Electricity theft analysis method, device and storage medium

Abstract

The present invention provides a method, device and storage medium for analyzing electricity theft, and relates to the technical field of safe electricity use. The present invention collects and analyzes historical and current current data, uses a deep neural network to train an electricity theft prediction model, and realizes real-time detection of abnormal power nodes. The first suspicious power node is identified according to the power loss deviation of the abnormal node, and the second suspicious node is screened out by the loss deviation of its adjacent nodes. A normal power value interval is set, and nodes not within this interval are identified as electricity theft risk nodes. Combined with historical electricity use behavior and temperature factors, an electricity calibration coefficient is generated, and the upper and lower thresholds of the power consumption of electricity theft are corrected. By comparing the actual loss deviation with the corrected threshold, the system can accurately determine the type of electricity theft. The present invention realizes accurate detection of electricity theft behavior and efficient management of the power system by constructing a topological network on the electricity consumption side and applying a deep neural network prediction model.

Description

Technical Field

[0001] The present invention relates to the technical field of safe electricity use, and in particular to an electricity theft analysis method, device and storage medium. Background Art

[0003] In the prior art, publication number CN107492043A discloses an electricity theft analysis method and device, which collects the user's electricity usage behavior data and the user's file data; the collected data is made into sample data, and the sample data includes training data, test data and prediction data, wherein the training data and the test data carry labels, and the labels are marked one by one according to the user's historical electricity theft records; the training data and the test data are loaded into a machine learning model established based on the Xgboost algorithm, and the model is trained and tested; the prediction data is loaded into the trained model to obtain the user's electricity theft behavior analysis results.

[0004] Insufficient existing technology:

[0005] First, the data collected is limited. The existing technology mainly relies on users' electricity usage behavior data and archival data, but these data cannot fully reflect users' actual electricity usage. Some electricity theft behaviors may be carried out through complex circuit modifications or the use of high-tech equipment. These behaviors may not be revealed in regular electricity usage data. In addition, the existing data cleaning and feature extraction have a low degree of automation. Although the existing technology mentions the steps of data cleaning and feature selection, these steps usually rely on manual intervention and have a low degree of automation, which will lead to inefficient data processing and easily introduce human errors, affecting the accuracy and reliability of the model.

[0006] Secondly, the model parameter settings used are fixed. In the prior art, the parameter settings of the Xgboost algorithm are relatively fixed, such as the model, loss function, and measurement method of each iteration of the classifier. These parameters need to be dynamically adjusted in different scenarios. Fixed parameter settings may cause the model to perform poorly when facing complex electricity theft behaviors, and the model optimization method is single. In the prior art, the model is optimized by modifying the feature items and model parameters of the data, but these optimization methods are relatively single, lacking diversified optimization strategies and lack of deep optimization of the model structure, such as introducing deep learning models or integrated learning methods to improve the generalization ability and prediction accuracy of the model.

[0007] Therefore, it is necessary to provide a method, device and storage medium for analyzing electricity theft to solve the above problems.

[0008] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0009] The object of the present invention is to provide a method, device and storage medium for analyzing electricity theft, so as to solve the problems raised in the above-mentioned background technology.

[0010] To achieve the above object, the present invention provides the following technical solutions:

[0011] A method for analyzing electricity theft, the specific steps comprising:

[0012] Step 1: Collect the topological structure data of the power transmission line, and construct the topological network of the power consumption side according to the topological structure data;

[0013] Step 2: Based on the topological network, the current data of the power nodes on the power consumption side are simulated in real time, including historical current data and current current data. A power theft prediction model is constructed based on a deep neural network. The historical current data is used as input, and whether the corresponding power node is an abnormal power node is used as a label to train the power theft prediction model. The current current data is input into the trained power theft prediction model to obtain the current abnormal power node.

[0014] Step 3: based on the power loss deviation of the abnormal power node, a first suspicious power node is selected from the abnormal power nodes, and based on the power loss deviation of the power nodes adjacent to the first suspicious power node, a second suspicious power node is selected from the first suspicious power node;

[0015] Step 4: Set a normal power value range, filter out the second suspicious power node whose power value is not within the normal power value range as a dangerous power theft node, and collect historical power consumption data, including holiday power consumption and temperature data, and generate a power consumption calibration coefficient based on the holiday power consumption and temperature data;

[0016] Step 5: Set the upper and lower thresholds of the power consumption of electricity theft, and correct the upper and lower thresholds according to the electricity calibration coefficient to obtain the upper and lower threshold correction values, and compare the loss deviation of the electricity theft risk node with the upper and lower threshold correction values ​​to determine the type of the electricity theft risk node.

[0017] Furthermore, the topological structure data of the power transmission line is collected, and the topological network of the power consumption side is constructed according to the topological structure data, and the method is based on:

[0018] The electric energy node information and connection information of the transmission line are collected. The electric energy node information includes substations, distribution stations, and user-end equipment. The connection information includes line length, reactance, and impedance. Based on the node information and connection information, the nodes and edges of the topological network are determined. The nodes represent various devices or users in the power system, and the edges represent the power transmission lines between the nodes. Then these nodes and edges are arranged according to the actual connection relationship to form a network diagram, which is represented by a mathematical model using graph theory methods, in which nodes are represented by vertices and lines are represented by edges. Using the one-to-one mapping relationship, the nodes in the power transmission line are mapped to the nodes on the power consumption side through node mapping, and then the edges in the power transmission line are mapped to the edges on the power consumption side based on edge mapping. Finally, the topological network on the power consumption side is constructed based on the nodes and edges in the mapped power transmission line.

[0019] Furthermore, a neural network electricity theft prediction model is constructed, and the neural network electricity theft prediction model is trained and optimized based on the following method:

[0020] Taking historical current data as input and whether the corresponding power node is an abnormal power node as a label, a convolutional neural network structure is used, including an input layer, a hidden layer and an output layer. The input layer is used to receive historical data and perform preprocessing. The hidden layer extracts features and performs nonlinear transformation through convolution and pooling operations. The output layer generates a detection result of whether the power node is an abnormal power node.

[0021] Select the cross entropy function as the loss function, calculate the loss function value according to the output result and the true label, and calculate the gradient through the back propagation algorithm to update the weights and bias of the neural network. Repeat the above operation until the model reaches the predetermined number of training rounds.

[0022] Further, a first suspicious power node is screened out, and based on the loss deviation of the power node adjacent to the first suspicious power node, a second suspicious power node is screened out from the first suspicious power node, according to the method:

[0023] The actual power of the power node at the current moment and the simulated power of the topology network simulation are obtained, and the absolute difference between the actual power and the simulated power is calculated. This absolute difference is defined as the power loss deviation, and the power loss threshold is set. If the power loss deviation of the abnormal power node exceeds the power loss threshold, the abnormal power node is defined as the first suspicious power node. The formula for calculating the power deviation is:

[0024] P diff =|P actual -P simulated |

[0025] Among them, P ditt Indicates the power loss deviation, P actualis the actual power of the power node at the current moment, P simulated simulated power for the topological network;

[0026] The logical formula for determining whether the abnormal power node is the first suspicious power node is:

[0027]

[0028] Wherein, Q represents the logic value for judging whether the abnormal power node is the first suspicious power node. When Q=0, it indicates that the power loss deviation of the abnormal power node does not exceed the power loss threshold and is not the first suspicious power node; when Q=1, it indicates that the power loss deviation of the abnormal power node exceeds the power loss threshold and is the first suspicious power node; P th is the set power loss threshold;

[0029] Based on the above method, the power loss deviation between the first suspicious electric energy node and the adjacent source electric energy node is calculated. Combined with the set power loss threshold, it is determined whether the power loss deviation between the first suspicious electric energy node and the adjacent source electric energy node exceeds the power loss threshold. If exceeded, the first suspicious electric energy node is defined as the second suspicious electric energy node.

[0030] Furthermore, a normal power value interval is set, and a second suspicious electric energy node whose power value is not within the normal power value interval is screened out as a power theft risk node, and the method is based on:

[0031] There is a normal power range [P min ,P max ], P min ,P max They represent the minimum and maximum values ​​of the normal range of power values ​​respectively. The screening principle is: to determine whether the power loss deviation of each second suspicious power node is within the normal range [P min ,P max ], if the second suspicious power node is not in this interval, the second suspicious power node is determined to be a power theft danger point P danger If the second suspicious power node is in this interval, it indicates that the node is a normal power node.

[0032] Furthermore, historical electricity consumption data is collected to generate electricity consumption calibration coefficients, based on the following method:

[0033] Collect the current day's holiday electricity consumption and temperature data. If the current day is a holiday, J q = a, if the current day is not a holiday, then J q =b, and a>b>0, the logical formula is:

[0034]

[0035] Among them, J z Indicates the electrical calibration coefficient, J q is the holiday data of the current day, and T is the temperature data of the current day.

[0036] Furthermore, an upper threshold and a lower threshold of the power consumption of electricity theft are set, and the upper threshold and the lower threshold are corrected according to the electricity calibration coefficient to obtain an upper threshold correction value and a lower threshold correction value, and the loss deviation of the electricity theft risk node is compared with the upper threshold correction value and the lower threshold correction value to determine the type of the electricity theft risk node, and the method is based on:

[0037] After determining the dangerous point of electricity theft, set the upper limit of the electricity theft threshold max With the lower limit y min , the obtained electricity calibration coefficient is used to correct the upper and lower limits of the electricity theft threshold, eliminating the abnormal electricity consumption caused by temperature differences and peak electricity consumption during holidays. The formula is:

[0038] Y MAX =y max *J z

[0039] Y MIN =y min *J z

[0040] Among them, Y MAX , Y MIN Respectively represent the upper threshold correction value and the lower threshold correction value;

[0041] Get the dangerous point P of electricity theft danger The corresponding power loss deviation Y danger , when Y danger >Y MAX When , it indicates that the power node loss suddenly increases, and there is no corresponding meter record, indicating that there is electricity theft by bypassing the meter or illegal access to informal lines; when Y danger <Y MIN When Y MIN ≤Y danger ≤Y MAX When the power is stolen, it means that the circuit is aging or additional power-consuming equipment is connected.

[0042] The present invention also provides an electricity theft analysis device, which is used to execute the above-mentioned electricity theft analysis method, including:

[0043] A topology network building module, which is used to collect topology data of power transmission lines and build a topology network on the power consumption side according to the topology data;

[0044] An electricity theft prediction model construction module, which simulates the current data of the power nodes on the power consumption side in real time based on the topological network, including historical current data and current current data, and constructs an electricity theft prediction model based on a deep neural network. The historical current data is used as input, and whether the corresponding power node is an abnormal power node is used as a label to train the electricity theft prediction model. The current current data is input into the trained electricity theft prediction model to obtain the current abnormal power node;

[0045] An abnormal range detection module, wherein the abnormal range detection module selects a first suspicious electric energy node from the abnormal electric energy nodes based on the power loss deviation of the abnormal electric energy node, and selects a second suspicious electric energy node from the first suspicious electric energy node based on the power loss deviation of the electric energy node adjacent to the first suspicious electric energy node;

[0046] A power theft danger point positioning module, wherein the power theft danger point positioning module sets a normal power value interval, selects a second suspicious electric energy node whose power value is not within the normal power value interval as a power theft danger node, and collects historical power consumption data, including holiday power consumption and temperature data, and generates a power consumption calibration coefficient according to the holiday power consumption and temperature data;

[0047] The electricity theft warning module is used to set the upper and lower thresholds of electricity theft power consumption, correct the upper and lower thresholds according to the electricity calibration coefficient to obtain the upper and lower threshold correction values, and compare the loss deviation of the electricity theft risk node with the upper and lower threshold correction values ​​to determine the type of the electricity theft risk node.

[0048] The present invention further provides a storage medium, on which a computer program is stored. When the computer program is executed by a processor, it is used to implement the above-mentioned electricity theft analysis method.

[0049] Compared with the prior art, the present invention has the following beneficial effects:

[0050] First, the present invention effectively improves the monitoring and management efficiency of the power system by constructing a topological network on the power consumption side and generating a power consumption calibration coefficient based on historical power consumption data and temperature information. The deep neural network model is used to predict and detect abnormal power nodes, which greatly improves the safety and operational efficiency of the power grid. By analyzing the current flow data and power loss deviation in real time, the system can automatically identify potential power theft behaviors, optimize power resource management, and reduce economic losses.

[0051] Secondly, the present invention sets and corrects the upper and lower thresholds of power consumption of electricity theft to ensure accurate identification and classification of electricity theft. By adjusting the thresholds with the electricity calibration coefficient, the system effectively filters out fluctuations in electricity consumption caused by seasonal changes or holidays, thereby focusing on real abnormal situations. This method improves the sensitivity and accuracy of detection, reduces false alarms, and quickly takes targeted measures, which strengthens the security and reliability of the power grid and lays the foundation for intelligent management of the power grid.

[0052] The present invention realizes accurate detection of electricity theft and efficient management of the power system by constructing a topological network on the electricity consumption side and applying a deep neural network prediction model. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 It is a schematic diagram of the overall method flow of the present invention.

[0054] Figure 2 It is a schematic diagram of the system module flow of the present invention. DETAILED DESCRIPTION

[0055] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments.

[0056] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0057] Example:

[0058] See also Figure 1 , a method for analyzing electricity theft, the specific steps comprising:

[0059] Step 1: Collect the topological structure data of the power transmission line, and construct the topological network of the power consumption side according to the topological structure data;

[0060] Step 2: Based on the topological network, the current data of the power nodes on the power consumption side are simulated in real time, including historical current data and current current data. A power theft prediction model is constructed based on a deep neural network. The historical current data is used as input, and whether the corresponding power node is an abnormal power node is used as a label to train the power theft prediction model. The current current data is input into the trained power theft prediction model to obtain the current abnormal power node.

[0061] Step 3: based on the power loss deviation of the abnormal power node, a first suspicious power node is selected from the abnormal power nodes, and based on the power loss deviation of the power nodes adjacent to the first suspicious power node, a second suspicious power node is selected from the first suspicious power node;

[0062] Step 4: Set a normal power value range, filter out the second suspicious power node whose power value is not within the normal power value range as a dangerous power theft node, and collect historical power consumption data, including holiday power consumption and temperature data, and generate a power consumption calibration coefficient based on the holiday power consumption and temperature data;

[0063] Step 5: Set the upper and lower thresholds of the power consumption of electricity theft, and correct the upper and lower thresholds according to the electricity calibration coefficient to obtain the upper and lower threshold correction values, and compare the loss deviation of the electricity theft risk node with the upper and lower threshold correction values ​​to determine the type of the electricity theft risk node.

[0064] It should be noted that by building a topological network on the power consumption side, the node and connection information of the power transmission line is collected and analyzed, and converted into a graph model, which greatly improves the monitoring and management efficiency of the power system. The automated machine settings use real-time data to update the topology structure, ensuring the accuracy and responsiveness of the system, which helps to optimize power dispatching, rapid fault detection, accurate load forecasting, and support the planning and expansion of the power grid. This process reduces manpower and time costs, and at the same time improves the intelligence level of the system through big data analysis, laying a solid foundation for the development of smart grids.

[0065] Therefore, it is necessary to collect the topological structure data of the power transmission line and construct the topological network of the power consumption side according to the topological structure data. The method is as follows:

[0066] The electric energy node information and connection information of the transmission line are collected. The electric energy node information includes substations, distribution stations, and user-end equipment. The connection information includes line length, reactance, and impedance. Based on the node information and connection information, the nodes and edges of the topological network are determined. The nodes represent various devices or users in the power system, and the edges represent the power transmission lines between the nodes. Then these nodes and edges are arranged according to the actual connection relationship to form a network diagram, which is represented by a mathematical model using graph theory methods, in which nodes are represented by vertices and lines are represented by edges. Using the one-to-one mapping relationship, the nodes in the power transmission line are mapped to the nodes on the power consumption side through node mapping, and then the edges in the power transmission line are mapped to the edges on the power consumption side based on edge mapping. Finally, the topological network on the power consumption side is constructed based on the nodes and edges in the mapped power transmission line.

[0067] It should be noted that the current data refers to the magnitude of the current in the network transmission line. A neural network electricity theft prediction model is constructed. The historical current data is analyzed through a convolutional neural network to detect anomalies, thereby improving the safety and efficiency of the power grid. The reason for setting up this model is that it can automatically identify potential electricity theft behaviors, reduce losses and optimize power resource management.

[0068] Therefore, it is necessary to build a neural network electricity theft prediction model, and train and optimize the neural network electricity theft prediction model based on the following methods:

[0069] Taking historical current data as input and whether the corresponding power node is an abnormal power node as a label, a convolutional neural network structure is used, including an input layer, a hidden layer and an output layer. The input layer is used to receive historical data and perform preprocessing. The hidden layer extracts features and performs nonlinear transformation through convolution and pooling operations. The output layer generates a detection result of whether the power node is an abnormal power node.

[0070] Select the cross entropy function as the loss function, calculate the loss function value according to the output result and the true label, and calculate the gradient through the back propagation algorithm to update the weights and bias of the neural network. Repeat the above operation until the model reaches the predetermined number of training rounds.

[0071] It should be noted that the loss deviation of the power node refers to the power loss deviation value, that is, the absolute difference between the actual power of the power node at the current moment and the simulated power simulated by the topological network. By calculating the loss deviation of the power node and gradually expanding the analysis, the first and second suspicious power nodes can be screened out. The importance of this process lies in that it improves the accuracy of power theft detection and reduces false alarms and missed alarms. By comparing the actual and expected power loss differences, the system can automatically identify potential abnormal nodes, and optimize resource allocation and maintenance strategies through topological structure expansion analysis, thereby enhancing the safety and reliability of the power grid. This process setting is based on logical formulas and topological analysis to ensure that the detection process is efficient and targeted.

[0072] Therefore, it is necessary to screen out the first suspicious power node, and based on the loss deviation of the power node adjacent to the first suspicious power node, screen out the second suspicious power node from the first suspicious power node, and the method is as follows:

[0073] The actual power of the power node at the current moment and the simulated power of the topology network simulation are obtained, and the absolute difference between the actual power and the simulated power is calculated. This absolute difference is defined as the power loss deviation, and the power loss threshold is set. If the power loss deviation of the abnormal power node exceeds the power loss threshold, the abnormal power node is defined as the first suspicious power node. The formula for calculating the power deviation is:

[0074] P diff =|P actual -P simulated |

[0075] Among them, P diff Indicates the power loss deviation, P actual is the actual power of the power node at the current moment, P simulated simulated power for the topological network;

[0076] The logical formula for determining whether the abnormal power node is the first suspicious power node is:

[0077]

[0078] Wherein, Q represents the logic value for judging whether the abnormal power node is the first suspicious power node. When Q=0, it indicates that the power loss deviation of the abnormal power node does not exceed the power loss threshold and is not the first suspicious power node; when Q=1, it indicates that the power loss deviation of the abnormal power node exceeds the power loss threshold and is the first suspicious power node; P th is the set power loss threshold;

[0079] Based on the above method, the power loss deviation between the first suspicious electric energy node and the adjacent source electric energy node is calculated. Combined with the set power loss threshold, it is determined whether the power loss deviation between the first suspicious electric energy node and the adjacent source electric energy node exceeds the power loss threshold. If exceeded, the first suspicious electric energy node is defined as the second suspicious electric energy node.

[0080] It should be noted that by setting the interval [P min ,P max ], the system can effectively identify those nodes whose loss deviation exceeds the normal range and mark them as power theft risk nodes P danger ,This method ensures the accuracy of detection, reduces the possibility of misjudgment, and can take targeted measures more quickly to maintain the normal operation and security of the power grid.,The reason for setting the normal interval is that it allows normal fluctuations under different time periods and conditions, thereby improving the flexibility and adaptability of detection.

[0081] Therefore, it is necessary to set a normal power value interval and screen out the second suspicious power node whose power value is not within the normal power value interval as a power theft risk node. The method is as follows:

[0082] There is a normal power range [P min ,P max ], P min ,P max They represent the minimum and maximum values ​​of the normal range of power values ​​respectively. The screening principle is: to determine whether the power loss deviation of each second suspicious power node is within the normal range [P min ,P max ], if the second suspicious power node is not in this interval, the second suspicious power node is determined to be a power theft danger point P danger If the second suspicious power node is in this interval, it indicates that the node is a normal power node.

[0083] It should be noted that in power load forecasting and power theft detection, collecting historical power consumption data to generate power consumption calibration coefficients is a key step. Its importance lies in adjusting and calibrating power consumption patterns by considering holidays and temperature factors to improve the accuracy of forecasts. The logical formula sets different calibration coefficients J in different temperature ranges. z , ensuring that electricity consumption can be appropriately adjusted under extreme temperature conditions. This method can more accurately reflect actual electricity consumption behavior, especially during holidays and under different temperature conditions, thereby improving the management efficiency and safety of the power system. The reason for setting this step is to capture the diversity and complexity of electricity consumption changes to support more flexible and accurate power distribution and anomaly detection.

[0084] Therefore, it is necessary to collect historical electricity consumption data to generate electricity consumption calibration coefficients based on the following method:

[0085] Collect the current day's holiday electricity consumption and temperature data. If the current day is a holiday, J q = a, if the current day is not a holiday, then J q =b, and a>b>0, the logical formula is:

[0086]

[0087] Among them, J z Indicates the electrical calibration coefficient, J q is the holiday data of the current day, and T is the temperature data of the current day; in the above logic formula, T≤5 means that in a low temperature environment, J q *ln(e+5-T) enhances the calibration factor to reflect the increase in electricity consumption due to increased heating demand; 5≤T≤35Under constant temperature conditions, Formula J q *lne reflects that at moderate temperatures, the power demand is basically stable and does not require additional adjustment; T≥35 is under high temperature conditions, J q *ln(e+T-35) enhances the calibration factor to capture the increase in electricity consumption caused by the increased use of cooling equipment such as air conditioners; the reason for setting a>b>0 is that holidays usually lead to an increase in electricity demand, which is reflected in the electricity calibration factor J z In terms of adjustment, a higher coefficient is needed to compensate for this change in demand. During holidays, people usually spend more time at home, which may lead to a significant increase in household electricity consumption, such as using more electrical appliances, air conditioners or heating equipment. This change needs to be reflected by using a higher calibration coefficient a to avoid misjudging normal electricity fluctuations as electricity theft; the electricity consumption pattern on non-holidays is more stable, so the electricity calibration coefficient b can be set lower. By distinguishing between electricity consumption on holidays and non-holidays and setting different calibration coefficients for each, the system can more accurately adjust the electricity theft threshold, eliminate electricity anomalies caused by changes in regular life patterns, and ensure the accuracy and sensitivity of detection. This setting helps focus on real electricity theft and improves the management efficiency and safety of the power system.

[0088] It should be noted that setting and correcting the upper and lower thresholds of power consumption for electricity theft is crucial for accurately identifying and classifying electricity theft. MAX and the lower limit Y MINThreshold, the system can effectively eliminate power consumption anomalies caused by temperature changes or holidays, so as to focus on the real power theft. When the loss deviation exceeds the upper limit, it may indicate illegal access or power theft by bypassing the meter; while below the lower limit may mean that the meter has been tampered with. This setting ensures the sensitivity and accuracy of detection, reduces misjudgment, and improves the safety and operational efficiency of the power grid.

[0089] Therefore, it is necessary to set the upper and lower thresholds of the power consumption of electricity theft, and correct the upper and lower thresholds according to the electricity calibration coefficient to obtain the upper and lower threshold correction values. The loss deviation of the electricity theft risk node is compared with the upper and lower threshold correction values ​​to determine the type of the electricity theft risk node. The method is as follows:

[0090] After determining the dangerous point of electricity theft, set the upper limit of the electricity theft threshold max With the lower limit y min , the obtained electricity calibration coefficient is used to correct the upper and lower limits of the electricity theft threshold, eliminating the abnormal electricity consumption caused by temperature differences and peak electricity consumption during holidays. The formula is:

[0091] Y MAX =y max *J z

[0092] Y MIN =y min *J z

[0093] Among them, Y MAX , Y MIN Respectively represent the upper threshold correction value and the lower threshold correction value; in the above formula Y MAX =y max *J z In the middle, it means that the upper limit of the threshold is expanded. This is because when the electrical equipment is in a high or low temperature environment, the power of heating or cooling will be increased, which will increase the upper limit of power. The calibration coefficient is used to increase the upper limit of the threshold to prevent the misjudgment of electricity theft due to temperature changes or increased electricity consumption during holidays.

[0094] Get the dangerous point P of electricity theft danger The corresponding power loss deviation Y danger , when Y danger >Y MAX When , it indicates that the power node loss suddenly increases, and there is no corresponding meter record, indicating that there is electricity theft by bypassing the meter or illegal access to informal lines; when Y danger <Y MINWhen Y MIN ≤Y danger ≤Y MAX When the power is stolen, it means that the circuit is aging or additional power-consuming equipment is connected.

[0095] The present invention also provides an electricity theft analysis device, which is used to execute the above-mentioned electricity theft analysis method, including:

[0096] A topology network building module, which is used to collect topology data of power transmission lines and build a topology network on the power consumption side according to the topology data;

[0097] An electricity theft prediction model construction module, which simulates the current data of the power nodes on the power consumption side in real time based on the topological network, including historical current data and current current data, and constructs an electricity theft prediction model based on a deep neural network. The historical current data is used as input, and whether the corresponding power node is an abnormal power node is used as a label to train the electricity theft prediction model. The current current data is input into the trained electricity theft prediction model to obtain the current abnormal power node;

[0098] An abnormal range detection module, wherein the abnormal range detection module selects a first suspicious electric energy node from the abnormal electric energy nodes based on the power loss deviation of the abnormal electric energy node, and selects a second suspicious electric energy node from the first suspicious electric energy node based on the power loss deviation of the electric energy node adjacent to the first suspicious electric energy node;

[0099] A power theft danger point positioning module, wherein the power theft danger point positioning module sets a normal power value interval, selects a second suspicious electric energy node whose power value is not within the normal power value interval as a power theft danger node, and collects historical power consumption data, including holiday power consumption and temperature data, and generates a power consumption calibration coefficient according to the holiday power consumption and temperature data;

[0100] The electricity theft warning module is used to set the upper and lower thresholds of electricity theft power consumption, correct the upper and lower thresholds according to the electricity calibration coefficient to obtain the upper and lower threshold correction values, and compare the loss deviation of the electricity theft risk node with the upper and lower threshold correction values ​​to determine the type of the electricity theft risk node.

[0101] The present invention further provides a storage medium, on which a computer program is stored. When the computer program is executed by a processor, it is used to implement the above-mentioned electricity theft analysis method.

[0102] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0103] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. Those skilled in the art may appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein may be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software methods depends on the specific application and design constraints of the technical solution.

[0104] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0105] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. A method for analyzing electricity theft, characterized in that: The specific steps include: Step 1: Collect the topological structure data of the power transmission line, and construct the topological network of the power consumption side according to the topological structure data; Step 2: Based on the topological network, the current data of the power nodes on the power consumption side are simulated in real time, including the current data at the historical moment and the current data at the current moment. The power theft prediction model is constructed based on the deep neural network. The historical current data is used as input, and whether the corresponding power node is an abnormal power node is used as a label to train the power theft prediction model. The current current data is input into the trained power theft prediction model to obtain the current abnormal power node. Step 3: based on the power loss deviation of the abnormal power node, a first suspicious power node is selected from the abnormal power nodes, and based on the power loss deviation of the power nodes adjacent to the first suspicious power node, a second suspicious power node is selected from the first suspicious power node; Step 4: Set a normal power value range, filter out the second suspicious power node whose power value is not within the normal power value range as a dangerous power theft node, and collect historical power consumption data, including holiday power consumption and temperature data, and generate a power consumption calibration coefficient based on the holiday power consumption and temperature data; Step 5: Set the upper and lower thresholds of the power consumption of electricity theft, and correct the upper and lower thresholds according to the electricity calibration coefficient to obtain the upper and lower threshold correction values, and compare the loss deviation of the electricity theft risk node with the upper and lower threshold correction values ​​to determine the type of the electricity theft risk node.

2. The electricity theft analysis method according to claim 1, characterized in that: The topological structure data of the power transmission line is collected, and the topological network of the power consumption side is constructed according to the topological structure data, according to the method: The electric energy node information and connection information of the transmission line are collected. The electric energy node information includes substations, distribution stations, and user-end equipment. The connection information includes line length, reactance, and impedance. Based on the electric energy node information and connection information, the nodes and edges of the topological network are determined. The nodes represent various devices or users in the power system, and the edges represent the power transmission lines between the nodes. Then these nodes and edges are arranged according to the actual connection relationship to form a network diagram, which is represented by a mathematical model using graph theory methods, in which nodes are represented by vertices and lines are represented by edges. Using the one-to-one mapping relationship, the nodes in the power transmission line are mapped to the nodes on the power consumption side through node mapping, and then the edges in the power transmission line are mapped to the edges on the power consumption side based on edge mapping. Finally, the topological network on the power consumption side is constructed based on the nodes and edges in the mapped power transmission line.

3. The electricity theft analysis method according to claim 2, characterized in that: Construct a neural network electricity theft prediction model, train and optimize the neural network electricity theft prediction model based on the following methods: Taking historical current data as input and whether the corresponding power node is an abnormal power node as a label, a convolutional neural network structure is used, including an input layer, a hidden layer and an output layer. The input layer is used to receive historical data and perform preprocessing. The hidden layer extracts features and performs nonlinear transformation through convolution and pooling operations. The output layer generates a detection result of whether the power node is an abnormal power node. Select the cross entropy function as the loss function, calculate the loss function value according to the output result and the true label, and calculate the gradient through the back propagation algorithm to update the weights and bias of the neural network. Repeat the above operation until the model reaches the predetermined number of training rounds.

4. The method for analyzing electricity theft according to claim 1, characterized in that , select the first suspicious power node, and select the second suspicious power node from the first suspicious power node based on the loss deviation of the power node adjacent to the first suspicious power node. The method is as follows: The actual power of the power node at the current moment and the simulated power of the topology network simulation are obtained, and the absolute difference between the actual power and the simulated power is calculated. This absolute difference is defined as the power loss deviation, and the power loss threshold is set. If the power loss deviation of the abnormal power node exceeds the power loss threshold, the abnormal power node is defined as the first suspicious power node. The formula for calculating the power deviation is: P diff =|P actual -P simulated | Among them, P diff Indicates the power loss deviation, P actual is the actual power of the power node at the current moment, P simulated simulated power for the topological network; The logical formula for determining whether the abnormal power node is the first suspicious power node is: Wherein, Q represents the logic value for judging whether the abnormal power node is the first suspicious power node. When Q=0, it indicates that the power loss deviation of the abnormal power node does not exceed the power loss threshold and is not the first suspicious power node; when Q=1, it indicates that the power loss deviation of the abnormal power node exceeds the power loss threshold and is the first suspicious power node; P th is the set power loss threshold; Based on the above method, the power loss deviation between the first suspicious electric energy node and the adjacent source electric energy node is calculated. Combined with the set power loss threshold, it is determined whether the power loss deviation between the first suspicious electric energy node and the adjacent source electric energy node exceeds the power loss threshold. If exceeded, the first suspicious electric energy node is defined as the second suspicious electric energy node.

5. The electricity theft analysis method according to claim 1, characterized in that: The normal power value interval is set, and the second suspicious power node whose power value is not within the normal power value interval is screened out as a power theft risk node, based on the following method: There is a normal power range [P min ,P mzx ], P min ,P mzx They represent the minimum and maximum values ​​of the normal range of power values ​​respectively. The screening principle is: to determine whether the power loss deviation of each second suspicious power node is within the normal range [P min ,P mzx ], if the second suspicious power node is not in this interval, the second suspicious power node is determined to be a power theft danger point P danger If the second suspicious power node is in this interval, it indicates that the node is a normal power node.

6. The electricity theft analysis method according to claim 1, characterized in that: The historical electricity usage data is collected to generate the electricity usage calibration coefficients based on the following method: Collect the current day's holiday electricity consumption and temperature data. If the current day is a holiday, J q = a, if the current day is not a holiday, then J q =b, and a>b>0, the logical formula is: Among them, J z Indicates the electrical calibration coefficient, J q is the holiday data of the current day, and T is the temperature data of the current day.

7. The electricity theft analysis method according to claim 6, characterized in that: The upper and lower thresholds of the power consumption of electricity theft are set, and the upper and lower thresholds are corrected according to the power calibration coefficient to obtain the upper and lower threshold correction values. The loss deviation of the electricity theft risk node is compared with the upper and lower threshold correction values ​​to determine the type of the electricity theft risk node. The method is as follows: After determining the dangerous point of electricity theft, set the upper limit of the electricity theft threshold max With the lower limit y min , the obtained electricity calibration coefficient is used to correct the upper and lower limits of the electricity theft threshold, eliminating the abnormal electricity consumption caused by temperature differences and peak electricity consumption during holidays. The formula is: Y MAX =y max *J z Y MIN =y min *J z Among them, Y MAX , Y MIN Respectively represent the upper threshold correction value and the lower threshold correction value; Get the electricity theft danger point P danger The corresponding power loss deviation Y danger , when Y danger >Y MAX When , it indicates that the power node loss suddenly increases, and there is no corresponding meter record, indicating that there is electricity theft by bypassing the meter or illegal access to informal lines; when Y danger <Y MIN When Y MIN ≤Y danger ≤Y MAX When the power is stolen, it means that the circuit is aging or additional power-consuming equipment is connected.

8. An electricity theft analysis device, characterized in that: The analysis device is used to execute the electricity theft analysis method according to any one of claims 1 to 7, comprising: A topology network building module, which is used to collect topology data of power transmission lines and build a topology network on the power consumption side according to the topology data; An electricity theft prediction model construction module, which simulates the current data of the power nodes on the power consumption side in real time based on the topological network, including historical current data and current current data, and constructs an electricity theft prediction model based on a deep neural network. The historical current data is used as input, and whether the corresponding power node is an abnormal power node is used as a label to train the electricity theft prediction model. The current current data is input into the trained electricity theft prediction model to obtain the current abnormal power node; An abnormal range detection module, wherein the abnormal range detection module selects a first suspicious electric energy node from the abnormal electric energy nodes based on the power loss deviation of the abnormal electric energy node, and selects a second suspicious electric energy node from the first suspicious electric energy node based on the power loss deviation of the electric energy node adjacent to the first suspicious electric energy node; A power theft danger point positioning module, wherein the power theft danger point positioning module sets a normal power value interval, selects a second suspicious electric energy node whose power value is not within the normal power value interval as a power theft danger node, and collects historical power consumption data, including holiday power consumption and temperature data, and generates a power consumption calibration coefficient according to the holiday power consumption and temperature data; The electricity theft warning module is used to set the upper and lower thresholds of electricity theft power consumption, correct the upper and lower thresholds according to the electricity calibration coefficient to obtain the upper and lower threshold correction values, and compare the loss deviation of the electricity theft risk node with the upper and lower threshold correction values ​​to determine the type of the electricity theft risk node.

9. A storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, it is used to implement the electricity theft analysis method according to any one of claims 1 to 7.

Images