mikey, Dunn & Clark in 1967, the group diagnostics proposed by Gentleman & Wilk in 1970, the statistics of k outliers of a single sample proposed by Tietjen & Moore in 1972, the single sample multiple outlier detection algorithm ESD (Generalized outlier quantized device) proposed by Rosner in 1989, and the method of GESR (Generalized outlier quantized device 2dual) for regression analysis in 1991 in Paul & Fung improving the subjectivity of the selection of the ESD method parameter k. In recent years, the outlier detection method of multiple samples has been developed to a certain extent, and most of them find a sample set of "abnormal users" first, and the sample set is also compared and analyzed by other users. And such sample sets are lacking or difficult to find in town gas enterprises.

The analysis of abnormal gas consumption is mainly based on gas consumption data of users and basic data of the users, including relevant parameters such as equipment, business state, scale, weather condition and the like, but because the relevant parameters change constantly and accurate information is difficult to obtain, the analysis of whether the gas consumption state of the users is normal is not facilitated.

Disclosure of Invention

The present invention aims to provide a method and a device for identifying abnormal gas consumption behaviors of town gas users, which overcome the above problems or at least partially solve the above problems.

In order to achieve the purpose, the technical scheme of the invention is realized as follows:

one aspect of the invention provides a method for identifying abnormal gas consumption behaviors of town gas users, which comprises the following steps: acquiring gas consumption data of urban gas public service users, and performing data cleaning on the gas consumption data of the users to obtain data to be analyzed; carrying out feature extraction on data to be analyzed to obtain feature data; analyzing the characteristic data by a single user to determine abnormal users; and carrying out same-class user outlier detection on the characteristic data to determine abnormal users.

Wherein, the characteristic extraction is carried out on the data to be analyzed, and the characteristic data obtaining comprises the following steps: acquiring a data variation index according to data to be analyzed, wherein the data variation index comprises: range, interquartile distance, mean deviation, standard deviation, variation coefficient, standard deviation, aggregate statistical characteristics of each order of autocorrelation coefficient, approximate entropy, time sequence data complexity, spectral statistics of Fourier transform, kurtosis and maximum fluctuation coefficient.

Wherein, the single user analysis of the characteristic data to determine the abnormal user comprises: and analyzing a single user according to the range, the interquartile distance, the mean deviation, the standard deviation, the variation coefficient, the standard deviation, the aggregation statistical characteristics of all-order autocorrelation coefficients, the approximate entropy, the time sequence data complexity, the spectrum statistics of Fourier transform, the kurtosis and the maximum fluctuation coefficient to determine the abnormal user.

Wherein, the single user analysis of the characteristic data to determine the abnormal user comprises: judging abnormal points of the gas use time sequence by utilizing the box-shaped graph, and determining users corresponding to the abnormal points as abnormal users; and/or fitting a gas consumption curve by adopting time series, regression analysis, neural network and/or wavelet analysis, comparing the gas consumption curve with the gas consumption data of the actual user, and determining the user exceeding the preset threshold value of the gas consumption curve as an abnormal user when the gas consumption data of the actual user exceeds the preset threshold value of the gas consumption curve.

The method for detecting the outliers of the similar users of the feature data and determining the abnormal users comprises the following steps: and performing outlier detection on the same type of users by utilizing the correlation coefficient of the time sequence data and the average temperature, the maximum fluctuation coefficient of the time sequence data, the standard deviation of the time sequence data, the kurtosis of the time sequence data, the aggregation statistical characteristics of the autocorrelation coefficients of all orders of the time sequence data, the approximate entropy of the time sequence data, the complexity of the time sequence data, the mass center of Fourier transform and the variance of the Fourier transform, and determining abnormal users.

The method for detecting the outliers of the similar users of the feature data and determining the abnormal users comprises the following steps: and clustering the gas users in the characteristic data into multiple classes by means of Mean-shift, GMM, hierarchical clustering and/or Birch, performing outlier detection to obtain a clustering result, and determining the outlier users in the clustering result as abnormal users.

Wherein, the method further comprises: and analyzing the abnormal users obtained by analyzing the single user and the abnormal users determined by detecting the outliers of the similar users to determine key abnormal users.

In another aspect, the present invention provides a device for identifying abnormal gas consumption behavior of town gas users, comprising: the acquisition module is used for acquiring gas consumption data of urban gas public service users, and performing data cleaning on the gas consumption data of the users to obtain data to be analyzed; the characteristic extraction module is used for extracting characteristics of the data to be analyzed to obtain characteristic data; the first determining module is used for analyzing the characteristic data by a single user and determining abnormal users; and the second determining module is used for detecting the outliers of the similar users of the characteristic data and determining abnormal users.

The feature extraction module performs feature extraction on data to be analyzed in the following mode to obtain feature data: the feature extraction module is specifically configured to obtain a data variation index according to data to be analyzed, where the data variation index includes: range, interquartile distance, mean deviation, standard deviation, variation coefficient, standard deviation, aggregate statistical characteristics of each order of autocorrelation coefficient, approximate entropy, time sequence data complexity, spectral statistics of Fourier transform, kurtosis and maximum fluctuation coefficient.

The first determining module is used for determining abnormal users by analyzing the characteristic data by a single user in the following mode: the first determining module is specifically used for analyzing a single user according to the range, the interquartile distance, the mean deviation, the standard deviation, the variation coefficient, the standard deviation, the aggregation statistical characteristics of each order of autocorrelation coefficients, the approximate entropy, the time sequence data complexity, the spectrum statistics of Fourier transform, the kurtosis and the maximum fluctuation coefficient, and determining the abnormal user.

The first determining module is used for determining abnormal users by analyzing the characteristic data by a single user in the following mode: the first determining module is specifically used for judging abnormal points of the gas consumption time sequence by utilizing the box-shaped graph and determining users corresponding to the abnormal points as abnormal users; and/or fitting a gas consumption curve by adopting time series, regression analysis, neural network and/or wavelet analysis, comparing the gas consumption curve with the gas consumption data of the actual user, and determining the user exceeding the preset threshold value of the gas consumption curve as an abnormal user when the gas consumption data of the actual user exceeds the preset threshold value of the gas consumption curve.

The second determining module performs similar user outlier detection on the feature data to determine abnormal users: the second determining module is specifically configured to perform outlier detection on similar users by using a correlation coefficient of the time series data and the average temperature, a maximum fluctuation coefficient of the time series data, a standard deviation of the time series data, a kurtosis of the time series data, an aggregation statistical characteristic of each order of autocorrelation coefficients of the time series data, an approximate entropy of the time series data, a complexity of the time series data, a centroid of fourier transform, and a variance of the fourier transform, and determine an abnormal user.

The second determining module performs similar user outlier detection on the feature data in the following mode to determine abnormal users: and the second determining module is specifically used for clustering the gas users in the characteristic data into multiple classes by means of Mean-shift, GMM, hierarchical clustering and/or Birch, performing outlier detection to obtain a clustering result, and determining the outlier users in the clustering result as abnormal users.

Wherein, the device still includes: and the third determining module is used for analyzing the abnormal users obtained by analyzing the single user and the abnormal users determined by the outlier detection of the similar users to determine key abnormal users.

Therefore, through the method and the device for identifying the abnormal gas consumption behavior of the town gas users, provided by the invention, the gas consumption data of the gas users are subjected to feature extraction, on one hand, a single user is analyzed to obtain the variation of the features, on the other hand, a clustering method is adopted to cluster similar users to find abnormal users, and the gas consumption condition of the abnormal users is further analyzed through comparison of the two aspects.

In addition, relevant characteristics of the users are abstracted through the gas consumption data of the users, and key abnormal users are found out for follow-up supervision by combining curve fitting of single users and cluster analysis of similar users.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on the drawings without creative efforts.

Fig. 1 is a flowchart of a method for identifying abnormal gas consumption behaviors of town gas users according to an embodiment of the present invention;

FIG. 2 is a fitted user usage profile provided by an embodiment of the present invention;

FIG. 3 is a diagram illustrating clustering results of different models according to an embodiment of the present invention;

fig. 4 is a schematic structural diagram of a device for identifying abnormal gas consumption behaviors of town gas users according to an embodiment of the present invention.

Detailed Description

Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

Fig. 1 shows a flowchart of a method for identifying an abnormal gas consumption behavior of a town gas user according to an embodiment of the present invention, and referring to fig. 1, the method for identifying an abnormal gas consumption behavior of a town gas user according to an embodiment of the present invention includes:

and S1, acquiring the gas consumption data of the urban gas public service users, and performing data cleaning on the gas consumption data of the users to obtain the data to be analyzed.

Specifically, half-hour gas usage data of the user is collected, hour, day, week and month data are accumulated, and corresponding temperatures at each moment of the date, including the highest, lowest and average temperatures, are collected and screened.

In order to reduce the influence caused by the size of the data level, the invention can standardize the data of different users.

In addition, when the daily data are analyzed, the daily data can be processed to obtain an uneven coefficient, and the calculation mode is the ratio of the daily gas consumption to the average daily gas consumption in the month and represents the change condition of the daily gas consumption.

And S2, performing feature extraction on the data to be analyzed to obtain feature data.

As an optional implementation manner of the embodiment of the present invention, performing feature extraction on data to be analyzed to obtain feature data includes: acquiring a data variation index according to data to be analyzed, wherein the data variation index comprises: range, interquartile distance, mean deviation, standard deviation, variation coefficient, standard deviation, aggregate statistical characteristics of each order of autocorrelation coefficient, approximate entropy, time sequence data complexity, spectral statistics of Fourier transform, kurtosis and maximum fluctuation coefficient.

Specifically, the invention further understands the overall characteristics of the data and the distribution of the data by analyzing the divergence condition of the statistical data, namely the data variation index, and further discovers the abnormal point data in the data by the data variation index. The data variation indexes adopted by the method have range difference, interquartile distance, mean deviation, standard deviation, variation coefficient and the like, and the large value of the variation indexes indicates large variation and wide spread; a small value indicates a small dispersion and is denser.

Since the data involved in the invention is label-free data, for this reason, an outlier detection algorithm based on statistics is the main reference for identification detection.

Besides basic indexes such as standard deviation, coefficient of variation and the like, the invention also adopts a plurality of statistical indexes as characteristic indexes, including the following seven indexes:

(1) standard deviation (standard _ deviation)

The standard deviation is the most common statistical index, and can reflect the fluctuation condition of a sequence data, and the formula is as follows:

wherein the value X₁，X₂，X₃，......X_N(all are real numbers), the mean (arithmetic mean) is μ, and the standard deviation is σ.

(2) Aggregate statistical characteristics of autocorrelation coefficients of respective orders (agg _ autocorrelation)

The statistical index reflects the aggregation (variance, mean) statistical characteristics among the step difference values of the time series data, and the formula is

Where x is the name of the aggregation function and n is the maximum difference order.

(3) Approximate entropy (approximate _ entropy)

Approximate entropy is a non-linear dynamic parameter for quantifying regularity, unpredictability and volatility of time series fluctuation, and represents the complexity of a time series by a non-negative number, reflects the possibility of new information in the time series, and the more complicated time series correspond to larger approximate entropy. It calculates the ratio between adjacent entropy values, which is a relative quantity. It is defined as:

ApEn＝Φ^m(r)-Φ^m+1(r)

x represents a vector formed by time sequence data, N represents the dimension of the time sequence data, m is an integer and represents the length of a comparison vector, and r is a real number and represents a metric value of 'similarity'.

(4) Time series data complexity (cid _ ce)

This index is used to assess the complexity of the time series, with more complex series having more troughs and peaks.

(5) Fourier transform spectral statistics (fft _ aggregated)

The indexes include spectral centroid (centroid), skewness (skew), kurtosis (kurtosis) and the like after Fourier transform.

(6) Kurtosis

Kurtosis is a statistic used to observe data distribution, represents the characteristic number of the probability density distribution curve with the peak value at the mean value, and is the ratio of the fourth-order central moment of a random variable to the square of the variance.

(7) Maximum fluctuation coefficient (max/ave)

The formula is as follows:

the invention selects the gas consumption as the characteristic of detecting the abnormal gas consumption of the user, and applies two strategies, namely a transaction aggregation strategy and a strategy based on rules to complete the behavior analysis facing the homogeneity, and extracts the characteristic variable on the historical transaction data. For a transaction aggregation policy, it can be broken down into three parts: aggregation period, transaction behavior metrics, and aggregation statistics. The aggregation period is the time range and granularity of transactions that need to be aggregated. For example, statistics for one year are summarized in units of one day, or are summarized in units of hours. The transaction behavior metric is the gas consumption of each user, and other quantities that can be directly observed can be used for measurement, such as the gas consumption of each user. This variable is selected because the precise data that the present invention can directly obtain is the gas usage by each user. Because the aggregation characteristics can be selected in a limited way, the invention mainly mines the commonalities existing in each time sequence according to the aggregation statistical information. The aggregated statistical information is statistical information applied to a numerical sequence extracted by gas usage time-series data according to the present invention. Common statistics are counts, mean and standard deviation. In order to fully extract features from the time series data, the invention can use an open-source toolkit tsfresh of Python, which is a module for feature mining of the time series data of Python, and the extracted features can be used for describing or clustering the time series based on the extracted features. Furthermore, they can also be used to build models that perform classification/regression tasks on a temporal sequence. Generally, these properties provide new insights into the time series and their dynamics. The invention extracts a plurality of statistical data, such as standard deviation, kurtosis, aggregation statistical characteristics of autocorrelation coefficients of various orders and the like.

The invention selects the gas consumption provided by the user as the characteristic for detecting the gas permeation of the user, selects two aggregation periods of the day degree and the month degree, and abstracts out the related characteristics as shown in the following table, wherein the related characteristics comprise 9 characteristics of 2 dimensions of data of the day degree and the month degree. In addition, these characteristic indexes can be extracted for both long-term and short-term times. The characteristics and their meanings are illustrated in table 1 below.

TABLE 1

And S3, analyzing the characteristic data by a single user to determine abnormal users.

As an optional implementation manner of the embodiment of the present invention, the analyzing the feature data by a single user, and determining an abnormal user includes: and analyzing a single user according to the range, the interquartile distance, the mean deviation, the standard deviation, the variation coefficient, the standard deviation, the aggregation statistical characteristics of all-order autocorrelation coefficients, the approximate entropy, the time sequence data complexity, the spectrum statistics of Fourier transform, the kurtosis and the maximum fluctuation coefficient to determine the abnormal user.

As an optional implementation manner of the embodiment of the present invention, the analyzing the feature data by a single user, and determining an abnormal user includes: judging abnormal points of the gas use time sequence by utilizing the box-shaped graph, and determining users corresponding to the abnormal points as abnormal users; and/or fitting a gas consumption curve by adopting time series, regression analysis, neural network and/or wavelet analysis, comparing the gas consumption curve with the gas consumption data of the actual user, and determining the user exceeding the preset threshold value of the gas consumption curve as an abnormal user when the gas consumption data of the actual user exceeds the preset threshold value of the gas consumption curve.

Specifically, for the gas consumption data of a single user, a box chart is adopted to identify abnormal and constant values, wherein the abnormal values are defined as values smaller than Q1-1.5IQR or larger than Q3+1.5 IQR. The gas usage curve of the user is fitted by methods such as time series, regression analysis, exponential smoothing, autoregressive moving average, wavelet analysis and the like, as shown in fig. 2: and marking users exceeding a certain error range as suspicious abnormal users. And analyzing divergence conditions of the statistical data, namely data variation indexes to further understand the overall characteristics of the data and understand the distribution conditions of the data, and further discovering abnormal point data in the data through the data variation indexes.

And S4, carrying out same-class user outlier detection on the feature data, and determining abnormal users.

As an optional implementation manner of the embodiment of the present invention, performing outlier detection on feature data of the same type of user, and determining an abnormal user includes: and performing outlier detection on the same type of users by utilizing the correlation coefficient of the time sequence data and the average temperature, the maximum fluctuation coefficient of the time sequence data, the standard deviation of the time sequence data, the kurtosis of the time sequence data, the aggregation statistical characteristics of the autocorrelation coefficients of all orders of the time sequence data, the approximate entropy of the time sequence data, the complexity of the time sequence data, the mass center of Fourier transform and the variance of the Fourier transform, and determining abnormal users.

As an optional implementation manner of the embodiment of the present invention, performing outlier detection on feature data of the same type of user, and determining an abnormal user includes: and clustering the gas users in the characteristic data into multiple classes by means of Mean-shift, GMM, hierarchical clustering and/or Birch, performing outlier detection to obtain a clustering result, and determining the outlier users in the clustering result as abnormal users.

Of course, the invention can also apply but not limit the following algorithm for the same kind of user outlier detection: waveCluster, ROCK, BIRCH, CURE, K-protocols, DENCLUE, Optigrid, CLIQUE, DBSCAN, CLARANS.

Specifically, aiming at the cleaned data set, in order to ensure the accuracy of the result, four clustering algorithms including Mean-shift, GMM, hierarchical clustering and Birch can be adopted to cluster the gas users into a plurality of classes to perform outlier detection results, and finally the data of which user is possibly abnormal is obtained according to the clustering results. According to the principle of ' good people's abundance ', one of the classes of data with fewer samples is regarded as a user with possible air stealing behavior.

In the selection of input variables, 9 characteristics of 2 dimensions including daily and monthly data and long-term and short-term trends of the characteristics are included, and the total number of the characteristics is 36.

The following analysis is performed using a boiler room as an example:

the final clustering results for the boiler users are shown in table 2.

TABLE 2

The 838 boiler gas samples are divided into 3 classes by the four clustering algorithms, and due to the temporary lack of abnormal labeled data, the method can take the principle of 'most boiler gas data are normal', the classes with more proportion are taken as positive classes, and the other classes are negative classes, namely abnormal data. From the results, see table 3: no.755, namely use _ NO: 01217030014214u9172013656 is identified as outliers by each of the four models and is in one class alone. Further, the boiler points judged as another type of abnormality by 3 or more models are:

TABLE 3

Serial number	Number of abnormal model to be discriminated	use_NO
			No.245	3	05113092117110u3200092079
No.246	3	05113092117217u3200092025
			No.247	3	05115090072413u3225013589
No.259	3	02317059745429u3262760883
			No.575	3	05115080040813u6769013324
No.577	3	05115090091315u6769013325
			No.609	3	01316050117627u7313590173
No.611	3	01316050118025u7313590163
			No.612	3	01316050118127u7313590113
No.614	3	01316050118322u7313590120
			No.615	3	05116091004118u7313590111

And manually checking the suspicious gas points according to the model voting result. Wherein, a schematic diagram of the clustering results of different models is shown in fig. 3.

The following analysis is performed by taking the catering user as an example:

the final clustering results for the dining users are shown in table 4.

TABLE 4

1156 catering gas samples are divided into 3 classes by four clustering algorithms, and due to the temporary lack of abnormal labeled data, the method takes the principle that most catering gas data are normal, and takes the classes with more proportion as positive classes, and the other classes are negative classes, namely abnormal data. The clustering results of the four models show obvious consistency, and 5 identified 10 sample points are abnormal for 3 models. Based on all model results, see table 5, the dining anomaly conditions are as follows:

TABLE 5

In addition, as for the result of the clustering algorithm, after the relevant department performs targeted survey and verification according to the clustering result, the effectiveness of the model needs to be verified according to the result, and the model is further screened or weighted so as to improve the accuracy of the clustering result.

As an optional implementation manner of the embodiment of the present invention, the method for identifying abnormal gas usage behavior of a town gas user provided by the embodiment of the present invention further includes: and analyzing the abnormal users obtained by analyzing the single user and the abnormal users determined by detecting the outliers of the similar users to determine key abnormal users.

Specifically, the abnormal statistics and clustering results of a single user are integrated, and the abnormal conditions of the user are comprehensively analyzed from two angles.

Therefore, the method for identifying the abnormal gas consumption behaviors of the town gas users, provided by the embodiment of the invention, aims at the problem that the conventional abnormal gas user identification needs data information such as basic information, business information and the like of the user, and the information changes rapidly and inaccurately.

In addition, the gas consumption data of the user can be analyzed, and the operability is strong; and by comparing the characteristic analysis of a single user with the clustering analysis of multiple users of the same type, abnormal users can be more easily identified.

Fig. 4 is a schematic structural diagram of an apparatus for identifying abnormal gas use behaviors of town gas users according to an embodiment of the present invention, in which the apparatus for identifying abnormal gas use behaviors of town gas users applies the above method, and the following description is made only for the structure of the apparatus for identifying abnormal gas use behaviors of town gas users, and for other inexhaustible matters, please refer to the related description in the above method for identifying abnormal gas use behaviors of town gas users, with reference to fig. 4, the apparatus for identifying abnormal gas use behaviors of town gas users according to an embodiment of the present invention includes:

the acquisition module is used for acquiring gas consumption data of urban gas public service users, and performing data cleaning on the gas consumption data of the users to obtain data to be analyzed;

the characteristic extraction module is used for extracting characteristics of the data to be analyzed to obtain characteristic data;

the first determining module is used for analyzing the characteristic data by a single user and determining abnormal users;

and the second determining module is used for detecting the outliers of the similar users of the characteristic data and determining abnormal users.

As an optional implementation manner of the embodiment of the present invention, the feature extraction module performs feature extraction on data to be analyzed in the following manner to obtain feature data: the feature extraction module is specifically configured to obtain a data variation index according to data to be analyzed, where the data variation index includes: range, interquartile distance, mean deviation, standard deviation, variation coefficient, standard deviation, aggregate statistical characteristics of each order of autocorrelation coefficient, approximate entropy, time sequence data complexity, spectral statistics of Fourier transform, kurtosis and maximum fluctuation coefficient.

As an optional implementation manner of the embodiment of the present invention, the first determining module determines the abnormal user by performing single user analysis on the feature data in the following manner: the first determining module is specifically used for analyzing a single user according to the range, the interquartile distance, the mean deviation, the standard deviation, the variation coefficient, the standard deviation, the aggregation statistical characteristics of each order of autocorrelation coefficients, the approximate entropy, the time sequence data complexity, the spectrum statistics of Fourier transform, the kurtosis and the maximum fluctuation coefficient, and determining the abnormal user.

An optional implementation manner of the embodiment of the present invention is characterized in that the first determining module determines the abnormal user through a single user analysis performed on the feature data in the following manner: the first determining module is specifically used for judging abnormal points of the gas consumption time sequence by utilizing the box-shaped graph and determining users corresponding to the abnormal points as abnormal users; and/or fitting a gas consumption curve by adopting time series, regression analysis, neural network and/or wavelet analysis, comparing the gas consumption curve with the gas consumption data of the actual user, and determining the user exceeding the preset threshold value of the gas consumption curve as an abnormal user when the gas consumption data of the actual user exceeds the preset threshold value of the gas consumption curve.

As an optional implementation manner of the embodiment of the present invention, the second determining module performs the same-class user outlier detection on the feature data, and determines an abnormal user: the second determining module is specifically configured to perform outlier detection on similar users by using a correlation coefficient of the time series data and the average temperature, a maximum fluctuation coefficient of the time series data, a standard deviation of the time series data, a kurtosis of the time series data, an aggregation statistical characteristic of each order of autocorrelation coefficients of the time series data, an approximate entropy of the time series data, a complexity of the time series data, a centroid of fourier transform, and a variance of the fourier transform, and determine an abnormal user.

As an optional implementation manner of the embodiment of the present invention, the second determining module performs the same-class user outlier detection on the feature data in the following manner, and determines an abnormal user: and the second determining module is specifically used for clustering the gas users in the characteristic data into multiple classes by means of Mean-shift, GMM, hierarchical clustering and/or Birch, performing outlier detection to obtain a clustering result, and determining the outlier users in the clustering result as abnormal users.

As an optional implementation manner of the embodiment of the present invention, the apparatus for identifying abnormal gas usage behavior of a town gas user provided in the embodiment of the present invention further includes: and the third determining module is used for analyzing the abnormal users obtained by analyzing the single user and the abnormal users determined by the outlier detection of the similar users to determine key abnormal users.

Therefore, the device for identifying the abnormal gas consumption behaviors of the town gas users, provided by the embodiment of the invention, aims at the problem that the conventional abnormal gas user identification needs data information such as basic information, business information and the like of the user, and the information changes rapidly and inaccurately.

As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.

The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

In a typical configuration, a computing device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.

The memory may include forms of volatile memory in a computer readable medium, Random Access Memory (RAM) and/or non-volatile memory, such as Read Only Memory (ROM) or flash memory (flash RAM). The memory is an example of a computer-readable medium.

Computer-readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), Read Only Memory (ROM), Electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), Digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information that can be accessed by a computing device. As defined herein, a computer readable medium does not include a transitory computer readable medium such as a modulated data signal and a carrier wave.

The above are merely examples of the present application and are not intended to limit the present application. Various modifications and changes may occur to those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A method for identifying abnormal gas consumption behaviors of town gas users is characterized by comprising the following steps:

acquiring gas consumption data of urban gas public service users, and performing data cleaning on the gas consumption data of the users to obtain data to be analyzed;

performing feature extraction on the data to be analyzed to obtain feature data;

analyzing the characteristic data by a single user to determine abnormal users;

and carrying out same-class user outlier detection on the characteristic data, and determining abnormal users.

2. The method according to claim 1, wherein the extracting the features of the data to be analyzed to obtain feature data comprises:

obtaining a data variation index according to the data to be analyzed, wherein the data variation index comprises: range, interquartile distance, mean deviation, standard deviation, variation coefficient, standard deviation, aggregate statistical characteristics of each order of autocorrelation coefficient, approximate entropy, time sequence data complexity, spectral statistics of Fourier transform, kurtosis and maximum fluctuation coefficient.

3. The method of claim 2, wherein the single user analysis of the feature data to determine anomalous users comprises:

and analyzing the single user according to the range, the interquartile distance, the mean deviation, the standard deviation, the variation coefficient, the standard deviation, the aggregation statistical characteristics of the autocorrelation coefficients of all orders, the approximate entropy, the time sequence data complexity, the spectrum statistic of the Fourier transform, the kurtosis and the maximum fluctuation coefficient to determine an abnormal user.

4. The method of any one of claims 1 to 3, wherein the single user analysis of the feature data to determine anomalous users comprises:

judging abnormal points of the gas use time sequence by utilizing the box-shaped graph, and determining users corresponding to the abnormal points as abnormal users; and/or

And fitting a gas usage curve by adopting a time sequence, regression analysis, a neural network and/or wavelet analysis, comparing the gas usage curve with the gas usage data of the actual user, and determining the user exceeding the preset threshold of the gas usage curve as an abnormal user when the gas usage data of the actual user exceeds the preset threshold of the gas usage curve.

5. The method according to claim 2, wherein the performing homogeneous user outlier detection on the feature data and determining abnormal users comprises:

and performing outlier detection on the similar users by utilizing the correlation coefficient of the time sequence data and the average temperature, the maximum fluctuation coefficient of the time sequence data, the standard deviation of the time sequence data, the kurtosis of the time sequence data, the aggregation statistical characteristics of the autocorrelation coefficients of all orders of the time sequence data, the approximate entropy of the time sequence data, the complexity of the time sequence data, the mass center of Fourier transform and the variance of the Fourier transform, and determining abnormal users.

6. A device for identifying abnormal gas use behaviors of town gas users is characterized by comprising:

the acquisition module is used for acquiring gas data for users of urban gas public service, and performing data cleaning on the gas data for the users to obtain data to be analyzed;

the characteristic extraction module is used for extracting the characteristics of the data to be analyzed to obtain characteristic data;

the first determining module is used for analyzing the single user of the characteristic data and determining abnormal users;

and the second determining module is used for carrying out similar user outlier detection on the characteristic data and determining abnormal users.

7. The device of claim 6, wherein the feature extraction module performs feature extraction on the data to be analyzed to obtain feature data by:

the feature extraction module is specifically configured to obtain a data variation index according to the data to be analyzed, where the data variation index includes: range, interquartile distance, mean deviation, standard deviation, variation coefficient, standard deviation, aggregate statistical characteristics of each order of autocorrelation coefficient, approximate entropy, time sequence data complexity, spectral statistics of Fourier transform, kurtosis and maximum fluctuation coefficient.

8. The apparatus of claim 7, wherein the first determining module determines the abnormal user by a single user analysis of the feature data by:

the first determining module is specifically configured to analyze the single user according to the range, the interquartile range, the mean square error, the standard deviation, the variation coefficient, the standard deviation, the aggregate statistical characteristic of each order of autocorrelation coefficient, the approximate entropy, the time series data complexity, the spectrum statistics of fourier transform, the kurtosis, and the maximum fluctuation coefficient, and determine an abnormal user.

9. The apparatus of any one of claims 6 to 8, wherein the first determining module determines the abnormal user by performing a single user analysis on the feature data by:

the first determining module is specifically used for judging abnormal points of the gas consumption time sequence by utilizing the box-shaped graph and determining users corresponding to the abnormal points as abnormal users; and/or fitting a gas usage curve by adopting time series, regression analysis, neural network and/or wavelet analysis, comparing the gas usage curve with gas usage data of an actual user, and determining the user exceeding the preset threshold of the gas usage curve as an abnormal user when the gas usage data of the actual user exceeds the preset threshold of the gas usage curve.

10. The apparatus of claim 7, wherein the second determining module performs homogeneous user outlier detection on the feature data to determine abnormal users:

the second determining module is specifically configured to perform outlier detection on the similar users by using a correlation coefficient of time series data and an average temperature, a maximum fluctuation coefficient of the time series data, a standard deviation of the time series data, a kurtosis of the time series data, an aggregation statistical characteristic of each order of autocorrelation coefficients of the time series data, an approximate entropy of the time series data, a complexity of the time series data, a centroid of fourier transform, and a variance of the fourier transform, and determine abnormal users.