A smart grid line loss prediction method based on deep learning

By using deep learning technology to build a line loss prediction sub-model in smart grids, the problem that traditional methods cannot capture nonlinear relationships and long-term dependencies is solved, real-time and accurate line loss prediction in smart grids is achieved, and efficient management and Quickly respond to demands.

CN118378544BActive Publication Date: 2024-11-08WUHUA COUNTY HUAYUAN ELECTRIC POWER DEVELOPMENT CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410814105.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-06-24
Publication Date
2024-11-08
Estimated Expiration
2044-06-24

AI Technical Summary

Technical Problem

Traditional line loss prediction methods cannot fully capture the complex nonlinear relationships and long-term dependencies in the power system, resulting in insufficient accuracy of the prediction results and cannot meet the smart grid's needs for real-time monitoring and rapid response.

Method used

Using the smart grid line loss prediction method based on deep learning, a line loss prediction sub-model is constructed by collecting operational data from each link of electricity consumption, a line loss prediction sub-model is constructed, using a long and short-term memory network for time series prediction, and optimized through k-fold cross-verification and performance evaluation. Model performance, and finally build a grid bus loss prediction model to achieve real-time and accurate line loss prediction.

Benefits of technology

It improves the accuracy and response speed of line loss prediction, can process large-scale data, provide real-time line loss prediction information, and enhances the real-time monitoring capabilities and resource allocation optimization of smart grids.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118378544B_ABST
    Figure CN118378544B_ABST
Patent Text Reader

Abstract

The present invention provides a method for predicting line loss of a smart grid based on deep learning, which relates to the field of line loss prediction of a power grid, and includes the following steps: S1: collecting operation-related data of line loss circuits in each power consumption link; S2: obtaining the actual line loss value of the line loss circuit in each power consumption link, and constructing a line loss prediction sub-model for each power consumption link; S3: constructing a power grid total line loss prediction model from all line loss prediction sub-models; S4: using the operation-related data of each power consumption link in the smart grid as input, providing it to the power grid total line loss prediction model, and outputting the predicted line loss value of the smart grid. Through the application of deep learning technology, the present invention can process large-scale power grid operation data, provide real-time and accurate line loss prediction, and optimize model performance through an automated method, thereby providing technical support for efficient management and operation and maintenance of smart grids.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of power grid line loss prediction, and in particular to a smart grid line loss prediction method based on deep learning. Background Art

[0002] The power loss of the power grid (referred to as line loss) is the power loss generated in the transmission, transformation, distribution and consumption of electricity from the power grid to the user. The line loss rate is the ratio of power loss to total power supply in a certain period of time. It is an important indicator to measure the technical and economic feasibility of power grid enterprises. It involves many disciplines such as planning, production, marketing, operation, finance, etc., and comprehensively reflects the technical and economic level of planning, design, production, operation and management of power grid enterprises.

[0003] With the continuous advancement of smart grid technology, the monitoring, control and optimization capabilities of the power grid have been significantly improved, becoming more refined and automated. By monitoring the flow of power in real time, smart grids can efficiently collect and process massive amounts of data, which provides a solid technical foundation for accurately predicting line losses. However, traditional line loss prediction methods have limitations and may not fully capture the complex nonlinear relationships and long-term dependencies that exist in the power system, resulting in insufficient accuracy in the prediction results. In addition, the power grid's growing demand for real-time monitoring and rapid response places higher demands on the accuracy and response speed of line loss prediction technology.

[0004] In this context, there is a lack of a method that can use deep learning and smart grid technology to accurately and real-time predict line loss. Existing prediction technologies have not yet fully adapted to the needs of high efficiency and high accuracy in the smart grid environment, especially in processing large-scale data and real-time prediction. Therefore, we urgently need a smart grid line loss prediction method based on deep learning to solve the above problems. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention provides a smart grid line loss prediction method based on deep learning to solve the problem raised in the above background technology that there is a lack of an accurate and real-time line loss prediction method that can utilize deep learning and smart grid technology.

[0006] The purpose of the present invention can be achieved by the following technical solution: comprising the following steps:

[0007] S1: Collect the operation data of the line loss circuit in each link of electricity consumption; the various links of electricity consumption include power generation, transmission, transformation, distribution, consumption and management line loss; the operation data include current, voltage, power factor, line load, power metering, power frequency, temperature and environmental parameters, and the environmental parameters include wind speed, humidity, temperature, light intensity and electromagnetic interference intensity;

[0008] S2: Obtain the actual line loss value of the line loss circuit in each power consumption link, and build a line loss prediction sub-model for each power consumption link. The model is expressed as ,in, represents the predicted line loss value of power consumption link i, represents the set of corresponding parameters in the operation-related data in the power consumption link i, Representation Model The parameter set of

[0009] Among them, the line loss prediction sub-model includes:

[0010] Use long short-term memory networks or gated recurrent units for time series forecasting;

[0011] The specific process of constructing the line loss prediction sub-model is as follows:

[0012] S21: Evaluate the difference between the actual line loss value of the line loss circuit and the historical predicted line loss value, and obtain the predicted line loss value predicted by the line loss prediction sub-model in the historical record, expressed as , t represents the time index predicted by the line loss prediction sub-model; the actual line loss value of the line loss circuit is calculated with the predicted line loss value in the historical records, and the formula is used , and obtain the historical predicted fluctuation value MS; where Lt represents the actual line loss value at time t, represents the predicted line loss value at the time of prediction t, N represents the total number of time indexes predicted by the line loss prediction sub-model in the historical records; the average of the absolute values ​​of the differences between the predicted line loss value and the actual line loss value is calculated, using the formula , and obtain the prediction error value MA; calculate the ratio of the predicted line loss value to the actual line loss value, and use the formula , and obtain the prediction error ratio MP; calculate the correlation between the predicted line loss value and the actual line loss value, and use the formula , and get the predicted correlation value MR; where, Represents the average value of all actual line loss values; marks the historical forecast fluctuation value, forecast error value, forecast error ratio and forecast correlation value as performance indicators;

[0013] S22: Divide the historical operation related data of the line loss circuit of each power consumption link, randomly divide the historical operation related data into k subsets of equal size according to their characteristics, and perform k-fold cross validation; in each round, use k-1 subsets as training sets and the remaining 1 subset as test sets; mark the training set and the test set as data sets;

[0014] S23: Train and verify the model, use the training set data to train the line loss prediction sub-model, and then use the test set to evaluate the performance of the line loss prediction sub-model, and calculate the predicted performance indicators according to step S21; perform fusion calculation on the predicted performance indicators, and use the formula , and obtain the performance evaluation value MZ; where v represents the error correction factor, and m1, m2, m3, and m4 represent the weights corresponding to the historical forecast fluctuation value, the forecast error value, the forecast error ratio, and the forecast correlation value, respectively;

[0015] S24: Evaluate the stability of the predicted performance indicators, and give priority to the data sets corresponding to the predicted performance indicators according to the size of the performance evaluation value; select the data set with the largest set of k compromise priorities as the retest data set, set the number of repeated tests, and use the retest data set to repeat step S23 according to the number of repeated tests to obtain the retest performance evaluation value ; Calculate the root mean square error of all retested performance evaluation values ​​using the formula , and obtain the fluctuation value FB of the retest data set; where p represents the total number of repeated tests, j represents the number of repeated tests, It represents the fluctuation value of repeated test number j. Represents the mean of the fluctuation value of the retest; calculate the mean of the fluctuation value of the retest data set to obtain the fluctuation mean FU; perform weighted calculation on the fluctuation value and the fluctuation mean of the retest data, and use the formula FC=FB×c1+FU×c2 to obtain the fluctuation assessment value FC; where c1 and c2 represent the weights corresponding to the fluctuation value and the fluctuation mean of the retest data respectively; select the data set corresponding to the smallest fluctuation assessment value as the data set for the final training model;

[0016] S3: Construct a total power grid loss prediction model from all line loss prediction sub-models. The model is expressed as ;in, It is expressed as the predicted line loss value of the entire power grid, n represents the total number of power consumption links, represents the weight of the line loss prediction sub-model corresponding to the power consumption link i;

[0017] S4: The operation-related data of each power consumption link in the smart grid is provided as input to the grid line loss prediction model, and the predicted line loss value of the smart grid is output; the predicted line loss value is displayed through a visual interface.

[0018] As a preferred embodiment of the present invention, the step of collecting operation-related data of each link line loss circuit includes obtaining data through smart meters, sensors and monitoring equipment.

[0019] As a preferred implementation of the present invention, it is also used to monitor abnormal line loss in the smart grid, specifically:

[0020] The operation-related data of the line loss circuit in each power consumption link are respectively input into the corresponding line loss prediction sub-model to obtain the predicted line loss value of the corresponding power consumption link; the line loss threshold of the power consumption link is set, and the predicted line loss value of the power consumption link is compared with its line loss threshold. If the predicted line loss value is greater than its line loss threshold, it means that the line loss of its power consumption link is abnormal, and the corresponding line loss abnormality detection signaling is generated, and the abnormal time, the location of the power consumption link, the predicted line loss value and the threshold are marked as abnormal detection process records; the line loss abnormality detection signaling is used to trigger the alarm mechanism, and the line loss abnormality detection signaling and the abnormality detection process record are sent to the operation and maintenance personnel.

[0021] As a preferred implementation of the present invention, the specific steps of obtaining the line loss threshold of the power consumption link are:

[0022] Step 1: Obtain line loss related information of the line loss circuit in the power consumption link; wherein the line loss related information includes historical operation data of the line loss circuit, line length and cross-section, and environmental information;

[0023] Step 2: Perform threshold impact analysis on line loss related information to obtain standard evaluation value of line loss circuit, line impact value and environmental evaluation value;

[0024] Step 3: Normalize the standard assessment value, line impact value and environmental assessment value to obtain the line loss assessment value of the power consumption link;

[0025] Step 4: Set a line loss threshold selection list for the power consumption link, match the line loss assessment value with the value range of any threshold in its line loss threshold selection list, and use the successfully matched line loss threshold as the line loss threshold for the power consumption link.

[0026] As a preferred implementation of the present invention, a threshold impact analysis is performed on line loss related information, and the specific analysis process is as follows:

[0027] Obtain the historical operation data of the line loss circuit in the power consumption link, calculate the difference between any parameter in the historical operation data and the corresponding parameter standard threshold, so as to obtain the standard deviation value corresponding to the parameter and express it as BZ; mark the moment when the power consumption link is first operated as the first moment, the current moment as the second moment, and the time area between the first moment and the second moment as the power consumption time zone; calculate the average value of all standard deviation values ​​to obtain the difference mean; calculate the standard deviation values ​​of all parameters in the power consumption time zone to obtain the difference wave value; calculate the difference between the maximum and minimum values ​​of the standard deviation value to obtain the floating difference; perform weighted calculation on the difference mean, difference wave value and floating difference to obtain the standard evaluation value;

[0028] Obtain the line length and cross-section values ​​of the line loss circuit, perform weighted calculation on the line length and cross-section values, and obtain the line impact value;

[0029] Obtain the environmental information of the line loss circuit in the power consumption link, set the standard environmental parameters of the line loss circuit, calculate the difference between any parameter in the environmental information and its standard environmental parameters, and obtain the deviation value corresponding to the environmental parameter; set a preset recent environmental change time zone, calculate the mean value, variance, and the difference between the maximum and minimum values ​​of the deviation values ​​corresponding to the environmental parameters in the recent environmental change time zone, and obtain the deviation mean value, deviation fluctuation value, and deviation difference value corresponding to the environmental parameters; perform weighted calculation on the deviation mean value, deviation fluctuation value, and deviation difference value corresponding to all environmental parameters to obtain the environmental assessment value.

[0030] Compared with the prior art, the present invention has the following beneficial effects:

[0031] 1. Through the application of deep learning technology, the present invention can process large-scale power grid operation data, provide real-time and accurate line loss prediction, and optimize model performance through automated methods, providing technical support for efficient management and operation of smart grids.

[0032] 2. The present invention constructs a dedicated line loss prediction sub-model for each power consumption link of the smart grid through a deep learning algorithm, thereby realizing refined management of power grid line losses, optimizing resource allocation, accurately capturing nonlinear relationships and long-term dependencies in the smart grid, and improving the accuracy of line loss prediction.

[0033] 3. The present invention collects and analyzes the operation-related data of the smart grid in real time and provides line loss prediction information in a timely manner, supports rapid response to changes in the grid status, and enhances the real-time monitoring capability of the smart grid line loss. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to facilitate understanding by those skilled in the art, the present invention is further described below with reference to the accompanying drawings.

[0035] Figure 1 It is a flow chart of a method for predicting line loss of a smart grid based on deep learning in the present invention.

[0036] Figure 2 It is a flowchart of the specific process of constructing a line loss prediction sub-model in a smart grid line loss prediction method based on deep learning in the present invention. DETAILED DESCRIPTION

[0037] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0038] See also Figure 1-Figure 2 As shown, a smart grid line loss prediction method based on deep learning includes the following steps:

[0039] S1: Collect the operation data of the line loss circuit in each link of electricity consumption; the various links of electricity consumption include power generation, transmission, transformation, distribution, consumption and management line loss; the operation data include current, voltage, power factor, line load, power metering, power frequency, temperature and environmental parameters, and the environmental parameters include wind speed, humidity, temperature, light intensity and electromagnetic interference intensity;

[0040] S2: Obtain the actual line loss value of the line loss circuit in each power consumption link, and build a line loss prediction sub-model for each power consumption link. The model is expressed as ,in, represents the predicted line loss value of power consumption link i, represents the set of corresponding parameters in the operation-related data in the power consumption link i, Representation Model The parameter set of

[0041] Among them, the line loss prediction sub-model includes:

[0042] Use long short-term memory networks or gated recurrent units for time series forecasting;

[0043] The specific process of constructing the line loss prediction sub-model is as follows:

[0044] S21: Evaluate the difference between the actual line loss value of the line loss circuit and the historical predicted line loss value, and obtain the predicted line loss value predicted by the line loss prediction sub-model in the historical record, expressed as , t represents the time index predicted by the line loss prediction sub-model; the actual line loss value of the line loss circuit is calculated with the predicted line loss value in the historical records, and the formula is used , and obtain the historical predicted fluctuation value MS; where Lt represents the actual line loss value at time t, represents the predicted line loss value at the time of prediction t, N represents the total number of time indexes predicted by the line loss prediction sub-model in the historical records; the average of the absolute values ​​of the differences between the predicted line loss value and the actual line loss value is calculated, using the formula , and obtain the prediction error value MA; calculate the ratio of the predicted line loss value to the actual line loss value, and use the formula , and obtain the prediction error ratio MP; calculate the correlation between the predicted line loss value and the actual line loss value, and use the formula , and get the predicted correlation value MR; where, Represents the average value of all actual line loss values; marks the historical forecast fluctuation value, forecast error value, forecast error ratio and forecast correlation value as performance indicators;

[0045] S22: Divide the historical operation related data of the line loss circuit of each power consumption link, randomly divide the historical operation related data into k subsets of equal size according to their characteristics, and perform k-fold cross validation; in each round, use k-1 subsets as training sets and the remaining 1 subset as test set; mark the training set and test set as data sets; for example, the historical operation related data is divided into 8 subsets, which are represented as subset 1, subset 2...subset 8, first, subsets 1 to 7 are used as training sets, and subset 8 is used as test set. In the first round of cross validation, the trained model is used to make predictions on the test set (subset 8); then in the second round of cross validation, subsets 1 to 6 and subset 8 are used as training sets, and subset 7 is used as test set; and this is repeated until each subset is used as a test set once;

[0046] S23: Train and verify the model, use the training set data to train the line loss prediction sub-model, and then use the test set to evaluate the performance of the line loss prediction sub-model, and calculate the predicted performance indicators according to step S21; perform fusion calculation on the predicted performance indicators, and use the formula , and obtain the performance evaluation value MZ; where v represents the error correction factor, and m1, m2, m3, and m4 represent the weights corresponding to the historical forecast fluctuation value, the forecast error value, the forecast error ratio, and the forecast correlation value, respectively;

[0047] S24: Evaluate the stability of the predicted performance indicators, and give priority to the data sets corresponding to the predicted performance indicators according to the size of the performance evaluation value; select the data set with the largest set of k compromise priorities as the retest data set, set the number of repeated tests, and use the retest data set to repeat step S23 according to the number of repeated tests to obtain the retest performance evaluation value ; Calculate the root mean square error of all retested performance evaluation values ​​using the formula , and obtain the fluctuation value FB of the retest data set; where p represents the total number of repeated tests, j represents the number of repeated tests, It represents the fluctuation value of repeated test number j. Represents the mean of the fluctuation value of the retest; calculate the mean of the fluctuation value of the retest data set to obtain the fluctuation mean FU; perform weighted calculation on the fluctuation value and the fluctuation mean of the retest data, and use the formula FC=FB×c1+FU×c2 to obtain the fluctuation assessment value FC; where c1 and c2 represent the weights corresponding to the fluctuation value and the fluctuation mean of the retest data respectively; select the data set corresponding to the smallest fluctuation assessment value as the data set for the final training model;

[0048] S3: Construct a total power grid loss prediction model from all line loss prediction sub-models. The model is expressed as ;in, It is expressed as the predicted line loss value of the entire power grid, n represents the total number of power consumption links, represents the weight of the line loss prediction sub-model corresponding to the power consumption link i;

[0049] S4: The operation-related data of each power consumption link in the smart grid is provided as input to the grid line loss prediction model, and the predicted line loss value of the smart grid is output; the predicted line loss value is displayed through a visual interface.

[0050] In the present application, the step of collecting operation-related data of each link line loss circuit includes obtaining it through smart meters, sensors and monitoring equipment.

[0051] It should be noted that smart meters can monitor the flow of electricity in real time and record power consumption and power quality parameters; sensors can measure various environmental parameters, such as temperature, humidity, wind speed, electromagnetic interference intensity, etc.; monitoring equipment provides more comprehensive and real-time data monitoring functions.

[0052] In this application, the present invention is also used to monitor abnormal line losses in smart grids, specifically:

[0053] The operation-related data of the line loss circuit in each power consumption link are respectively input into the corresponding line loss prediction sub-model to obtain the predicted line loss value of the corresponding power consumption link; the line loss threshold of the power consumption link is set, and the predicted line loss value of the power consumption link is compared with its line loss threshold. If the predicted line loss value is greater than its line loss threshold, it means that the line loss of its power consumption link is abnormal, and the corresponding line loss abnormality detection signaling is generated, and the abnormal time, the location of the power consumption link, the predicted line loss value and the threshold are marked as abnormal detection process records; the line loss abnormality detection signaling is used to trigger the alarm mechanism, and the line loss abnormality detection signaling and the abnormality detection process record are sent to the operation and maintenance personnel.

[0054] It should be noted that by utilizing real-time data and prediction models, abnormal line losses in smart grids can be discovered in a timely manner, thus improving the stability and efficiency of the power system.

[0055] In this application, the specific steps for obtaining the line loss threshold of the power consumption link are:

[0056] Step 1: Obtain line loss related information of the line loss circuit in the power consumption link; wherein the line loss related information includes historical operation data of the line loss circuit, line length and cross-section, and environmental information;

[0057] Step 2: Perform threshold impact analysis on line loss related information to obtain the standard evaluation value BF of the line loss circuit, the line impact value UG and the environmental evaluation value LG;

[0058] Step 3: Normalize the standard assessment value, line impact value and environmental assessment value, and use the formula , obtain the line loss assessment value PG of the power consumption link; wherein w represents the number of the node or line of the line loss circuit in the power consumption link, wBF, wUG, and wLG represent the standard assessment value, line impact value, and environmental assessment value of the node or line number w in the line loss circuit, respectively, and d1, d2, and d3 represent the weighted impact factors of the standard assessment value, line impact value, and environmental assessment value on the line loss threshold, respectively;

[0059] Step 4: Set a line loss threshold selection list for the power consumption link, match the line loss assessment value with the value range of any threshold in its line loss threshold selection list, and use the successfully matched line loss threshold as the line loss threshold for the power consumption link.

[0060] It should be noted that by determining the appropriate line loss threshold based on the historical operating data, actual conditions and environmental conditions of the line loss circuit, so as to timely detect and respond to abnormal line loss conditions, it will help improve the operating efficiency, stability and reliability of the smart grid.

[0061] In this application, the threshold impact analysis is performed on the line loss related information. The specific analysis process is as follows:

[0062] Obtain the historical operation data of the line loss circuit in the power consumption link, calculate the difference between any parameter in the historical operation data and the corresponding parameter standard threshold, and obtain the standard deviation value corresponding to the parameter, which is expressed as BZ; mark the moment when the power consumption link is first operated as the first moment, the current moment as the second moment, and the time area between the first moment and the second moment as the power consumption time zone; calculate the average value of all standard deviation values ​​to obtain the difference mean BS; calculate the standard deviation values ​​of all parameters in the power consumption time zone , and obtain the difference wave value CB; where C represents the total number of times in the electricity consumption time zone, represents the standard deviation of parameter a at the zth moment in the electricity consumption time zone C, It represents the difference mean of parameter a in the power consumption time zone in the historical operation data; the maximum and minimum values ​​of the standard deviation are calculated, and the floating difference is marked as FD; the difference mean, difference wave value and floating difference are weighted and calculated using the formula , and obtain the standard evaluation value BF; where b1, b2, and b3 represent the weights corresponding to the difference mean, difference wave value, and floating difference respectively;

[0063] The line length and cross-section values ​​of the line loss circuit are expressed as UG1 and UG2 respectively. The line length and cross-section values ​​are weighted and calculated using the formula , and obtain the line influence value UG; where u1 and u2 represent the weights corresponding to the line length and cross-section of the line loss circuit respectively;

[0064] Obtain the environmental information of the line loss circuit in the power consumption link, set the standard environmental parameters of the line loss circuit, calculate the difference between any parameter in the environmental information and its standard environmental parameters, and obtain the deviation value corresponding to the environmental parameter; set a preset recent environmental change time zone, calculate the mean value, variance, and the difference between the maximum and minimum values ​​of the deviation values ​​corresponding to the environmental parameters in the recent environmental change time zone, and obtain the deviation mean, deviation fluctuation value, and deviation difference value corresponding to the environmental parameters, which are marked as LG1, LG2, and LG3 respectively; perform weighted calculation on the deviation mean, deviation fluctuation value, and deviation difference value corresponding to all environmental parameters, and use the formula , and obtain the environmental assessment value LG; where g represents the index of the environmental parameter g in the environmental information, gLG1, gLG2, and gLG3 represent the deviation mean, deviation fluctuation, and deviation difference values ​​corresponding to the environmental parameter g, respectively, and l1, l2, and l3 represent the weights corresponding to the deviation mean, deviation fluctuation, and deviation difference values, respectively.

[0065] It should be noted that by analyzing the line loss related information, the standard assessment value of the line loss circuit, the line impact value and the environmental assessment value are determined, providing a basis for the subsequent setting of the line loss threshold.

[0066] Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments, or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A smart grid line loss prediction method based on deep learning, It is characterized in that The following steps are involved: S1: Collect the operation data of the line loss circuit in each link of electricity consumption; the various links of electricity consumption include power generation, transmission, transformation, distribution, consumption and management line loss; the operation data include current, voltage, power factor, line load, power metering, power frequency, temperature and environmental parameters, and the environmental parameters include wind speed, humidity, temperature, light intensity and electromagnetic interference intensity; S2: Obtain the actual line loss value of the line loss circuit in each power consumption link, and build a line loss prediction sub-model for each power consumption link. The model is expressed as ,in, represents the predicted line loss value of power consumption link i, represents the set of corresponding parameters in the operation-related data in the power consumption link i, Representation Model The parameter set of Among them, the line loss prediction sub-model includes: Use long short-term memory networks or gated recurrent units for time series forecasting; The specific process of constructing the line loss prediction sub-model is as follows: S21: Evaluate the difference between the actual line loss value of the line loss circuit and the historical predicted line loss value, and obtain the predicted line loss value predicted by the line loss prediction sub-model in the historical record, expressed as , t represents the time index predicted by the line loss prediction sub-model; the actual line loss value of the line loss circuit is calculated with the predicted line loss value in the historical records, and the formula is used , and obtain the historical predicted fluctuation value MS; where Lt represents the actual line loss value at time t, represents the predicted line loss value at the time of prediction t, N represents the total number of time indexes predicted by the line loss prediction sub-model in the historical records; the average of the absolute values ​​of the differences between the predicted line loss value and the actual line loss value is calculated, using the formula , and obtain the prediction error value MA; calculate the ratio of the predicted line loss value to the actual line loss value, and use the formula , and obtain the prediction error ratio MP; calculate the correlation between the predicted line loss value and the actual line loss value, and use the formula , and get the predicted correlation value MR; where, Represents the average value of all actual line loss values; marks the historical forecast fluctuation value, forecast error value, forecast error ratio and forecast correlation value as performance indicators; S22: Divide the historical operation related data of the line loss circuit of each power consumption link, randomly divide the historical operation related data into k subsets of equal size according to their characteristics, and perform k-fold cross validation; in each round, use k-1 subsets as training sets and the remaining 1 subset as test sets; mark the training set and the test set as data sets; S23: Train and verify the model, use the training set data to train the line loss prediction sub-model, and then use the test set to evaluate the performance of the line loss prediction sub-model, and calculate the predicted performance indicators according to step S21; perform fusion calculation on the predicted performance indicators, and use the formula , and obtain the performance evaluation value MZ; where v represents the error correction factor, and m1, m2, m3, and m4 represent the weights corresponding to the historical forecast fluctuation value, the forecast error value, the forecast error ratio, and the forecast correlation value, respectively; S24: Evaluate the stability of the predicted performance indicators, and give priority to the data sets corresponding to the predicted performance indicators according to the size of the performance evaluation value; select the data set with the largest set of k compromise priorities as the retest data set, set the number of repeated tests, and use the retest data set to repeat step S23 according to the number of repeated tests to obtain the retest performance evaluation value ; Calculate the root mean square error of all retested performance evaluation values ​​using the formula , and obtain the fluctuation value FB of the retest data set; where p represents the total number of repeated tests, j represents the number of repeated tests, It represents the fluctuation value of repeated test number j. Represents the mean of the fluctuation value of the retest; calculate the mean of the fluctuation value of the retest data set to obtain the fluctuation mean FU; perform weighted calculation on the fluctuation value and the fluctuation mean of the retest data, and use the formula FC=FB×c1+FU×c2 to obtain the fluctuation assessment value FC; where c1 and c2 represent the weights corresponding to the fluctuation value and the fluctuation mean of the retest data respectively; select the data set corresponding to the smallest fluctuation assessment value as the data set for the final training model; S3: Construct a total power grid loss prediction model from all line loss prediction sub-models. The model is expressed as ;in, It is expressed as the predicted line loss value of the entire power grid, n represents the total number of power consumption links, represents the weight of the line loss prediction sub-model corresponding to the power consumption link i; S4: The operation-related data of each power consumption link in the smart grid is provided as input to the grid line loss prediction model, and the predicted line loss value of the smart grid is output; the predicted line loss value is displayed through a visual interface.

2. According to the method for predicting line loss of smart grid based on deep learning in claim 1, It is characterized in that The steps for collecting operation-related data of line loss circuits in each link include obtaining data from smart meters, sensors and monitoring equipment.

3. According to the method for predicting line loss of smart grid based on deep learning in claim 1, It is characterized in that It is also used to monitor abnormal line losses in smart grids, specifically: The operation-related data of the line loss circuit in each power consumption link are respectively input into the corresponding line loss prediction sub-model to obtain the predicted line loss value of the corresponding power consumption link; the line loss threshold of the power consumption link is set, and the predicted line loss value of the power consumption link is compared with its line loss threshold. If the predicted line loss value is greater than its line loss threshold, it means that the line loss of its power consumption link is abnormal, and the corresponding line loss abnormality detection signaling is generated, and the abnormal time, the location of the power consumption link, the predicted line loss value and the threshold are marked as abnormal detection process records; the line loss abnormality detection signaling is used to trigger the alarm mechanism, and the line loss abnormality detection signaling and the abnormality detection process record are sent to the operation and maintenance personnel.

4. According to the method for predicting line loss of smart grid based on deep learning in claim 3, It is characterized in that The specific steps for obtaining the line loss threshold of the power consumption link are as follows: Step 1: Obtain line loss related information of the line loss circuit in the power consumption link; wherein the line loss related information includes historical operation data of the line loss circuit, line length and cross-section, and environmental information; Step 2: Perform threshold impact analysis on line loss related information to obtain the standard evaluation value BF of the line loss circuit, the line impact value UG and the environmental evaluation value LG; Step 3: Normalize the standard assessment value, line impact value and environmental assessment value, and use the formula , obtain the line loss assessment value PG of the power consumption link; wherein w represents the number of the node or line of the line loss circuit in the power consumption link, wBF, wUG, and wLG represent the standard assessment value, line impact value, and environmental assessment value of the node or line number w in the line loss circuit, respectively, and d1, d2, and d3 represent the weighted impact factors of the standard assessment value, line impact value, and environmental assessment value on the line loss threshold, respectively; Step 4: Set a line loss threshold selection list for the power consumption link, match the line loss assessment value with the value range of any threshold in its line loss threshold selection list, and use the successfully matched line loss threshold as the line loss threshold for the power consumption link.

5. According to claim 4, a method for predicting line loss in a smart grid based on deep learning, It is characterized in that The threshold impact analysis is performed on the line loss related information. The specific analysis process is as follows: Obtain the historical operation data of the line loss circuit in the power consumption link, calculate the difference between any parameter in the historical operation data and the corresponding parameter standard threshold, and obtain the standard deviation value corresponding to the parameter, which is expressed as BZ; Mark the moment when the power consumption link is first operated as the first moment, the current moment as the second moment, and the time area between the first moment and the second moment as the power consumption time zone; calculate the average value of all standard deviations to obtain the mean difference BS; calculate the standard deviation values ​​of all parameters in the power consumption time zone , and obtain the difference wave value CB; where C represents the total number of times in the electricity consumption time zone, represents the standard deviation of parameter a at the zth moment in the electricity consumption time zone C, It represents the difference mean of parameter a in the power consumption time zone in the historical operation data; the maximum and minimum values ​​of the standard deviation are calculated, and the floating difference is marked as FD; the difference mean, difference wave value and floating difference are weighted and calculated using the formula , and obtain the standard evaluation value BF; where b1, b2, and b3 represent the weights corresponding to the difference mean, difference wave value, and floating difference respectively; The line length and cross-section values ​​of the line loss circuit are expressed as UG1 and UG2 respectively. The line length and cross-section values ​​are weighted and calculated using the formula , and obtain the line influence value UG; where u1 and u2 represent the weights corresponding to the line length and cross-section of the line loss circuit respectively; Obtain the environmental information of the line loss circuit in the power consumption link, set the standard environmental parameters of the line loss circuit, calculate the difference between any parameter in the environmental information and its standard environmental parameters, and obtain the deviation value corresponding to the environmental parameter; set a preset recent environmental change time zone, calculate the mean value, variance, and the difference between the maximum and minimum values ​​of the deviation values ​​corresponding to the environmental parameters in the recent environmental change time zone, and obtain the deviation mean, deviation fluctuation value, and deviation difference value corresponding to the environmental parameters, which are marked as LG1, LG2, and LG3 respectively; perform weighted calculation on the deviation mean, deviation fluctuation value, and deviation difference value corresponding to all environmental parameters, and use the formula , and obtain the environmental assessment value LG; where g represents the index of the environmental parameter g in the environmental information, gLG1, gLG2, and gLG3 represent the deviation mean, deviation fluctuation, and deviation difference values ​​corresponding to the environmental parameter g, respectively, and l1, l2, and l3 represent the weights corresponding to the deviation mean, deviation fluctuation, and deviation difference values, respectively.