An adaptive current waveform detection method for smart grid
By building circuit topology diagrams and neural network architectures, the problem of low accuracy of traditional current waveform detection methods is solved, and the adaptive detection and robustness of current waveforms in smart grids are achieved.
Patent Information
- Application Number
- CN202411244378.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-06
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2044-09-06
AI Technical Summary
Traditional current waveform detection methods rely on fixed statistical rules, resulting in low detection accuracy and poor adaptability, making it difficult to effectively deal with the uncertainty caused by current fluctuations in complex power grids.
By treating the distribution transformer as a power supply in the station area, building a circuit topology diagram of the station area, traversing the load nodes for correlation analysis of current fluctuations, building a neural network architecture for the station area circuit diagram, conducting positive sample training, obtaining an instantaneous current predictor, and predicting and splitting The circuit current waveform is performed and abnormal analysis is performed.
Adaptive detection of current waveforms in smart grids is realized, the accuracy of detection and system adaptability are improved, and robustness is enhanced.
Smart Images

Figure CN118779808B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of current detection, and in particular to an adaptive current waveform detection method for a smart grid. Background Art
[0002] Grid current waveform detection is a technical means for monitoring and analyzing current fluctuations in the grid. Its purpose is to detect possible anomalies and faults in the grid system by monitoring and analyzing the current waveform, so as to improve the stability and safety of grid operation. Traditional current detection methods mainly rely on statistical rules to determine the current threshold. This method often relies on a large amount of historical data and has high requirements for data accuracy. Once the data is inaccurate, the accuracy of the detection results is difficult to guarantee. In addition, traditional methods are usually unable to effectively deal with the uncertainty caused by current fluctuations in complex grids. Summary of the invention
[0003] The present application provides an adaptive current waveform detection method for smart grids, which is used to solve the technical problems in the prior art that traditional current waveform detection relies on fixed statistical rules and data accuracy, has low detection accuracy and poor adaptability.
[0004] The present application provides an adaptive current waveform detection method for smart grids, the method comprising: regarding the distribution transformer as a substation power supply, constructing a substation circuit topology diagram, wherein the substation circuit topology diagram comprises a first load node, a second load node, and up to the Nth load node; traversing the first load node, the second load node, and up to the Nth load node to perform current fluctuation element correlation analysis, and obtain a first load control element, a second load control element, and up to the Nth load control element; performing current shunting and splitting on the substation circuit topology diagram, and obtaining a first split circuit, a second split circuit, and up to the Mth split circuit; based on the substation circuit topology diagram, building a substation circuit diagram neural network architecture, wherein the first load node, the second load node, and up to the Nth load node, and the substation power supply are input nodes of the substation circuit diagram neural network architecture, and the first split circuit, the second split circuit, and up to the Mth split circuit are output nodes of the substation circuit diagram neural network architecture; using the instantaneous output voltage of the substation power supply, the first load control element, the second load control element, and up to the Nth load control element as input node input parameters, and using the first split circuit current, the second split circuit current, and the second split circuit current as the input parameters. The circuit current until the Mth split circuit current is the output node output parameter, and the neural network architecture of the substation circuit diagram is trained with positive samples to obtain an instantaneous current predictor; the substation power supply output voltage timing information, the first load control element monitoring timing information, the second load control element monitoring timing information until the Nth load control element monitoring timing information, the first split circuit current monitoring waveform, the second split circuit current monitoring waveform until the Mth split circuit current monitoring waveform are obtained; through the instantaneous current predictor, the substation power supply output voltage timing information, the first load control element monitoring timing information, the second load control element monitoring timing information until the Nth load control element monitoring timing information are multi-timing predicted to obtain the first split circuit current prediction waveform, the second split circuit current prediction waveform until the Mth split circuit current prediction waveform; according to the first split circuit current prediction waveform, the second split circuit current prediction waveform until the Mth split circuit current prediction waveform, the first split circuit current monitoring waveform, the second split circuit current monitoring waveform until the Mth split circuit current monitoring waveform are abnormally analyzed to obtain the current waveform abnormal circuit, and the current waveform detection result is added.
[0005] One or more technical solutions provided in this application have at least the following technical effects or advantages:
[0006] The present application provides an adaptive current waveform detection method for smart grids, which relates to the field of current detection technology. The method regards the distribution transformer as a substation power source, constructs a substation circuit topology diagram, traverses multiple load nodes in the substation circuit topology diagram to perform current fluctuation factor correlation analysis, obtains multiple load control factors, and uses this to perform positive sample training to obtain an instantaneous current predictor, predicts multiple split circuit current prediction waveforms, and uses this to perform abnormal analysis on multiple split circuit current monitoring waveforms to obtain current waveform abnormal circuits. The method solves the technical problems in the prior art that traditional current waveform detection relies on fixed statistical rules and data accuracy, has low detection accuracy and poor adaptability, realizes adaptive detection of current waveforms in smart grids, improves detection accuracy, and enhances the technical effects of the system's adaptability and robustness. BRIEF DESCRIPTION OF THE DRAWINGS
[0007] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.
[0008] Figure 1 A schematic flow chart of an adaptive current waveform detection method for smart grids provided in an embodiment of the present application;
[0009] Figure 2 A schematic diagram of a process for performing correlation analysis of current fluctuation elements in an adaptive current waveform detection method for smart grids provided in an embodiment of the present application;
[0010] Figure 3 A schematic diagram of a flow chart of a circuit for obtaining an abnormal current waveform in an adaptive current waveform detection method for a smart grid provided in an embodiment of the present application. DETAILED DESCRIPTION
[0011] The present application provides an adaptive current waveform detection method for smart grids, which is used to solve the technical problems in the prior art that traditional current waveform detection relies on fixed statistical rules and data accuracy, has low detection accuracy and poor adaptability.
[0012] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.
[0013] It should be noted that the terms "first", "second", etc. in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products, or devices. Example
[0014] like Figure 1 As shown, the present application provides an adaptive current waveform detection method for smart grids, the method comprising:
[0015] P10: Consider the distribution transformer as a substation power source and construct a substation circuit topology diagram, wherein the substation circuit topology diagram includes a first load node, a second load node, and up to an Nth load node.
[0016] Specifically, the distribution transformer is first regarded as the core of the substation power system, that is, the substation power supply, which is the key link in the distribution and conversion of the entire substation power. Then, a detailed substation circuit topology diagram is constructed. The substation circuit topology diagram is a graphical representation method used to intuitively display the flow path and connection relationship of the power inside the substation.
[0017] In the process of constructing the substation circuit topology diagram, graph theory and power system analysis methods can be used. Graph theory provides a mathematical method to describe the network structure composed of nodes (representing devices or loads in the circuit) and edges (representing wires or transmission paths connecting these devices). For substation circuits, these nodes are concretized as various load devices, such as electrical equipment of residential users, commercial users or industrial users, which are marked as the first load node, the second load node, and up to the Nth load node, where N represents the total number of load nodes in the substation.
[0018] Each load node represents one or more power-consuming units in the substation, and is directly or indirectly connected to the distribution transformer (i.e., the substation power supply) through wires, forming a complex circuit network, i.e., the substation circuit topology diagram, which can clearly reflect the position, connection relationship, and power transmission path of each load node in the circuit.
[0019] The construction of the substation circuit topology diagram usually depends on the actual configuration data of the distribution system. The specific node location, line connection relationship, etc. can be obtained through the geographic information system (GIS) of the substation, or can be deduced through the power grid operation monitoring data. To ensure the accuracy and real-time performance of the topology diagram, an automated network analysis tool can be used to automatically draw and update it in combination with the on-site load data. In this way, the substation circuit topology diagram can intuitively display the power relationship between the distribution transformer and all load nodes, providing a data basis and network structure support for subsequent current waveform detection.
[0020] P20: Traverse the first load node, the second load node, and up to the Nth load node to perform current fluctuation factor correlation analysis to obtain the first load control factor, the second load control factor, and up to the Nth load control factor.
[0021] Further, such as Figure 2 As shown, step P20 of the embodiment of the present application also includes:
[0022] P21: When the first load node is a capacitor, perform capacitance correlation analysis to obtain the first load control factor; P22: When the first load node is an inductor, perform inductance correlation analysis to obtain the first load control factor; P23: When the first load node is other components, perform resistance correlation analysis to obtain the first load control factor; P24: Wherein, the acquisition process of the second load control factor up to the Nth load control factor is the same as that of the first load control factor.
[0023] It should be understood that all nodes from the first load node, the second load node to the Nth load node are traversed to perform correlation analysis of current fluctuation elements. The purpose of the current fluctuation element correlation analysis is to extract the control elements that affect the current fluctuation of these nodes by analyzing the current change characteristics of each load node, thereby providing basic data support for subsequent current prediction and waveform detection.
[0024] Exemplarily, when the first load node is a capacitor, a capacitance correlation analysis is performed to obtain the control factor of the load node. A capacitor is an energy storage element that usually affects the phase and amplitude of the current, especially in transient response and frequency changes. Through capacitance correlation analysis, the impact of the capacitor on the current fluctuation of the entire station area can be identified, and the control factor of the impact can be extracted.
[0025] When the first load node is an inductor, an inductor correlation analysis is performed to obtain the control factors of the load node. The inductor is another energy storage element, but it stores magnetic energy and has a greater resistance to current changes, especially in the case of circuit switching or short circuit, the inductor will produce a large current impact. Through inductor correlation analysis, the influence of inductance on the current waveform can be determined and the control factors of this influence can be extracted.
[0026] When the first load node is another element (such as a resistor), resistance correlation analysis is performed to obtain the control factor of the load node. The role of the resistor is relatively simple. It only affects the magnitude of the current but not the phase of the current. Through resistance correlation analysis, the load characteristics of the resistor can be extracted, thereby obtaining the control factor affecting the current fluctuation.
[0027] Finally, referring to the acquisition process of the first load control factor, the second load control factor is acquired in sequence until the Nth load control factor, and all the first load control factors, second load control factors until the Nth load control factors obtained through the above analysis are summarized and associated. These control factors will constitute the core parameter library for the current prediction and waveform detection of the entire substation, ensuring that the subsequent current fluctuation prediction can accurately reflect the impact of each load node.
[0028] In terms of technical support, the correlation analysis of load elements such as capacitors, inductors and resistors relies on accurate electrical parameter measurement and historical data analysis. By using high-precision sensors to measure the voltage and current waveforms of each load node, combined with machine learning algorithms or correlation analysis models, the control elements of each load node can be dynamically adjusted and identified to ensure that the current detection system in the substation has adaptive accuracy and reliability.
[0029] Furthermore, step P21 of the embodiment of the present application further includes:
[0030] P21-1: Configure the initial load control factor; P21-2: Take the initial load control factor and the capacitor as monitoring variables, and collect the historical operation data of the first load node, wherein the historical operation data of the first load node includes the initial load control factor characteristic value data set and the first load node capacitance characteristic value data set; P21-3: Normalize the first load node capacitance characteristic value data set to construct a benchmark data sequence; P21-4: Normalize the initial load control factor characteristic value data set to construct a comparison data sequence; P21-5: Perform grey correlation analysis based on the benchmark data sequence and the comparison data sequence to obtain the initial load control factor correlation set, wherein the initial load control factor correlation set corresponds to the initial load control factor one-to-one; P21-6: Traverse the initial load control factor correlation set, and compare them with the correlation threshold respectively. When it is greater than or equal to the correlation threshold, add the corresponding initial load control factor to the first load control factor, wherein the correlation threshold is a user-defined correlation threshold value.
[0031] Optionally, for the case where the capacitor is used as the first load node, a more detailed data processing and analysis process is designed to accurately extract the control factors of the capacitor. First, based on the physical properties and circuit theory of the capacitor, a series of parameters that may affect its current fluctuation characteristics are preliminarily set as initial load control factors. These factors include but are not limited to capacitance value, operating voltage range, temperature coefficient, etc.
[0032] Subsequently, the historical operation data of the first load node (i.e., capacitor) is obtained through the data acquisition system using the initial load control factor and the capacitor itself as monitoring variables. These data include two parts: one is the characteristic value data set of the initial load control factor, which records the specific values of each initial control factor at different time points; the other is the characteristic value data set of the first load node capacitance, which records the characteristic values that can be directly observed or calculated during the operation of the capacitor, such as the voltage across the capacitor, the current flowing through, etc.
[0033] In order to eliminate the influence of different dimensions and magnitudes on data analysis, the first load node capacitance characteristic value data set is processed by a normalization method, and converted into a dimensionless value in the interval [0,1], thereby constructing a reference data sequence. Similarly, the characteristic value data set of the initial load control element is normalized to construct a comparison data sequence. In this way, the reference data sequence and the comparison data sequence are under the same measurement, which is convenient for subsequent comparison and analysis.
[0034] Next, the reference data sequence and the comparison data sequence are compared using the grey correlation analysis method in the grey system theory. The grey correlation analysis is a method for measuring the degree of correlation between system factors, and the degree of correlation is determined by analyzing the similarity or difference between the factors. In this step, the initial load control factor correlation set is obtained by calculating the grey correlation between each initial load control factor and the capacitor characteristic value, and the initial load control factor correlation set corresponds to the initial load control factor one by one.
[0035] Finally, obtain the preset correlation threshold, which is a user-defined correlation threshold value, and can be adaptively adjusted by the user according to the actual situation. And traverse the correlation of all initial load control elements in the initial load control element correlation set, and compare them with the correlation threshold respectively. When the correlation of any initial load control element is greater than or equal to the correlation threshold, the corresponding initial load control element is added to the first load control element. These elements contained in the first load control element are considered to be the main factors affecting the capacitor current fluctuation characteristics, and will be used as key control parameters in subsequent analysis.
[0036] P30: splitting and dividing the substation circuit topology diagram to obtain a first split circuit, a second split circuit, and finally an Mth split circuit.
[0037] Specifically, the previously constructed substation circuit topology diagram is split and divided. That is, the circuit of the entire substation is divided according to different branch paths to form multiple split circuits, such as the first split circuit, the second split circuit, and the Mth split circuit. In this way, the current flow path can be further refined, which is convenient for independent monitoring and analysis of the current waveforms of different branch circuits.
[0038] The core of current splitting is to determine the direction of current flow by analyzing the branch structure of the circuit. For each split circuit, identify its independent load path and separate these paths from the overall circuit. The split circuit can contain multiple load nodes or devices to form a local small power system. For example, some load nodes in the substation circuit may have a common power supply path. In this case, it is necessary to separate these nodes from their common path to form a split circuit. These split circuits may be based on regional divisions, such as the power supply path of a building or a community, or they may be based on function, such as a circuit dedicated to powering industrial equipment.
[0039] In terms of technical support, current splitting relies on a deep understanding of the distribution network and the support of real-time monitoring data. Circuit simulation tools and geographic information systems (GIS) can be combined to automatically identify and divide branch paths in the circuit. Load flow analysis technology can also be applied to analyze the current flow characteristics of each node to determine the best splitting solution. The split circuit not only helps to simplify the analysis of the current waveform, but also improves the overall monitoring efficiency of the power grid, ensuring that abnormal waveforms in different split circuits can be identified and processed earlier.
[0040] P40: Based on the substation circuit topology diagram, a substation circuit diagram neural network architecture is constructed, wherein the first load node, the second load node until the Nth load node, and the substation power supply are input nodes of the substation circuit diagram neural network architecture, and the first splitting circuit, the second splitting circuit until the Mth splitting circuit are output nodes of the substation circuit diagram neural network architecture.
[0041] In a possible embodiment of the present application, an innovative neural network architecture for the substation circuit diagram is constructed based on the constructed substation circuit topology diagram. This architecture aims to use the powerful data processing and learning capabilities of the neural network to model and analyze the complex relationships in the substation circuit to achieve more accurate and efficient current waveform prediction, load management, and fault detection functions.
[0042] First, the basic structure of the neural network architecture of the substation circuit diagram is clarified. In this architecture, each load node in the substation circuit topology diagram (including the first load node, the second load node to the Nth load node) and the substation power supply are used as input nodes. These input nodes represent the basic elements and energy sources in the substation circuit, and their state changes will directly affect the behavior of the entire circuit.
[0043] Secondly, according to the shunt splitting results obtained in the above steps, the first splitting circuit, the second splitting circuit, and the Mth splitting circuit are set as output nodes of the neural network architecture of the station circuit diagram. These output nodes reflect the performance of the circuit under different shunt paths and are important targets for learning and prediction by the neural network.
[0044] For example, when building a neural network architecture, the technical advantages of graph neural networks are utilized. Graph neural networks are a type of neural network specifically designed to process graph structured data, which can effectively capture the complex relationships and features between nodes. By treating the substation circuit topology as a special graph structure and applying the relevant algorithms of the graph neural network, accurate modeling of the relationship between each load node and split circuit in the substation circuit can be achieved.
[0045] Specifically, each input node in the substation circuit graph neural network architecture aggregates information from its neighboring nodes through a series of graph convolutional layers or graph attention layers. After nonlinear transformation and feature extraction, this information will be used to generate the output representation of each split circuit. Finally, by comparing the difference between the actual output and the predicted output and optimizing the neural network parameters using the back-propagation algorithm, the network's prediction accuracy and generalization ability for substation circuit behavior can be continuously improved.
[0046] P50: Using the instantaneous output voltage of the substation power supply, the first load control factor, the second load control factor up to the Nth load control factor as input node input parameters, and using the first split circuit current, the second split circuit current up to the Mth split circuit current as output node output parameters, perform positive sample training on the substation circuit diagram neural network architecture to obtain an instantaneous current predictor.
[0047] Furthermore, step P50 of the embodiment of the present application also includes:
[0048] P51: Construct an input space according to the instantaneous output voltage, the first load control factor, the second load control factor, and up to the Nth load control factor; P52: Collect a positive sample data set of the first substation power grid, wherein the positive sample is data collected when the circuit is running healthily, and the first substation power grid is a complete positive sample data set for initial training; P53: Distribute the positive sample data set of the first substation power grid in the input space to obtain the input particle distribution coordinates; P54: According to the particle distribution distance threshold, perform cluster analysis on the input particle distribution coordinates to obtain the input particle clustering result, wherein the particle distribution distance threshold is the distance threshold preset by the model configuration personnel to identify whether the particles belong to the same class; P55: According to the intra-class particle number threshold, traverse the input particle clustering results and perform random deletion to obtain the second substation power grid positive sample data set, wherein the second substation power grid positive sample data set includes the instantaneous output voltage record data set, the first load control factor eigenvalue record data set, the second The load control factor characteristic value record data set until the Nth load control factor characteristic value record data set, the first split circuit current characteristic value record data set, the second split circuit current characteristic value record data set until the Mth split circuit current characteristic value record data set, wherein the intra-class particle number threshold is the maximum number of intra-class particles preset by the model configuration personnel, and the second substation area power grid positive sample data set is a representative positive sample data set obtained after clustering and screening the first substation area power grid positive sample data set; P56: Taking the first split circuit current characteristic value record data set, the second split circuit current characteristic value record data set until the Mth split circuit current characteristic value record data set as supervision examples, taking the instantaneous output voltage record data set, the first load control factor characteristic value record data set, the second load control factor characteristic value record data set until the Nth load control factor characteristic value record data set as input examples, train the substation circuit diagram neural network architecture to obtain the instantaneous current predictor.
[0049] It should be understood that the instantaneous output voltage of the power supply of the substation and each load control element are used as input parameters of the input node, and the current of the split circuit is used as the output parameter of the output node to train the substation circuit diagram neural network architecture, and finally obtain an instantaneous current predictor. The function of this predictor is to predict the current fluctuation in the split circuit according to the changes in the power supply and load.
[0050] First, a multidimensional input space is constructed based on the instantaneous output voltage of the power supply in the substation area and the first to Nth load control factors (these factors may include load type, power demand, switch status, etc., which together affect the current distribution in the circuit). This input space is the basis for subsequent data processing and model training, and it can fully reflect the various changes in conditions during the operation of the power grid.
[0051] Subsequently, a positive sample data set of the first substation power grid is collected. Positive samples specifically refer to data collected when the circuit is running healthily. These data reflect various parameters of the power grid under normal working conditions, including instantaneous output voltage, various load control elements, and current values of each split circuit. The first substation power grid positive sample data set is a complete positive sample data set for initial training, which is usually historical data under normal operating conditions of the power grid. This data set contains the normal current waveform of the power grid under different loads and power supply outputs. These data are mainly used to train the neural network architecture to help the system learn the relationship between the voltage change of the substation power supply and the current change of each load node. The collected first substation power grid positive sample data set is mapped to the previously constructed input space to form the input particle distribution coordinates. This step is to convert abstract data into specific points in space to facilitate subsequent data analysis and processing.
[0052] Furthermore, a cluster analysis method is used to group the input particle distribution coordinates according to the particle distribution distance threshold. The particle distribution distance threshold is a distance threshold preset by the model configuration personnel to identify whether the particles belong to the same category, which can be set based on empirical data. Cluster analysis is an unsupervised learning algorithm that can classify similar data points into the same category, thereby revealing the inherent structure and laws of the data. Through cluster analysis, potential patterns in the data can be discovered, providing guidance for subsequent data processing and model training.
[0053] Next, the clustering results are traversed according to the intra-class particle number threshold, and random deletion operations are performed. The intra-class particle number threshold is the maximum number of intra-class particles preset by the model configuration personnel. In order to reduce data redundancy, the system will limit the number of particles in each class to ensure that the number of particles in each class does not exceed the intra-class particle number threshold. On this basis, the particles that exceed the intra-class particle number threshold are randomly deleted to obtain the second substation power grid positive sample data set. The purpose of this step is to improve the generalization ability and robustness of the model through data enhancement and balancing processing. By randomly deleting some data points, different scenarios in power grid operation are simulated to make the model more adaptable to the complex and changeable power grid environment. Therefore, the second substation power grid is a representative positive sample data set obtained by clustering and screening the positive sample data set of the first substation power grid. This data set contains the characteristic value records of the instantaneous output voltage, load control elements, and split circuit current in the substation power grid, and is a representative data set after screening. After optimization and filtering, the quality of the training data is ensured and redundant data is removed, which can be used for more in-depth training and optimization of neural networks to ensure that the model can accurately predict the current waveform in the face of complex power grids.
[0054] Finally, the current characteristic value record data set of each split circuit in the second substation power grid positive sample data set is used as the supervision case (i.e., the target variable), and the instantaneous output voltage and the characteristic value record data set of each load control element are used as the input case (i.e., the independent variable) to train the substation circuit diagram neural network architecture. Through multiple iterations and optimizations, an instantaneous current predictor that can accurately predict the instantaneous current of the substation power grid is obtained. This predictor can monitor the operation status of the power grid in real time, predict future current changes, and provide strong guarantees for the safe and stable operation of the power grid.
[0055] Furthermore, step P56 of the embodiment of the present application also includes:
[0056] P56-1: The ratio of the predicted deviation of the current characteristic values of M split circuits that is greater than the current characteristic value predicted deviation threshold is used as the first loss evaluation function, wherein the first loss evaluation function has a first convergence loss threshold, and the first convergence loss threshold is the first loss evaluation function convergence threshold preset by the model configuration personnel; P56-2: The average of the predicted deviation of the current characteristic values of M split circuits is used as the second loss evaluation function, wherein the second loss evaluation function has a second convergence loss threshold, and the second convergence loss threshold is the second loss evaluation function convergence threshold preset by the model configuration personnel; P56-3: The first loss evaluation function and the second loss evaluation function are The evaluation function configures a logical AND relationship to construct a joint convergence condition; P56-4: Using the first split circuit current characteristic value record data set, the second split circuit current characteristic value record data set, and the Mth split circuit current characteristic value record data set as supervision examples, and using the instantaneous output voltage record data set, the first load control factor characteristic value record data set, the second load control factor characteristic value record data set, and the Nth load control factor characteristic value record data set as input examples, train the substation circuit diagram neural network architecture, and when at least the preset number of trainings in Q consecutive trainings meets the joint convergence condition, output the instantaneous current predictor.
[0057] Specifically, the training of the neural network not only relies on traditional supervised learning, but also introduces a loss evaluation function and a joint convergence condition to ensure that the neural network achieves the ideal prediction accuracy during the training process. First, a first loss evaluation function is defined, which measures the accuracy of the prediction by taking the proportion of the current characteristic value prediction deviation of M split circuits greater than the set prediction deviation threshold as a standard. Specifically, if the current characteristic value prediction deviation of a split circuit exceeds the preset threshold, it means that a large error has occurred in the current prediction of the circuit, so it is included in the calculation range of the first loss evaluation function. The function has a first convergence loss threshold, which is the first loss evaluation function convergence threshold preset by the model configuration personnel. When the proportion of the model current characteristic value prediction deviation greater than the preset value is less than the threshold, the convergence state of the model is considered acceptable. The first convergence loss threshold can be obtained by statistically analyzing the proportion of historical current characteristic value prediction deviation greater than the preset value, that is, by analyzing the historical data of the power grid, calculating the deviation between the current prediction and the actual value of each split circuit in multiple trainings, and statistically calculating the proportion of these deviations greater than a preset threshold, and determining a suitable proportion threshold through multiple experiments as the convergence standard of the model.
[0058] Next, the second loss evaluation function is introduced. The second loss evaluation function evaluates the overall prediction error by calculating the mean of the current characteristic value prediction deviations of the M split circuits. The calculation of this mean can better reflect the accuracy of the global prediction, rather than just focusing on individual circuits that exceed the deviation threshold. Similar to the first loss evaluation function, the second loss evaluation function also has a second convergence loss threshold, which is used to measure the degree of training convergence. Among them, the second convergence loss threshold is the convergence threshold of the second loss evaluation function preset by the model configuration personnel. When the mean of the model current characteristic value prediction deviation is less than the threshold, the convergence state of the model is considered acceptable. The threshold can be calculated by statistically analyzing the mean of the historical current characteristic value prediction deviations, that is, by performing statistical analysis on the deviation mean in multiple experiments, an acceptable deviation mean threshold is determined as the standard for model convergence.
[0059] In order to comprehensively utilize the advantages of the first loss evaluation function and the second loss evaluation function, we configure the logical and relationship for them and construct the joint convergence condition. This means that during the training process, only when the two loss evaluation functions simultaneously meet their respective convergence loss thresholds, the current training state is considered to meet the convergence condition. This setting of joint convergence conditions helps to improve the comprehensiveness and rigor of model training and ensure that the model achieves high prediction accuracy in all aspects.
[0060] Finally, the M split circuit current characteristic value record data sets are used as the supervision examples (i.e., the target value in the training process), and the instantaneous output voltage record data sets and the characteristic value record data sets of each load control element are used as the input examples (i.e., the independent variables in the training process) to train the neural network architecture of the substation circuit diagram. During the training process, the satisfaction of the joint convergence conditions is continuously monitored. If at least the preset number of trainings in Q consecutive trainings meets the joint convergence conditions, the model is considered to have been trained to maturity, and the final instantaneous current predictor is output. This step ensures the high performance and stability of the instantaneous current predictor through strict convergence judgment and training process control.
[0061] P60: Obtain the output voltage timing information of the power supply in the substation, the monitoring timing information of the first load control element, the monitoring timing information of the second load control element until the Nth load control element monitoring timing information, the current monitoring waveform of the first split circuit, the current monitoring waveform of the second split circuit until the Mth split circuit.
[0062] Optionally, comprehensively collect and integrate key timing information and monitoring waveforms during the operation of the substation power grid to provide basic data support for subsequent analysis, prediction or optimization.
[0063] Specifically, first obtain the output voltage timing information of the power supply in the substation. This timing information refers to the change of the output voltage of the power supply in the substation at different time points. These change data are usually monitored in real time by voltage sensors, which can reflect the dynamic characteristics and voltage fluctuation behavior of the power supply.
[0064] Next, the monitoring timing information of the first load control element, the second load control element, and the Nth load control element is obtained. These load control elements represent the electrical characteristics of different load nodes in the substation (such as load current, voltage, power, etc.), and their timing information records the changes of these characteristics over time. By monitoring this timing information, the system can analyze the contribution of each load node to the current waveform and identify the impact of load changes on the overall current waveform.
[0065] In addition, the current monitoring waveform of the first split circuit, the current monitoring waveform of the second split circuit, and the current monitoring waveform of the Mth split circuit are obtained. These monitoring waveforms are collected in real time by the current sensors in the split circuits, reflecting the changes in the current in each split circuit. The current monitoring waveform can reveal the fluctuation pattern of the current over time and provide training data for the subsequent prediction model.
[0066] The collection of this data relies on high-precision monitoring equipment, including voltage sensors and current sensors. These sensors upload the collected timing information and monitoring waveforms to the system's central processing unit for real-time analysis and storage. At the same time, the system combines data preprocessing techniques such as filtering and noise reduction to ensure that the collected data accurately reflects the actual operation of the power grid. Through this step, the system can obtain complete substation circuit operation data, providing data support for subsequent current prediction and waveform analysis. Technically, the real-time monitored timing information and waveform data provide training samples for the current predictor model and lay the foundation for the detection of current anomalies.
[0067] P70: Through the instantaneous current predictor, multi-timing predictions are performed on the substation power supply output voltage timing information, the first load control element monitoring timing information, the second load control element monitoring timing information, and up to the Nth load control element monitoring timing information to obtain the first split circuit current prediction waveform, the second split circuit current prediction waveform, and up to the Mth split circuit current prediction waveform.
[0068] Furthermore, step P70 of the embodiment of the present application further includes:
[0069] P71: Obtain the output voltage of the power supply of the substation at the first moment, the monitoring value of the first load control factor at the first moment, the monitoring value of the second load control factor at the first moment, and the monitoring value of the Nth load control factor at the first moment; P72: Fit the output voltage of the power supply of the substation at the first moment, the monitoring value of the first load control factor at the first moment, the monitoring value of the second load control factor at the first moment, and the monitoring value of the Nth load control factor at the first moment through the instantaneous current predictor, and obtain the predicted current value of the first split circuit at the first moment, the predicted current value of the second split circuit at the first moment, and the predicted current value of the Mth split circuit at the first moment; P73: Until the output voltage of the power supply of the substation at the Yth moment, the monitoring value of the first load control factor at the Yth moment, the monitoring value of the second load control factor at the Yth moment, and the monitoring value of the Nth load control factor at the Yth moment are obtained. Control factor monitoring value; P74: Fit the output voltage of the substation power supply at the Yth moment, the first load control factor monitoring value at the Yth moment, the second load control factor monitoring value at the Yth moment until the Nth load control factor monitoring value at the Yth moment through the instantaneous current predictor, and obtain the first split circuit current prediction value at the Yth moment, the second split circuit current prediction value at the Yth moment until the Mth split circuit current prediction value at the Yth moment; P75: Construct the first split circuit current prediction waveform according to the first split circuit current prediction value at the first moment until the first split circuit current prediction value at the Yth moment; P76: Construct the Mth split circuit current prediction waveform according to the Mth split circuit current prediction value at the first moment until the Mth split circuit current prediction value at the Yth moment.
[0070] It should be understood that the trained instantaneous current predictor is used to perform multi-time series prediction on the output voltage time series information of the power supply in the substation and the monitoring time series information of each load control element, thereby generating the current prediction waveform of different split circuits. This step aims to predict the future current waveform in advance by modeling and fitting the voltage and load changes at each time point.
[0071] First, the output voltage of the power supply in the substation and the monitoring values of each load control element at the first moment are obtained. These monitoring values reflect the state of the voltage output of the power supply and the current, power and other control elements of each load node at a specific moment.
[0072] Next, these first-moment monitoring values are fitted through the instantaneous current predictor. This fitting process takes the monitored voltage and load control data as input to predict the current value of each split circuit at this moment. In this way, the predicted current value of the first split circuit at the first moment, the predicted current value of the second split circuit at the first moment, and the predicted current value of the Mth split circuit at the first moment can be calculated. Repeat this process to obtain the monitoring values of the power supply output voltage of the station area and each load control element at the Yth moment. The data at the Yth moment is also used for subsequent fitting and prediction.
[0073] Furthermore, the voltage and load control factors at the Yth moment are fitted by the instantaneous current predictor to generate the current prediction value at that moment, namely the first split circuit current prediction value, the second split circuit current prediction value at the Yth moment, until the Mth split circuit current prediction value.
[0074] The current prediction waveform of the first split circuit is constructed point by point from all the current prediction values between the first moment and the Yth moment. This waveform reflects how the current of the first split circuit fluctuates with the changes of the power supply voltage and load control elements of the station area at different time points.
[0075] Similarly, a current prediction waveform of the Mth split circuit is constructed, which is based on the current prediction value from the first moment to the Yth moment. In this way, the current fluctuation of each split circuit is grasped in advance, providing data support for subsequent current waveform detection and abnormality analysis.
[0076] P80: Based on the first split circuit current prediction waveform, the second split circuit current prediction waveform, and up to the Mth split circuit current prediction waveform, perform abnormal analysis on the first split circuit current monitoring waveform, the second split circuit current monitoring waveform, and up to the Mth split circuit current monitoring waveform, obtain the current waveform abnormal circuit, and add the current waveform detection result.
[0077] Further, such as Figure 3 As shown, step P80 of the embodiment of the present application also includes:
[0078] P81: Compare the current prediction waveform of the first splitting circuit and the current monitoring waveform of the first splitting circuit to obtain the current characteristic value deviation from the first moment until the current characteristic value deviation at the Yth moment; P82: Count the proportion of the current characteristic value deviation from the first moment until the current characteristic value deviation at the Yth moment that is greater than or equal to the current characteristic value deviation threshold, and set it as the current waveform abnormality coefficient; P83: When the current waveform abnormality coefficient is greater than or equal to the current waveform abnormality coefficient threshold, add the first splitting circuit into the current waveform abnormality circuit, wherein the current waveform abnormality coefficient threshold is a threshold value of whether the current waveform is abnormal and is custom-configured by professionals; P84: Add the current waveform abnormality circuit to the current waveform detection result.
[0079] Optionally, a previously generated split circuit current prediction waveform is compared with an actual current monitoring waveform to identify and locate a circuit with an abnormal current waveform.
[0080] First, the current prediction waveforms of the first to Mth split circuits are compared with the corresponding current monitoring waveforms one by one. By extracting and comparing the current characteristic values (such as amplitude, phase, etc.) of the two at each moment (from the first moment to the Yth moment), the corresponding current characteristic value deviations are calculated. These deviations reflect the difference between the predicted value and the actual value, which is an important basis for judging whether the current waveform is abnormal.
[0081] Next, the current characteristic value deviation of each split circuit is statistically analyzed. In particular, the proportion of deviation values greater than or equal to the preset current characteristic value deviation threshold is calculated, and this proportion is set as the current waveform abnormality coefficient. This coefficient intuitively reflects the degree to which the current waveform deviates from the normal state and is an important indicator for evaluating the health of the circuit.
[0082] Then, the current waveform abnormality coefficient is compared with the preset current waveform abnormality coefficient threshold. If the current waveform abnormality coefficient of a split circuit is greater than or equal to the threshold, it is considered that the circuit has a current waveform abnormality, and it is marked as a current waveform abnormality circuit to achieve accurate identification of the abnormal circuit. Among them, the current waveform abnormality coefficient threshold is a threshold value of whether the current waveform is abnormal, which is customized by professionals. It can be obtained by analyzing the distribution of the current waveform abnormality coefficient under different operating conditions, especially the difference between normal operation and failure. Exemplarily, the system first collects the current monitoring data of the power grid during normal operation and abnormal conditions, and performs statistical analysis on these data to obtain the current waveform abnormality coefficient under normal and abnormal conditions. Then, by performing a centralized trend analysis on the current waveform abnormality coefficient under normal and abnormal conditions, a critical point is determined, which can more accurately distinguish between normal current waveforms and abnormal waveforms. The critical point is the current waveform abnormality coefficient threshold, which will be used as a standard for judging whether the current waveform of a split circuit is abnormal.
[0083] Finally, all identified circuits with abnormal current waveforms are added to the current waveform detection results. This result not only provides clear fault indications for power grid operation and maintenance personnel, but also facilitates subsequent targeted fault detection and repair work.
[0084] Furthermore, step P82 of the embodiment of the present application also includes:
[0085] P82-1: Obtain the positive sample current monitoring data of the first split circuit, wherein the positive sample current monitoring data of the first split circuit includes a theoretical current recording data set and a monitoring current recording data set, and the positive sample current monitoring data of the first split circuit is the current data of the first split circuit under normal working conditions; P82-2: Calculate the excessive deviation modulus value set of the theoretical current recording data set and the monitoring current recording data set; P82-3: Calculate the excessive deviation modulus value set of the theoretical current recording data set and the monitoring current recording data set; P82-4: Perform a central tendency analysis on the excessive deviation modulus value set to obtain an excessive deviation threshold; P82-5: Perform a central tendency analysis on the excessive deviation modulus value set to obtain an excessive deviation threshold; P82-6: Construct the current characteristic value deviation threshold based on the excessive deviation threshold and the excessive deviation threshold.
[0086] Specifically, the specific construction process of the current characteristic value deviation threshold can be to process the positive sample current monitoring data, calculate the modulus of excessive and insufficient deviations, and perform central trend analysis to construct the deviation threshold of the current characteristic value. These deviation thresholds are key parameters for current waveform anomaly detection, which help to more accurately determine whether there is an abnormality in current fluctuations.
[0087] First, the current data of the first split circuit under normal working conditions (i.e., positive samples) are collected, i.e., the positive sample current monitoring data of the first split circuit. These data include two parts: a theoretical current recording data set, which is calculated based on the ideal state of the circuit design, and a monitoring current recording data set, which is collected in real time by sensors actually installed in the circuit. The theoretical current recording data set represents the ideal current value calculated based on the load and power supply status, while the monitoring current recording data set is the current value collected by the actual monitoring equipment.
[0088] Furthermore, by comparing the two sets of data point by point, the current deviation at each corresponding point (i.e., the difference between the measured value and the theoretical value) is calculated, and the deviation modulus (i.e., the absolute value of the deviation) of all deviation values exceeding the preset preliminary threshold (such as a smaller value set based on experience or preliminary statistics) is screened out to form a set of excessive deviation modulus values. This process helps to identify those current fluctuation points that deviate significantly from theoretical expectations. Similarly, a set of excessively small deviation modulus values is calculated. Excessively small deviation refers to a situation where the monitored current value is significantly lower than the theoretical current value. Calculating the modulus values of these deviations can capture the situation of insufficient current.
[0089] For the set of excessive deviation modulus values, a central tendency analysis is performed. The central tendency analysis is used to identify the main trend in the deviation modulus value set, determine its concentrated distribution area, and thus calculate a reasonable excessive deviation threshold. This threshold represents the critical value at which the system considers the deviation to be too large. Similarly, a central tendency analysis is performed on the set of excessive deviation modulus values to obtain an excessively small deviation threshold. This threshold is used to identify the critical value of the current being too small.
[0090] Finally, a unified current characteristic value deviation threshold is constructed based on the excessive deviation threshold and the too small deviation threshold obtained by analysis. This system not only takes into account the positive anomaly of current fluctuation (excessive deviation), but also covers the negative anomaly (too small deviation), providing comprehensive data support for the health monitoring and fault warning of the circuit system. By monitoring the current data in real time and comparing it with these thresholds, abnormal conditions in the circuit can be discovered and handled in a timely manner to ensure the stable operation of the system.
[0091] In summary, the embodiments of the present application have at least the following technical effects:
[0092] The present application regards the distribution transformer as the substation power source, constructs the substation circuit topology map, traverses multiple load nodes of the substation circuit topology map to perform current fluctuation factor correlation analysis, obtains multiple load control factors, and uses this to perform positive sample training to obtain an instantaneous current predictor, predicts and obtains multiple split circuit current prediction waveforms, and uses this to perform abnormal analysis on the multiple split circuit current monitoring waveforms to obtain current waveform abnormal circuits.
[0093] The technical effect of achieving adaptive detection of current waveforms in smart grids, improving detection accuracy, and enhancing the adaptability and robustness of the system is achieved.
[0094] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
[0095] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.
[0096] This specification and the drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.
Claims
1. An adaptive current waveform detection method for smart grid, characterized in that: include: The distribution transformer is regarded as a substation power source, and a substation circuit topology diagram is constructed, wherein the substation circuit topology diagram includes a first load node, a second load node, and up to an Nth load node; Traversing the first load node, the second load node, and finally the Nth load node to perform current fluctuation factor correlation analysis, and obtain a first load control factor, a second load control factor, and finally the Nth load control factor; Performing shunt splitting on the area circuit topology diagram to obtain a first split circuit, a second split circuit, and up to an Mth split circuit; Based on the station area circuit topology diagram, a station area circuit diagram neural network architecture is constructed, wherein the first load node, the second load node until the Nth load node, and the station area power supply are input nodes of the station area circuit diagram neural network architecture, and the first splitting circuit, the second splitting circuit until the Mth splitting circuit are output nodes of the station area circuit diagram neural network architecture; Taking the instantaneous output voltage of the power supply of the station area, the first load control factor, the second load control factor to the Nth load control factor as input node input parameters, taking the first split circuit current, the second split circuit current to the Mth split circuit current as output node output parameters, performing positive sample training on the neural network architecture of the station area circuit diagram to obtain an instantaneous current predictor; Obtaining the output voltage timing information of the power supply in the substation, the first load control element monitoring timing information, the second load control element monitoring timing information until the Nth load control element monitoring timing information, the first split circuit current monitoring waveform, the second split circuit current monitoring waveform until the Mth split circuit current monitoring waveform; Through the instantaneous current predictor, multi-time prediction is performed on the output voltage timing information of the substation power supply, the first load control element monitoring timing information, the second load control element monitoring timing information, and the Nth load control element monitoring timing information to obtain the first split circuit current prediction waveform, the second split circuit current prediction waveform, and the Mth split circuit current prediction waveform; According to the first split circuit current prediction waveform, the second split circuit current prediction waveform, and up to the Mth split circuit current prediction waveform, an abnormal analysis is performed on the first split circuit current monitoring waveform, the second split circuit current monitoring waveform, and up to the Mth split circuit current monitoring waveform to obtain a current waveform abnormal circuit and add the current waveform detection result.
2. The method according to claim 1, characterized in that Traversing the first load node, the second load node, and up to the Nth load node to perform current fluctuation factor correlation analysis, and obtaining the first load control factor, the second load control factor, and up to the Nth load control factor, including: When the first load node is a capacitor, performing capacitance correlation analysis to obtain the first load control factor; When the first load node is an inductor, performing inductance correlation analysis to obtain the first load control factor; When the first load node is other components, performing resistance correlation analysis to obtain the first load control factor; Among them, the acquisition process of the second load control element to the Nth load control element is the same as that of the first load control element.
3. The method according to claim 2, characterized in that When the first load node is a capacitor, performing capacitance correlation analysis to obtain the first load control factor includes: Configure initial load control elements; Taking the initial load control factor and the capacitor as monitoring variables, collecting historical operation data of the first load node, wherein the historical operation data of the first load node includes a characteristic value data set of the initial load control factor and a characteristic value data set of the first load node capacitance; Normalizing the first load node capacitance characteristic value data set to construct a reference data sequence; Normalizing the initial load control factor characteristic value data set to construct a comparison data sequence; Performing grey correlation analysis on the reference data sequence and the comparison data sequence to obtain an initial load control element correlation set, wherein the initial load control element correlation set corresponds one-to-one to the initial load control element; Traverse the initial load control element association set and compare them with the association threshold respectively. When it is greater than or equal to the association threshold, add the corresponding initial load control element into the first load control element, wherein the association threshold is a user-defined association threshold value.
4. The method according to claim 1, characterized in that The instantaneous output voltage of the power supply in the area, the first load control factor, the second load control factor to the Nth load control factor are used as input node input parameters, and the first split circuit current, the second split circuit current to the Mth split circuit current are used as output node output parameters, and the neural network architecture of the area circuit diagram is trained with positive samples to obtain an instantaneous current predictor, including: constructing an input space according to the instantaneous output voltage, the first load control factor, the second load control factor, and up to the Nth load control factor; Collecting a positive sample data set of the power grid in the first substation area, wherein the positive sample is data collected when the circuit is running healthily, and the positive sample data set of the power grid in the first substation area is a complete positive sample data set used for initial training; Distribute the first substation area power grid positive sample data set in the input space to obtain input particle distribution coordinates; According to the particle distribution distance threshold, cluster analysis is performed on the input particle distribution coordinates to obtain the input particle clustering result, wherein the particle distribution distance threshold is a distance threshold preset by the model configuration personnel to identify whether particles belong to the same type; According to the intra-class particle number threshold, the input particle clustering results are traversed and randomly deleted to obtain a second substation power grid positive sample data set, wherein the second substation power grid positive sample data set includes an instantaneous output voltage record data set, a first load control factor characteristic value record data set, a second load control factor characteristic value record data set until the Nth load control factor characteristic value record data set, a first split circuit current characteristic value record data set, a second split circuit current characteristic value record data set until the Mth split circuit current characteristic value record data set, wherein the intra-class particle number threshold is the maximum number of intra-class particles preset by the model configuration personnel, and the second substation power grid positive sample data set is a representative positive sample data set obtained after clustering and screening the first substation power grid positive sample data set; The first split circuit current characteristic value record data set, the second split circuit current characteristic value record data set, and the Mth split circuit current characteristic value record data set are used as supervision examples, and the instantaneous output voltage record data set, the first load control factor characteristic value record data set, the second load control factor characteristic value record data set, and the Nth load control factor characteristic value record data set are used as input examples to train the substation circuit diagram neural network architecture and obtain the instantaneous current predictor.
5. The method according to claim 4, characterized in that The method uses the first split circuit current characteristic value record data set, the second split circuit current characteristic value record data set, and the Mth split circuit current characteristic value record data set as supervision examples, and uses the instantaneous output voltage record data set, the first load control factor characteristic value record data set, the second load control factor characteristic value record data set, and the Nth load control factor characteristic value record data set as input examples to train the station area circuit diagram neural network architecture to obtain the instantaneous current predictor, including: The ratio of the current characteristic value prediction deviation of the M split circuits that is greater than the current characteristic value prediction deviation threshold is used as the first loss evaluation function, wherein the first loss evaluation function has a first convergence loss threshold, and the first convergence loss threshold is a first loss evaluation function convergence threshold preset by a model configuration personnel; Taking the average of the predicted deviations of the current characteristic values of the M split circuits as the second loss evaluation function, wherein the second loss evaluation function has a second convergence loss threshold, and the second convergence loss threshold is a second loss evaluation function convergence threshold preset by a model configuration personnel; Configuring a logical AND relationship for the first loss evaluation function and the second loss evaluation function to construct a joint convergence condition; The first split circuit current characteristic value record data set, the second split circuit current characteristic value record data set, and the Mth split circuit current characteristic value record data set are used as supervision examples, and the instantaneous output voltage record data set, the first load control factor characteristic value record data set, the second load control factor characteristic value record data set, and the Nth load control factor characteristic value record data set are used as input examples to train the neural network architecture of the substation circuit diagram. When at least a preset number of trainings in Q consecutive trainings meet the joint convergence condition, the instantaneous current predictor is output.
6. The method according to claim 1, characterized in that By means of the instantaneous current predictor, multi-time sequence prediction is performed on the output voltage timing information of the power supply in the substation area, the monitoring timing information of the first load control element, the monitoring timing information of the second load control element, and up to the monitoring timing information of the Nth load control element, to obtain the first split circuit current prediction waveform, the second split circuit current prediction waveform, and up to the Mth split circuit current prediction waveform, including: Obtaining the output voltage of the power supply in the first moment, the first load control factor monitoring value at the first moment, the second load control factor monitoring value at the first moment, and the Nth load control factor monitoring value at the first moment; The instantaneous current predictor is used to fit the output voltage of the power supply of the metro area at the first moment, the first load control factor monitoring value at the first moment, the second load control factor monitoring value at the first moment, and the Nth load control factor monitoring value at the first moment, to obtain the first split circuit current prediction value at the first moment, the second split circuit current prediction value at the first moment, and the Mth split circuit current prediction value at the first moment; Until the output voltage of the power supply in the area at the Yth moment, the first load control factor monitoring value at the Yth moment, the second load control factor monitoring value at the Yth moment, and the Nth load control factor monitoring value at the Yth moment are obtained; The instantaneous current predictor is used to fit the output voltage of the power supply of the substation at the Yth moment, the first load control factor monitoring value at the Yth moment, the second load control factor monitoring value at the Yth moment, and the Nth load control factor monitoring value at the Yth moment, to obtain the first split circuit current prediction value at the Yth moment, the second split circuit current prediction value at the Yth moment, and the Mth split circuit current prediction value at the Yth moment; Constructing the first split circuit current prediction waveform according to the first split circuit current prediction value at the first moment to the first split circuit current prediction value at the Yth moment; Until the Mth split circuit current prediction waveform is constructed based on the Mth split circuit current prediction value at the first moment until the Mth split circuit current prediction value at the Yth moment.
7. The method according to claim 1, characterized in that According to the first split circuit current prediction waveform, the second split circuit current prediction waveform, and up to the Mth split circuit current prediction waveform, the first split circuit current monitoring waveform, the second split circuit current monitoring waveform, and up to the Mth split circuit current monitoring waveform are subjected to abnormal analysis to obtain a current waveform abnormal circuit, and added into the current waveform detection result, including: Comparing the first split circuit current prediction waveform and the first split circuit current monitoring waveform, obtaining the current characteristic value deviation at the first moment until the current characteristic value deviation at the Yth moment; Counting the proportion of the current characteristic value deviation at the first moment until the current characteristic value deviation at the Yth moment is greater than or equal to the current characteristic value deviation threshold, and setting it as the current waveform abnormality coefficient; When the current waveform abnormality coefficient is greater than or equal to the current waveform abnormality coefficient threshold, the first splitting circuit is added to the current waveform abnormality circuit, wherein the current waveform abnormality coefficient threshold is a threshold value of whether the current waveform is abnormal and is configured by professionals; The current waveform abnormality circuit is added to the current waveform detection result.
8. The method according to claim 7, characterized in that The current characteristic value deviation at the first moment is counted until the current characteristic value deviation at the Yth moment is greater than or equal to the current characteristic value deviation threshold, and is set as the current waveform abnormality coefficient, which includes: Obtaining positive sample current monitoring data of the first split circuit, wherein the positive sample current monitoring data of the first split circuit includes a theoretical current recording data set and a monitoring current recording data set, and the positive sample current monitoring data of the first split circuit is current data of the first split circuit under normal working state; Calculating a set of excessive deviation modulus values between the theoretical current recording data set and the monitored current recording data set; Calculating a set of too small deviation modulus values between the theoretical current recording data set and the monitored current recording data set; Performing a central tendency analysis on the excessive deviation modulus value set to obtain an excessive deviation threshold; Performing a central tendency analysis on the set of too small deviation modulus values to obtain an too small deviation threshold; The current characteristic value deviation threshold is constructed according to the excessively large deviation threshold and the excessively small deviation threshold.
