A line loss determination method, device and equipment of a power grid system and a storage medium

Abstract

The application discloses a kind of line loss determination method, device and equipment of power grid system, equipment and storage medium, including according to the historical load electric energy data of each target node to determine the predicted value and normal fluctuation range of load electric energy actual data;According to normal fluctuation range and load electric energy actual data to determine abnormal load electric energy data;According to the first correction of abnormal load electric energy data to specific hierarchical relationship;According to the second correction of abnormal load electric energy data to node load type;According to the third correction of abnormal load electric energy data of target node to long time continuous abnormal data according to adjacent node;According to the total input load electric energy and total output load electric energy of each layer to be obtained according to the corrected load electric energy actual data;According to total input and total output load electric energy to determine the line loss of each layer.The application solves the problem that the existing line loss determination method can only carry out overall analysis on line loss, cannot in-depth analyze local loss, so it is difficult to accurately locate the loss source.

Description

Technical Field

[0001] This invention relates to the field of power systems, and more particularly to a method, apparatus, equipment, and storage medium for determining line losses in a power grid system. Background Technology

[0002] In power system line loss management, line loss analysis is a crucial step, directly impacting the grid's operational efficiency and economic benefits. Currently, commonly used methods for determining line losses primarily involve comparing the total electricity input to the grid with the total electricity metered at the user side, using the difference as the basis for line loss determination. This reflects energy loss during power transmission at a macro level. However, this method only provides an overall analysis of line losses and struggles to deeply assess losses in specific local areas, accurately pinpointing the sources of loss and hindering the development of targeted loss reduction measures. Furthermore, this method is highly dependent on the accuracy and completeness of measurement data; any errors or missing data will significantly affect the reliability of the calculation results. Summary of the Invention

[0003] This invention provides a method, apparatus, device, and storage medium for determining line losses in a power grid system, to address the problem that existing line loss determination methods can only perform overall line loss analysis and cannot deeply analyze local losses, thus making it difficult to accurately locate the source of loss.

[0004] A method for determining line losses in a power grid system includes: Based on the power grid topology structure of the current power grid system, historical load power data of each node in the historical time period and actual load power data of each target node in the target time period are obtained. Based on the historical load power data of each target node, the predicted value of the actual load power data is determined; based on the predicted value of the actual load power data, the normal fluctuation range of the actual load power data is determined; based on the normal fluctuation range and the actual load power data of each target node within the target time period, the abnormal load power data of each target node within the target time period is determined. Based on the specific hierarchical relationship of each target node, the abnormal load power data of each target node within the target time period is corrected for the first time. After the first correction, the method further includes: based on the node load type of the target node, the abnormal load power data of the target node corresponding to the preset node load type within the target time period is corrected for the second time. And / or based on the historical load power data of the adjacent nodes of each target node, the abnormal load power data of the target node with long-term continuous abnormal data within the target time period is corrected for the third time. Based on the corrected actual load power data and the power grid topology, the total input load power and total output load power of each layer are obtained; based on the total input load power and total output load power of each layer, the line loss of each layer is determined, and the line loss value of each layer in the power grid topology is obtained.

[0005] A line loss determination device for a power grid system, comprising: The load data acquisition module is used to acquire historical load power data of each node within a historical time period, and actual load power data of each target node within a target time period, based on the specific hierarchical relationship of the current power grid system and the power grid topology. An abnormal load judgment module is used to determine the predicted value of the actual load power data based on the historical load power data of each target node; determine the normal fluctuation range of the actual load power data based on the predicted value of the actual load power data; and determine the abnormal load power data of each target node within the target time period based on the normal fluctuation range and the actual load power data of each target node within the target time period. The abnormal data correction module is used to perform a first correction on the abnormal load power data of each target node within a target time period based on the specific hierarchical relationship of each target node; after the first correction, it further includes: performing a second correction on the abnormal load power data of the target node corresponding to a preset node load type within a target time period based on the node load type of the target node; and / or performing a third correction on the abnormal load power data of the target node with long-term continuous abnormal data within a target time period based on the historical load power data of the adjacent nodes of each target node; The line loss determination module is used to obtain the total input load energy and total output load energy of each layer based on the corrected actual load energy data and the power grid topology; and to determine the line loss of each layer based on the total input load energy and total output load energy of each layer, thereby obtaining the line loss value of each layer in the power grid topology.

[0006] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described method for determining line losses in a power grid system.

[0007] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for determining line losses in a power grid system.

[0008] This invention provides a method, apparatus, device, and storage medium for determining line losses in a power grid system. The method involves acquiring historical and actual load power data. First, based on historical data, a predicted value of the actual load power is determined, thereby defining its normal fluctuation range and accurately identifying abnormal load power data at target nodes within a target time period. Then, an initial correction is performed based on the hierarchical relationship between nodes, making preliminary adjustments to the abnormal data from the perspective of the power grid topology. Next, a second correction is conducted based on the load type corresponding to the node, fully considering the characteristic differences of different load types to make the correction more targeted and accurate. For abnormal data that occurs continuously over a long period, a third correction is performed using historical load data from adjacent nodes, further improving data quality by leveraging the correlation between node data. This multi-level correction mechanism effectively ensures the reliability of actual load power data. Based on the obtained high-quality corrected data and combined with the power grid topology, the total input and total output load power at each level can be accurately calculated, thereby determining the line loss value at each level. This not only helps to fully understand the energy loss of the power grid at different levels, but also enables precise identification of the source of loss by analyzing the local losses at each level. This provides key data support for the optimized operation of the power grid, energy conservation and consumption reduction, and equipment maintenance, significantly improving the overall operating efficiency and economic benefits of the power grid. Attached Figure Description

[0009] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 This is a flowchart of a method for determining line losses in a power grid system according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the hierarchical structure of a method for determining line losses in a power grid system according to an embodiment of the present invention; Figure 3 This is a flowchart of a method for determining line losses in a power grid system according to an embodiment of the present invention; Figure 4 This is a flowchart of a method for determining line losses in a power grid system according to an embodiment of the present invention; Figure 5 This is a schematic diagram of a line loss determination device for a power grid system according to an embodiment of the present invention; Figure 6 This is a schematic diagram of a computer device according to an embodiment of the present invention. Detailed Implementation

[0011] To make the technical problems solved, the technical solutions, and the beneficial effects of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0012] In one embodiment, such as Figure 1 As shown, a method for determining line losses in a power grid system is provided, comprising the following steps: S01: Based on the power grid topology structure of the current power grid system's specific hierarchical relationships, obtain the historical load power data of each node within a historical time period, as well as the actual load power data of each target node within a target time period.

[0013] In this embodiment, the power grid topology refers to the electrical connection relationships between various components in the power system (such as high-voltage transmission networks, transformers, transmission lines, busbars, loads, etc.). Based on factors such as the voltage level of the power grid and the power transmission process, the nodes in the topology are hierarchically divided. For example, power transmission in the power grid follows a sequence from high voltage to low voltage. High-voltage electricity generated by generators is first transmitted to substations via the high-voltage transmission network for voltage reduction, then distributed to various distribution areas via transmission lines, and finally further reduced in voltage within the distribution areas before being supplied to users. This high-voltage-to-low-voltage transmission process determines the hierarchical relationship of the power grid equipment: the high-voltage transmission network is at the highest level, followed by substations, then transmission lines, then distribution areas, and finally users are at the lowest level. A distribution area, also known as a distribution transformer area, typically refers to an area centered on a distribution transformer that supplies power to users within a certain surrounding range. It is the direct connection link between the power grid and users, responsible for converting medium-voltage electricity into low-voltage electricity and distributing it to various users. Each distribution area has its specific power supply range and user group, and possesses relatively independent operating characteristics.

[0014] In this embodiment, a tree data structure is selected to store the power grid topology information based on the scale and complexity of the power grid. Tree structures are suitable for power grid systems with clear hierarchical relationships and well-defined branches, such as… Figure 2 As shown, consider a simple radial power grid. The high-voltage transmission network serves as the root node (level 1); substations are its child nodes (level 2); lines are substation child nodes (level 3); distribution transformers are further substation child nodes (level 4); and users are distribution transformer child nodes (level 5). The entire structure forms a multi-branch tree.

[0015] In this embodiment, the target node is the node directly related to the determination of line loss. It can cover all nodes or a portion of specific designated nodes.

[0016] In this embodiment, historical and real-time load power data are acquired through a dispatch automation system configured in the power grid. These systems can collect and store load power data from each node of the power grid in real time. By interfacing with the data interface of the dispatch automation system and following a predetermined data format and transmission protocol, the system periodically reads the historical and real-time load power data required by the target node (the real-time load power data is the actual load power data). Missing values ​​in the actual load power data are then filled using a set constant value, for example, by filling missing parts with 0.

[0017] S02: Based on the historical load power data of each target node, determine the predicted value of the actual load power data; based on the predicted value of the actual load power data, determine the normal fluctuation range of the actual load power data; based on the normal fluctuation range and the actual load power data of each target node within the target time period, determine the abnormal load power data of each target node within the target time period.

[0018] In this embodiment, the normal fluctuation range of actual load power data is used to quantify the natural variation amplitude of actual load power data over time. If the variation amplitude of actual load power data is within the normal fluctuation range, the actual load power data is considered normal. If the variation amplitude of actual load power data exceeds the upper and lower limits of the normal fluctuation range, the actual load power data is considered abnormal load power data. Abnormal load power data may manifest as missing data or abnormal data values.

[0019] S03: Based on the specific hierarchical relationship of each target node, perform a first correction on the abnormal load power data of each target node within the target time period; after the first correction, the correction further includes: based on the node load type of the target node, perform a second correction on the abnormal load power data of the target node corresponding to the preset node load type within the target time period; and / or based on the historical load power data of the adjacent nodes of each target node, perform a third correction on the abnormal load power data of the target node with long-term continuous abnormal data within the target time period.

[0020] In this embodiment, the hierarchical position refers to the specific hierarchical level of the node in the power grid topology, for example, such as Figure 2 As shown, the hierarchical position of C1 distribution area is 4. Based on the hierarchical position of each target node in the power grid topology, a suitable basic prediction model is selected, and this model is used to perform the first correction of missing or abnormal values ​​in the abnormal load power data of nodes at different levels; the correction here includes completing the missing data and correcting the abnormal data values.

[0021] In this embodiment, node load types include industrial loads, commercial loads, residential loads, and agricultural loads. Preset node load types refer to several load types manually set by the user; these can cover all node load types or only a portion of them. Based on the abnormal load power data that has undergone the first correction, and according to the specific load type of the target node, other applicable correction algorithms are further selected for different types of loads, or the basic prediction model used in the first correction is optimized. Therefore, based on the newly selected correction algorithm or the optimized prediction model, the abnormal load power data is further corrected to improve the accuracy of the data correction.

[0022] In this embodiment, the adjacent nodes of the target node refer to the parent node or sibling nodes of the current target node. For example, such as Figure 2 As shown, the parent node of transformer C1 is the direct superior node that has a power supply relationship with it, such as feeder B node; while its sibling nodes are other nodes at the same level as transformer C1, such as transformer C2 node and transformer C3 node.

[0023] In this embodiment, based on the abnormal load power data that has been corrected for the first or second time, the load power data of adjacent nodes of each target node is further corrected for the third time by combining the historical load power data.

[0024] S04: Based on the corrected actual load power data and the power grid topology, obtain the total input load power and total output load power of each layer; determine the line loss of each layer based on the total input load power and total output load power of each layer, and obtain the line loss value of each layer in the power grid topology.

[0025] In this embodiment, as Figure 3 As shown, the total input load of each layer refers to the total electrical energy flowing into a specific power grid level (such as a substation, feeder, or distribution substation), which is usually collected in real time by smart meters or metering devices at the entrance of that level. The total output load of each layer refers to the total electrical energy output from that power grid level and actually used by the next level of equipment or end users, and recorded by the metering device.

[0026] In this embodiment, line loss refers to the collective term for energy loss caused by physical losses and metering differences during the transmission of electrical energy from the generation end to the user end in a power system. Layered line loss calculation is performed to decompose the total line loss of the entire network into specific layers based on the power grid topology and specific levels, thereby achieving precise location of losses and determination of responsibility. The line loss value of each layer in the power grid topology is the difference between the total input load and the total output load of each layer. For example, as... Figure 2As shown, Level 1 represents the high-voltage transmission network, corresponding to the main transmission lines. The line loss value at this level equals the power output of the power plant minus the total input power of all regional substations. Level 2 represents regional substations, with a large substation as a unit. The line loss value at this level equals the total input power of the substation minus the total power output of all its feeder lines. Level 3 represents medium-voltage distribution feeders, with a 10kV or 35kV distribution line as a unit. The line loss value at this level equals the metered power output at the starting point of the line minus the total metered power output on the high-voltage side of all transformers along the line. Level 4 represents distribution substations, the most common basic unit, referring to an area powered by a single distribution transformer. The line loss value at this level equals the power output of the low-voltage output meter of the substation minus the total power output of all end-user meters within the substation. Level 5 represents indoor lines. The line loss value at this level equals the power output of the user's incoming line meter minus the total power consumption of all indoor electrical equipment.

[0027] This embodiment of a method for determining line losses in a power grid system includes acquiring historical and actual load power data. First, based on historical data, the predicted value of the actual load power is determined, thereby defining its normal fluctuation range and achieving accurate identification of abnormal load power data at target nodes within a target time period. Then, an initial correction is performed based on the hierarchical relationship between nodes, making preliminary adjustments to the abnormal data from the perspective of the power grid topology. Next, a second correction is conducted based on the load type corresponding to the node, fully considering the characteristic differences of different load types to make the correction more targeted and accurate. For abnormal data that occurs continuously over a long period, a third correction is performed using historical load data from adjacent nodes, further improving data quality by leveraging the correlation between node data. Through this multi-level correction mechanism, the reliability of actual load power data is effectively guaranteed. Based on the high-quality corrected data and combined with the power grid topology, the total input and total output load power at each level can be accurately calculated, thereby determining the line loss value at each level. This not only helps to fully understand the energy loss of the power grid at different levels, but also enables precise identification of the source of loss by analyzing the local losses at each level. This provides key data support for the optimized operation of the power grid, energy conservation and consumption reduction, and equipment maintenance, significantly improving the overall operating efficiency and economic benefits of the power grid.

[0028] Optionally, in step S02, determining the predicted value of the actual load power data based on the historical load power data of each target node specifically includes the following steps: S201: Determine the load baseline prediction model based on the time series characteristics and external influencing factors of the historical load power data of each target node.

[0029] In this embodiment, time series features are timestamp information extracted at different time scales, such as one day (24 hours), one week (7 days), or one year (season). External influencing factors include weather (such as temperature and humidity), holiday markers, and other factors that affect load values. The load baseline prediction model uses a Random Forest Regressor (RFR) model. Historical load energy data includes fields such as timestamps, temperature, humidity, holiday markers, and load values. The historical load energy data is divided into a load energy training dataset and a load energy validation dataset. The load energy training dataset is input into the RFR model, which learns the complex nonlinear relationship between load values ​​and factors such as temperature, humidity, time, and holidays in the historical data. Subsequently, the load energy validation dataset is input into the trained RFR model, which outputs predicted load values. The mean squared error between the predicted and actual load values ​​is used as the objective function, and the parameters of the RFR model are tuned accordingly. When the objective function reaches its minimum value, the corresponding model parameters are the optimal parameter configuration. Configure the model based on the optimal parameter settings to obtain a trained random forest regression model.

[0030] S202: Determine the predicted value of the actual load power data based on the load baseline prediction model of each target node.

[0031] In this embodiment, the timestamp, temperature, humidity, and holiday identifiers from the actual load power data of each target node are input into the corresponding trained random forest regression model in step S201 to obtain the predicted value of the actual load power data. Specifically, this predicted value refers to the predicted load value, and different target nodes correspond to different random forest regression models.

[0032] The step of determining the normal fluctuation range of the actual load power data based on the predicted value of the actual load power data specifically includes the following steps: S203: Based on the predicted value of the actual load power data and the statistical distribution characteristics of the historical load power data, the normal fluctuation range of the actual load power data is obtained.

[0033] In this embodiment, the predicted value of the actual load power data refers to the predicted load value in step S202, and the statistical distribution characteristic of the historical load power data refers to the standard deviation of the load data; the standard deviation is a measure of the volatility of historical load data, specifically referring to the degree of dispersion of historical load sample values ​​around their mean. Therefore, the formula for calculating the normal fluctuation range is: in, This is the predicted value; k is the two-sided 95th percentile of the standard normal distribution, with a value of 1.96. This represents the standard deviation of historical load power data.

[0034] This embodiment presents a method for determining line losses in a power grid system. By constructing a load baseline prediction model that integrates time-series characteristics and external influencing factors, and defining a dynamic fluctuation range based on statistical distribution, it achieves accurate prediction of load power data and intelligent anomaly monitoring. The load baseline prediction model not only fully considers the inherent patterns in historical load data, such as trends, periodicity, and seasonality, but also incorporates external influencing factors such as temperature, humidity, and holidays, making the prediction results closer to actual load changes under complex environments. Simultaneously, by dynamically defining the fluctuation range by combining the predicted values ​​with the statistical distribution characteristics (such as standard deviation) of historical data, the threshold can be automatically adjusted according to factors such as load level, season, and time period. This not only effectively avoids misjudging normal fluctuations but also significantly improves the sensitivity to identifying true anomalies, thereby significantly enhancing the accuracy and robustness of anomaly detection.

[0035] Optionally, in step S02, determining the abnormal load power data of each target node within the target time period based on the normal fluctuation range and the actual load power data of each target node within the target time period specifically includes the following steps: S301: If any sample point in the actual load power data exceeds the normal fluctuation range, and the actual load power data value recovers to the normal fluctuation range in one or two subsequent consecutive sampling periods, then the actual load power data value of the sample point exceeding the normal fluctuation range is determined to be an abnormal peak value fluctuation.

[0036] In this embodiment, the value of the sampling point refers to the actual load value in the actual load power data. In the actual load power data, when the actual load value at a certain moment instantaneously and significantly exceeds the upper or lower limit of the normal fluctuation range (for example, more than 3 times the standard deviation), and then quickly recovers to the normal fluctuation range within one or two subsequent sampling points, the actual load value of the sampling point that exceeds the normal fluctuation range is determined to be an abnormal peak value fluctuation.

[0037] S302: If the actual load power data of N consecutive sampling points exceeds the normal fluctuation range, then the actual load power data of the N consecutive sampling points is determined to be a continuous fluctuation abnormality; where N is a preset positive integer threshold.

[0038] In this embodiment, when the actual load values ​​of multiple consecutive sampling points (e.g., N, where N can be set according to business needs, such as N > 5) are consistently higher or lower than the normal fluctuation range, even if the deviation of a single sampling point is not large, it will be determined that the actual load power data of these N consecutive sampling points have continuous abnormal fluctuations. Such abnormalities usually indicate that the metering device has malfunctioned or that there are persistent problems such as data transmission interruption.

[0039] S303: Determine the curve similarity between the time series curve of the actual load power data and the preset standard curve. If the curve similarity is greater than the preset similarity threshold, it is determined that there is an abnormal morphological fluctuation in the actual load power data. The abnormal load power data includes the peak value fluctuation abnormality, the continuous fluctuation abnormality, and the morphological fluctuation abnormality.

[0040] In this embodiment, the time series curve is a broken line or smooth curve formed by connecting continuous points with time on the horizontal axis and the actual load power value on the vertical axis. It is used to visually present the data change pattern over time. The actual load time series curve is compared with a preset standard time series curve, and the similarity between the two curves is calculated. When the similarity between the two curves is greater than a preset similarity threshold, it indicates that there is a significant difference in the shape of the two curves. For example, if there is an unnatural load peak during the load off-peak period such as early morning, even if the peak value does not exceed the upper limit of the range, it is judged as an abnormal shape, which may indicate theft of electricity or sudden large load at night.

[0041] This embodiment of a method for determining line losses in a power grid system can more accurately and intelligently identify abnormal data that requires intervention by performing multi-dimensional anomaly judgment on peak fluctuation anomalies, continuous fluctuation anomalies, and morphological fluctuation anomalies in actual load power data, effectively avoiding misjudgments and omissions caused by traditional fixed threshold methods.

[0042] Optionally, in step S03, the first correction of the abnormal load power data of each target node within the target time period according to the specific hierarchical relationship of each target node specifically includes the following steps: S401: Based on the level of each target node in the power grid topology, if the level of the target node is greater than a preset level threshold, a first load prediction model for extracting the periodic global features of the data is selected to perform the first correction on the abnormal load power data of each target node within the target time period.

[0043] In this embodiment, the global periodicity of the data refers to the strong regularity and periodicity exhibited by the historical data of the target node. The first load forecasting model can be a time series forecasting model, which can extract the global periodicity of the data, such as the random forest model, the Autoregressive Integrated Moving Average (ARIMA) model, and the Prophet model. The level of each target node in the power grid topology is determined. If the level of a target node is greater than a preset level threshold (e.g., the level threshold is set to 3), then nodes with a level greater than 3 are classified as high-level nodes (such as high-voltage transmission networks and substations). These nodes aggregate a large amount of downstream load, and their overall behavior exhibits strong regularity and periodicity. Therefore, when data for such nodes is missing, a time series forecasting model is preferred. These models can effectively capture their inherent trends and seasonality (daily, weekly, quarterly), thereby performing high-precision data filling. The process of training the time series forecasting model using the historical load power data of the target node is existing technology and not an improvement of this invention; therefore, it will not be elaborated here.

[0044] In this embodiment, the time information of the abnormal load power data of the target node within the target time period is input into the trained time series prediction model, and the predicted value of the load value in the abnormal load power data is output. Based on the predicted value of the load value, the abnormal load power data of the target node within the target time period is corrected for the first time.

[0045] S402: If the level of the target node is not greater than the preset level threshold and the number of abnormal load power data is less than the preset number threshold, then the second load prediction model used to extract local change features of the data is selected to perform the first correction on the abnormal load power data of each target node in the target time period.

[0046] In this embodiment, local data variation characteristics refer to the continuity, smoothness, and correlation exhibited by data within a local range. The second load prediction model can employ local smoothing algorithms (such as cubic spline interpolation). When the level of the target node does not exceed a preset level threshold (e.g., the level threshold is set to 3), the target node and its nodes with a level not higher than 3 are classified as low-level nodes (such as transformer areas or user sides). The load of such nodes has stronger randomness and volatility. If data is missing, and the missing time is short (e.g., 1 to 2 sampling points), local smoothing algorithms are preferred to ensure the smoothness of data changes.

[0047] This embodiment of a method for determining line losses in a power grid system improves the accuracy of abnormal load power data by selecting appropriate load prediction models based on target nodes at different levels, and then correcting the abnormal data based on the prediction results. This method effectively enhances the accuracy and completeness of actual load power data and significantly improves the reliability of line loss calculation results.

[0048] Optionally, in step S03, namely, the second correction of the abnormal load power data of the target node corresponding to the preset node load type within the target time period based on the node load type of the target node, specifically includes the following steps: S501: Based on the node load type of the current target node, if the current target node is determined to be an industrial load or a commercial load, then extract the working days or rest days and production shifts as external variable features and add them to the first load prediction model to obtain the third load prediction model; based on the third load prediction model, perform a second correction on the abnormal load power data of the target node corresponding to the industrial load or commercial load within the target time period.

[0049] In this embodiment, load types include industrial load, commercial load, residential load, and agricultural load. Industrial and commercial loads exhibit strong start-stop regularity and are closely related to production and business activities. A third load forecasting model is constructed by introducing workdays and rest days, typical production shifts, etc., as external regression variables into the first load forecasting model (time series model). This model not only effectively captures the inherent trend and seasonal characteristics of load (such as daily, weekly, and quarterly variations), but also further reflects factors closely related to production and business activities, such as workdays or rest days, and typical production shifts, thereby improving the accuracy of load forecasting. The process of training the third load forecasting model using historical load power data of the target node is existing technology and not an improvement of this invention; therefore, it will not be elaborated here.

[0050] In this embodiment, the time information of the abnormal load power data of the target node within the target time period is input into the trained third load prediction model, and the predicted value of the load value in the abnormal load power data is output. Based on the predicted value of the load value, the abnormal load power data of the target node within the target time period is corrected for the second time.

[0051] S502: If the current target node is determined to be a residential load, then the similarity between the time series curve of the historical load power data of the current target node and the time series curve of the historical load power data of each neighboring node is compared; if the similarity is greater than a preset first similarity threshold, then the abnormal load power data of the target node determined to be a residential load in the target time period is corrected a second time based on the average load curve of the historical load power data of the neighboring nodes whose similarity is greater than the preset first similarity threshold.

[0052] In this embodiment, residential load refers to the total amount of electricity consumed by residential users in their daily lives and production activities. Although individual electricity consumption behavior has a high degree of randomness, it generally exhibits a clear bimodal characteristic in the morning and evening across the region as a whole. Based on the template matching method of similar load curves, the load value curve of the target node's historical load data is used to identify a group of adjacent nodes with the most similar load value curves from nodes of the same type. For example, by calculating the similarity between the load curve of the target node's historical load data and the load curves of other adjacent nodes of the same type, if the similarity is greater than a first similarity threshold, the adjacent node is added to the similar adjacent node group. Finally, the load value on the average load curve of the adjacent nodes in these similar adjacent node groups is calculated and used to perform a second correction on the abnormal load power data of the target node identified as residential load in the target time period.

[0053] This embodiment of a method for determining line losses in a power grid system involves introducing new external variables (such as weekdays or shifts) to correct the model for nodes with industrial or commercial load types. For nodes with residential load types, data correction is performed based on the similarity between the node and its neighboring nodes with similar load types. According to the characteristics of different load types, the correction method most suitable for their electricity consumption characteristics is adopted, and abnormal data undergoes secondary processing. This targeted correction strategy significantly improves the accuracy of the second correction, effectively avoiding the systematic bias that may result from using a single model to correct all types of data, thereby improving the overall accuracy of data correction and the robustness of the system.

[0054] Optionally, in step S03, which involves performing a third correction on the abnormal load power data of the target nodes within the target time period based on the historical load power data of the adjacent nodes of each target node, the specific steps include the following: S601: Obtain the load correlation coefficient between each target node and its neighboring nodes.

[0055] In this embodiment, the load correlation coefficient between the load curve of the historical load power data of each target node and the load curve of the historical load data of its neighboring nodes will be calculated and stored in advance.

[0056] S602: If the load correlation coefficient is greater than the preset correlation threshold, then based on the historical load power data of the corresponding adjacent nodes, the abnormal load power data of the target node with long-term continuous abnormal data within the target time period is corrected for the third time.

[0057] In this embodiment, the relationship between the load correlation coefficient obtained in step S601 and the preset correlation threshold is first determined. If the load correlation coefficient is greater than the preset correlation threshold (e.g., load correlation coefficient > 0.8), it indicates that the two are highly correlated, meaning that the historical load power data of the target node with abnormal data is highly synchronized with its neighboring nodes (parent or sibling nodes). At this time, a multiple regression model can be constructed based on the historical load power data of these highly synchronized neighboring nodes. The real-time, accurate parameter variables (e.g., timestamps, temperature, humidity, etc.) of the neighboring nodes at the same time and highly correlated with the load value are used as input data and input into the multiple regression model to output the predicted value of the load value that the target node needs to correct. Using this predicted value, a third correction is performed on the long-term continuous abnormal load power data of the target node within the target time period. This method makes full use of the spatial correlation between nodes and is usually more accurate than correction methods that only rely on the historical data of the target node itself.

[0058] If the correlation between the two is not significant (e.g., load correlation coefficient < 0.8), the data of adjacent nodes are not used for direct estimation. Instead, the model is regressed to one that relies solely on the target node's own attributes (such as hierarchical location and load type) to avoid introducing noise interference.

[0059] This embodiment of a method for determining line losses in a power grid system employs a spatial correlation correction strategy based on load correlation coefficients. For target nodes with long-term continuous abnormal data, a third correction is performed using historical load data from their neighboring nodes. By introducing historical load data from strongly correlated neighboring nodes as a correction benchmark, spatial reference information is effectively integrated, reducing the impact of single-node data noise and thus improving the robustness and accuracy of abnormal data correction.

[0060] Optionally, such as Figure 3 and Figure 4 As shown, the method for determining line loss also specifically includes the following steps: S701: Based on the line loss values ​​of each layer in the power grid topology and the preset line loss threshold, locate the high-loss line loss layer in the power grid topology.

[0061] In this embodiment, the relationship between the line loss value of each layer in the power grid topology and the preset line loss threshold is determined. If the line loss value of a certain layer is greater than the preset line loss threshold, then that layer is classified as a high-loss line loss layer.

[0062] S702: Configure a switching switch in the area corresponding to the high-loss line loss layer of the power grid topology, set the on / off state of the switching switch and the equipment selection in the area as independent variables, and redetermine and judge the line loss corresponding to the high-loss line loss layer. When the redetermined line loss value corresponding to the high-loss line loss layer meets the preset target conditions, determine the current on / off state of the switching switch and the equipment selection as the optimal decision variable value.

[0063] In this embodiment, the high-loss line loss layer is optimized according to a preset optimization model, and corresponding improvement suggestions are proposed. For example, the load distribution of the high-loss line loss layer can be adjusted through simulation calculations, or the loss parameters of different models of equipment can be compared to propose equipment replacement suggestions. These optimization measures all rely on optimization algorithms to provide decision support.

[0064] In this embodiment, the preset optimization model is a mathematical programming model aimed at minimizing regional line loss. The construction of this optimization model mainly includes three core steps: defining the objective function, decision variables, and constraints. It aims to provide strong decision support for generating optimization suggestions such as load allocation adjustments and equipment replacement. Specifically, the optimization model construction process includes the following steps: 1) Define the objective function: The core objective of the optimization model is to minimize the total active power loss within a specified power grid area. P loss The objective function is expressed as: in, L This represents the set of all lines and transformer branches in the region corresponding to the high-loss line loss layer. I i For flow through branch road i The current, R i branch road i The resistance.

[0065] Decision Variables: To minimize the objective function, the optimization model needs to adjust a series of controllable parameters, which are the decision variables. These mainly include: Network topology variables (Sk): For areas corresponding to high-loss line loss layers, tie switches or sectionalizing switches k are configured. A binary variable Sk∈{0,1} is defined for each switch, where 1 represents a closed switch and 0 represents an open switch. By changing the switch combination, load transfer between different feeders can be achieved. Equipment selection variables (Ej): For equipment points that can be upgraded (such as old transformer j), one or more binary variables Ej are defined to indicate whether to replace it with a new model of equipment. For example, Ej_new=1 indicates replacing the original equipment model, and Ej_old=1 indicates not replacing the original equipment model. Different equipment models correspond to different loss parameters, such as resistance Rj and reactance Xj.

[0066] Constraints: Any optimization scheme must be implemented within the boundaries of power grid safety and stability. Therefore, the optimization model includes the following key constraints: ① Power flow balance constraint: The injected power of each node must be equal to the sum of the outflow power, which ensures that the solution of the model satisfies Kirchhoff's laws.

[0067] ② Voltage constraint: Voltage amplitude at all nodes ( Vi It must be maintained within the safety limits stipulated by the state or industry, for example: in, This indicates the rated voltage.

[0068] ③ Branch capacity constraint: The load current (or power) of all lines and transformers must not exceed their rated maximum capacity. Ii_max ), that is, | Ii |≤ Ii_max .

[0069] ④ Network connectivity and radial constraints: The optimized network topology must ensure that all users receive reliable power (network connectivity), and for distribution networks, it is usually necessary to maintain a radial structure, that is, there should be no loops.

[0070] After locating the high-loss line loss layer, the system inputs the power grid topology, line parameters, and load data of the corresponding area into the optimization model. By solving the mixed-integer nonlinear programming problem, a set of optimal decision variable values ​​that minimize the objective function (line loss) can be obtained. The preset objective condition is the aforementioned constraint condition.

[0071] S703: Optimize and adjust the high-loss line loss layer according to the optimal decision variable value.

[0072] In this example, the high-loss line loss layer is optimized and adjusted based on the optimal decision variable values ​​obtained in step S702. For example, if the solution shows that a certain switch state variable Sk has changed, a suggestion to adjust the load distribution is generated, such as: "It is recommended to close the XX tie switch and open the YY section switch to transfer part of the load in the high-load area to the light-load line, which is expected to reduce the line loss by Z%."

[0073] If the equipment selection variable Ej in the solution results points to a new type of equipment, the system generates equipment replacement suggestions, such as: "It is recommended to replace the S7 transformer in XX distribution area with the S13 energy-saving transformer. Considering its no-load and load loss characteristics, it is expected to reduce line loss by W kW under typical load conditions in this area."

[0074] This embodiment presents a method for determining line losses in a power grid system. It constructs a mathematical optimization model for high-loss areas, aiming to minimize regional line losses. This model transforms the line loss problem into a mathematical optimization problem with a definite solution by defining an objective function, setting decision variables, and establishing constraints. This allows optimization recommendations to move beyond empirical judgment and instead rely on precise decisions calculated by the model, thus providing scientific and reliable decision support for optimization measures such as load allocation adjustments and equipment replacement.

[0075] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0076] In one embodiment, a line loss determination device for a power grid system is provided, which corresponds one-to-one with the line loss determination method for the power grid system described in the above embodiments. For example... Figure 5 As shown, the line loss determination device for this power grid system includes a load data acquisition module 101, an abnormal load judgment module 102, an abnormal data correction module 103, and a line loss determination module 104. Detailed descriptions of each functional module are as follows: The load data acquisition module 101 is used to acquire historical load power data of each node in a historical time period, and actual load power data of each target node in a target time period, based on the power grid topology structure of the current power grid system and the specific hierarchical relationship. The abnormal load judgment module 102 is used to determine the predicted value of the actual load power data based on the historical load power data of each target node; determine the normal fluctuation range of the actual load power data based on the predicted value of the actual load power data; and determine the abnormal load power data of each target node within the target time period based on the normal fluctuation range and the actual load power data of each target node within the target time period.

[0077] The abnormal data correction module 103 is used to perform a first correction on the abnormal load power data of each target node within a target time period according to the specific hierarchical relationship of each target node; after the first correction, it further includes: performing a second correction on the abnormal load power data of the target node corresponding to a preset node load type within a target time period according to the node load type of the target node; and / or performing a third correction on the abnormal load power data of the target node with long-term continuous abnormal data within a target time period according to the historical load power data of the adjacent nodes of each target node.

[0078] The line loss determination module 104 is used to obtain the total input load energy and total output load energy of each layer based on the corrected actual load energy data and the power grid topology; and to determine the line loss of each layer based on the total input load energy and total output load energy of each layer, thereby obtaining the line loss value of each layer in the power grid topology.

[0079] Optionally, the above-mentioned abnormal load judgment module 102 specifically includes: The model building submodule is used to determine the load baseline prediction model based on the time series characteristics and external influencing factors of the historical load power data of each target node.

[0080] The prediction submodule is used to determine the predicted value of the actual load power data based on the load baseline prediction model of each target node.

[0081] The normal fluctuation determination submodule is used to determine the normal fluctuation range of the actual load power data based on the predicted value of the actual load power data and the statistical distribution characteristics of the historical load power data.

[0082] Optionally, the above-mentioned abnormal load judgment module 102 specifically includes: The peak value fluctuation anomaly judgment unit is used to determine that if any sampling point in the actual load power data exceeds the normal fluctuation range, and the actual load power data value recovers to the normal fluctuation range in one or two subsequent consecutive sampling periods, the actual load power data value of the sampling point exceeding the normal fluctuation range is an abnormal peak value fluctuation.

[0083] The continuous fluctuation anomaly judgment unit is used to determine that the load power actual data value of the N consecutive sampling points is a continuous fluctuation anomaly if the load power actual data value of N consecutive sampling points exceeds the normal fluctuation range; where N is a preset positive integer threshold.

[0084] The morphological fluctuation anomaly judgment unit is used to judge the curve similarity between the time series curve of the actual load power data and the preset standard curve. If the curve similarity is greater than the preset similarity threshold, it is determined that there is a morphological fluctuation anomaly in the actual load power data. The abnormal load power data includes the peak value fluctuation anomaly, the continuous fluctuation anomaly and the morphological fluctuation anomaly.

[0085] Optionally, the above-mentioned abnormal data correction module 103 specifically includes: The first type of hierarchical correction submodule is used to perform the first correction on the abnormal load power data of each target node in the target time period based on the hierarchical level of each target node in the power grid topology. If the hierarchical level of the target node is greater than a preset hierarchical threshold, the first load prediction model for extracting the periodic global features of the data is selected.

[0086] The second type of hierarchical correction submodule is used to select a second load prediction model for extracting local change features of data and perform the first correction on the abnormal load data of each target node within the target time period if the hierarchical level of the target node is not greater than a preset hierarchical threshold and the number of abnormal load power data is less than a preset number threshold.

[0087] Optionally, the above-mentioned abnormal data correction module 103 specifically includes: The first load type submodule is used to determine whether the current target node is an industrial load or a commercial load based on the node load type of the current target node. If the current target node is determined to be an industrial load or a commercial load, the module will extract the working days or rest days and production shifts as external variable features and add them to the first load prediction model to obtain the third load prediction model. Based on the third load prediction model, the module will perform a second correction on the abnormal load power data of the target node corresponding to the industrial load or commercial load within the target time period.

[0088] The second load type submodule is used to compare the similarity between the time series curve of the historical load power data of the current target node and the time series curve of the historical load power data of each neighboring node if the current target node is determined to be a residential load; if the similarity is greater than a preset first similarity threshold, then based on the average load curve of the historical load power data of the neighboring nodes whose similarity is greater than the preset first similarity threshold, the abnormal load power data of the target node determined to be a residential load in the target time period is corrected for the second time.

[0089] Optionally, the above-mentioned abnormal data correction module 103 specifically includes: The load correlation coefficient acquisition submodule is used to obtain the load correlation coefficient between each target node and its neighboring nodes.

[0090] The load correlation coefficient correction submodule is used to perform a third correction on the abnormal load energy data of the target node with long-term continuous abnormal data within the target time period based on the historical load energy data of the corresponding adjacent nodes if the load correlation coefficient is greater than the preset correlation threshold.

[0091] Optionally, the line loss determination device for the power grid system further includes: The high-loss line loss layer location submodule is used to locate the high-loss line loss layer in the power grid topology based on the line loss values ​​of each layer in the power grid topology and a preset line loss threshold.

[0092] The optimization submodule is used to configure switching switches in the area corresponding to the high-loss line loss layer of the power grid topology, set the on / off state of the switching switches and the equipment selection in the area as independent variables, and redetermine and judge the line loss corresponding to the high-loss line loss layer. When the redetermined line loss value corresponding to the high-loss line loss layer meets the preset target conditions, the on / off state of the current switching switches and the equipment selection are determined as the optimal decision variable values. The optimization and adjustment submodule is used to optimize and adjust the high-loss line loss layer based on the optimal decision variable value.

[0093] This invention provides a line loss determination device for a power grid system. The device involves acquiring historical and actual load power data. First, it determines the predicted value of the actual load power based on historical data, thereby defining its normal fluctuation range and accurately identifying abnormal load power data of target nodes within a target time period. Then, it performs an initial correction based on the hierarchical relationship between nodes, making preliminary adjustments to the abnormal data from the perspective of the power grid topology. Next, it performs a second correction based on the load type corresponding to the node, fully considering the characteristic differences of different types of loads to make the correction more targeted and accurate. For abnormal data that occurs continuously over a long period, a third correction is performed using historical load data from adjacent nodes, further improving data quality by leveraging the correlation between node data. This multi-level correction mechanism effectively ensures the reliability of the actual load power data. Based on the high-quality corrected data and combined with the power grid topology, the total input and total output load power of each level can be accurately calculated, thereby determining the line loss value of each level. This not only helps to fully understand the energy loss of the power grid at different levels, but also enables precise identification of the source of loss by analyzing the local losses at each level. This provides key data support for the optimized operation of the power grid, energy conservation and consumption reduction, and equipment maintenance, significantly improving the overall operating efficiency and economic benefits of the power grid.

[0094] Specific limitations regarding the line loss determination device for power grid systems can be found in the limitations of the line loss determination method for power grid systems described above, and will not be repeated here. Each module in the aforementioned line loss determination device for power grid systems can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0095] In one embodiment, such as Figure 6 As shown, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the line loss determination method for the power grid system described in the above embodiments, for example... Figure 1 As shown in S01-S04, or Figures 2 to 4 As shown, to avoid repetition, it will not be described again here. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in this embodiment of the power grid system line loss determination device, for example... Figure 5 The functions of the load data acquisition module 101, abnormal load judgment module 102, abnormal data correction module 103, and line loss determination module 104 shown are not described again here to avoid repetition.

[0096] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When executed by a processor, the computer program implements the line loss determination method for the power grid system described in the above embodiment, for example... Figure 1 As shown in S01-S04, or Figures 2 to 4 As shown, to avoid repetition, it will not be described again here. Alternatively, when the computer program is executed by the processor, it implements the functions of each module / unit in this embodiment of the power grid system line loss determination device, for example... Figure 5 The functions of the load data acquisition module 101, abnormal load judgment module 102, abnormal data correction module 103, and line loss determination module 104 shown are not described again here to avoid repetition.

[0097] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for determining line losses in a power grid system, characterized in that, The method for determining line loss includes: Based on the power grid topology structure of the current power grid system, historical load power data of each node in the historical time period and actual load power data of each target node in the target time period are obtained. Based on the historical load power data of each target node, the predicted value of the actual load power data is determined; based on the predicted value of the actual load power data, the normal fluctuation range of the actual load power data is determined; based on the normal fluctuation range and the actual load power data of each target node within the target time period, the abnormal load power data of each target node within the target time period is determined. Based on the specific hierarchical relationship of each target node, the abnormal load power data of each target node within the target time period is corrected for the first time. After the first correction, the method further includes: based on the node load type of the target node, the abnormal load power data of the target node corresponding to the preset node load type within the target time period is corrected for the second time. And / or based on the historical load power data of the adjacent nodes of each target node, the abnormal load power data of the target node with long-term continuous abnormal data within the target time period is corrected for the third time. Based on the corrected actual load power data and the power grid topology, the total input load power and total output load power of each layer are obtained; based on the total input load power and total output load power of each layer, the line loss of each layer is determined, and the line loss value of each layer in the power grid topology is obtained; The first correction, based on the specific hierarchical relationship of each target node, of the abnormal load power data of each target node within the target time period, includes: Based on the level of each target node in the power grid topology, if the level of the target node is greater than a preset level threshold, a first load prediction model for extracting the periodic global features of the data is selected to perform the first correction on the abnormal load power data of each target node in the target time period. If the level of the target node is not greater than the preset level threshold, and the number of abnormal load power data is less than the preset number threshold, then the second load prediction model for extracting local change features of the data is selected to perform the first correction on the abnormal load power data of each target node within the target time period. The second correction, based on the node load type of the target node, involves performing a second correction on the abnormal load power data of the target node within the target time period corresponding to the preset node load type, including: Based on the load type of the current target node, if the current target node is determined to be an industrial load or a commercial load, then extracting workdays or rest days and production shifts as external variable features and adding them to the first load prediction model to obtain a third load prediction model; based on the third load prediction model, the abnormal load power data of the target node corresponding to the industrial load or commercial load within the target time period is corrected for the second time. If the current target node is determined to be a residential load, the similarity between the time series curve of the historical load power data of the current target node and the time series curve of the historical load power data of each neighboring node is compared. If the similarity is greater than a preset first similarity threshold, the abnormal load power data of the target node determined to be a residential load in the target time period is corrected a second time based on the average load curve of the historical load power data of the neighboring nodes whose similarity is greater than the preset first similarity threshold.

2. The method for determining line losses in a power grid system according to claim 1, characterized in that, The step of determining the predicted value of the actual load power data based on the historical load power data of each target node includes: Based on the time series characteristics and external influencing factors of the historical load power data of each target node, a load baseline prediction model is determined. Based on the load baseline prediction model for each target node, the predicted value of the actual load power data is determined; The step of determining the normal fluctuation range of the actual load power data based on the predicted value of the actual load power data includes: Based on the predicted value of the actual load power data and the statistical distribution characteristics of the historical load power data, the normal fluctuation range of the actual load power data is determined.

3. The method for determining line losses in a power grid system according to claim 1 or 2, characterized in that, The step of determining the abnormal load power data of each target node within the target time period based on the normal fluctuation range and the actual load power data of each target node within the target time period includes: If any sampling point in the actual load power data exceeds the normal fluctuation range, and the actual load power data value recovers to the normal fluctuation range in one or two subsequent consecutive sampling periods, then the actual load power data value of the sampling point that exceeds the normal fluctuation range is determined to be an abnormal peak value fluctuation. If the actual load power data of N consecutive sampling points exceeds the normal fluctuation range, then the actual load power data of the N consecutive sampling points is determined to be a continuous fluctuation abnormality; where N is a preset positive integer threshold. The similarity between the time series curve of the actual load power data and the preset standard curve is determined. If the curve similarity is greater than the preset similarity threshold, it is determined that there is an abnormal morphological fluctuation in the actual load power data. The abnormal load power data includes the peak value fluctuation anomaly, the continuous fluctuation anomaly, and the morphological fluctuation anomaly.

4. The method for determining line losses in a power grid system according to claim 1, characterized in that, The third correction is performed on the abnormal load power data of the target nodes within the target time period based on the historical load power data of the adjacent nodes of each target node, including: Obtain the load correlation coefficient between each target node and its neighboring nodes; If the load correlation coefficient is greater than the preset correlation threshold, then based on the historical load power data of the corresponding adjacent nodes, the abnormal load power data of the target node with long-term continuous abnormal data within the target time period will be corrected for the third time.

5. The method for determining line losses in a power grid system according to claim 1, characterized in that, The method for determining line loss also includes: Based on the line loss values ​​of each layer in the power grid topology and the preset line loss threshold, locate the high-loss line loss layer in the power grid topology; In the region corresponding to the high-loss line loss layer of the power grid topology, a switching switch is configured. The on / off state of the switching switch and the equipment selection in the region are set as independent variables. The line loss corresponding to the high-loss line loss layer is re-determined and judged. When the re-determined line loss value corresponding to the high-loss line loss layer meets the preset target conditions, the on / off state of the current switching switch and the equipment selection are determined as the optimal decision variable values. The high-loss line-loss layer is optimized and adjusted based on the optimal decision variable values.

6. A device for determining line losses in a power grid system, characterized in that, The line loss determination device includes: The load data acquisition module is used to acquire historical load power data of each node within a historical time period, and actual load power data of each target node within a target time period, based on the specific hierarchical relationship of the current power grid system and the power grid topology. An abnormal load judgment module is used to determine the predicted value of the actual load power data based on the historical load power data of each target node; determine the normal fluctuation range of the actual load power data based on the predicted value of the actual load power data; and determine the abnormal load power data of each target node within the target time period based on the normal fluctuation range and the actual load power data of each target node within the target time period. The abnormal data correction module is used to perform a first correction on the abnormal load power data of each target node within a target time period based on the specific hierarchical relationship of each target node; after the first correction, it further includes: performing a second correction on the abnormal load power data of the target node corresponding to a preset node load type within a target time period based on the node load type of the target node; and / or performing a third correction on the abnormal load power data of the target node with long-term continuous abnormal data within a target time period based on the historical load power data of the adjacent nodes of each target node; The line loss determination module is used to obtain the total input load energy and total output load energy of each layer based on the corrected actual load energy data and the power grid topology; and to determine the line loss of each layer based on the total input load energy and total output load energy of each layer, thereby obtaining the line loss value of each layer in the power grid topology. The abnormal data correction module includes: The first type of hierarchical correction submodule is used to perform the first correction on the abnormal load power data of each target node in the target time period based on the hierarchical level of each target node in the power grid topology. If the hierarchical level of the target node is greater than the preset hierarchical threshold, the first load prediction model for extracting the periodic global features of the data is selected. The second type of hierarchical correction submodule is used to select the second load prediction model for extracting local change features of data and perform the first correction on the abnormal load power data of each target node in the target time period if the hierarchical level of the target node is not greater than the preset hierarchical threshold and the number of abnormal load power data is less than the preset number threshold. The abnormal data correction module includes: The first load type submodule is used to determine whether the current target node is an industrial load or a commercial load based on the node load type of the current target node. If the current target node is determined to be an industrial load or a commercial load, the extracted workday or rest day and production shift are added as external variable features to the first load prediction model to obtain a third load prediction model. Based on the third load prediction model, the abnormal load power data of the target node corresponding to the industrial load or commercial load in the target time period are corrected for the second time. The second load type submodule is used to compare the similarity between the time series curve of the historical load power data of the current target node and the time series curve of the historical load power data of each neighboring node if the current target node is determined to be a residential load; if the similarity is greater than a preset first similarity threshold, then based on the average load curve of the historical load power data of the neighboring nodes whose similarity is greater than the preset first similarity threshold, the abnormal load power data of the target node determined to be a residential load in the target time period is corrected for the second time.

7. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method for determining line losses in the power grid system according to any one of claims 1 to 5.

8. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the method for determining line losses in the power grid system according to any one of claims 1 to 5.

Images