A method for verifying an electricity stealing prevention early warning model based on an electricity stealing prevention simulation experiment platform
By reproducing public transformer substations or lines on an anti-electricity theft simulation experimental platform, and carrying out electricity theft and metering anomaly modifications, the anti-electricity theft early warning model is verified using simulation data. This solves the problem of poor model universality and scalability in existing technologies, and realizes the verification of the model's reliability and accuracy.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHINA ELECTRIC POWER RESEARCH INSTITUTE CO LTD
- Filing Date
- 2022-08-09
- Publication Date
- 2026-05-29
AI Technical Summary
Existing anti-electricity theft algorithm models have differences in their operating environment, input source data, and model task management, which makes it impossible to effectively assess their universality and generalizability, thus affecting the accuracy of the early warning model.
By using an anti-electricity theft simulation experimental platform, public transformer substations or lines are reproduced to carry out electricity theft and metering anomaly repairs. Simulated data is obtained to replace real data, and correlation analysis is performed by combining historical electricity theft data and normal electricity user data to verify the performance and reliability of the anti-electricity theft early warning model.
The universality, recall, precision, and timeliness of the anti-electricity theft early warning model have been effectively verified, ensuring the reliability and scalability of the model.
Smart Images
Figure CN115510781B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of anti-electricity theft technology, and in particular to a method for verifying an anti-electricity theft early warning model based on an anti-electricity theft simulation experimental platform. Background Technology
[0002] With continuous economic development, the demand for electricity has increased significantly. However, along with this increase, electricity theft has also gradually increased. Electricity theft poses a significant social hazard; power outages of varying sizes caused by transformer and line burnout due to theft occur frequently. Besides affecting the operation of the power grid, it can also lead to incidents affecting public order, fires, and other social instability. To reduce electricity theft, power companies have begun to explore anti-theft technologies. As anti-theft efforts deepen, many power grid companies have chosen to develop anti-theft algorithm models. However, due to differences in the operating environment, input data, model task management, and output of anti-theft algorithm models developed by different power grid companies, it is impossible to effectively evaluate the universality and scalability of anti-theft early warning models when selecting them. Consequently, the accuracy of the results obtained when using these models cannot be guaranteed. Summary of the Invention
[0003] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method for verifying an anti-electricity theft early warning model based on an anti-electricity theft simulation experimental platform.
[0004] The objective of this invention is achieved through the following technical solution:
[0005] A method for verifying an anti-electricity theft early warning model based on an anti-electricity theft simulation experimental platform includes the following steps:
[0006] Step 1: Select a public transformer substation or line, extract the electricity users on the selected public transformer substation or line, and reproduce them on the anti-electricity theft simulation test platform.
[0007] Step 2: On the anti-electricity theft simulation test platform, the public transformer area or line is modified to prevent electricity theft and metering anomalies. The power consumption data of the public transformer area or line after the modification is obtained by simulation on the anti-electricity theft simulation test platform, and the power consumption data obtained by simulation on the anti-electricity theft simulation test platform replaces the original power consumption data of the public transformer area or line.
[0008] Step 3: Retrieve several historical electricity theft users and normal electricity users from different regions, and extract all electricity consumption data and event data of historical electricity theft users and normal electricity users during the electricity theft period, as well as line loss data of the lines or transformer areas where the electricity theft users and normal users are located. Perform correlation analysis on the extracted data, combine the related data, and obtain combined verification data for verifying the anti-electricity theft early warning model. Validate the anti-electricity theft early warning model based on the obtained combined verification data and the electricity consumption data of the replaced public transformer area or line.
[0009] Furthermore, in step one, when screening public transformer substations or lines, the screening criteria for public transformer substations include the substation's monthly average line loss rate, the duration for which no electricity theft or metering abnormalities have occurred, whether the mutual transformation relationship has changed, and the user composition of the substation. The screening criteria for lines include the duration for which no electricity theft users or users with metering abnormalities have occurred, whether the user-transformer relationship has changed, and the number of users using three-phase four-wire meters under the line.
[0010] Furthermore, when replicating electricity users, electricity users are divided into single-phase meter users and three-phase meter users. On the anti-electricity theft simulation test platform, the electricity consumption of single-phase meter users is determined by average power and average current, and the electricity consumption of three-phase meter users is determined by phase average power and phase average current. The electricity consumption of single-phase meter users and three-phase meter users simulated on the anti-electricity theft simulation test platform matches the corresponding actual electricity consumption.
[0011] Furthermore, in step three, the verification of the anti-electricity theft early warning model based on the combined verification data and the electricity consumption data of the replaced public transformer area or line includes performance efficiency verification, early warning reliability verification, and early warning timeliness verification.
[0012] Furthermore, the reliability verification of the anti-electricity theft early warning model includes precision verification, recall verification, and MAP index verification.
[0013] Furthermore, the precision rate is the proportion of actual electricity thieves among the total number of suspected electricity thieves output by the anti-electricity theft early warning model, the recall rate is the proportion of real electricity thieves output by the anti-electricity theft early warning model among all electricity thieves, and the MAP index is the ranking of real electricity thieves among several output results of the anti-electricity theft early warning model.
[0014] Furthermore, in step two, after replacing the original electricity consumption data of the public transformer substation or line with the electricity consumption data obtained from the anti-electricity theft simulation test platform, the line loss rate of the replaced public transformer substation or line is also corrected.
[0015] Furthermore, in step two, when modifying the reproduced public transformer area or line for electricity theft and metering anomalies on the anti-electricity theft simulation test platform, the types of electricity theft events to be modified include bypassing, undervoltage, undercurrent, phase shift, differential expansion and rectification, and the types of metering anomalies include common zero, wiring error, data acquisition failure, clock anomaly and meter counting anomaly.
[0016] The beneficial effects of this invention are:
[0017] Leveraging the ability of an anti-electricity theft simulation platform to flexibly simulate and reliably reproduce various real-world electricity theft scenarios, this study modifies existing anti-electricity theft data using the platform. Real-world electricity consumption data is then replaced with simulated data, which is used as validation data. This ensures that the validation results effectively reflect the universality of the anti-electricity theft early warning model. Furthermore, the model is validated in terms of comprehensiveness, accuracy, target users, and detection timeliness, and its generalizability is evaluated. Attached Figure Description
[0018] Figure 1 This is a schematic diagram of a process of the present invention;
[0019] Figure 2 This is a graph showing the actual electricity consumption of a single-phase meter user in a public transformer substation or line according to an embodiment of the present invention.
[0020] Figure 3 This is a graph showing the actual electricity consumption of a three-phase meter user in a public transformer substation or line according to an embodiment of the present invention.
[0021] Figure 4 This is a targeted user radar map of an embodiment of the present invention, which has a generalizable anti-electricity theft early warning model. Detailed Implementation
[0022] The present invention will be further described below with reference to the accompanying drawings and embodiments.
[0023] Example:
[0024] A method for validating an anti-electricity theft early warning model based on an anti-electricity theft simulation experimental platform, such as... Figure 1 As shown, it includes the following steps:
[0025] Step 1: Select a public transformer substation or line, extract the electricity users on the selected public transformer substation or line, and reproduce them on the anti-electricity theft simulation test platform.
[0026] Step 2: On the anti-electricity theft simulation test platform, the public transformer area or line is modified to prevent electricity theft and metering anomalies. The power consumption data of the public transformer area or line after the modification is obtained by simulation on the anti-electricity theft simulation test platform, and the power consumption data obtained by simulation on the anti-electricity theft simulation test platform replaces the original power consumption data of the public transformer area or line.
[0027] Step 3: Retrieve several historical electricity theft users and normal electricity users from different regions, and extract all electricity consumption data and event data of historical electricity theft users and normal electricity users during the electricity theft period, as well as line loss data of the lines or transformer areas where the electricity theft users and normal users are located. Perform correlation analysis on the extracted data, combine the related data, and obtain combined verification data for verifying the anti-electricity theft early warning model. Validate the anti-electricity theft early warning model based on the obtained combined verification data and the electricity consumption data of the replaced public transformer area or line.
[0028] The anti-electricity theft early warning model was validated using simulation and historical data. The electricity consumption data modified by simulation data can reflect the situation of most transformer areas where electricity theft occurs, and effectively verify the accuracy, comprehensiveness, and target users of the anti-electricity theft early warning model. The massive amount of historical data can effectively verify the universality of the anti-electricity theft early warning model, as well as its accuracy and operational efficiency.
[0029] When retrieving a number of historical electricity theft users and normal electricity users from different regions, special transformer users and low-voltage users were screened separately.
[0030] The screening criteria for dedicated transformer users are as follows:
[0031] For users who steal electricity from dedicated transformers, it is necessary to screen out those users whose electricity theft start and end times are recorded in the marketing system;
[0032] For users with normal electricity consumption via dedicated transformers, priority should be given to users with normal electricity consumption who share the same power line as those who steal electricity.
[0033] Furthermore, all users to be selected must have data collected in the data collection system.
[0034] The screening criteria for low-voltage users are as follows:
[0035] For low-voltage electricity theft users, it is necessary to screen low-voltage users who have electricity theft start and end times in the marketing system and whose travel meter reading data exists in the data collection system.
[0036] For normal low-voltage electricity users, priority will be given to normal low-voltage electricity users who are in the same transformer area as the electricity theft user.
[0037] Furthermore, priority is given to screening low-voltage users who have relevant voltage and current data.
[0038] When extracting electricity consumption data and event data from historical electricity theft users and normal electricity users during the theft period, the data for dedicated transformer electricity theft users needs to include electricity consumption data such as current, voltage, load, power factor, and energy meter readings, as well as related abnormal event data. Data for low-voltage electricity theft users needs to include energy meter readings and abnormal event data. If load-related data is available, current and voltage data should be extracted first. For all electricity theft user data, data from three time periods should be extracted: two months before the start of the theft, during the theft, and two months after the end of the theft. For normal electricity user data, electricity consumption data from normal users in the same transformer area and on the same line as the electricity theft user should be extracted. All data should be identical to the electricity theft user data, and the extraction time period should be the same as that of the electricity theft user in the same transformer area and on the same line.
[0039] The extracted line loss data specifically includes the daily line loss data of the lines and transformer areas where the electricity theft users are located. This data mainly includes the power supply, power consumption, power loss, line loss rate, total number of meters under the lines (transformers), data collection success rate, and other relevant data of other users under the lines (transformers) where the electricity theft users are located, as well as the year corresponding to the period of electricity theft by the electricity theft users, the daily line loss data for the whole year, and the daily line loss details for the whole year.
[0040] The extracted data are subjected to correlation analysis. When combining related data, the data of electricity theft users and normal users are combined at a ratio of approximately 1:500. The data of electricity theft users are injected into the data of normal users, and user-related abnormal event data and line loss data of relevant lines or transformer areas are added to obtain combined verification data.
[0041] In step one, when screening public transformer substations or lines, the screening criteria for public transformer substations include the substation's monthly average line loss rate, the duration for which no electricity theft or metering abnormalities have occurred, whether the mutual transformation relationship has changed, and the user composition of the substation. The screening criteria for lines include the duration for which no electricity theft users or users with metering abnormalities have occurred, whether the user-transformer relationship has changed, and the number of users using three-phase four-wire meters under the line.
[0042] In this embodiment, when screening public transformer substations, the following substations are selected: those with a monthly average line loss rate below 3%, no electricity theft or metering abnormalities for at least six consecutive months, unchanged customer-transformer relationships, and a user composition that includes both low-voltage single-phase and low-voltage three-phase users, where the low-voltage single-phase meters can provide neutral and live wire current data. Similarly, when screening power lines, the following lines are selected: those with no electricity theft or metering abnormalities for at least one consecutive month, unchanged customer-transformer relationships during the period without electricity theft or metering abnormalities, and at least three dedicated transformer users using three-phase four-wire meters.
[0043] When replicating electricity users, electricity users are divided into single-phase meter users and three-phase meter users. On the anti-electricity theft simulation test platform, the electricity consumption of single-phase meter users is determined by average power and average current, and the electricity consumption of three-phase meter users is determined by phase average power and phase average current. The electricity consumption of single-phase meter users and three-phase meter users simulated on the anti-electricity theft simulation test platform matches the corresponding actual electricity consumption.
[0044] For single-phase meter users in a public transformer area or line, the actual electricity consumption change of one single-phase meter user is as follows: Figure 2 As shown.
[0045] First, calculate the average power consumption during a specific time period based on the electricity consumption during that period. Because the power supply voltage of the anti-electricity theft simulation experimental platform is stable at 220V, the average power... And power data from the gate meter in the distribution area. Calculate the average current during this time period Finally, since the anti-electricity theft simulation platform operates for 8 hours daily, with a sampling interval one-third that of normal users, the average current of the simulated users is set at 3 to ensure power consumption matching. 220V, and 3 As operating parameters for single-phase meter users. Figure 2 Taking the T2 time period as an example, the average power during the T2 time period is The average current during time period T2 The calculation is performed using the following formula:
[0046] ;
[0047] in: The average power during time period T2. The average current during time period T2. This is the power data of the gate meter in the transformer area.
[0048] For three-phase meter users in the public transformer area or line, the actual electricity consumption change of one of the three-phase meter users is as follows: Figure 3 As shown.
[0049] First, based on the ratio of three-phase power at a given moment, the distribution of three-phase power consumption in the previous time period is calculated for that moment. Then, the three-phase power consumption in the previous time period is calculated. The remaining processing is similar to that for single-phase meter users. Finally, the operating parameters of the three-phase meter user are obtained as U. A =U B =U C =220V, 3 3 and 3 , , as well as .by Figure 3 Taking the electricity consumption during time period T2 as an example, the three-phase power at time t3 is P. 2A P 2B and P 2C The three-phase power consumption during time period T2 is W. 2A W 2B and W 2C And W 2A :W 2B :W 2C =P 2A :P 2B :P 2C The average current of phase A during time period T2 The calculation is performed using the following formula; the other two are calculated similarly:
[0050] ;
[0051] in: The average power of phase A during time period T2. This represents the electricity consumption of phase A during time period T2. The average current during time period T2. This is the power data of the gate meter in the transformer area.
[0052] Step three involves verifying the anti-electricity theft early warning model based on the combined verification data and the electricity consumption data of the replaced public transformer area or line. This verification includes performance efficiency verification, early warning reliability verification, and early warning timeliness verification.
[0053] Performance efficiency verification mainly includes verification of resource utilization, program running stability, running result stability, and computation time.
[0054] The verification method for resource utilization is as follows: When the anti-electricity theft warning model is running normally, the server resource usage is scanned every five minutes. If the CPU utilization exceeds the threshold K (K=90%) twice consecutively, it is recorded as one instance of overload operation. After each anti-electricity theft warning model analysis, the average CPU utilization during the model operation is recorded to determine whether it exceeds 70%.
[0055] The verification method for program stability is as follows: whether errors or crashes occur during a single run; whether errors or crashes occur after running continuously for 10 days and performing more than 30 repeated calculations; and whether the anti-electricity theft early warning model can operate normally by providing source data from different scales (5 million, 10 million, 20 million, and 30 million users respectively) during the operation.
[0056] The verification method for the stability of the running results is as follows: run the model 5 times under the same operating environment, input the same data each time, check the consistency of the results, and judge that the verification is passed when the consistency of the results is greater than 99%.
[0057] The verification method for computation time is as follows: In a standard verification environment, inject 10 million user data and test whether the model can complete the processing within the specified time and output the results normally.
[0058] The reliability verification of the anti-electricity theft early warning model includes precision verification, recall verification, and MAP index verification.
[0059] The precision rate is the proportion of actual electricity thieves among the total number of suspected electricity thieves output by the anti-electricity theft early warning model. The recall rate is the proportion of real electricity thieves output by the anti-electricity theft early warning model among all electricity thieves. The MAP index is the ranking of real electricity thieves among several output results of the anti-electricity theft early warning model.
[0060] The precision verification method is as follows: the model is verified by using the electricity consumption data replaced by simulation data and historical electricity consumption data as data inputs, respectively. The total number of suspected electricity theft users output by the model calculation and the actual amount of electricity theft in the model calculation output are recorded. The weights of the precision rate obtained by using the electricity consumption data replaced by simulation data and the precision rate obtained by using historical electricity consumption data are set, and the total precision rate of the anti-electricity theft early warning model is calculated. Specifically, the weight of the precision rate obtained by using the electricity consumption data replaced by simulation data is 0.7, and the weight of the precision rate obtained by using historical electricity consumption data is 0.3.
[0061] The recall rate is achieved by using electricity consumption data replaced with simulated data as input for model verification. The higher the recall rate, the more thorough the detection of electricity theft users.
[0062] In step two, after replacing the original electricity consumption data of the public transformer substation or line with the electricity consumption data obtained from the anti-electricity theft simulation test platform, the line loss rate of the replaced public transformer substation or line is also corrected.
[0063] Because the principle of constant electricity consumption was followed when replicating the data for electricity users, the changes in line loss in the public transformer area or line are solely caused by electricity theft. Therefore, the corrected line loss data is as follows:
[0064] ;
[0065] The lost electricity mentioned in the formula is the stolen electricity. The stolen electricity can be calculated in two ways. One way is to subtract the electricity consumption after the modification from the original normal electricity consumption of the user. The other way is to retrieve the rail meter configured for the corresponding electricity meter of the modified user from the anti-theft simulation test platform, obtain the actual electricity consumption recorded by the modified user, and subtract the electricity consumption after the modification from the electricity consumption recorded by the rail meter to obtain the stolen electricity.
[0066] In step two, when modifying the reproduced public transformer area or line for electricity theft and metering anomalies on the anti-electricity theft simulation test platform, the types of electricity theft events to be modified include bypass, undervoltage, undercurrent, phase shift, differential expansion and rectification, and the types of metering anomalies include common zero, wiring error, data acquisition failure, clock anomaly and meter counting anomaly.
[0067] Specifically, in this embodiment, when modifying the reproduced public transformer area or line for electricity theft and metering anomalies on the anti-electricity theft simulation experimental platform, the modified electricity theft methods are shown in Table 1:
[0068] Table 1. Types of Electricity Theft Methods (including modifications)
[0069]
[0070] After setting the electricity theft method, in order to verify the timeliness of the anti-electricity theft early warning model, different electricity theft times were set for the users who were modified, and the electricity theft times were set to continuous and intermittent. For the modification of continuous electricity theft time, the 1st, 15th, 24th and 27th days were selected as the electricity theft start time, and for intermittent electricity theft time, the electricity theft start time was randomly selected.
[0071] To verify the accuracy of the anti-electricity theft early warning model in identifying different electricity theft methods, the types of electricity theft modifications were varied. Following the principle of controlling variables, the model was tested against seven different types of electricity theft users: undervoltage, undercurrent, phase shift, differential voltage expansion, bypassing, continuous electricity theft, and intermittent electricity theft. The precision rate of the model for each type of theft was calculated and presented as a radar chart. If the precision rate of the model for identifying one or more electricity theft methods is above 0.9, or if the precision rate for identifying all seven methods is above 0.7, then the model is considered to be generalizable. Specifically, the radar chart of the target users for a generalizable anti-electricity theft early warning model is shown below. Figure 4 As shown.
[0072] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any way. Other variations and modifications are possible without departing from the technical solutions described in the claims.
Claims
1. A method for verifying an anti-electricity theft early warning model based on an anti-electricity theft simulation experimental platform, characterized in that, Includes the following steps: Step 1: Select a public transformer substation or line, extract the electricity users on the selected public transformer substation or line, and reproduce them on the anti-electricity theft simulation test platform. Step 2: On the anti-electricity theft simulation test platform, the public transformer area or line is modified to prevent electricity theft and metering anomalies. The power consumption data of the public transformer area or line after the modification is obtained by simulation on the anti-electricity theft simulation test platform, and the power consumption data obtained by simulation on the anti-electricity theft simulation test platform replaces the original power consumption data of the public transformer area or line. Step 3: Retrieve several historical electricity theft users and normal electricity users from different regions, and extract all electricity consumption data and event data of historical electricity theft users and normal electricity users during the electricity theft period, as well as line loss data of the lines or transformer areas where the electricity theft users and normal users are located. Perform correlation analysis on the extracted data, combine the related data, and obtain combined verification data for verifying the anti-electricity theft early warning model. Validate the anti-electricity theft early warning model based on the obtained combined verification data and the electricity consumption data of the replaced public transformer area or line. In step two, after replacing the original electricity consumption data of the public transformer substation or line with the electricity consumption data obtained from the anti-electricity theft simulation test platform, the line loss rate of the replaced public transformer substation or line is also corrected. In step two, when modifying the reproduced public transformer area or line for electricity theft and metering anomalies on the anti-electricity theft simulation test platform, the types of electricity theft events to be modified include bypass, undervoltage, undercurrent, phase shift, differential expansion and rectification, and the types of metering anomalies include common zero, wiring error, data acquisition failure, clock anomaly and meter counting anomaly.
2. The method for verifying an anti-electricity theft early warning model based on an anti-electricity theft simulation experimental platform according to claim 1, characterized in that, In step one, when screening public transformer substations or lines, the screening criteria for public transformer substations include the substation's monthly average line loss rate, the duration for which no electricity theft or metering abnormalities have occurred, whether the mutual transformation relationship has changed, and the user composition of the substation. The screening criteria for lines include the duration for which no electricity theft users or users with metering abnormalities have occurred, whether the user-transformer relationship has changed, and the number of users using three-phase four-wire meters under the line.
3. The method for verifying an anti-electricity theft early warning model based on an anti-electricity theft simulation experimental platform according to claim 1, characterized in that, When replicating electricity users, electricity users are divided into single-phase meter users and three-phase meter users. On the anti-electricity theft simulation test platform, the electricity consumption of single-phase meter users is determined by average power and average current, and the electricity consumption of three-phase meter users is determined by phase average power and phase average current. The electricity consumption of single-phase meter users and three-phase meter users simulated on the anti-electricity theft simulation test platform matches the corresponding actual electricity consumption.
4. The method for verifying an anti-electricity theft early warning model based on an anti-electricity theft simulation experimental platform according to claim 1, characterized in that, Step three involves verifying the anti-electricity theft early warning model based on the combined verification data and the electricity consumption data of the replaced public transformer area or line. This verification includes performance efficiency verification, early warning reliability verification, and early warning timeliness verification.
5. The method for verifying an anti-electricity theft early warning model based on an anti-electricity theft simulation experimental platform according to claim 4, characterized in that, The reliability verification of the anti-electricity theft early warning model includes precision verification, recall verification, and MAP index verification.
6. The method for verifying an anti-electricity theft early warning model based on an anti-electricity theft simulation experimental platform according to claim 5, characterized in that, The precision rate is the proportion of actual electricity thieves among the total number of suspected electricity thieves output by the anti-electricity theft early warning model. The recall rate is the proportion of real electricity thieves output by the anti-electricity theft early warning model among all electricity thieves. The MAP index is the ranking of real electricity thieves among several output results of the anti-electricity theft early warning model.