10kV line and transformer area line loss monitoring method

Abstract

The invention relates to the technical field of power supply, and particularly discloses a 10kV line and transformer area line loss monitoring method, which comprises the following steps of: deploying an electric energy acquisition device at a key node and a transformer area of a power grid to obtain differential power consumption data and voltage drop data in real time, and respectively extracting a line loss fluctuation characteristic value and a voltage stability characteristic value; the feature robustness and identification capability are enhanced through principal component analysis and a K-means clustering algorithm, the two types of features are further fused to construct a comprehensive line loss risk feature vector, the comprehensive line loss risk feature vector is input into an intelligent analysis model based on a graph neural network, end-to-end risk score prediction is carried out, and the risk is divided into high, middle and low risk levels according to a dynamic threshold value. The system automatically triggers a differentiated early warning mechanism according to the risk level, and continuously optimizes an early warning strategy through feedback learning, thereby achieving the global perception, local recognition and dynamic response of the line loss risk, improving the accuracy, intelligent level and operation and maintenance efficiency of line loss monitoring, and having a good engineering application prospect.

Description

Technical Field

[0001] The present invention relates to the technical field of power supply, and in particular to a method for monitoring line loss of a 10kV line and a transformer substation. Background Art

[0002] Line loss refers to the loss incurred by the power grid during the transmission, distribution, and management of electric energy. As an important component of the power grid, the line loss of 10kV lines and substations directly affects the economic benefits of power supply companies. However, traditional line loss monitoring methods often suffer from inaccurate monitoring and low efficiency. In modern power distribution systems, the 10kV distribution network is a key layer connecting the backbone power grid and user loads. Its operating efficiency and energy loss level directly affect the economic efficiency and safety of the power system. Line loss is one of the important indicators for measuring the operating quality of the distribution network. Especially in complex substation structures, due to factors such as large load fluctuations, three-phase imbalance, metering errors, and potential electricity theft, line loss exhibits nonlinear, time-varying, and spatially correlated characteristics, posing great challenges to traditional line loss monitoring and risk identification.

[0003] The existing technology has the following deficiencies: This paper addresses the problem that existing technologies fail to fully consider the topological relationships and multi-dimensional feature interactions between power grid nodes, resulting in low line loss risk identification accuracy and delayed early warning responses. This paper proposes a method that integrates line loss fluctuation characteristics with voltage stability characteristics, and uses a graph neural network to perform end-to-end risk scoring prediction and dynamic graded early warning. This problem has received little attention in traditional line loss monitoring, particularly in the lack of mature technical means to effectively model the spatial dependencies and feature coupling between the "point-line-network" structure of the power grid. Summary of the Invention

[0004] The purpose of the present invention is to provide a 10kV line and substation line loss monitoring method to solve the above-mentioned problems.

[0005] The purpose of the present invention can be achieved through the following technical solutions: A 10kV line and substation line loss monitoring method comprises the following steps: S1: Install electric energy collection devices at nodes and substations on 10kV lines to monitor differential power consumption and voltage drop data in real time; S2: Analyze the differential power consumption data and calculate the line loss fluctuation characteristic value based on the power difference between adjacent time periods to assess whether there is abnormal energy loss in the line; S3: Analyze the voltage drop data and calculate the voltage stability characteristic value based on the voltage difference from the power supply end to the user end to evaluate the power quality of the end users of the power grid; S4: Combine the line loss fluctuation characteristic value with the voltage stability characteristic value to construct a comprehensive line loss risk characteristic vector. This vector is then input into a pre-trained intelligent analysis model for fusion analysis to determine the current line loss risk level. S5: Dynamically adjust the early warning mechanism based on the risk level output by the intelligent analysis model.

[0006] As a further solution of the present invention, the step of evaluating whether there is abnormal energy loss in the circuit specifically includes: During the monitoring period, the differential power consumption data of the line between the 10kV line node and the substation is obtained in real time, and the differential power consumption data is analyzed. According to the power difference value between adjacent times, the line loss fluctuation characteristic value is calculated to determine whether the line loss fluctuation characteristic value is greater than or equal to the preset threshold. If so, there is abnormal energy loss in the line; if not, there is no abnormal energy loss in the line.

[0007] As a further solution of the present invention: the process of obtaining the line loss fluctuation characteristic value is: Real-time acquisition of differential power consumption data between 10kV line nodes and substations; Standardize the differential electricity consumption data and calculate the covariance matrix of the standardized data; Solving for the covariance matrix The eigenvalues ​​and corresponding eigenvectors of The projection matrix is ​​used to reduce the dimension of the normalized differential electricity consumption data; Calculate the energy of the first principal component to obtain the line loss fluctuation eigenvalue. Specifically, extract the column vector corresponding to the first principal component from the reduced-dimensional data matrix. For each sample, calculate the square of its column vector on the first principal component. Add the square values ​​of the column vectors of all samples on the first principal component and take the square root to obtain the line loss fluctuation eigenvalue.

[0008] As a further solution of the present invention, the evaluation of the power quality of the end users of the power grid specifically includes: During the monitoring period, the voltage drop data from the power supply end to the user end is obtained in real time; The voltage drop data is analyzed, and the voltage stability characteristic value is calculated based on the voltage difference from the power supply end to the user end. It is determined whether the voltage stability characteristic value is greater than or equal to the preset threshold. If so, the power quality of the end user of the power grid is normal. If not, the power quality of the end user of the power grid is abnormal.

[0009] As a further solution of the present invention: the process of obtaining the voltage stability characteristic value is: Real-time acquisition of voltage drop data from the power supply end to the user end to form a voltage drop data set; Choose the number of clusters , and randomly select the initial cluster centers, each cluster center represents a hypothetical voltage drop pattern; For each data point, by calculating its distance from all cluster centers, it is assigned to the category of the cluster center closest to it; Recalculate the new center position of each cluster until the maximum number of iterations is reached; Based on the final clustering results, the voltage stability characteristic value of the entire system is calculated.

[0010] As a further solution of the present invention, the line loss fluctuation characteristic value and the voltage stability characteristic value are combined to construct a comprehensive line loss risk characteristic vector, and the vector is input into a pre-trained intelligent analysis model for fusion analysis, specifically including: The line loss fluctuation characteristic value and the voltage stability characteristic value are obtained, and the line loss fluctuation characteristic value and the voltage stability characteristic value are constructed into a comprehensive line loss risk characteristic vector as the input of the intelligent analysis model to minimize the error between the risk score of the predicted line loss and the risk score of the actual line loss. The intelligent analysis model is trained as the training target, and the line loss risk score is output according to the trained intelligent analysis model. The intelligent analysis model is a graph neural network model.

[0011] As a further solution of the present invention: the training process of the intelligent analysis model is: Multiple sets of historical comprehensive line loss risk feature vectors are obtained to construct a training data set, and the graph neural network model is trained with the goal of minimizing the error between the risk score of the predicted line loss and the risk score of the actual line loss. The model parameters are adjusted through the mean square error loss function. The graph neural network model is trained with the goal of minimizing the error between the risk score of the predicted line loss and the risk score of the actual line loss. The model parameters are adjusted through the mean square error loss function. By inputting the currently collected comprehensive line loss risk feature vector into the trained graph neural network model, the risk score of the line loss corresponding to the 10kV line node or substation is output.

[0012] As a further solution of the present invention, determining the risk level of the current line loss specifically includes: Determine whether the line loss risk score is greater than or equal to the preset first threshold. If so, it is recorded as a high-risk line loss level. If not, determine whether the line loss risk score is less than or equal to the preset second threshold. If so, it is recorded as a low-risk line loss level. If not, it is recorded as a medium-risk line loss level.

[0013] As a further solution of the present invention, the early warning mechanism is dynamically adjusted according to the risk level output by the intelligent analysis model, specifically including: For 10kV lines or substations with high-risk levels, the system automatically triggers a red warning signal and sends an emergency alarm notification to the operation and maintenance dispatch center through the communication network, and pushes it to the mobile terminal devices of relevant responsible personnel; for areas with medium-risk levels, the system triggers a yellow warning signal, starts enhanced monitoring mode and records trend change data, and regularly generates risk assessment reports; for areas with low-risk levels, maintains a green status indicator and a regular inspection mechanism.

[0014] Beneficial effects of the present invention: (1) This invention innovatively proposes a 10kV line and substation line loss monitoring method that integrates multi-source feature information and combines it with graph neural network modeling. By systematically collecting differential power consumption and voltage drop data, the statistically robust line loss fluctuation characteristic values ​​and the voltage stability characteristic values ​​based on cluster analysis are extracted respectively. The two are organically integrated to construct a structured and highly representative comprehensive line loss risk feature vector as the input of the intelligent analysis model. On this basis, the graph neural network is introduced to model the power grid topology, and the electrical connection relationship between nodes is abstracted into a graph structure. By utilizing its powerful nonlinear mapping and neighborhood information aggregation capabilities, the end-to-end scoring prediction of the line loss risk of each node is achieved. This method not only breaks through the limitations of traditional line loss monitoring methods that rely on a single indicator and isolated analysis nodes, but also achieves a leap from local anomaly identification to global situation awareness by deeply exploring the multi-dimensional coupling relationship between "points-lines-networks" in the power grid, significantly improving the accuracy, stability and generalization ability of line loss risk assessment. Compared with existing technologies, this solution has stronger data-driven characteristics and spatial reasoning capabilities, which promotes the transformation of distribution network line loss management from empirical judgment to intelligent decision-making. It has high technological advancement and engineering application value.

[0015] (2) The present invention innovatively constructs a dynamic hierarchical warning mechanism based on line loss risk scoring, adopts a high, medium and low risk classification strategy, and combines differentiated and scenario-based response measures to achieve refined and intelligent monitoring of the line loss status of 10kV distribution lines and substations. Based on the risk score output by the graph neural network model and combined with the preset dynamic threshold, the system automatically determines the risk level of the current node or area and triggers the corresponding warning action: for example, a high risk level triggers a red emergency alarm and pushes it to the operation and maintenance terminal, a medium risk initiates enhanced monitoring and trend analysis, and a low risk maintains the regular inspection mode, thereby achieving closed-loop management from abnormality identification to disposal response. More importantly, the present invention introduces a self-learning optimization mechanism based on historical warning events and actual feedback data. By continuously collecting the actual operating status and fault repair information after the warning response, the risk classification threshold and warning parameter configuration are dynamically adjusted, so that the entire warning system has good adaptive capabilities and evolutionary characteristics. This mechanism not only effectively improves the robustness and sensitivity of the line loss monitoring system, but also significantly reduces the false alarm rate and missed alarm risk brought about by the traditional reliance on manual experience to set thresholds, greatly reduces the frequency of on-site inspections and operation and maintenance costs, and promotes the transformation of distribution network operation and maintenance from "passive response" to "active prevention and control", comprehensively improving the operational safety, management efficiency and intelligence level of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The present invention will be further described below with reference to the accompanying drawings.

[0017] Figure 1 The present invention is a flowchart of a 10kV line and substation line loss monitoring method. DETAILED DESCRIPTION

[0018] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0019] See also Figure 1 As shown, the present invention is a 10kV line and substation line loss monitoring method, comprising the following steps: S1: Install electric energy collection devices at nodes and substations on 10kV lines to monitor differential power consumption and voltage drop data in real time; S2: Analyze the differential power consumption data and calculate the line loss fluctuation characteristic value based on the power difference between adjacent time periods to assess whether there is abnormal energy loss in the line; S3: Analyze the voltage drop data and calculate the voltage stability characteristic value based on the voltage difference from the power supply end to the user end to evaluate the power quality of the end users of the power grid; S4: Combine the line loss fluctuation characteristic value with the voltage stability characteristic value to construct a comprehensive line loss risk characteristic vector. This vector is then input into a pre-trained intelligent analysis model for fusion analysis to determine the current line loss risk level. S5: Dynamically adjust the early warning mechanism based on the risk level output by the intelligent analysis model.

[0020] In S1: Install electric energy collection devices at nodes and substations on the 10kV line to monitor differential power consumption and voltage drop data in real time; In the 10kV line and substation line loss monitoring method described in this invention, electric energy collection devices are first installed at key nodes and substations in the 10kV distribution network to collect and transmit differential power consumption and voltage drop data in real time. Deployed at key locations such as substation outlets, branch nodes, and the low-voltage side of distribution transformers, these electric energy collection devices possess high-precision metering capabilities and can simultaneously collect basic electrical parameters such as active power, reactive power, current, and voltage. The collected data is precisely aligned using timestamps to ensure the accuracy and timeliness of subsequent analysis results.

[0021] The electric energy collection device includes but is not limited to a smart meter, a multifunctional power monitoring terminal and an edge computing collection unit; wherein the smart meter is used to collect the input and output power data of each node to achieve accurate acquisition of differential power consumption; the multifunctional power monitoring terminal is used to collect three-phase voltage, current waveforms and their effective values, and supports dynamic analysis of voltage drop; the edge computing collection unit has local data preprocessing capabilities and can complete preliminary data cleaning, feature extraction and communication upload functions. The above devices form a data collection network through optical fiber communication or wireless public network, and transmit the collected raw data to the main station system or cloud platform to provide data support for the subsequent extraction of line loss fluctuation characteristic values ​​and voltage stability characteristic values.

[0022] In S2, the differential power consumption data is analyzed. Based on the power difference between adjacent time periods, the line loss fluctuation characteristic value is calculated to assess whether there is abnormal energy loss in the line. Specifically, the following are performed: During the monitoring period, the differential power consumption data of the line between the 10kV line node and the substation is obtained in real time, and the differential power consumption data is analyzed. According to the power difference value between adjacent times, the line loss fluctuation characteristic value is calculated to determine whether the line loss fluctuation characteristic value is greater than or equal to the preset threshold. If so, there is abnormal energy loss in the line; if not, there is no abnormal energy loss in the line.

[0023] The process of obtaining the line loss fluctuation characteristic value is as follows: Real-time acquisition of differential power consumption data between 10kV line nodes and substations; The differential electricity consumption data is standardized and the covariance matrix of the standardized data is calculated. The calculation expression is: ; Where, represents the covariance matrix, represents the number of samples, represents the normalized differential electricity consumption data matrix, express The transposed matrix of Represents a transpose operation; Solving for the covariance matrix The eigenvalues ​​and corresponding eigenvectors are calculated as follows: ; Where, represents the feature vector, represents the eigenvalue; The projection matrix is ​​used to reduce the dimension of the normalized differential power consumption data. The calculation expression is: ; Where, represents the differential electricity consumption data matrix after dimensionality reduction, represents the projection matrix; Calculate the energy of the first principal component to obtain the line loss fluctuation eigenvalue. Specifically, extract the column vector corresponding to the first principal component from the reduced-dimensional data matrix. For each sample, calculate the square of its column vector on the first principal component. Add the square values ​​of the column vectors of all samples on the first principal component and take the square root to obtain the line loss fluctuation eigenvalue.

[0024] It should be noted that: by standardizing the differential electricity consumption data, modeling the covariance matrix, eigenvalue decomposition and dimensionality reduction projection, the energy of the first principal component is finally extracted as the line loss fluctuation eigenvalue, thereby achieving accurate identification of line energy loss anomalies. This method not only effectively eliminates the dimensional differences in electricity data between different nodes or stations, and improves the consistency and comparability of the data, but also mines the potential structural information of the data through eigenvalue decomposition, thereby enhancing the sensitivity and identification ability to abnormal line loss fluctuations. Compared with the traditional method of directly using the original differential value or simple statistical indicators, this solution can more accurately capture the dynamic change trend of line loss fluctuations, reduce noise interference, and improve the robustness and stability of the evaluation results. It has significant technological advancement and innovation.

[0025] In S3, the voltage drop data is analyzed and the voltage stability characteristic value is calculated based on the voltage difference from the power supply end to the user end to assess the power quality of the end users of the power grid. Specifically, the following are performed: The evaluation of the power quality of the end users of the power grid specifically includes: During the monitoring period, the voltage drop data from the power supply end to the user end is obtained in real time; The voltage drop is defined as follows: Voltage drop refers to the phenomenon that when current passes through a wire, the voltage drops due to the impedance (including resistance and reactance) of the wire itself, indicating the actual reduction in voltage between the power source point and the load point; The voltage drop data is analyzed, and the voltage stability characteristic value is calculated based on the voltage difference from the power supply end to the user end. It is determined whether the voltage stability characteristic value is greater than or equal to the preset threshold. If so, the power quality of the end user of the power grid is normal. If not, the power quality of the end user of the power grid is abnormal.

[0026] The process of obtaining the voltage stability characteristic value is as follows: Real-time acquisition of voltage drop data from the power supply end to the user end to form a voltage drop data set; Choose the number of clusters , and randomly select the initial cluster centers, each cluster center represents a hypothetical voltage drop pattern; For each data point, by calculating its distance from all cluster centers, it is assigned to the category of the cluster center closest to it. The calculation expression is: ; Where, Indicates the The voltage drop measurements at multiple time points of each monitoring point are consistent with the voltage drop measurements at the first The Euclidean distance between cluster centers, represents the number of data points, represents the distance from the center, Indicates the number of time points, Indicates the total number of time points, Indicates the The monitoring point is The voltage drop at a point in time, Indicates the The cluster center is in the The voltage drop at a point in time, Indicates the The voltage drop measurement value of each monitoring point at multiple time points, Indicates the The location vectors of cluster centers; Recalculate the new center position of each cluster until the maximum number of iterations is reached. The calculation expression is: ; Where, Indicates the The set of all data points of the class, is the number of elements in the corresponding set; Based on the final clustering results, the voltage stability characteristic value of the entire system is calculated, and the calculation expression is: ; Where, represents the voltage stability characteristic value, Indicates the number of clusters.

[0027] It should be noted that: by collecting voltage drop data from the power supply end to the user end, a voltage drop data set is constructed, and the K-means clustering algorithm is used to classify and model the voltage change patterns of multiple time periods and multiple nodes, so as to calculate the voltage stability characteristic value, which is used to evaluate the power quality of the end users of the power grid. This method makes full use of the distribution characteristics between cluster centers and the discrete degree of data points to effectively capture the spatiotemporal correlation of voltage fluctuations, breaking through the limitations of traditional judgments that rely only on a single moment or average voltage value, and can more comprehensively and accurately reflect the dynamic stability of power grid operation. Compared with existing technologies, this solution not only improves the recognition accuracy and response speed of voltage abnormalities, but also has good adaptability and scalability. It is especially suitable for intelligent perception and risk warning of power quality in complex distribution network environments, and has significant technical innovation and practical application value.

[0028] In S4, the line loss fluctuation characteristic value is combined with the voltage stability characteristic value to construct a comprehensive line loss risk characteristic vector. This vector is then input into a pre-trained intelligent analysis model for fusion analysis to determine the current line loss risk level. Specifically, the following are included: The line loss fluctuation characteristic value and the voltage stability characteristic value are obtained, and the line loss fluctuation characteristic value and the voltage stability characteristic value are constructed into a comprehensive line loss risk characteristic vector as the input of the intelligent analysis model to minimize the error between the risk score of the predicted line loss and the risk score of the actual line loss. The intelligent analysis model is trained as the training target, and the line loss risk score is output according to the trained intelligent analysis model. The intelligent analysis model is a graph neural network model.

[0029] The training process of the intelligent analysis model is as follows: Multiple sets of historical comprehensive line loss risk feature vectors are obtained to construct a training data set, and the graph neural network model is trained with the goal of minimizing the error between the risk score of the predicted line loss and the risk score of the actual line loss. The model parameters are adjusted through the mean square error loss function. The graph neural network model is trained with the goal of minimizing the error between the risk score of the predicted line loss and the risk score of the actual line loss. The model parameters are adjusted through the mean square error loss function. By inputting the currently collected comprehensive line loss risk feature vector into the trained graph neural network model, the risk score of the line loss corresponding to the 10kV line node or substation can be output.

[0030] Determining the risk level of the current line loss specifically includes: Determine whether the line loss risk score is greater than or equal to the preset first threshold. If so, it is recorded as a high-risk line loss level. If not, determine whether the line loss risk score is less than or equal to the preset second threshold. If so, it is recorded as a low-risk line loss level. If not, it is recorded as a medium-risk line loss level.

[0031] It should be noted that: the present invention innovatively proposes to fuse the line loss fluctuation characteristic value with the voltage stability characteristic value to construct a comprehensive line loss risk characteristic vector, and performs end-to-end risk score prediction and level division based on the graph neural network model, fully exploring the topological structure relationship and multi-dimensional feature interaction information between 10kV lines and substations. This method breaks through the limitations of traditional line loss analysis that only relies on a single indicator or independent node evaluation. By modeling the spatial dependency relationship between power grid nodes through graph neural networks, it realizes the global perception and local refined identification of line loss risks. The mean square error loss function is used to optimize the model parameters during the training process, which improves the accuracy and stability of the score prediction; and the three-level risk level mechanism based on dynamic threshold division further enhances the practicality and operability of the system. Overall, this step significantly improves the intelligence level, adaptability and early warning capabilities of line loss risk identification, and has outstanding technical innovation and engineering application value.

[0032] In S5, the early warning mechanism is dynamically adjusted based on the risk level output by the intelligent analysis model, including: For high-risk 10kV lines or substations, the system automatically triggers a red warning signal and sends an emergency alert notification to the operation and maintenance dispatch center via the communication network, and simultaneously pushes it to the mobile terminals of relevant responsible personnel. For medium-risk areas, the system triggers a yellow warning signal, activates enhanced monitoring mode, records trend change data, and regularly generates risk assessment reports. For low-risk areas, the system maintains a green status indicator and regular inspection mechanism. Based on historical warning events and subsequent actual line loss feedback, the warning strategy is adaptively optimized and adjusted, and the feedback learning mechanism is used to update the threshold range and warning response parameters corresponding to each risk level, so that the warning mechanism can dynamically adapt to changes in the power grid operating environment and improve the accuracy and practicality of the warning.

[0033] Working Principle: This invention aims to achieve intelligent, refined identification and dynamic early warning of line loss risks in distribution networks. By deploying electric energy collection devices at key nodes and substations, it acquires real-time differential power consumption and voltage drop data, and extracts line loss fluctuation and voltage stability eigenvalues. The line loss fluctuation eigenvalues ​​are constructed using principal component analysis, improving the robustness and accuracy of anomaly identification. The voltage stability eigenvalues ​​are analyzed using a K-means clustering algorithm to explore the spatiotemporal characteristics of voltage variations, enhancing the comprehensiveness and timeliness of voltage quality assessment. Furthermore, these two features are fused to form a comprehensive line loss risk feature vector, which is fed into a pre-trained graph neural network model to predict line loss risk scores for each node or substation. The risk is then categorized into three risk levels: high, medium, and low, based on a set threshold. Ultimately, the system automatically triggers the corresponding early warning mechanism based on the output risk level, including red emergency warning, yellow enhanced monitoring, and green routine inspection modes. A feedback learning mechanism continuously optimizes the early warning strategy, improving the overall system's adaptability and early warning accuracy. Compared with traditional line loss monitoring methods, the present invention not only integrates multi-dimensional electrical parameter information, but also introduces the idea of ​​grid topology modeling. It fully mines the spatial dependency between nodes through graph neural networks, and realizes global risk perception from "point" to "network", breaking through the limitations of isolated node analysis in existing technologies. At the same time, combined with the dynamic hierarchical early warning mechanism and self-learning optimization function, it significantly improves the response speed and operation and maintenance efficiency of line loss anomaly identification, and has good engineering application prospects and promotion value. This solution is suitable for complex and changeable distribution network operating environments, and provides strong technical support for power companies to achieve intelligent line loss management and proactive risk prevention and control.

[0034] The above is a detailed description of an embodiment of the present invention. However, the content described is only a preferred embodiment of the present invention and should not be considered to limit the scope of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. A 10kV line and area line loss monitoring method, characterized in that: The following steps are involved: S1: Install electric energy collection devices at nodes and substations on 10kV lines to monitor differential power consumption and voltage drop data in real time; S2: Analyze the differential power consumption data and calculate the line loss fluctuation characteristic value based on the power difference between adjacent time periods to assess whether there is abnormal energy loss in the line; S3: Analyze the voltage drop data and calculate the voltage stability characteristic value based on the voltage difference from the power supply end to the user end to evaluate the power quality of the end users of the power grid; S4: Combine the line loss fluctuation characteristic value with the voltage stability characteristic value to construct a comprehensive line loss risk characteristic vector. This vector is then input into a pre-trained intelligent analysis model for fusion analysis to determine the current line loss risk level. S5: Dynamically adjust the early warning mechanism based on the risk level output by the intelligent analysis model.

2. A 10kV line and substation line loss monitoring method according to claim 1, characterized in that: The evaluation of whether there is abnormal energy loss in the circuit specifically includes: During the monitoring period, the differential power consumption data of the line between the 10kV line node and the substation is obtained in real time, and the differential power consumption data is analyzed. According to the power difference value between adjacent times, the line loss fluctuation characteristic value is calculated to determine whether the line loss fluctuation characteristic value is greater than or equal to the preset threshold. If so, there is abnormal energy loss in the line; if not, there is no abnormal energy loss in the line.

3. A 10kV line and substation line loss monitoring method according to claim 2, characterized in that: The process of obtaining the line loss fluctuation characteristic value is as follows: Real-time acquisition of differential power consumption data between 10kV line nodes and substations; Standardize the differential electricity consumption data and calculate the covariance matrix of the standardized data; Solving for the covariance matrix The eigenvalues ​​and corresponding eigenvectors of The projection matrix is ​​used to reduce the dimension of the normalized differential electricity consumption data; Calculate the energy of the first principal component to obtain the line loss fluctuation eigenvalue. Specifically, extract the column vector corresponding to the first principal component from the reduced-dimensional data matrix. For each sample, calculate the square of its column vector on the first principal component. Add the square values ​​of the column vectors of all samples on the first principal component and take the square root to obtain the line loss fluctuation eigenvalue.

4. A 10kV line and substation line loss monitoring method according to claim 1, characterized in that: The evaluation of the power quality of the end users of the power grid specifically includes: During the monitoring period, the voltage drop data from the power supply end to the user end is obtained in real time; The voltage drop data is analyzed, and the voltage stability characteristic value is calculated based on the voltage difference from the power supply end to the user end. It is determined whether the voltage stability characteristic value is greater than or equal to the preset threshold. If so, the power quality of the end user of the power grid is normal. If not, the power quality of the end user of the power grid is abnormal.

5. A 10kV line and substation line loss monitoring method according to claim 4, characterized in that: The process of obtaining the voltage stability characteristic value is as follows: Real-time acquisition of voltage drop data from the power supply end to the user end to form a voltage drop data set; Choose the number of clusters , and randomly select the initial cluster centers, each cluster center represents a hypothetical voltage drop pattern; For each data point, by calculating its distance from all cluster centers, it is assigned to the category of the cluster center closest to it; Recalculate the new center position of each cluster until the maximum number of iterations is reached; Based on the final clustering results, the voltage stability characteristic value of the entire system is calculated.

6. A 10kV line and substation line loss monitoring method according to claim 1, characterized in that: The line loss fluctuation characteristic value and the voltage stability characteristic value are combined to construct a comprehensive line loss risk characteristic vector, which is then input into a pre-trained intelligent analysis model for fusion analysis. Specifically, the following steps are involved: The line loss fluctuation characteristic value and the voltage stability characteristic value are obtained, and the line loss fluctuation characteristic value and the voltage stability characteristic value are constructed into a comprehensive line loss risk characteristic vector as the input of the intelligent analysis model to minimize the error between the risk score of the predicted line loss and the risk score of the actual line loss. The intelligent analysis model is trained as the training target, and the line loss risk score is output according to the trained intelligent analysis model. The intelligent analysis model is a graph neural network model.

7. A 10kV line and substation line loss monitoring method according to claim 6, characterized in that: The training process of the intelligent analysis model is as follows: Multiple sets of historical comprehensive line loss risk feature vectors are obtained to construct a training data set, and the graph neural network model is trained with the goal of minimizing the error between the risk score of the predicted line loss and the risk score of the actual line loss. The model parameters are adjusted through the mean square error loss function. The graph neural network model is trained with the goal of minimizing the error between the risk score of the predicted line loss and the risk score of the actual line loss. The model parameters are adjusted through the mean square error loss function. By inputting the currently collected comprehensive line loss risk feature vector into the trained graph neural network model, the risk score of the line loss corresponding to the 10kV line node or substation is output.

8. A 10kV line and substation line loss monitoring method according to claim 1, characterized in that: Determining the risk level of the current line loss specifically includes: Determine whether the line loss risk score is greater than or equal to the preset first threshold. If so, it is recorded as a high-risk line loss level. If not, determine whether the line loss risk score is less than or equal to the preset second threshold. If so, it is recorded as a low-risk line loss level. If not, it is recorded as a medium-risk line loss level.

9. A 10kV line and substation line loss monitoring method according to claim 1, characterized in that: The dynamic adjustment of the early warning mechanism based on the risk level output by the intelligent analysis model specifically includes: For 10kV lines or substations with high-risk levels, the system automatically triggers a red warning signal and sends an emergency alarm notification to the operation and maintenance dispatch center through the communication network, and pushes it to the mobile terminal devices of relevant responsible personnel; for areas with medium-risk levels, the system triggers a yellow warning signal, starts enhanced monitoring mode and records trend change data, and regularly generates risk assessment reports; for areas with low-risk levels, maintains a green status indicator and a regular inspection mechanism.

Images