CN110929226B - Power distribution network power failure prediction method, device and system - Google Patents

Abstract

The application relates to a power failure prediction method, device and system for a power distribution network, wherein the method comprises the following steps: acquiring current memory node fact data of the distribution transformer equipment and a training data set of the distribution transformer equipment; processing current memory node fact data according to the training data set to obtain a power failure feature vector; and searching a memory tree model of the corresponding distribution equipment based on the power failure feature vector to obtain a power failure probability prediction result of the corresponding distribution equipment. When a new fact appears, the application firstly utilizes the memory tree to recall, calculates an internal evaluation outage prior probability, completes outage prediction, improves the accuracy of outage prediction, ensures the optimal running state of the power distribution network on the premise of safety and stability, provides reliable outage early warning for operation and maintenance personnel by the outage prediction result, makes related work in advance, prevents the occurrence of the outage, and reduces economic loss caused by outage.

Description

Power outage prediction method, device and system for distribution network

Technical Field

The present application relates to the technical field of power outage prediction, and in particular to a method, device and system for predicting power outages in a distribution network.

Background technique

Power supply reliability is one of the most important indicators for measuring the management level of power supply enterprises. It can show the operating status of the power system and the management level of power enterprises. It also plays an important guiding role in the planning, design, inspection and maintenance of power grids and equipment operation. The distribution network is the intermediate link between users and the power system. The power supply reliability of the distribution network is a very important factor in measuring the quality of power. Through the analysis and research of power supply reliability, we can find the relatively weak links of the distribution network, link it with the development of the local economy, and formulate measures to improve power supply reliability, so as to provide better power quality and create higher social benefits.

During the peak summer season, the power consumption of residential users in the distribution area increased rapidly, and the power load reached the highest peak of the year. With the increase in social power consumption, it is easy to cause frequent power outages in the distribution network. Frequent power outages in the distribution network not only lead to users' dissatisfaction with power services, but also bring great inconvenience to residents' daily lives and even cause serious commercial economic losses.

During the implementation process, the inventors found that there are at least the following problems in the traditional technology: the traditional power outage prediction of the distribution network is based on manual experience, which has large power outage prediction errors, large workload and high labor costs.

Summary of the invention

Based on this, it is necessary to provide a distribution network power outage prediction method, device and system to address the problems that traditional power outage prediction of distribution network is based on manual experience judgment, large power outage prediction error, large workload and high labor cost.

In order to achieve the above object, an embodiment of the present invention provides a power distribution network outage prediction method, comprising the following steps:

Obtain the current memory node fact data of the distribution transformer and the training data set corresponding to the distribution transformer;

According to the training data set, the current memory node fact data is processed to obtain the power outage feature vector;

Based on the power outage feature vector, the memory tree model of the corresponding distribution transformer is searched to obtain the power outage probability prediction result of the corresponding distribution transformer; wherein, the memory tree model is established based on the historical memory node fact data of the corresponding distribution transformer; and the training data set is obtained by training the historical memory node fact data.

In one embodiment, the memory node fact data includes any one or any combination of the following data: heavy overload duration, heavy three-phase imbalance duration, daily maximum active load rate, daily maximum three-phase imbalance, daily average active load rate, average three-phase imbalance and power outage event data of the day.

In one embodiment, based on the power outage feature vector, searching the memory tree model of the corresponding distribution transformer to obtain the power outage probability prediction result of the corresponding distribution transformer includes:

Based on the power outage probability prediction results, the memory node fact data of the prediction period is obtained;

The current memory node fact data in the memory tree model is updated to the memory node fact data of the prediction period.

In one embodiment, the step of updating the current memory node fact data in the memory tree model to the memory node fact data of the prediction period includes:

Delete the current memory node fact data in the memory tree model; and add the memory node fact data of the prediction period to the corresponding position in the current memory node fact data.

In one embodiment, the step of deleting the current memory node fact data in the memory tree model includes:

In the memory tree model, the parent node corresponding to the memory node to be deleted is pointed to the child node of the memory node to be deleted; the memory node to be deleted is the memory node corresponding to the current memory node fact data.

In one embodiment, the step of adding the memory node fact data of the prediction period to the corresponding position in the corresponding current memory node fact data includes:

Create a new memory node for the memory node fact data corresponding to the forecast period;

Set the newly added memory node as the memory root of the memory tree model, and point the newly added memory node to the previous node of the memory tree model.

In one embodiment, the steps are also included:

Detect the memory node impression value and memory decay value of each memory node fact data in turn;

When the memory node impression value is less than the memory impression threshold and the memory decay value is less than the forget threshold, the corresponding memory node fact data in the memory tree model is deleted.

On the other hand, an embodiment of the present invention further provides a power outage prediction device for a distribution network, comprising:

A data acquisition unit, used to acquire current memory node fact data of the distribution transformer device and a training data set corresponding to the distribution transformer device;

A data processing unit, used for processing the current memory node fact data according to the training data set to obtain a power outage feature vector;

The power outage prediction unit is used to search the memory tree model of the corresponding distribution transformer equipment based on the power outage feature vector to obtain the power outage probability prediction result of the corresponding distribution transformer equipment; wherein the memory tree model is established based on the historical memory node fact data of the corresponding distribution transformer equipment; and the training data set is obtained by training the historical memory node fact data.

On the other hand, an embodiment of the present invention further provides a distribution network power outage prediction system, including a memory and a processor, the memory storing a computer program, and the processor implementing the steps of any of the above-mentioned distribution network power outage prediction methods when executing the computer program.

On the other hand, an embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of any one of the above-mentioned methods for predicting power outages in a distribution network are implemented.

One of the above technical solutions has the following advantages and beneficial effects:

In each embodiment of the above-mentioned distribution network power outage prediction method, the current memory node fact data of the distribution transformer and the training data set of the corresponding distribution transformer are obtained; according to the training data set, the current memory node fact data is processed to obtain the power outage feature vector; based on the power outage feature vector, the memory tree model of the corresponding distribution transformer is searched to obtain the power outage probability prediction result of the corresponding distribution transformer, so as to realize the power outage prediction of the distribution network. The present application can establish a memory tree model for each distribution transformer, select the corresponding memory node fact data of the distribution transformer as the fact, and establish the distribution transformer memory tree. Whenever a new fact appears, first use its own memory tree to recall, calculate an internal evaluation power outage prior probability, complete the power outage prediction, improve the accuracy of the power outage prediction, so that the distribution network can ensure the best operating state under the premise of safety and stability, and the power outage prediction result provides reliable power outage warning for operation and maintenance personnel, so that relevant work can be done in advance to prevent problems before they happen and reduce the economic losses caused by power outages.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a schematic diagram of an application environment of a method for predicting power outages in a distribution network according to an embodiment;

FIG2 is a schematic diagram of a first flow chart of a method for predicting power outages in a distribution network according to an embodiment;

FIG3 is a schematic diagram of a second flow chart of a method for predicting power outages in a distribution network according to an embodiment;

FIG4 is a schematic diagram of a third flow chart of a method for predicting power outages in a distribution network in one embodiment;

FIG5 is a schematic diagram of establishing a memory tree model in one embodiment;

FIG6 is a schematic diagram of a memory tree model after adding a node in one embodiment;

FIG7 is a schematic diagram of the structure of a power distribution network outage prediction device in one embodiment;

FIG8 is a schematic diagram of the structure of a power distribution network outage prediction system in one embodiment.

Detailed ways

In order to facilitate understanding of the present application, the present application will be described more fully below with reference to the relevant drawings. The preferred embodiments of the present application are given in the drawings. However, the present application can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the disclosure of the present application more thorough and comprehensive.

Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein in the specification of this application are only for the purpose of describing specific embodiments and are not intended to limit this application. The term "and/or" used herein includes any and all combinations of one or more of the related listed items.

The power outage prediction method for distribution network provided in the present application can be applied in the application environment shown in FIG1 . The terminal 102 communicates with the distribution transformer 104 through the network. The terminal 102 can be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, and portable wearable devices, and the distribution transformer 104 can be implemented by an independent distribution transformer or a distribution transformer cluster composed of multiple distribution transformers.

In one embodiment, as shown in FIG2 , a power distribution network outage prediction method is provided, and the method is applied to the terminal 102 in FIG1 as an example for description, including the following steps:

Step S210, obtaining the current memory node fact data of the distribution transformer and the training data set corresponding to the distribution transformer.

The distribution transformer equipment may be, but is not limited to, overhead lines, poles, cables, distribution transformers, switchgear, and reactive compensation capacitors. The memory node fact data refers to the fact data of the corresponding memory node in the memory tree. For example, the memory node fact data may be, but is not limited to, the three-phase imbalance data of the corresponding distribution transformer equipment. Training data refers to the data used to train the data mining model in the data mining process; the training data set refers to a collection of multiple training data.

Specifically, a training data set can be obtained by training the historical memory node fact data in advance, and then the training data set of the corresponding distribution transformer can be obtained; a memory tree model is obtained by establishing the historical memory node fact data of the corresponding distribution transformer, and then the current memory node fact data of the distribution transformer can be obtained by querying the memory tree model.

Step S220, processing the current memory node fact data according to the training data set to obtain a power outage feature vector.

The power outage feature vector may be used to indicate a feature vector indicating that a power outage factor exists in the corresponding distribution transformer.

Specifically, according to the acquired training data set, the current memory node fact data of the distribution transformer is processed to obtain the power outage feature vector.

Step S230, based on the power outage feature vector, search the memory tree model of the corresponding distribution transformer to obtain the power outage probability prediction result of the corresponding distribution transformer; wherein the memory tree model is established based on the historical memory node fact data of the corresponding distribution transformer; and the training data set is obtained by training the historical memory node fact data.

The power outage probability prediction result can be used to indicate the probability of a power outage occurring in the distribution transformer in the next time period (eg, the next day).

Specifically, according to the processed power outage feature vector, the memory tree model corresponding to the distribution transformer is searched to obtain the power outage probability prediction result of the corresponding distribution transformer in the next time period.

Specifically, in the above-mentioned embodiment of the power outage prediction method for the distribution network, the current memory node fact data of the distribution transformer and the training data set of the corresponding distribution transformer are obtained; according to the training data set, the current memory node fact data is processed to obtain the power outage feature vector; based on the power outage feature vector, the memory tree model of the corresponding distribution transformer is searched to obtain the power outage probability prediction result of the corresponding distribution transformer, so as to realize the power outage prediction of the distribution network. By establishing a memory tree model for each distribution transformer, the corresponding memory node fact data of the distribution transformer is selected as the fact, and the distribution transformer memory tree is established. Whenever a new fact appears, first use its own memory tree to recall, calculate an internal evaluation power outage prior probability, complete the power outage prediction, improve the accuracy of the power outage prediction, so that the distribution network can ensure the best operating state under the premise of safety and stability, and the power outage prediction result provides reliable power outage warning for operation and maintenance personnel, so that relevant work can be done in advance, prevent problems before they happen, and reduce the economic losses caused by power outages.

In a specific embodiment, the memory node fact data includes any one or any combination of the following data: heavy overload duration, heavy three-phase imbalance duration, daily maximum active load rate, daily maximum three-phase imbalance, daily average active load rate, average three-phase imbalance and power outage event data of the day.

The power outage event data of the day can be used to indicate whether a power outage occurs in the distribution transformer on the day. For example, the power outage event data of the day can be 0 or 1, where when the power outage event data of the day is 0, it indicates that there is no power outage in the distribution transformer, and when the power outage event data of the day is 1, it indicates that there is a power outage in the distribution transformer.

In an example, taking the distribution transformer equipment of the district bureau as an example, the following variables and characteristics are adopted: the memory decay coefficient is α, the forget threshold is δ, the memory period is T, and the memory node fact data are the heavy overload duration f _zgzsc (hereinafter referred to as f ₁ ), the heavy three-phase imbalance duration f _zscbphsc (hereinafter referred to as f ₂ ), the daily maximum active load rate f _{max_ygfzl} (hereinafter referred to as f ₃ ), the daily maximum three-phase imbalance f _{max_sxbphd} (hereinafter referred to as f ₄ ), the daily average active load rate f _{ave_ygfzl} (hereinafter referred to as f ₅ ), the average three-phase imbalance f _{ave_sxbphd} (hereinafter referred to as f ₆ ) and the power outage event data of the day _{fis_poweroff} (hereinafter referred to as f ₇ ).

In one embodiment, as shown in FIG3 , a power distribution network outage prediction method is provided, and the method is applied to the terminal 102 in FIG1 as an example for description, including the following steps:

Step S310, obtaining the current memory node fact data of the distribution transformer and the training data set corresponding to the distribution transformer.

Step S320, processing the current memory node fact data according to the training data set to obtain a power outage feature vector.

Step S330, based on the power outage feature vector, search the memory tree model of the corresponding distribution transformer to obtain the power outage probability prediction result of the corresponding distribution transformer; wherein the memory tree model is established based on the historical memory node fact data of the corresponding distribution transformer; and the training data set is obtained by training the historical memory node fact data.

Step S340, based on the power outage probability prediction result, obtain the memory node fact data of the prediction period.

The memory node fact data of the prediction period refers to the memory node fact data occurring in the time period corresponding to the power outage probability prediction result.

For example, if the power outage probability prediction result is for the distribution transformer equipment prediction for the next day, the memory node fact data of the prediction period corresponds to the memory node fact data occurring the next day.

Step S350, updating the current memory node fact data in the memory tree model to the memory node fact data of the prediction period.

Among them, the specific content and process of the above steps S310, S320 and S330 can be referred to the above content and will not be repeated here.

Specifically, a memory tree model is established for the distribution transformer equipment, and the memory node fact data of the distribution transformer equipment is selected to establish the distribution transformer memory tree. Whenever a new fact appears, the memory tree recall is used to calculate an internal evaluation of the prior probability of power outage and complete the power outage prediction. When the power outage fact is confirmed in the later period, the memory node fact data of the prediction period is used to dynamically update the memory tree model structure and the memory impression of the memory node, so that the distribution network can ensure the best operating state under the premise of safety and stability. The power outage prediction results provide reliable power outage warnings for operation and maintenance personnel, so that relevant work can be done in advance to prevent problems before they happen and reduce the economic losses caused by power outages. The power outage prediction implementation process is simplified and has good robustness.

In a specific embodiment, the step of updating the current memory node fact data in the memory tree model to the memory node fact data of the prediction period includes:

Delete the current memory node fact data in the memory tree model; and add the memory node fact data of the prediction period to the corresponding position in the current memory node fact data.

Specifically, the current memory node fact data of the distribution transformer is deleted from the memory tree, and the memory node fact data of the prediction period is added to the root of the memory tree, thereby completing the update of the memory tree structure.

Furthermore, a copy Temp of the memory node fact data of the prediction period is saved, the current memory node fact data is deleted from the memory tree, and Temp is added to the root of the memory tree, thus completing the memory tree structure update.

In a specific embodiment, the step of deleting the current memory node fact data in the memory tree model includes:

In the memory tree model, the parent node corresponding to the memory node to be deleted is pointed to the child node of the memory node to be deleted; the memory node to be deleted is the memory node corresponding to the current memory node fact data.

Specifically, the successor node of the parent node of the memory node to be deleted is pointed to the child node of the memory node to be deleted, thereby completing the node deletion operation.

In a specific embodiment, the step of adding the memory node fact data of the prediction period to the corresponding position in the corresponding current memory node fact data includes:

Create a new memory node for the memory node fact data corresponding to the forecast period;

Set the newly added memory node as the memory root of the memory tree model, and point the newly added memory node to the previous node of the memory tree model.

Specifically, a new memory node for the memory node fact data corresponding to the prediction period is newly created, an initial node memory impression is set, the new memory node is set as a memory root and points to the previous node, thereby obtaining a memory tree after the new memory node is added.

In one embodiment, as shown in FIG4 , a power distribution network outage prediction method is provided, and the method is applied to the terminal 102 in FIG1 as an example for description, including the following steps:

Step S410, obtaining current memory node fact data of the distribution transformer and a training data set corresponding to the distribution transformer.

Step S420: Process the current memory node fact data according to the training data set to obtain a power outage feature vector.

Step S430, based on the power outage feature vector, search the memory tree model of the corresponding distribution transformer to obtain the power outage probability prediction result of the corresponding distribution transformer; wherein the memory tree model is established based on the historical memory node fact data of the corresponding distribution transformer; and the training data set is obtained by training the historical memory node fact data.

Step S440, sequentially detect the memory node impression value and memory decay value of each memory node fact data.

Among them, the memory node impression value can be used to indicate the impression level of the corresponding memory node in the memory tree; the memory decay value can be used to indicate the decay level of the corresponding memory node in the memory tree.

Step S450, when the memory node impression value is less than the memory impression threshold, and the memory decay value is less than the forget threshold, the corresponding memory node fact data in the memory tree model is deleted.

Among them, the specific content and process of the above steps S410, S420 and S430 can be referred to the above content and will not be repeated here.

Specifically, a memory tree model is established for the distribution transformer equipment, and the memory node fact data of the distribution transformer equipment is selected to establish the distribution transformer memory tree. Whenever a new fact appears, the memory tree itself is used to recall and calculate an internal evaluation of the prior probability of power outage to complete the power outage prediction. In addition, the memory nodes in the memory tree can be processed for memory forgetting. When the memory node impression value is less than the memory impression threshold, and the memory decay value is less than the forget threshold, that is, the memory node meets the memory forgetting condition, the corresponding memory node fact data in the memory tree model can be deleted, and then the memory forgetting process is completed, which optimizes the memory tree model and has better robustness.

In order to solve the problem that the traditional power outage prediction of the distribution network is based on manual experience judgment, the power outage prediction error is large, the workload is large and the labor cost is high, in one example, the working process of the power outage prediction of the distribution network is specifically described.

Step 1: Assume that a district bureau has a total of N = 1 distribution transformers, the memory decay coefficient is α = 0.05, the forget threshold is δ = 0.01, the memory period is T = 24h, and the memory node fact data are the heavy overload duration of the distribution transformer (hereinafter referred to as _f1 ), the heavy three-phase imbalance duration (hereinafter referred to as _f2 ), the daily maximum active load rate (hereinafter referred to as _f3 ), the daily maximum three-phase imbalance (hereinafter referred to as _f4 ), the daily average active load rate (hereinafter referred to as _f5 ), the average three-phase imbalance (hereinafter referred to as _f6 ) and the power outage event data of the day (hereinafter referred to as _f7 ).

Step 2: Establish a memory tree model, as shown in Figure 5. The dotted line is the memory fact node, _f7 is the fact result, _f1 ~ _f6 are fact factors, α is the memory decay coefficient, β is the memory node impression, and t is the fact occurrence event.

Step 3: Generate training data set. Assume that the original data set of the distribution transformer equipment in the district is T = {t ₁ ,t ₂ ,…,t _n }, where _ti (i = 1,2,…,n) is the operating data of the i-th distribution transformer equipment, and _ti = {s ₁ ,s ₂ ,…,s _m }, where s _j (j = 1,2,…,m) is the operating data of the distribution transformer equipment at time j. Among them, x _j is the timestamp of time j, /> is the active load rate at time j, /> is the current of phase A at time j,/> is the B phase current at time j,/> is the C phase current at time j. The training data set is calculated as follows:

Calculate the date d _k : d _k = date(x _j );

Calculate the maximum active load factor f ₃ ^(k) on date d _k :

Calculate the heavy overload duration f ₁ ^(k) of date d _k : Where I(x) is the indicator function;

Calculate the average active load factor f ₅ ^(k) on date d _k :

Calculate the three-phase unbalance _xsxbphd ^(j) at time j:

Calculate the maximum three-phase unbalance f ₄ ^(k) on date d _k :

Calculate the heavy three-phase unbalance f ₂ ^(k) on date d _k : f ₂ ^(k) = Δt∑I(x _sxbphd ^(j) > 60,jind _k );

Calculate the average three-phase unbalance f ₆ ^(k) on date d _k :

After the above calculation, the training data set T'={ _t'1 , _t'2 ,…, _t'n } is formed, where _t'i (i=1, 2,…, n) is the training data generated for the i-th distribution transformer. Let _t'i ={ _f1 , _f2 ,…, _fm }, where _fk (j=1, 2,…, m) is the training sample _fk =[ _dk , _f1 ^(k) , _f2 ^(k) , _f3 ^(k) , _f4 ^(k) , _f5 ^(k) , _f6 ^(k) ] of the distribution transformer on date j. Assume that the data label is L = {l ₁ ,l ₂ ,…,l _n }, where l _i (i = 1,2,…,n) is the label of the i-th distribution transformer, and l _i = {y ₁ ,y ₂ ,…,y _m }, where y _k (j = 1,2,…,m) is the label of the distribution transformer on date k, indicating whether a power outage occurred the next day. The training data set is used for original memory generation. The following is the training data set generated by the distribution transformer for 10 consecutive days, as shown in the following table.

Serial number d _k f ₁ ^(k) f ₂ ^(k) f ₃ ^(k) f ₄ ^(k) f ₅ ^(k) _f6 ^(k) f ₇ ^(k) 1 2018/5/1 0.00 21.32 12.50 0.50 63.16 33.73 0 2 2018/5/2 0.00 14.52 7.73 0.50 62.00 34.69 0 3 2018/5/3 0.00 11.38 8.01 1.50 67.95 32.55 0 4 2018/5/4 0.00 9.78 6.37 0.25 66.04 29.50 0 5 2018/5/5 0.00 10.37 6.85 1.00 72.86 40.08 0 6 2018/5/6 0.00 12.65 7.85 0.00 63.54 31.52 0 7 2018/5/7 0.00 13.15 9.12 0.00 59.26 32.59 0 8 2018/5/8 0.00 12.17 7.53 0.00 57.75 28.07 0 9 2018/5/9 0.00 11.18 6.95 0.25 60.00 32.93 0 10 2018/5/10 0.00 10.75 6.28 0.25 67.57 37.02 0 … … … … … … … … … 59 2.18/6/28 0.00 20.47 14.86 0.00 42.11 20.39 0

Step 4: Select 1 to 58 groups of data as training samples, input sample data, and search the memory tree. Suppose the sample of the i-th distribution transformer on date k is fi _,k = [d _i,k , _fi,1 ^(k) ,fi _,2 ^(k) ,fi _,3 ^(k) ,fi _,4 ^(k) , _fi,5 ^(k) ,fi _,6 ^(k) ],li _,k = _yi,k . Find the memory node node that is closest to the training sample and the memory tree tree and the power outage label label corresponding to the memory node. Among them, nodes = {(f _j ,dis)|dis = || _fi,k -f _j || ² ,f _j in tree},

Then the minimum distance dis _i,k can be obtained: dis _i,k =||f _i,k -f _node || ^2. Assume that in the nodes set, the node set whose training sample distance from the memory tree is less than thr = 0.8 is Bnodes and the elements have been sorted from small to large by distance: Bnodes = {(f _j ,label _j )|(f _j ,dis)∈nodes and dis＜thr}. If dis _i,k >thr, go to step 5. If dis _i,k ≤thr and label _i,k ＞label _node , go to step 6. If dis _i,k ≤thr and label _i,k ≠label _node , update the memory impression of the memory node node, where the value w _thr =3: Otherwise, update the memory impression of the node Node according to the following formula, with the value ω _dis = 2.5:/> After executing the above jump, return to step 4 until the creation of the memory tree is completed.

Step 5: Add a new memory node. Create a new memory node t _n+1 and set the initial node memory impression to: The new memory node is set at the memory root and points to the previous node, and the memory tree after the new memory node is added is as shown in FIG6 .

Step 6: Update the memory tree. Save a copy of the node to be updated as Temp, delete the node from the memory tree using step 7, and add Temp to the root of the memory tree using step 5. This completes the update of the memory tree structure. Use the following method to update the memory impression of the node:

Step 7: Delete the memory node. Point the successor node of the parent node of the memory node to be deleted to the child node of the node to complete the node deletion operation.

Step 8: Forecast power outages for the next day. When the i-th distribution transformer generates a new day's operating data, calculate the power outage feature vector, _fi,new = [2018/6/28, 0.00, 20.47, 14.86, 0.00, 42.11, 20.39], search the memory tree, and find the memory node set Result and the power outage label label of the corresponding node whose feature vector is short to the memory tree tree and has a good impression.

The power outage probability prediction result is:

p(label＝1)＝1-p(label＝0)＝1-0.9358＝0.0642

Among them, label = 1 means power outage, label = 0 means no power outage, so the predicted date is 2018/6/28. The probability of power outage is 0.0642, and the probability of no power outage is 0.9358. The power outage prediction is completed.

Step 9: From the data set, we know that the actual result in step 9 is label'=1, that is, no power outage. Then we can get a new data sample, that is, the data sample with sequence number 59 in the data set. Using this sample, according to step 4, update the memory tree of the distribution transformer.

Step 10: Memory forgetting: If a node in the memory tree β _j <β _thr = 0.001 and α ^ωj <δ = 0.0001, ω = 0.1, then the node is deleted using step 7 to complete the memory forgetting process.

After the above process, the power outage prediction of the distribution transformer equipment can be carried out, and the result of the prediction date of 2018/6/28 can be obtained: the power outage probability is 0.0642, and the power outage probability is 0.9358. The prediction results can be provided to the distribution transformer operation and maintenance staff as a reliable reference for operation and maintenance, so that timely measures can be taken for the distribution transformer with a high probability of power outage to prevent it before it happens, or to ensure power supply, ensure the power supply reliability of the substation, and ensure the economic benefits of the substation.

It should be understood that, although the various steps in the flowcharts of Figures 2-4 are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a portion of the steps in Figures 2-4 may include a plurality of sub-steps or a plurality of stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these sub-steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of the sub-steps or stages of other steps.

In one embodiment, as shown in FIG7 , a power distribution network outage prediction device is provided, comprising:

The data acquisition unit 710 is used to acquire the current memory node fact data of the distribution transformer and the training data set corresponding to the distribution transformer;

The data processing unit 720 is used to process the current memory node fact data according to the training data set to obtain the power outage feature vector;

The power outage prediction unit 730 is used to search the memory tree model of the corresponding distribution transformer device based on the power outage feature vector to obtain the power outage probability prediction result of the corresponding distribution transformer device; wherein the memory tree model is established based on the historical memory node fact data of the corresponding distribution transformer device; and the training data set is obtained by training the historical memory node fact data.

For the specific definition of the power outage prediction device for distribution network, please refer to the definition of the power outage prediction method for distribution network mentioned above, which will not be repeated here. Each module in the above-mentioned power outage prediction device for distribution network can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the power outage prediction system for distribution network in the form of hardware, or can be stored in the memory in the power outage prediction system for distribution network in the form of software, so that the processor can call and execute the operations corresponding to the above-mentioned modules.

In one embodiment, as shown in FIG8 , a distribution network power outage prediction system is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any of the above-mentioned distribution network power outage prediction methods when executing the computer program.

The processor may be configured to perform the following steps:

Obtain the current memory node fact data of the distribution transformer and the training data set corresponding to the distribution transformer;

According to the training data set, the current memory node fact data is processed to obtain the power outage feature vector;

Based on the power outage feature vector, the memory tree model of the corresponding distribution transformer is searched to obtain the power outage probability prediction result of the corresponding distribution transformer; wherein, the memory tree model is established based on the historical memory node fact data of the corresponding distribution transformer; and the training data set is obtained by training the historical memory node fact data.

In one embodiment, a computer-readable storage medium is further provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of any of the above-mentioned methods for predicting power outages in a distribution network are implemented.

Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned division operation methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in the present application can include non-volatile and/or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims (10)

1. A method for predicting power outages in a distribution network, comprising the following steps: Acquire current memory node fact data of the distribution transformer and a training data set corresponding to the distribution transformer; According to the training data set, processing the current memory node fact data to obtain a power outage feature vector; Based on the power outage feature vector, a memory tree model corresponding to the distribution transformer is searched to obtain a power outage probability prediction result corresponding to the distribution transformer; wherein the memory tree model is established based on historical memory node fact data corresponding to the distribution transformer; and the training data set is obtained by training the historical memory node fact data; The step of searching the memory tree model corresponding to the distribution transformer device based on the power outage feature vector to obtain the power outage probability prediction result corresponding to the distribution transformer device includes: When the i-th distribution transformer generates a new day's operation data, the power outage feature vector is calculated, and the memory tree model is searched to find the memory node set Result with a shorter distance between the power outage feature vector and the memory tree model and the power outage label label of the corresponding memory node, and the power outage probability prediction result is determined as: p(label＝1)＝1-p(label＝0) Wherein, label＝1 indicates power outage, label＝0 indicates no power outage, i is the i-th distribution transformer, _fj is the fact data of the j-th memory node, dis indicates the distance between the memory node and the memory tree model, _βi is the memory node impression of the i-th distribution transformer, k is the date, ( _fj , dis, _βi , k) indicates the power outage feature vector, I(( _fj , dis, _βi , k)in Result and label＝1) indicates the memory node set label whose distance from the power outage feature vector of the i-th distribution transformer on date k to the memory tree model is shorter than the preset distance threshold, _βj is the memory impression of the j-th memory node of the memory tree model, _βjI is the memory impression of the memory node j corresponding to the memory node set label I, p(label＝1) indicates the probability of power outage, p(label＝0) indicates the probability of no power outage, e indicates the exponential function, I(( _fj , dis, _βi , k)in Result and label＝0) represents the memory node set label whose power outage feature vector and the memory tree model distance are shorter than the preset distance threshold when the i-th distribution transformer is out of power on date k, Σ represents the sum of the memory impressions of all memory nodes j corresponding to the memory node set label I, and N is the total number of distribution transformers, It represents the sum of the memory impressions of power outage events occurring in all distribution and transformer equipment. 2. The distribution network power outage prediction method according to claim 1 is characterized in that the memory node fact data includes any one or any combination of the following data: heavy overload duration, heavy three-phase imbalance duration, daily maximum active load rate, daily maximum three-phase imbalance, daily average active load rate, average three-phase imbalance and power outage event data of the day. 3. The method for predicting power outages in a distribution network according to claim 1, characterized in that after the step of searching the memory tree model corresponding to the distribution transformer device based on the power outage feature vector to obtain the power outage probability prediction result corresponding to the distribution transformer device, the method comprises: Based on the power outage probability prediction result, obtaining memory node fact data of the prediction period; The current memory node fact data in the memory tree model is updated to the memory node fact data of the prediction period. 4. The power distribution network outage prediction method according to claim 3, characterized in that the step of updating the current memory node fact data in the memory tree model to the memory node fact data of the prediction period comprises: The current memory node fact data in the memory tree model is deleted; and the memory node fact data of the prediction period is added to the corresponding position in the current memory node fact data. 5. The power distribution network outage prediction method according to claim 4, characterized in that the step of deleting the current memory node fact data in the memory tree model comprises: In the memory tree model, the parent node corresponding to the memory node to be deleted is pointed to the child node of the memory node to be deleted; the memory node to be deleted is the memory node corresponding to the current memory node fact data. 6. The method for predicting power outages in a distribution network according to claim 4, wherein the step of adding the memory node fact data of the prediction period to the corresponding position in the current memory node fact data comprises: Establishing a new memory node corresponding to the memory node fact data of the prediction period; The newly added memory node is set as the memory root of the memory tree model, and the newly added memory node points to the previous node of the memory tree model. 7. The method for predicting power outages in a distribution network according to claim 1, further comprising the steps of: sequentially detecting the memory node impression value and the memory decay value of each of the memory node fact data; When the memory node impression value is less than the memory impression threshold, and the memory decay value is less than the forget threshold, the corresponding memory node fact data in the memory tree model is deleted. 8. A power outage prediction device for a distribution network, comprising: A data acquisition unit, used to acquire current memory node fact data of the distribution transformer and a training data set corresponding to the distribution transformer; A data processing unit, configured to process the current memory node fact data according to the training data set to obtain a power outage feature vector; A power outage prediction unit, configured to search a memory tree model corresponding to the distribution transformer device based on the power outage feature vector to obtain a power outage probability prediction result corresponding to the distribution transformer device; wherein the memory tree model is established based on historical memory node fact data corresponding to the distribution transformer device; and the training data set is obtained by training the historical memory node fact data; The step of searching the memory tree model corresponding to the distribution transformer device based on the power outage feature vector to obtain the power outage probability prediction result corresponding to the distribution transformer device includes: When the i-th distribution transformer generates a new day's operation data, the power outage feature vector is calculated, and the memory tree model is searched to find the memory node set Result with a short distance between the power outage feature vector and the memory tree model and the power outage label label of the corresponding memory node, and the power outage probability prediction result is determined as: p(label＝1)＝1-p(label＝0) Wherein, label＝1 indicates power outage, label＝0 indicates no power outage, i is the i-th distribution transformer, _fj is the fact data of the j-th memory node, dis indicates the distance between the memory node and the memory tree model, _βi is the memory node impression of the i-th distribution transformer, k is the date, ( _fj , dis, _βi , k) indicates the power outage feature vector, I(( _fj , dis, _βi , k)in Result and label＝1) indicates the memory node set label whose distance from the power outage feature vector of the i-th distribution transformer on date k to the memory tree model is shorter than the preset distance threshold, _βj is the memory impression of the j-th memory node of the memory tree model, _βjI is the memory impression of the memory node j corresponding to the memory node set label I, p(label＝1) indicates the probability of power outage, p(label＝0) indicates the probability of no power outage, e indicates the exponential function, I(( _fj , dis, _βi , k)in Result and label＝0) represents the memory node set label whose power outage feature vector and the memory tree model distance are shorter than the preset distance threshold when the i-th distribution transformer is out of power on date k, Σ represents the sum of the memory impressions of all memory nodes j corresponding to the memory node set label I, and N is the total number of distribution transformers, It represents the sum of the memory impressions of power outage events occurring in all distribution and transformer equipment. 9. A distribution network power outage prediction system, comprising a memory and a processor, wherein the memory stores a computer program, and wherein the processor implements the steps of the distribution network power outage prediction method according to any one of claims 1 to 7 when executing the computer program. 10. A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the steps of the power distribution network outage prediction method according to any one of claims 1 to 7 are implemented.