A method for analyzing the main causes of line loss and related equipment

Abstract

This invention discloses a method and related equipment for analyzing the root causes of line losses. The method includes: acquiring abnormal data related to line losses to form a first set, the first set containing causal abnormal data and result abnormal data with causal relationships; removing result abnormal data from the first set to obtain a second set containing causal abnormal data; processing the second set to determine the contribution of the abnormal data in the second set to line losses; and determining the root cause of the line losses based on the contribution. In embodiments of this invention, the abnormal data is first filtered using causal relationships, and then sorted using contribution, this two-stage abnormal data analysis method can quickly locate the root cause. When calculating the contribution, interference from fixed line losses is reduced. The random forest algorithm is used to construct the regression model, taking into account both the independence and correlation between abnormal data in the power industry, making the analysis results more accurate.

Description

Technical Field

[0001] This invention generally relates to the power industry. More specifically, this invention relates to a method for analyzing the root causes of line losses and related equipment. Background Technology

[0002] In the power industry, managing distribution network line losses is a crucial task for power supply companies. Line loss refers to the loss of electricity within distribution lines, and many factors influence it, such as abnormal user electricity consumption, abnormal data acquisition, and abnormal equipment metering. Reducing line losses by addressing these anomalies is a common line loss management method.

[0003] Currently, power distribution network information acquisition systems can automatically identify abnormal line conditions such as abnormal user electricity consumption, abnormal data acquisition, and abnormal equipment metering by monitoring and analyzing electricity consumption data such as voltage, power, and power. The focus is on analyzing the relationship between anomalies and line loss data, such as anomaly correlation analysis and decision tree path analysis. The invention with publication number CN112597704A uses a deep learning attention mechanism to obtain the correlation between anomalies and data, and determines the cause of the anomaly through the correlation. The invention patent with publication number CN112598234A uses the AdaBoost ensemble tree algorithm to analyze the cause of anomalies. However, these technologies have certain drawbacks: correlation analysis is easily affected by confounding variables, leading to incorrect outputs; primary and secondary factors cannot be distinguished; model features are disconnected from actual anomalies; and the mapping relationship between features and maintenance work needs to be manually created.

[0004] In view of this, there is an urgent need to provide a method for analyzing the main causes of line loss. Summary of the Invention

[0005] To address at least one or more of the technical problems mentioned above, this invention provides a method for analyzing the root causes of line loss, comprising: acquiring abnormal data to obtain a first set, the first set including causal abnormal data and result abnormal data with causal relationships; removing result abnormal data from the first set to obtain a second set; acquiring the contribution of the abnormal data in the second set to line loss; and determining the root cause of line loss based on the contribution.

[0006] By first removing abnormal data, the abnormal data causing line loss becomes more concentrated, thus performing a screening at the level of abnormal data. Then, a second screening is performed using contribution, allowing the main cause of line loss to be quickly located. This two-stage main cause analysis ensures the accuracy of the output results.

[0007] Furthermore, in some embodiments, the causal anomaly data and the result anomaly data in the first set are determined by training a causal association rule model between the anomaly data to determine the causal relationship between the anomaly data:

[0008]

[0009] Where X and Y are the outlier data; UpLift X,Y∈group(a),Z∈others(o) (X→Y) represents the degree to which abnormal data in X leads to abnormal data in Y. Let Y be the probability of X occurring. Let X be the probability of Y if X does not occur; X,Y∈group(a) means that X and Y belong to the same group a; Z∈others(o) means other abnormal data outside group a; when uplift is greater than a preset threshold, it means that there is a causal relationship X→Y, where X is the cause of the abnormal data and Y is the result of the abnormal data.

[0010] Furthermore, in some embodiments, when calculating the contribution of abnormal data to line loss, the line loss is obtained by subtracting the fixed line loss from the actual line loss.

[0011] By removing fixed line losses before calculating the contribution, the impact of abnormal data on line loss can be highlighted.

[0012] Furthermore, in some embodiments, the fixed line loss is obtained by performing linear regression on the actual line loss using the electricity consumption data, removing the electricity consumption data whose residual exceeds a preset residual threshold, and obtaining the first data; performing linear regression on the first data against the actual line loss, and obtaining the intercept as the fixed line loss.

[0013] Furthermore, in some embodiments, the contribution of abnormal data in the second set to line loss is obtained by: establishing a mapping relationship between abnormal data and line loss, and calculating the contribution value of each abnormal data to line loss.

[0014] Furthermore, in some embodiments, a regression model of abnormal data on line loss is constructed using the random forest algorithm to establish a mapping relationship between abnormal data and line loss; the contribution value of abnormal data to line loss is calculated based on the regression model using the TreeShap algorithm.

[0015] Furthermore, in some embodiments, the line loss cause analysis method further includes obtaining the shap values ​​of abnormal data in multiple transformer areas, taking the average value as the default contribution value; and sorting all abnormal data according to the default contribution value as the default abnormal data sort.

[0016] Furthermore, in some embodiments, the line loss cause analysis method further includes obtaining the shap values ​​of abnormal data in multiple transformer areas and taking the average value as the default contribution value; the importance value S(i)′ of abnormal data i is calculated using the following formula:

[0017] S(i)′=p*S(i) o+(1-p)*S(i), where, S(i) o S(i) represents the default contribution value of abnormal data i; S(i) represents the contribution value of abnormal data i in the current transformer area, and P is the weighting value.

[0018] Furthermore, according to another aspect of the present invention, a processing apparatus is provided, comprising: a processor configured to execute program instructions; and a memory configured to store the program instructions, which, when loaded and executed by the processor, cause the processor to perform the line loss cause analysis method described in any of the preceding claims.

[0019] According to another aspect of the present invention, a computer-readable storage medium is provided, characterized in that it stores program instructions that, when loaded and executed by a processor, cause the processor to perform the line loss cause analysis method described in any of the preceding claims.

[0020] In embodiments of this invention, abnormal data are first filtered using causal relationships, and then ranked using contribution rates. This two-stage abnormal data analysis method can quickly locate the primary cause of abnormal data. When calculating the contribution rate, linear regression is used to calculate the intercept term to represent the linear fixed loss of the transformer area, and then the SHAP algorithm is used to estimate the remaining line loss, reducing the interference caused by fixed line losses and making the results more accurate. The random forest algorithm is used to construct the regression model, taking into account both the independence and correlation between abnormal data in the power industry, further improving the accuracy of the analysis results. Attached Figure Description

[0021] The above and other objects, features, and advantages of exemplary embodiments of the present invention will become readily apparent upon reading the following detailed description with reference to the accompanying drawings. In the drawings, several embodiments of the invention are illustrated by way of example and not limitation, and like or corresponding reference numerals denote like or corresponding parts, wherein:

[0022] Figure 1 A schematic diagram of a line loss cause analysis method according to an embodiment of the present invention is shown;

[0023] Figure 2 A schematic diagram illustrating the steps of calculating the contribution value based on the regression of abnormal data on line loss according to an embodiment of the present invention is shown.

[0024] Figure 3 A schematic block diagram of a system for analyzing the main causes of line loss according to an embodiment of the present invention is shown. Detailed Implementation

[0025] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] It should be understood that the terms "comprising" and "including" as used in the specification and claims of this invention indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0027] It should also be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this specification and claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes such combinations.

[0028] As used in this specification and claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if [described condition or event] is detected" may be interpreted, depending on the context, as "once determined," "in response to determination," "once [described condition or event] is detected," or "in response to detection of [described condition or event]."

[0029] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0030] Figure 1This paper illustrates a method for analyzing the root causes of line loss using embodiments of the present invention. The method includes: performing operation 101 to acquire abnormal data regarding line loss to form a first set, which includes causal abnormal data and result abnormal data with causal relationships. According to one embodiment of the invention, abnormal data refers to events related to electricity consumption, data collection, metering, etc., that reflect abnormal data status of the line and are associated with line loss. Abnormal data is acquired by collecting data on a per-transformer-area basis. Abnormal data occurring simultaneously in each transformer area each day constitutes an abnormal data group. A transformer area refers to the coverage area of ​​each transformer. A causal relationship rule model between abnormal data is trained, and a causal relationship graph is constructed based on the causal relationships between abnormal data. In step 102, the result abnormal data in the first set is removed using the causal relationship graph to obtain a list of causal abnormal data, which is the second set. The second set includes causal abnormal data and abnormal data that do not have a causal relationship with other abnormal data.

[0031] Then, operation 103 is executed to process the second set, obtain the contribution of abnormal data in the second set to line loss, and then operation 104 is executed to determine the main cause of line loss based on the contribution. The contribution of abnormal data to line loss in the transformer area can be calculated using machine learning model interpretability algorithms.

[0032] In the above embodiments of the present invention, a list of main causes of abnormal data in a distribution area is obtained through the causal relationship between abnormal data, and then the ranking of the list of main cause abnormal data is determined by the contribution value to the line loss of the distribution area. Based on the mutual influence between abnormal data that cause line loss in the power industry, the causal relationship between abnormal data is used for initial screening, reducing the processing workload in contribution calculation. The obtained main causes of line loss can more accurately and directly reflect the relevant elements for adjusting and reducing line loss, making it easier for users to adjust line loss in a targeted manner.

[0033] According to one embodiment of the present invention, the cause anomaly data and the result anomaly data in the first set are determined in the following manner:

[0034] Train a causal association rule model among outlier data to determine the causal relationships between them.

[0035] Where X and Y are the outlier data; UpLift X,Y∈group(a),Z∈others(o) (X→Y) represents the degree to which abnormal data in X leads to abnormal data in Y. It is calculated by subtracting the probability of Y when X occurs from the probability of Y when X does not occur. Let Y be the probability of X occurring. Let X be the probability of Y when X does not occur; X, Y∈group(a) means that X and Y belong to the same group a; Z∈others(o) means that abnormal data from other groups are not in group a; when uplift is greater than a preset threshold, it means that there is a causal relationship between X and Y, where X is the abnormal data of the cause and Y is the abnormal data of the result.

[0036] In other words, the causal effect value, conditional average treatment effect (CATE), is calculated using an uplift-type causal analysis model. This is the value calculated using uplift, where "treatment" refers to the abnormal data in X in the formula, and "effect" refers to the probability of abnormal data occurring in Y in the formula. A threshold is set for CATE. Relationships exceeding the threshold are defined as causal relationships. Uplift is a type of causal analysis method. Because CATE estimation involves enumerating all groupings for all conditions (group(a) and others(o), it has very high computational complexity. Therefore, a common solution is to use machine learning algorithms to estimate the probability of Y occurring under conditional groupings. Common algorithms include improved vertical score matching algorithms, such as dual machine learning and vertical random forests.

[0037] Figure 2 The diagram illustrates the steps for calculating the contribution value based on the regression of abnormal data on line loss.

[0038] According to one embodiment of the present invention, such as Figure 2 Operation 203: When calculating the contribution of abnormal data to line loss, the line loss is obtained by subtracting the fixed line loss from the actual line loss.

[0039] In analyzing the main factors affecting line loss, besides the impact of abnormal data states such as abnormal user electricity consumption data, abnormal data collection data, and abnormal equipment metering data, there is also the fixed loss of the line transmission itself, also known as fixed line loss. This part of the loss is usually unrelated to abnormal data and is generally unaffected by it. However, the existence of fixed line loss will affect the calculation results when calculating the contribution of abnormal data to the line loss of the distribution area. According to one embodiment of the present invention, as... Figure 2 Step 202: Before calculating the contribution, remove the fixed line loss to highlight the impact of abnormal data on line loss.

[0040] According to one embodiment of the present invention, the fixed line loss is obtained in the following manner: Figure 2 Step 201: Perform linear regression on the electricity consumption data against the actual line loss, and remove the electricity consumption data whose residual exceeds the preset residual threshold to obtain the first data; Step 202: Perform linear regression on the first data against the actual line loss, and calculate the intercept as the fixed line loss.

[0041] For example, before calculating the fixed line loss, the daily electricity consumption data of users in the transformer area and the daily line loss data of the transformer area are obtained. The data is rotated to form a table with the electricity consumption data in rows by date and columns by user number. Finally, a column for the daily line loss of the transformer area is added. Noisy data affecting the accuracy of the regression model is removed: a 30-day sliding window regression is used, and outlier values ​​are removed from the regression residuals using twice the standard deviation. Then, linear regression is performed on the line loss data using the user electricity consumption data to determine the intercept term as the theoretical fixed line loss of the transformer area. If multiple transformer areas exist, the corresponding fixed line loss is calculated using the data for each area. The intercept term of the linear regression is easy to calculate and quantify; other regression methods can also be used to determine the fixed line loss, which is not limited in this invention.

[0042] According to one embodiment of the present invention, the contribution of abnormal data in the second set to line loss is obtained by: establishing a mapping relationship between abnormal data and line loss, and calculating the contribution value of each abnormal data point to line loss. To calculate the contribution value of abnormal data to line loss, firstly, an interpretability model is used to construct a mapping relationship between abnormal data and line loss; then, based on the interpretability model, a machine learning model interpretability algorithm is used to calculate the contribution of abnormal data to the line loss of the transformer area.

[0043] Commonly used model interpretability algorithms include linear regression coefficients, tree model feature importance, and partial dependency graphs.

[0044] According to one embodiment of the present invention, such as Figure 2 Step 203 involves constructing a regression model of outlier data on line loss using the random forest algorithm, establishing a mapping relationship between outlier data and line loss; then, using the TreeShap algorithm, based on the regression model, the contribution of outlier data to line loss is calculated. The TreeShap algorithm is an approximate calculation method of the Shapley algorithm. The Shapley value is an estimation variable designed based on joint game theory, and the algorithm for calculating the contribution of outlier data to the target value is a theoretically well-established algorithm for interpretability analysis of machine learning models. Similar to CATE, the Shapley algorithm has high computational complexity; therefore, tree-based models such as random forests are used to approximate the Shapley value. The Shapley value, instead of the regression term, is used as the contribution of outlier data to line loss because, according to the arguments in Shapley-related papers, the Shapley value provides a more accurate and reasonable estimate of the contribution.

[0045] For example: First, process the abnormal data into a table, where rows represent transformer substations and columns represent abnormal data. In the table cells, if the abnormal data occurred in a transformer substation, the value is the substation's electricity consumption; otherwise, the value is 0. Add a column for the daily line loss of the transformer substation.

[0046] In one embodiment of the present invention, such as Figure 2 Step 204 in the process involves calculating the SHAP value for each transformer area using a single-sample SHAP value calculation method. Calculating the SHAP value at the smallest granularity transformer area level balances the interpretation needs of both local and global data. Global and local SHAP importance values ​​are obtained through global or grouped averaging methods. In establishing the mapping between outlier data and line loss, a random forest model is used to consider both the mutual influence between outlier data points and the impact of individual outlier data points on line loss. Individual decision trees isolate the mutual influence between outlier data points, emphasizing the impact of individual outlier data points, while the ensemble learning model further reflects the mutual influence between outlier data points.

[0047] According to one embodiment of the present invention, the method for analyzing the main causes of line loss further includes: obtaining the shap values ​​of abnormal data in multiple transformer substations, and taking the average value as a default contribution value; sorting all abnormal data according to the default contribution value, and using this as the default abnormal data sort. When there are multiple transformer substations, the average shap value of each abnormal data in each transformer substation is taken as the default contribution value of each abnormal data, thereby determining the main cause of line loss.

[0048] According to an embodiment of the present invention, the line loss main cause analysis method further includes obtaining the shap values ​​of abnormal data in multiple transformer areas and taking the average value as the default contribution value;

[0049] The formula for calculating the shap importance value S(i)′ of outlier data i is as follows: S(i)′=p*S(i) o +

[0050] (1-p)*S(i), where S(i) o S(i) represents the default contribution value of abnormal data i. S(i) represents the contribution value of abnormal data i in the current transformer area. P is the weighting value.

[0051] For example, during online use, when dealing with abnormal data in a certain transformer area, the first step is to obtain a list of cause-related abnormal data for that area. Specifically, by tracing back the parent node of each abnormal data node in the area according to the cause-effect graph, the initial parent node's abnormal data is found recursively. Then, the line loss contribution of each abnormal data in the cause-related abnormal data list is calculated using random forest and the Treeshap algorithm. To address the uncertainty of abnormal occurrences in the transformer area, i.e., sporadic anomalies, a weighted integration method of online calculated abnormal data and default abnormal data is used to determine the final contribution and ranking of the abnormal data. The Treeshap algorithm is chosen because it has relatively high approximate accuracy and higher computational efficiency than kernel methods.

[0052] Figure 3A schematic block diagram of a system 300 for analyzing the root causes of line loss according to an embodiment of the present invention is shown. The analysis system 300 may include a device 301 according to an embodiment of the present invention, as well as its peripheral devices and an external network, wherein the device 301 is used to perform operations such as data cleaning and processing.

[0053] like Figure 3 As shown, device 301 may include CPU 3011, which may be a general-purpose CPU, a dedicated CPU, or other information processing and program execution unit. Furthermore, device 301 may also include a mass storage device 3012 and a read-only memory (ROM) 3013. The mass storage device 3012 can be configured to store various types of data, and the ROM 3013 can be configured to store data required for power-on self-test of device 301, initialization of various functional modules in the system, drivers for basic input / output of the system, and data required to boot the operating system.

[0054] Furthermore, the test device 300 also includes other hardware platforms or components, such as the TPU 3014, GPU 3015, FPGA 3016, and MLU 3017 shown. It is understood that although various hardware platforms or components are shown in the test device 300, they are merely exemplary and not limiting; those skilled in the art can add or remove corresponding hardware as needed. For example, device 301 may include only a CPU as a known hardware platform and another hardware platform as the test hardware platform of this invention.

[0055] The device 301 of the present invention also includes a communication interface 3018, through which it can connect to a local area network / wireless local area network (LAN / WLAN) 305, and further through the LAN / WLAN to connect to a local server 306 or to the Internet (“Internet”) 307. Alternatively or additionally, the device 301 of the present invention can also connect directly to the Internet or a cellular network via the communication interface 3018 based on wireless communication technology, such as third-generation (“3G”), fourth-generation (“4G”), or fifth-generation (“5G”) wireless communication technology. In some application scenarios, the device 301 of the present invention can also access a server 308 on an external network and, possibly, a database 309, as needed to obtain various known neural network models, data, and modules, and can remotely store various measured data.

[0056] Peripherals of device 301 may include a display device 302, an input device 303, and a data transmission interface 304. In one embodiment, the display device 302 may include, for example, one or more speakers and / or one or more visual displays, configured to provide voice prompts and / or display images and videos of the computational process or final results of the test device of the present invention. The input device 303 may include, for example, a keyboard, mouse, microphone, gesture capture camera, or other input buttons or controls, configured to receive input of test data or user instructions. The data transmission interface 304 may include, for example, a serial interface, parallel interface, or Universal Serial Bus interface (“USB”), Small Computer System Interface (“SCSI”), Serial ATA, FireWire (“FireWire”), PCI Express, and High Definition Multimedia Interface (“HDMI”), configured for data transmission and interaction with other devices or systems. According to the present invention, the data transmission interface 304 can receive raw data or data after type conversion for analysis and transmit various types of data and results to device 301.

[0057] The CPU 3011, mass storage 3012, read-only memory ROM 3013, TPU 3014, GPU 3015, FPGA 3016, MLU 3017, and communication interface 3018 of the device 301 of the present invention can be interconnected via bus 3019, and can interact with peripheral devices through this bus. In one embodiment, the CPU 3011 can control other hardware components in the device 301 and its peripheral devices through this bus 3019.

[0058] It should also be understood that any module, unit, component, server, computer, terminal, or device that executes the instructions in this invention may include or otherwise access computer-readable media, such as storage media, computer storage media, or data storage devices (removable) and / or non-removable) such as disks, optical discs, or magnetic tapes. Computer storage media may include volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information, such as computer-readable instructions, data structures, program modules, or other data.

[0059] Based on the above, the present invention also discloses a computer-readable storage medium storing program instructions adapted for loading and execution by a processor of a high-voltage user electricity theft analysis method based on current loss anomaly data.

[0060] Computer-readable storage media can be any suitable magnetic or magneto-optical storage medium, such as resistive random access memory (RRAM), dynamic random access memory (DRAM), static random access memory (SRAM), enhanced dynamic random access memory (EDRAM), high-bandwidth memory (HBM), hybrid memory cube (HMC), etc., or any other medium that can be used to store desired information and can be accessed by an application, module, or both. Any such computer storage medium can be part of a device or accessible to or connected to a device. Any application or module described in this invention can be implemented using computer-readable / executable instructions that can be stored or otherwise retained by such a computer-readable medium.

[0061] It should be understood that the possible terms "first" or "second," etc., in the claims, specification, and drawings of this invention are used to distinguish different objects, rather than to describe a specific order. The terms "comprising" and "including" as used in the specification and claims of this invention indicate the presence of the described features, integrals, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or collections thereof.

[0062] It should also be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this specification and claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes such combinations.

[0063] While the embodiments of the present invention are described above, these are merely examples for the purpose of facilitating understanding of the invention and are not intended to limit the scope or application scenarios of the invention. Any person skilled in the art can make any modifications and changes in form and detail of the implementation without departing from the spirit and scope disclosed herein; however, the patent protection scope of the present invention shall still be determined by the scope defined in the appended claims.

[0064] It should also be understood that any module, unit, component, server, computer, terminal, or device that executes the instructions executorized herein may include or otherwise access computer-readable media, such as storage media, computer storage media, or data storage devices (removable) and / or non-removable) such as disks, optical discs, or magnetic tapes. Computer storage media may include volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information, such as computer-readable instructions, data structures, program modules, or other data.

[0065] While numerous embodiments of the invention have been shown and described herein, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Many modifications, alterations, and alternatives will occur to those skilled in the art without departing from the spirit and essence of the invention. It should be understood that various alternatives to the embodiments of the invention described herein may be employed in the practice of the invention. The appended claims are intended to define the scope of protection of the invention and therefore cover equivalents or alternatives within the scope of these claims.

Claims

1. A line loss main cause analysis method, characterized by, The method comprises: obtaining abnormal data related to line loss to form a first set, wherein the first set contains cause abnormal data and result abnormal data having a causal relationship; removing the result abnormal data in the first set to obtain a second set containing cause abnormal data; processing the second set to determine the contribution of abnormal data in the second set to line loss; and determining the main cause of line loss according to the contribution. The cause abnormal data and result abnormal data in the first set are determined by:

2. The line loss root cause analysis method of claim 1, wherein, training a causal relationship rule model between abnormal data to determine the causal relationship between abnormal data: wherein, X and Y are abnormal data; X, Y ∈ group(a) indicates that X and Y belong to the same group a; UpLift X,Y∈group(a),Z∈others(o) (X→Y) is: the degree of promotion of X abnormal data leading to Y abnormal data; Probability of Y occurring given X; P(Y|X) is the probability of Y given X; Z ∈ others(o) indicates all other abnormal data not in a; When the uplift is greater than a preset threshold, it indicates that there is an X→Y causal relationship, X is the cause abnormal data, and Y is the result abnormal data. In calculating the contribution of abnormal data to line loss, the line loss is obtained by subtracting the line fixed loss from the actual line loss.

3. The line loss root cause analysis method of claim 1, wherein, The line fixed loss is obtained by:

4. The line loss main cause analysis method according to claim 3, characterized by, performing linear regression on the actual line loss with power consumption data, removing power consumption data with residual error exceeding a preset residual error threshold to obtain first data; performing linear regression on the actual line loss with the first data to obtain the intercept as the line fixed loss. The contribution of abnormal data in the second set to line loss is obtained by:

5. The line loss root cause analysis method of claim 1, wherein, establishing a mapping relationship between abnormal data and line loss, and calculating the contribution value of each abnormal data to line loss.

6. The line loss main cause analysis method according to claim 5, wherein: a regression model of abnormal data to line loss is constructed by a random forest algorithm to establish a mapping relationship between abnormal data and line loss; a contribution value of abnormal data to line loss is calculated based on the regression model by a treeshap algorithm. Further comprising:

7. The line loss root cause analysis method of claim 6, wherein, obtaining the shap value of abnormal data in multiple areas, and taking the mean value as the default contribution value; sorting all abnormal data according to the default contribution value as the default abnormal data sorting. Further comprising:

8. The line loss root cause analysis method of claim 6, wherein, obtaining the shap value of abnormal data in multiple areas, and taking the mean value as the default contribution value; The importance value S(i)' of abnormal data i is calculated according to the following formula: wherein, S(i)' = p * S(i) o + (1 - p) * S(i) S(i) represents the contribution value of abnormal data i in the current area; S(i) o default contribution value for abnormal data i; P is a weighting value. The method comprises:

9. A processing device, characterized by a processor configured to execute program instructions; and a memory configured to store the program instructions, which, when loaded and executed by the processor, cause the processor to execute the line loss main cause analysis method according to any one of claims 1 to 8. The program instructions are stored in the memory, which, when loaded and executed by the processor, cause the processor to execute the line loss main cause analysis method according to any one of claims 1 to 8. ​ 10. A computer-readable storage medium, characterized in that, ​

Images