Power grid risk identification method and system based on big data

Abstract

The present invention relates to the field of power system safety technology, and in particular to a method and system for identifying power grid risks based on big data. When the present invention supplements missing data in power big data, the missing data is classified according to the type of missing data, and different data supplement operations are used according to different categories, which can effectively improve the pertinence and accuracy of data supplementation. For time series data such as power load data, the present invention has many potential connections between data at different times, and it is not periodic data. Therefore, when using a deep learning model to predict power load data, when the prediction result error does not meet the requirements, the data set is amplified and the amplified data set is input into the deep learning model to obtain more accurate prediction data.

Description

Technical Field

[0001] The present invention relates to the field of power system safety technology, and in particular to a power grid risk identification method and system based on big data. Background Art

[0002] With the rapid development of science and technology in recent years, many modern new technologies have been continuously invested in the power industry, and the risk identification of power grids can certainly be changed with the help of new forces. There are technical solutions for power grid risk identification based on big data in the prior art. For example, the Chinese invention patent application (CN112565255A) discloses a safety early warning method for power Internet of Things equipment based on BP neural network, including the following steps S1. Building an asset information function module, extracting power Internet of Things equipment data information, and preprocessing the data information; S2. Establishing a BP neural network; S3. Training the BP neural network with the data information preprocessed in step S1, updating the weights and thresholds; S4. Establishing a BP neural network power Internet of Things asset identification platform to warn power Internet of Things equipment. However, when the above scheme preprocesses the data, the processing process is relatively simple, which makes the data preprocessing effect poor and affects the accuracy of risk identification. Summary of the invention

[0003] In order to solve the above technical problems, the present invention provides a power grid risk identification method and system based on big data. According to the characteristics of power load data, a highly targeted preprocessing method is designed to improve the accuracy of big data preprocessing.

[0004] The present invention provides a power grid risk identification method based on big data, comprising:

[0005] S1: Obtain big data for grid risk identification;

[0006] S2: performing preprocessing operations on the big data;

[0007] S2.1: performing a data deduplication operation on the big data;

[0008] S2.2: Deleting invalid data in the big data;

[0009] S2.3: Supplement the missing data in the big data;

[0010] The S2.3 is specifically:

[0011] S2.3.1: Determine the missing data type of the big data. If it is the first missing data type, proceed to S2.3.2. If it is the second missing data type, proceed to S2.3.3.

[0012] S2.3.2: Use the first complementation method to complement the missing data;

[0013] S2.3.3: Use the second filling method to fill in the missing data;

[0014] S3: Input the big data pre-processed by S2 into the risk identification model to obtain risk identification results.

[0015] Preferably, the first supplementation method is specifically:

[0016] S2.3.2.1: Calculate the change rate r of the two data before and after the missing data to be supplemented 1 ;

[0017] S2.3.2.2: Calculate the change rate r of the second preset number of data before and after the missing data to be supplemented 2 ;

[0018] S2.3.2.3: Based on the data before and after the missing data to be supplemented and the change rate r 1 and r 2 Perform fill operations on the missing data to be filled.

[0019] Preferably, the operation of using the second complementation method to complement the missing data is specifically:

[0020] S2.3.3.1: Establish a deep learning model for completing the missing data;

[0021] S2.3.3.2: The deep learning model is trained by forming a data set with manually annotated historical data of the equipment in the power grid, each data set containing n data;

[0022] S2.3.3.3: Input the n data before the missing data into the deep learning model to obtain a first result;

[0023] S2.3.3.4: Calculate the prediction error between the predicted data and the actual data at the moment when data exists in the first result;

[0024] S2.3.3.5: Determine the prediction error. If the prediction error is less than a second preset threshold, use the prediction value of the time to be supplemented in the first result to supplement the big data; otherwise, proceed to the next step;

[0025] S2.3.3.6: The deep learning model is trained by forming a data set with manually annotated historical data of the equipment in the power grid, each data set containing 2n data;

[0026] S2.3.3.7: Input the 2n data before the missing data into the deep learning model to obtain a second result;

[0027] S2.3.3.8: Perform a supplement operation on the big data according to the predicted value of the time to be supplemented in the second result.

[0028] Preferably, in S2.3.2.3, the formula for the supplementation operation is:

[0029] ;

[0030] Where, d n is the data to be supplemented, n is the sampling point time of the data to be supplemented, d t The data of the first t sampling points with values ​​of the data to be supplemented.

[0031] Preferably, in S2.3.3.1, the deep learning model is a long short-term memory network model.

[0032] Preferably, in S2.3.3.2, the value of n is greater than the number of consecutive missing data in the big data.

[0033] Preferably, in S2.3.3.4, the prediction error is the average error between the predicted value and the actual value.

[0034] Preferably, in S1, the big data specifically includes: bus voltage data where the transformer is located, feeder voltage data where the transformer is located, transformer voltage data, and transformer power data; wherein the transformer power data is transformer active power data.

[0035] Preferably, in S3, the risk identification model is a BP neural network model, and the risk identification results are primary risk, secondary risk, tertiary risk, and no risk.

[0036] According to another aspect of the present invention, a power grid risk identification system based on big data is provided, using the above-mentioned power grid risk identification method based on big data, the system comprising:

[0037] A big data acquisition module, used to acquire big data for power grid risk identification;

[0038] A big data preprocessing module, used to perform preprocessing operations on the big data;

[0039] The risk identification module is used to input the big data preprocessed by the big data preprocessing module into the risk identification model to obtain the risk identification results.

[0040] The present invention has the following beneficial effects:

[0041] 1. When the present invention supplements the missing data in the electric power big data, the missing data is classified according to the type of the missing data, and different data supplement operations are used according to different categories, which can effectively improve the pertinence and accuracy of the data supplement.

[0042] 2. For time series data such as power load data, the present invention has many potential connections between data at different times, and it is not periodic data. Therefore, when using a deep learning model to predict power load data, when the prediction result error does not meet the requirements, the data set is amplified and the amplified data set is input into the deep learning model to obtain more accurate prediction data. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0044] Figure 1 is a flow chart of a method for identifying power grid risks based on big data provided in Example 1 of the present invention;

[0045] Figure 2 is a flowchart of preprocessing operations for the big data provided by Embodiment 1 of the present invention;

[0046] Figure 3 is a flowchart of an operation of supplementing missing data in the big data provided by Embodiment 1 of the present invention;

[0047] Figure 4 is a flow chart of a first complementation method provided in Example 1 of the present invention;

[0048] Figure 5 It is a flow chart of the second complementation method provided in Example 1 of the present invention. DETAILED DESCRIPTION

[0049] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work belong to the scope of protection of the present invention.

[0050] Embodiment 1, Figure 1: is a flow chart of a method for identifying power grid risks based on big data provided by Example 1 of the present invention. The method for identifying power grid risks based on big data specifically includes:

[0051] S1: Obtain big data for grid risk identification;

[0052] The big data includes three categories: bus data, feeder data, and transformer data;

[0053] In this embodiment, the risk of overload of each transformer in the power grid is identified, and therefore, the big data specifically includes: bus voltage data of the transformer, feeder voltage data of the transformer, transformer voltage data, and transformer power data; wherein, the transformer power data is transformer active power data.

[0054] S2: performing preprocessing operations on the big data;

[0055] In the prior art, generally, the same data preprocessing method is used to preprocess big data, especially abnormal data. On the one hand, using the same data preprocessing method for different types of abnormal data leads to poor targeted data preprocessing. On the other hand, using the same data preprocessing method to preprocess big data, especially abnormal data, will also lead to reduced accuracy of data processing. Especially for big data, if the accuracy of data preprocessing is not high, the subsequent process will further amplify the inaccuracy, thereby greatly affecting the accuracy of power grid risk identification.

[0056] Based on this, in this embodiment, if Figure 2 As shown, the S2 is specifically:

[0057] S2.1: performing a data deduplication operation on the big data;

[0058] Among them, in power big data, a major type of abnormal data is duplicate data, mainly duplicate data instances. Duplicate data instances refer to duplicate records generated by repeated actions of the acquisition system.

[0059] Among them, the S2.1 specifically includes: identifying duplicate data in the big data and deleting redundant duplicate data.

[0060] S2.2: Deleting invalid data in the big data;

[0061] Among them, in the field of power system big data, invalid data includes obviously invalid data and illogical data. Obviously invalid data mainly refers to invalid or hugely different character types in format, which are generally caused by abnormalities in the acquisition system or fluctuations. Illogical data mainly refers to instances that exceed the reasonable value range of the attribute, or instances that violate the established rule pattern, which are generally caused by human operation errors or accidental events. When invalid data appears, there are generally many consecutive data points that are invalid. Illogical data generally appears in the form of a single or a few outliers.

[0062] In this embodiment, the Z-score method is used to identify and delete the invalid data;

[0063] Specifically, the operation of identifying and deleting the invalid data using the Z-score method is as follows: for each data point, the difference between the data point and the average value of the data points at the sampling point within a preset time is calculated, and then the difference is divided by the standard deviation to obtain the Z-score;

[0064] If the absolute value of the Z-score is greater than a first preset threshold, the data point is considered invalid data and is deleted.

[0065] S2.3: Supplement the missing data in the big data;

[0066] Generally, the abnormal data deletion operation in the above steps will cause the data of some data points to be deleted, resulting in the missing of the data. From the above analysis, it can be seen that when invalid data appears, there are generally many consecutive data points with invalid data. Illogical data generally appears in the form of single or a few outlier data. Therefore, when supplementing, there will be two types of supplementation situations, one is to supplement the data of more consecutive data points, and the other is to supplement the data of a single or a few outlier points. The supplemented data has a relatively important relationship with the quality of big data. Therefore, this embodiment adopts different methods to supplement the missing data in the big data.

[0067] Specifically, Figure 3 As shown, the S2.3 is specifically:

[0068] S2.3.1: Determine the missing data type of the big data. If it is the first missing data type, proceed to S2.3.2. If it is the second missing data type, proceed to S2.3.3.

[0069] If the large data contains a continuous lack of data exceeding a first preset amount, it is a first missing data type, otherwise it is a second missing data type;

[0070] In this step, the first preset number is 3.

[0071] S2.3.2: Use the first complementation method to complement the missing data;

[0072] Among them, Figure 4 As shown, the first supplementation method is specifically:

[0073] S2.3.2.1: Calculate the change rate r of the two data before and after the missing data to be supplemented 1 ;

[0074] For example, if t 1 The sampling point data is missing. In this case, calculate t 0 Sampling point d 1 and t 2 Sampling point data d 2 The rate of change r, r = (d 2 -d 1 ) / d 1 If t 1 , t 2 The sampling point data is missing. In this case, calculate t 0 Sampling point d 1 and t 3 Sampling point d 3 The rate of change r, r = (d 3 -d 1 ) / (2×d 1 );

[0075] S2.3.2.2: Calculate the change rate r of the second preset number of data before and after the missing data to be supplemented 2 ;

[0076] In this step, the second preset number is 7;

[0077] For example, if the data of sampling points t7 and t8 are missing, then the calculation of t 0 -t 7 Sampling point data and t 9 -t 15 The rate of change of the sampling point data is r; at this time, if t 0 -t 7 and t 9 -t 15 The sampling point data also has missing data points. The mean of the data before and after the missing data point is used to replace the missing data point data. 0 -t 7 and t 9 -t 15If there are missing consecutive data points in the sampling point data, the mean of the data before and after the missing data point is used to replace the missing data point data;

[0078] S2.3.2.3: Based on the data before and after the missing data to be supplemented and the change rate r 1 and r 2 Perform filling operations on the missing data to be filled;

[0079] Specifically, the formula for the supplementation operation is:

[0080] ;

[0081] Where, d n is the data to be supplemented, n is the sampling point time of the data to be supplemented, d t The data of the first t sampling points with values ​​of the data to be supplemented;

[0082] In this step, when supplementing the data, not only the data of the sampling points before and after the data are considered, but also the data of the sampling points at multiple times before and after the data are considered. By calculating the data changes in different periods, the data to be supplemented is supplemented, which improves the accuracy of data supplementation and the validity of the data.

[0083] S2.3.3: Use the second filling method to fill in the missing data;

[0084] Among them, Figure 5 As shown, the operation of using the second complement method to complement the missing data is specifically as follows:

[0085] S2.3.3.1: Establish a deep learning model for completing the missing data;

[0086] The deep learning model is a long short-term memory network model (LSTM); the long short-term memory network model (LSTM) is a special recurrent neural network that can be used to solve the gradient vanishing and gradient exploding problems that traditional recurrent neural networks may encounter when processing long time series data. LSTM introduces a gating mechanism, including a forget gate, an input gate, and an output gate, and can autonomously select which information to retain and which information to forget, thereby better capturing the long-term dependencies in time series data.

[0087] S2.3.3.2: The deep learning model is trained by forming a data set with manually annotated historical data of the equipment in the power grid, each data set containing n data;

[0088] In this embodiment, the deep learning model is trained by forming a data set with manually annotated historical data of the equipment in the power grid, and the training is stopped when the loss function in the validation set of the data set no longer decreases;

[0089] Wherein, the manual labeling is labeling the historical data according to the actual operation of the equipment in the power grid, and the labeling results are level one risk, level two risk, level three risk, and no risk;

[0090] Among them, the first-level risk is a relatively large safety risk, the risk is in a completely uncontrollable state, the probability of a safety accident is high, and once an accident occurs, it will cause great harm; the second-level risk is poor system operation safety, there are local safety hazards, and the risk is within a controllable range. Once a risk occurs, the risk area and the cause of the risk can be determined in time, and it can be controlled to avoid further expansion of the risk; the third-level risk is a relatively small risk, the system is in good condition, and no major risk will be generated during operation; the no risk is the occurrence of no risk events in the power grid.

[0091] Furthermore, the value of n is greater than the number of consecutive missing data in the big data.

[0092] S2.3.3.3: Input the n data before the missing data into the deep learning model to obtain a first result;

[0093] Wherein, the first result is n predicted data predicted by the deep learning model;

[0094] S2.3.3.4: Calculate the prediction error between the predicted data and the actual data at the moment when data exists in the first result;

[0095] Among them, it can be known from the above record that the first result includes the predicted data of some moments when data exists; therefore, the predicted data of the moments when data exists in the first result and the actual data can be calculated to obtain the prediction error of the first result;

[0096] For example, if n is 8, the number of sampling points in a data set is 8, that is, there are 8 sampling data, that is, d 1 -d 8 At this time, d 1 -d 5 Data missing, d 6 -d 8 The data exists, and after prediction by the above deep learning model, we can get d 1测 -d 8测 The value of d 6测 -d 8测 The value of d 6 -d 8The prediction error is calculated by comparing the values ​​of

[0097] In this embodiment, the prediction error is the average error between the predicted value and the actual value;

[0098] S2.3.3.5: Determine the prediction error. If the prediction error is less than a second preset threshold, use the prediction value of the time to be supplemented in the first result to supplement the big data; otherwise, proceed to the next step;

[0099] S2.3.3.6: The deep learning model is trained by forming a data set with manually annotated historical data of the equipment in the power grid, each data set containing 2n data;

[0100] In fact, for time series data such as power load data, there are many potential connections between data at different times, and it is not periodic data. Therefore, in this embodiment, the data set is amplified and the amplified data set is input into the deep learning model to obtain more accurate prediction data.

[0101] S2.3.3.7: Input the 2n data before the missing data into the deep learning model to obtain a second result;

[0102] In this step, since the expanded data set is used to train the deep learning model, the obtained result is more accurate than the first result. In fact, when using deep learning to predict data, the larger the data set, the more complex the deep learning calculation process and the higher the requirements for the computing equipment. Therefore, in this step, the data set is amplified once and then stopped.

[0103] S2.3.3.8: Perform a supplement operation on the big data according to the predicted value of the time to be supplemented in the second result.

[0104] When supplementing missing data in electric power big data, this embodiment classifies the missing data according to the type of missing data, and adopts different data supplement operations according to different categories, which can effectively improve the pertinence and accuracy of data supplementation.

[0105] S3: inputting the big data pre-processed by S2 into the risk identification model to obtain risk identification results;

[0106] Among them, the risk identification model is a BP neural network model, which includes at least an input layer, a hidden layer and an output layer; the input of the BP neural network model is preprocessed big data, and the output of the BP neural network model is a risk identification result.

[0107] Among them, the risk identification results are level one risk, level two risk, level three risk, and no risk.

[0108] Embodiment 2: This embodiment discloses a power grid risk identification system based on big data, adopting the power grid risk identification method based on big data of embodiment 1, and the system includes:

[0109] A big data acquisition module, used to acquire big data for power grid risk identification;

[0110] A big data preprocessing module, used to perform preprocessing operations on the big data;

[0111] The risk identification module is used to input the big data preprocessed by the big data preprocessing module into the risk identification model to obtain the risk identification results.

[0112] Embodiment 3: This embodiment discloses an electronic device, which includes one or more processors and a memory.

[0113] The processor may be a central processing unit (CPU) or other forms of processing units having data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device to perform desired functions.

[0114] The memory may include one or more computer program products, and the computer program product may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory (cache), etc. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor may run the program instructions to implement a power grid risk identification method based on big data and / or other desired functions of any embodiment of the present application described above. Various contents such as initial external parameters, thresholds, etc. may also be stored in the computer-readable storage medium.

[0115] In addition, the electronic device may further include any other appropriate components according to specific application scenarios.

[0116] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the technical solutions of the embodiments of the present invention.

Claims

1. A power grid risk identification method based on big data, characterized in that: include: S1: Obtain big data for grid risk identification; S2: performing preprocessing operations on the big data; S2.1: performing a data deduplication operation on the big data; S2.2: Deleting invalid data in the big data; S2.3: Supplement the missing data in the big data; The S2.3 is specifically: S2.3.1: Determine the missing data type of the big data. If it is the first missing data type, proceed to S2.3.2; if it is the second missing data type, proceed to S2.3.

3. If the big data is missing data exceeding the first preset number of consecutive data, it is the first missing data type; otherwise, it is the second missing data type. S2.3.2: Use a first complementation method to complement the missing data; the first complementation method is specifically: S2.3.2.1: Calculate the change rate r1 of the two data before and after the missing data to be supplemented; S2.3.2.2: Calculate the change rate r2 of the second preset number of data before and after the missing data to be supplemented; S2.3.2.3: Perform a supplement operation on the missing data to be supplemented according to the data before and after the missing data to be supplemented and the change rates r1 and r2; the formula for the supplement operation is: ； Where, d n is the data to be supplemented, n is the sampling point time of the data to be supplemented, d t The data of the first t sampling points with values ​​of the data to be supplemented; S2.3.3: Use the second filling method to fill in the missing data; S3: Input the big data pre-processed by S2 into the risk identification model to obtain risk identification results.

2. The method for identifying power grid risks based on big data according to claim 1 is characterized in that: The operation of using the second complementation method to complement the missing data is specifically as follows: S2.3.3.1: Establish a deep learning model for completing the missing data; S2.3.3.2: The deep learning model is trained by forming a data set with manually annotated historical data of the equipment in the power grid, each data set containing n data; S2.3.3.3: Input the n data before the missing data into the deep learning model to obtain a first result; S2.3.3.4: Calculate the prediction error between the predicted data and the actual data at the moment when data exists in the first result; S2.3.3.5: Determine the prediction error. If the prediction error is less than a second preset threshold, use the prediction value of the time to be supplemented in the first result to supplement the big data; otherwise, proceed to the next step; S2.3.3.6: The deep learning model is trained by forming a data set with manually annotated historical data of the equipment in the power grid, each data set containing 2n data; S2.3.3.7: Input the 2n data before the missing data into the deep learning model to obtain a second result; S2.3.3.8: Perform a supplement operation on the big data according to the predicted value of the time to be supplemented in the second result.

3. The method for identifying power grid risks based on big data according to claim 2 is characterized in that: In S2.3.3.1, the deep learning model is a long short-term memory network model.

4. The method for identifying power grid risks based on big data according to claim 2 is characterized in that: In S2.3.3.2, the value of n is greater than the number of consecutive missing data in the big data.

5. The method for identifying power grid risks based on big data according to claim 2 is characterized in that: In S2.3.3.4, the prediction error is the average error between the predicted value and the actual value.

6. The method for identifying power grid risks based on big data according to claim 1, characterized in that: In S1, the big data specifically includes: bus voltage data where the transformer is located, feeder voltage data where the transformer is located, transformer voltage data, and transformer power data; wherein the transformer power data is transformer active power data.

7. The method for identifying power grid risks based on big data according to claim 1, characterized in that: In S3, the risk identification model is a BP neural network model, and the risk identification results are primary risk, secondary risk, tertiary risk, and no risk.

8. A power grid risk identification system based on big data, adopting a power grid risk identification method based on big data as described in any one of claims 1 to 7, characterized in that: The system comprises: A big data acquisition module, used to acquire big data for power grid risk identification; A big data preprocessing module, used to perform preprocessing operations on the big data; The risk identification module is used to input the big data preprocessed by the big data preprocessing module into the risk identification model to obtain the risk identification results.

Images