Intelligent power grid supervision method and system based on power grid dispatching graph

Abstract

The invention discloses an intelligent power grid supervision method and system based on a power grid dispatching graph, and the method comprises the steps: monitoring the power parameters of a transformer substation, a power plant and a power transmission line node in real time through a voltage sensor and a current sensor, and obtaining voltage data and current data; based on the power grid dispatching graph, using an isolated forest algorithm to construct a power grid anomaly detection model for anomaly detection and early warning, and identifying and responding to an early warning event; based on the power grid dispatching graph, constructing an LSTM neural network model to obtain a power grid health condition score and a predicted health result; providing power grid dispatching decision support based on the predicted health result; and taking the power grid health condition score and the predicted health result as optimization targets of a genetic algorithm to obtain an optimized power grid operation plan. By adopting the isolated forest algorithm and the LSTM neural network model, an early warning event can be recognized and responded, the health condition of the power grid can be evaluated, the trend can be predicted, scientific and accurate scheduling decision support is provided, and the stability and safety of the power grid are improved.

Description

Technical Field

[0001] The present invention relates to the field of power system automation and intelligent technology, and in particular to a smart grid supervision method and system based on a grid dispatching diagram. Background Art

[0002] The power grid supervision system, or power monitoring system, is an indispensable component of the power industry. It plays a crucial role in ensuring the safe, stable, and efficient operation of power production, transmission, and distribution. Based on computers, communications equipment, and measurement and control units, this system leverages advanced computer, network communication, and automation technologies to achieve comprehensive, multi-dimensional, and real-time management of the power system. It collects key parameters of the power system, such as voltage, current, and power factor, in real time and displays the system's operating status through an intuitive visual interface. Furthermore, the power grid supervision system supports remote control, allowing personnel to remotely operate power equipment, significantly improving work efficiency and responsiveness. In short, the power grid supervision system is a comprehensive management platform that integrates data acquisition, transmission, processing, display, and remote control, providing strong support for optimized management and decision-making within the power system.

[0003] To address the challenges of real-time monitoring, early warning response, and optimized dispatch decision support in power grid regulatory systems, existing technologies rely on traditional grid monitoring software and data analysis tools. While these tools can collect and process grid operational data, they can also suffer from delayed early warning responses and inaccurate decision support when faced with complex and changing grid environments and large amounts of power data. This is especially true during peak loads or emergencies, when traditional systems struggle to make timely and effective early warnings and dispatch decisions, leading to unstable grid operations, energy waste, and even safety incidents. Summary of the Invention

[0004] In order to solve the above-mentioned problems, the present invention provides a smart grid supervision method and system based on a power grid dispatching diagram.

[0005] In a first aspect, the present invention provides a smart grid supervision method based on a grid dispatch diagram, comprising:

[0006] Through voltage sensors and current sensors, the power parameters of substations, power plants and transmission line nodes are monitored in real time to obtain voltage and current data, and the power grid dispatch diagram is obtained based on the voltage and current data;

[0007] Based on the power grid dispatch graph, the isolation forest algorithm is used to build a power grid anomaly detection model;

[0008] Perform anomaly detection and early warning based on the power grid anomaly detection model, identify and respond to early warning events;

[0009] Based on the power grid dispatch diagram, a long short-term memory (LSTM) neural network model is constructed;

[0010] Use LSTM neural network model to obtain grid health score and predict health results;

[0011] Providing grid dispatch decision support based on the predicted health results provided by the LSTM model;

[0012] On the basis of grid dispatch decision support, the grid health score and predicted health results are used as the optimization objectives of the genetic algorithm to obtain the optimized grid operation plan.

[0013] Furthermore, voltage and current sensors are used to monitor power parameters at substations, power plants, and transmission line nodes in real time, obtaining voltage and current data, including:

[0014] Sense power parameters through voltage sensors and current sensors installed at substations, power plants and transmission line nodes;

[0015] Convert the power parameters into electrical signals, enhance the electrical signals through the amplification circuit, remove the interference components in the electrical signals through filtering, and then convert the enhanced and interference-free electrical signals into digital signals through the analog-to-digital converter;

[0016] Preprocess the digital signal, calculate the maximum value, minimum value, average value, effective value, variance and standard deviation as feature data, and obtain voltage data and current data based on the feature data;

[0017] A power grid dispatching diagram is constructed based on voltage and current data, which visually displays the power grid topology and real-time operating status.

[0018] Furthermore, based on the power grid dispatch graph, the isolation forest algorithm is used to build a power grid anomaly detection model, including:

[0019] The voltage data and current data in the power grid dispatch diagram are used as the model input of the power grid anomaly detection model;

[0020] Extract subsamples from the model input and construct an isolation tree for each subsample; the process of constructing the isolation tree is to select the voltage RMS value and the current peak value as input features, select a split value between the minimum and maximum values ​​of the voltage RMS value and the current peak value respectively, and divide the data of the current node into two left and right child nodes according to the split value. The data points with the voltage RMS value and the current peak value less than the split value are placed in the left child node, and the data points with the voltage RMS value and the current peak value greater than or equal to the split value are placed in the right child node. If the number of data points in the child node is less than the predetermined threshold and all available features have been used, the splitting is stopped. If the number of data points in the child node exceeds the predetermined threshold and there are available features that have not been used, the splitting is continued;

[0021] After the isolation tree is constructed, the path length and average path length of each voltage data point and current data point in the isolation tree are calculated, and the average path length is used as an indicator to measure whether the data point is abnormal; thus, a power grid anomaly detection model is constructed;

[0022] The path length h(x) is the number of edges from the root node to the leaf node of the data point x;

[0023] Average path length ; T represents the total number of isolated trees, is the path length of the data point x in the i-th tree.

[0024] Furthermore, anomaly detection and early warning are performed based on the power grid anomaly detection model to identify and respond to early warning events, including:

[0025] Converting average path length into anomaly score based on power grid anomaly detection model , The expression is:

[0026] ;

[0027] in, represents the anomaly score of data point x, c is the average path length of voltage data and current data minus two standard deviations, and represents the preset threshold;

[0028] The newly collected new voltage data and new current data are input into the isolation forest of the power grid anomaly detection model to evaluate their isolation degree. The data are then compared with c. If the average path length of the first data point is not less than c, the first data point is considered a normal point. If the average path length of the second data point is less than c, the second data point is considered an anomaly.

[0029] Mark the abnormal area corresponding to the abnormal point in the power grid dispatch diagram, generate early warning information including the time, location and impact range of the abnormality, send the early warning information to relevant management personnel, and record the early warning event and response results.

[0030] Furthermore, based on the power grid dispatch diagram, an LSTM neural network model is constructed, including:

[0031] The pre-configured LSTM neural network model has an input layer, a hidden layer, and an output layer;

[0032] The input layer receives the voltage and current data of the power grid dispatch diagram, and corresponds the time step including the four features of voltage RMS, current peak, variance and standard deviation to the voltage and current feature vectors at each time point. The input vector of each time step is represented as , represents the effective value of the voltage at the t-th time step, Indicates the peak current. represents the variance, represents the standard deviation;

[0033] The hidden layer uses the gating mechanism in the internal long short-term memory unit state to capture its long-term dependencies over time. The gating mechanism includes a forget gate, an input gate, a cell state update, and an output gate.

[0034] The forget gate ignores outdated voltage fluctuations and abnormal conditions that have been resolved. The calculation process is:

[0035] ;

[0036] in, represents the input vector at the current time step, represents the output state of the previous time step, represents the weight matrix corresponding to the forget gate, represents the bias term, Represents the sigmoid activation function, which maps the result to between 0 and 1, indicating the degree of retention and discarding. If it is close to 1, the information is retained; if If it is close to 0, the information is discarded;

[0037] The input gate introduces new voltage and current features. These new voltage and current features include sudden high current peaks and low voltage RMS values, which determine the new information added to the cell state. The calculation process is:

[0038] ;

[0039] ;

[0040] in, Represents the input factor, using the sigmoid activation function to determine whether each element is added to the cell state. Use the Tanh activation function to generate new candidate cell states ranging from -1 to 1, and Determine the selection mechanism of the input gate, and Control the generation of candidate cell states;

[0041] Combining the results of the forget gate and the input gate, the cell state at the previous moment is bitwise multiplied by the output of the forget gate, and then added to the candidate state generated by the input gate to update the cell state that stores long-term information. The calculation process is:

[0042] ;

[0043] The output gate determines the output state of the current time step , as the final health score output, the calculation process is:

[0044] ;

[0045] ;

[0046] in, Represents the output factor, and uses the Sigmoid activation function to determine whether each element is output. Make the output value fall between -1 and 1 to reflect the change of the current cell state;

[0047] The output layer receives the output state from the hidden layer .

[0048] Furthermore, the LSTM neural network model is used to obtain the grid health score and predict the health results, including:

[0049] The output layer of the LSTM neural network model uses an internal fully connected layer to convert the output state into a linear activation function Maps to a scalar value , It represents the grid health score at the current time point, and the expression is:

[0050] ;

[0051] in, represents the weight matrix of the output layer, represents the bias term;

[0052] A health score threshold is set. When the grid health score is lower than the health score threshold, the early warning mechanism is triggered and the predicted health result is obtained.

[0053] Furthermore, based on the predicted health results provided by the LSTM model, grid dispatch decision support is provided, including:

[0054] Using a continuous hidden state sequence as the input of the LSTM model, the hidden state sequence is mapped to the grid health score of the future time step through the fully connected layer within the output layer of the LSTM model;

[0055] The grid health score provided by the output layer assists in trend analysis and identifies deterioration trends. If the health of the target transmission line node is predicted to deteriorate, power distribution is optimized in advance to reduce the load pressure on the target transmission line node.

[0056] Furthermore, based on the grid dispatch decision support, the grid health score and predicted health results are used as the optimization objectives of the genetic algorithm to obtain the optimized grid operation plan, including:

[0057] The grid health score and predicted health results output by the LSTM model are used as optimization targets for the genetic algorithm. Initial individuals containing different operating parameter combinations are generated to form the initial population. The different operating parameter combinations include the output power setpoints of different power plants and the load distribution ratios of each transmission line.

[0058] Based on the grid health score and predicted health results provided by the LSTM model, a fitness function is defined to evaluate the performance of each individual, taking into account the grid health status, power supply and demand balance, and cost-effectiveness factors. The calculation formula is:

[0059] ;

[0060] Where p represents an individual, represents the grid health score under an individual, Indicates the degree of balance between electricity supply and demand under individual conditions, represents the individual cost, 、 and represents the weight coefficient;

[0061] Individuals with higher fitness are selected from the current population through roulette to enter the next generation. Single-point crossover and multi-point crossover are performed on the selected individuals. Information is exchanged between different operation strategies to explore potential optimization solutions, generate new offspring individuals, change the operation parameter values ​​of the newly generated offspring individuals to mutate, introduce new gene mutations, and add the new individuals after selection, crossover and mutation to the new generation population to replace the old individuals with lower fitness. The selection, crossover and mutation process is repeated until the fitness no longer increases significantly. The individual with the highest fitness is selected as the optimal solution, and an optimized power grid operation plan is proposed based on the optimal solution.

[0062] In a second aspect, a smart grid supervision system based on a power grid dispatching map is provided, which is used to implement the smart grid supervision method based on a power grid dispatching map of the first aspect, including:

[0063] The power parameter monitoring module is used to monitor the power parameters of substations, power plants and transmission line nodes in real time through voltage sensors and current sensors to obtain voltage and current data; based on the voltage and current data, a grid dispatch diagram is obtained;

[0064] The power grid anomaly warning analysis module is used to build a power grid anomaly detection model based on the power grid dispatch diagram using the isolation forest algorithm; perform anomaly detection and warning based on the power grid anomaly detection model, and identify and respond to warning events; build a long short-term memory (LSTM) neural network model based on the power grid dispatch diagram; use the LSTM neural network model to obtain a power grid health score and predicted health results; provide power grid dispatch decision support based on the predicted health results provided by the LSTM model; and, based on the power grid dispatch decision support, use the power grid health score and predicted health results as the optimization targets of the genetic algorithm to obtain an optimized power grid operation plan.

[0065] Furthermore, the system also includes:

[0066] Network communication center and integrated management module;

[0067] The network communication center is connected with the power parameter monitoring module, the power grid abnormality early warning analysis module and the comprehensive management module;

[0068] The comprehensive management module is used to integrate the functions of the power parameter monitoring module and the power grid anomaly warning and analysis module.

[0069] The beneficial effects achieved by the present invention are:

[0070] Using voltage and current sensors, the power parameters of substations, power plants, and transmission line nodes are monitored in real time to obtain voltage and current data, from which a grid dispatch diagram is derived. Based on the grid dispatch diagram, an isolation forest algorithm is used to construct a grid anomaly detection model. This model is used to detect anomalies and provide early warnings, identifying and responding to early warning events. Based on the grid dispatch diagram, a long short-term memory (LSTM) neural network model is constructed. The LSTM neural network model is used to obtain a grid health score and predicted health results. Based on the predicted health results provided by the LSTM model, grid dispatch decision support is provided. Based on this grid dispatch decision support, the grid health score and predicted health results are used as optimization targets for the genetic algorithm to obtain an optimized grid operation plan.

[0071] Power parameter monitoring enables high-precision, real-time monitoring of voltage and current data at substations, power plants, and transmission line nodes, ensuring the accuracy and real-time nature of grid operating parameters and providing a reliable data basis for subsequent abnormal warnings and dispatching decisions.

[0072] The power grid anomaly warning analysis uses the advanced isolation forest algorithm and LSTM neural network model, which can not only effectively identify potential problems in the power grid and issue early warnings, but also assess the health of the power grid and predict future trends, thereby providing scientific and accurate scheduling decision support and significantly improving the stability and security of the power grid. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] Figure 1 Flowchart of the smart grid supervision method based on the power grid dispatching diagram of the present invention;

[0074] Figure 2 This is a structural diagram of the smart grid monitoring system based on the power grid dispatching diagram of the present invention. DETAILED DESCRIPTION

[0075] The following is an explanation of the embodiments of the present invention by specific specific examples. Those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. Although the description of the present invention will be introduced in conjunction with the preferred embodiment, this does not mean that the features of this invention are limited to this embodiment. On the contrary, the purpose of introducing the invention in conjunction with the embodiment is to cover other options or modifications that may be extended based on the claims of the present invention. In order to provide a deep understanding of the present invention, the following description will contain many specific details. The present invention can also be implemented without using these details. In addition, in order to avoid confusion or blurring the focus of the present invention, some specific details will be omitted in the description. It should be noted that the embodiments of the present invention and the features in the embodiments can be combined with each other without conflict.

[0076] It should be noted that in this specification, similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, it does not need to be further defined or explained in subsequent drawings.

[0077] To make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0078] like Figure 1 As shown, an embodiment of the present invention discloses a smart grid supervision method based on a power grid dispatching diagram, comprising:

[0079] 101, using voltage sensors and current sensors, monitor power parameters of substations, power plants, and transmission line nodes in real time to obtain voltage data and current data, and obtain a power grid dispatch diagram based on the voltage data and current data;

[0080] In this embodiment, power parameters are sensed by high-precision voltage sensors and current sensors installed at substations, power plants, and transmission line nodes;

[0081] Convert the power parameters into electrical signals, enhance the electrical signals through the amplification circuit, remove the interference components in the electrical signals through filtering, and then convert the enhanced and interference-free electrical signals into digital signals through the analog-to-digital converter;

[0082] Preprocessing of digital signals involves preliminary screening of digital signals to remove erroneous data points and outliers, using filtering techniques to reduce high-frequency noise and unwanted frequency components, normalizing the data to map it to a value between 0 and 1 through linear transformation, and performing detrending on data with periodic fluctuations to eliminate the impact of long-term changes on short-term fluctuations.

[0083] The maximum, minimum, average, effective value, variance, and standard deviation are calculated from the preprocessed data as feature data, the frequency components are analyzed, the main harmonic content is identified, the power quality status is evaluated, and a power grid dispatch diagram is constructed to visually display the power grid topology and real-time operating status. The extracted voltage and current data are transmitted to the network communication center via a communication link, and the power parameters of each node are mapped to the power grid dispatch diagram; the power grid dispatch diagram visually displays the power grid topology and real-time operating status.

[0084] 102, based on the power grid dispatch graph, a power grid anomaly detection model is constructed using the isolation forest algorithm;

[0085] In this embodiment, the voltage data and current data that have been preprocessed and feature extracted in the power grid dispatch diagram are used as the model input of the power grid anomaly detection model;

[0086] Extract subsamples from the model input and construct an isolation tree for each subsample; the process of constructing the isolation tree is to select the voltage RMS value and the current peak value as input features, select a split value between the minimum and maximum values ​​of the voltage RMS value and the current peak value respectively, and divide the data of the current node into two left and right child nodes according to the split value. The data points with the voltage RMS value and the current peak value less than the split value are placed in the left child node, and the data points with the voltage RMS value and the current peak value greater than or equal to the split value are placed in the right child node. If the number of data points in the child node is less than the predetermined threshold and all available features have been used, the splitting is stopped. If the number of data points in the child node exceeds the predetermined threshold and there are available features that have not been used, the splitting is continued;

[0087] After the isolation tree is constructed, the path length and average path length of each voltage data point and current data point in the isolation tree are calculated, and the average path length is used as an indicator to measure whether the data point is abnormal; thus, a power grid anomaly detection model is constructed;

[0088] The path length h(x) is the number of edges from the root node to the leaf node of the data point x;

[0089] Average path length ; T represents the total number of isolated trees, is the path length of the data point x in the i-th tree.

[0090] 103, perform anomaly detection and early warning based on the power grid anomaly detection model, identify and respond to early warning events;

[0091] In this embodiment, the average path length is converted into anomaly score based on the power grid anomaly detection model. , The expression is:

[0092] ;

[0093] in, represents the anomaly score of data point x, c is the average path length of voltage data and current data minus two standard deviations, and represents the preset threshold;

[0094] The newly collected new voltage data and new current data are input into the isolation forest of the power grid anomaly detection model to evaluate their isolation degree. The data are then compared with c. If the average path length of the first data point is not less than c, the first data point is considered a normal point. If the average path length of the second data point is less than c, the second data point is considered an anomaly.

[0095] Mark the abnormal area corresponding to the abnormal point in the power grid dispatch diagram, generate early warning information including the time, location and impact range of the abnormality, send the early warning information to relevant management personnel, and record the early warning event and response results.

[0096] 104. Based on the power grid dispatch diagram, a long short-term memory (LSTM) neural network model is constructed;

[0097] In this embodiment, the pre-configured LSTM neural network model has an input layer, a hidden layer, and an output layer;

[0098] The input layer receives the voltage and current data of the power grid dispatch diagram, and corresponds the time step including the four features of voltage RMS, current peak, variance and standard deviation to the voltage and current feature vectors at each time point. The input vector of each time step is represented as , represents the effective value of the voltage at the t-th time step, Indicates the peak current. represents the variance, represents the standard deviation;

[0099] The hidden layer uses the gating mechanism in the internal long short-term memory unit state to capture its long-term dependencies over time. The gating mechanism includes a forget gate, an input gate, a cell state update, and an output gate.

[0100] The forget gate ignores outdated voltage fluctuations and abnormal conditions that have been resolved. The calculation process is:

[0101] ;

[0102] in, represents the input vector at the current time step, represents the output state of the previous time step, represents the weight matrix corresponding to the forget gate, represents the bias term, Represents the sigmoid activation function, which maps the result to between 0 and 1, indicating the degree of retention and discarding. If it is close to 1, the information is retained; if If it is close to 0, the information is discarded;

[0103] The input gate introduces new voltage and current features. These new voltage and current features include sudden high current peaks and low voltage RMS values, which determine the new information added to the cell state. The calculation process is:

[0104] ;

[0105] ;

[0106] in, Represents the input factor, using the sigmoid activation function to determine whether each element is added to the cell state. Use the Tanh activation function to generate new candidate cell states ranging from -1 to 1, and Determine the selection mechanism of the input gate, and Control the generation of candidate cell states;

[0107] Combining the results of the forget gate and the input gate, the cell state at the previous moment is bitwise multiplied by the output of the forget gate, and then added to the candidate state generated by the input gate to update the cell state that stores long-term information. The calculation process is:

[0108] ;

[0109] The output gate determines the output state of the current time step , as the final health score output, the calculation process is:

[0110] ;

[0111] ;

[0112] in, Represents the output factor, and uses the Sigmoid activation function to determine whether each element is output. Make the output value fall between -1 and 1 to reflect the change of the current cell state;

[0113] The output layer receives the output state from the hidden layer .

[0114] 105, using LSTM neural network model to obtain grid health score and predict health results;

[0115] In this embodiment, the output layer of the LSTM neural network model is used to convert the output state into a linear activation function through the internal fully connected layer. Maps to a scalar value , It represents the grid health score at the current time point, and the expression is:

[0116] ;

[0117] in, represents the weight matrix of the output layer, represents the bias term;

[0118] A health score threshold is set. When the grid health score is lower than the health score threshold, the early warning mechanism is triggered and the predicted health result is obtained.

[0119] 106, based on the predicted health results provided by the LSTM model, provides grid dispatch decision support;

[0120] In this embodiment, a continuous hidden state sequence is used as the input of the LSTM model, and the hidden state sequence is mapped to the grid health score of the future time step through the fully connected layer within the output layer of the LSTM model;

[0121] To predict the grid health three time steps into the future, the output layer outputs a three-dimensional vector, which is calculated as follows;

[0122] ;

[0123] in, Represents the three-dimensional vector output by the output layer, represents a continuous hidden state sequence;

[0124] The grid health score provided by the output layer assists in trend analysis and identifies deterioration trends. If the health of the target transmission line node is predicted to deteriorate, power distribution is optimized in advance to reduce the load pressure on the target transmission line node.

[0125] 107. Based on the grid dispatch decision support, the grid health score and predicted health results are used as the optimization objectives of the genetic algorithm to obtain the optimized grid operation plan.

[0126] In this embodiment, the grid health score and predicted health results output by the LSTM model are used as optimization targets of the genetic algorithm to generate initial individuals containing different operating parameter combinations to form an initial population. The different operating parameter combinations include the output power setpoints of different power plants and the load distribution ratios of each transmission line.

[0127] Based on the grid health score and predicted health results provided by the LSTM model, a fitness function is defined to evaluate the performance of each individual, taking into account the grid health status, power supply and demand balance, and cost-effectiveness factors. The calculation formula is:

[0128] ;

[0129] Where p represents an individual, represents the grid health score under an individual, Indicates the degree of balance between electricity supply and demand under individual conditions, represents the individual cost, 、 and represents the weight coefficient;

[0130] Individuals with higher fitness are selected from the current population through a roulette wheel to enter the next generation. Single-point and multi-point crossovers are performed on the selected individuals. Information is exchanged between different operation strategies to explore potential optimal solutions, generating new offspring individuals. The operational parameter values ​​of the newly generated offspring individuals are changed to introduce new genetic mutations. The new individuals after selection, crossover, and mutation are added to the new generation population to replace the old individuals with lower fitness. The selection, crossover, and mutation process is repeated until fitness no longer significantly improves. The individual with the highest fitness is selected as the optimal solution, and an optimized grid operation plan is proposed based on this optimal solution. The optimized operation suggestions are then displayed on the grid dispatch diagram, including adjusting the output power of power plants and changing the load distribution ratio of transmission lines.

[0131] In summary, the beneficial effects of the smart grid supervision method based on the power grid dispatch diagram provided by the present invention are:

[0132] Using voltage and current sensors, the power parameters of substations, power plants, and transmission line nodes are monitored in real time to obtain voltage and current data, from which a grid dispatch diagram is derived. Based on the grid dispatch diagram, an isolation forest algorithm is used to construct a grid anomaly detection model. This model is used to detect anomalies and provide early warnings, identifying and responding to early warning events. Based on the grid dispatch diagram, a long short-term memory (LSTM) neural network model is constructed. The LSTM neural network model is used to obtain a grid health score and predicted health results. Based on the predicted health results provided by the LSTM model, grid dispatch decision support is provided. Based on this grid dispatch decision support, the grid health score and predicted health results are used as optimization targets for the genetic algorithm to obtain an optimized grid operation plan.

[0133] Power parameter monitoring enables high-precision, real-time monitoring of voltage and current data at substations, power plants, and transmission line nodes, ensuring the accuracy and real-time nature of grid operating parameters and providing a reliable data basis for subsequent abnormal warnings and dispatching decisions.

[0134] The power grid anomaly warning analysis uses the advanced isolation forest algorithm and LSTM neural network model, which can not only effectively identify potential problems in the power grid and issue early warnings, but also assess the health of the power grid and predict future trends, thereby providing scientific and accurate scheduling decision support and significantly improving the stability and security of the power grid.

[0135] The above embodiments describe the specific process of the smart grid supervision method based on the power grid dispatching diagram. The following embodiments illustrate the smart grid supervision system based on the power grid dispatching diagram. Figure 2 As shown, an embodiment of the present invention provides a smart grid monitoring system based on a power grid dispatching diagram, comprising:

[0136] The power parameter monitoring module 201 is used to monitor the power parameters of substations, power plants, and transmission line nodes in real time through voltage sensors and current sensors to obtain voltage data and current data; and to obtain a power grid dispatch diagram based on the voltage data and current data;

[0137] The power grid anomaly warning analysis module 202 is used to build a power grid anomaly detection model based on the power grid dispatch diagram using the isolation forest algorithm; perform anomaly detection and warning based on the power grid anomaly detection model, identify and respond to warning events; build a long short-term memory (LSTM) neural network model based on the power grid dispatch diagram; use the LSTM neural network model to obtain a power grid health score and predicted health results; provide power grid dispatch decision support based on the predicted health results provided by the LSTM model; and, based on the power grid dispatch decision support, use the power grid health score and predicted health results as optimization targets of the genetic algorithm to obtain an optimized power grid operation plan.

[0138] The system also includes:

[0139] Network communication center 203 and integrated management module 204;

[0140] The network communication center 203 is connected to the power parameter monitoring module 201, the power grid abnormality early warning analysis module 202 and the comprehensive management module 204;

[0141] Comprehensive management module 204, used to integrate the functions of the power parameter monitoring module and the power grid anomaly warning analysis module;

[0142] The integrated management platform realizes the collaborative work of various modules, provides a human-computer interaction interface, opens an API interface to connect with external systems, and develops an integrated management platform to integrate sensor data collection, isolation forest algorithm anomaly detection, LSTM model health status assessment and prediction, and genetic algorithm optimization suggestion generation functions. The integrated management platform integrates and pre-processes high-precision voltage and current data from substations, power plants and transmission line nodes, performs anomaly detection through the isolation forest model, triggers warnings and records events for abnormal data points, and uses the LSTM model to evaluate the current health status and predict future trends for normal data points, optimizes power distribution in advance, and based on the LSTM prediction results, the genetic algorithm proposes suggestions for optimizing grid operations and selects the optimal solution as the final solution. The platform provides real-time data collection, model loading and updating, task scheduling, visualization and decision support functions, achieving full process coverage from data processing to early warning response and optimization suggestion implementation tracking, so as to ensure the safe and stable operation of the power grid. The integrated management platform realizes interaction with users and seamless docking with external systems through the human-computer interaction interface and open API interface. The human-computer interaction interface displays real-time voltage and current data, health status score, early warning information and optimization operation suggestions, supports users to monitor, make decisions and track the implementation of solutions, integrates the power grid dispatch map, displays the power grid topology, real-time operation status and abnormal areas, and the API interface allows external systems to access data streams, call isolation forest and LSTM models for anomaly detection and health assessment, obtain optimization suggestions and dispatch map information, and receive early warning information.

[0143] First, the power parameter monitoring module's high-precision voltage and current sensors monitor power parameters at substations, power plants, and transmission line nodes in real time, automatically collecting voltage and current data for real-time transmission to the network communication center. Second, the grid anomaly warning and analysis module receives power parameter data from the network communication center and uses the isolation forest algorithm to detect anomalies and issue warnings, identifying potential issues and enabling timely responses. Furthermore, this module constructs an LSTM neural network model to assess the grid's health and predict future trends, providing decision support for grid dispatch. Based on the status assessment and prediction results, a genetic algorithm is used to provide recommendations for optimizing grid operations. Finally, the integrated management module, through a developed integrated management platform, enables collaboration among various modules to ensure the smooth operation of the entire monitoring system. Users can conveniently view real-time grid data, warning information, and dispatch decision recommendations through a human-computer interface. The module also provides an API for easy integration and data exchange with external systems. Through these steps, users can efficiently utilize this smart grid monitoring system to achieve comprehensive grid monitoring and optimized dispatch.

[0144] Although the present invention has been illustrated and described with reference to certain preferred embodiments thereof, it should be understood by those skilled in the art that the above description is provided as a further detailed description of the present invention in conjunction with specific embodiments thereof, and that the specific implementation of the present invention is not limited to these descriptions. Those skilled in the art may make various changes in form and details, including simple deductions or substitutions, without departing from the spirit and scope of the present invention.

Claims

1. A smart grid supervision method based on a grid dispatching diagram, characterized in that: include: Using voltage sensors and current sensors, real-time monitoring of power parameters of substations, power plants, and transmission line nodes is performed to obtain voltage and current data, and a grid dispatch diagram is obtained based on the voltage and current data; Based on the power grid dispatch graph, an isolation forest algorithm is used to construct a power grid anomaly detection model; Perform anomaly detection and early warning based on the power grid anomaly detection model, and identify and respond to early warning events; Based on the power grid dispatch diagram, a long short-term memory (LSTM) neural network model is constructed; Using the LSTM neural network model to obtain a grid health score and predict health results; Providing grid dispatch decision support based on the predicted health results provided by the LSTM model; On the basis of the power grid dispatch decision support, the power grid health score and the predicted health result are used as optimization targets of a genetic algorithm to obtain an optimized power grid operation plan.

2. The smart grid supervision method based on the power grid dispatching diagram according to claim 1, characterized in that: The voltage and current sensors are used to monitor the power parameters of substations, power plants, and transmission line nodes in real time to obtain voltage and current data, including: Sense power parameters through voltage sensors and current sensors installed at substations, power plants and transmission line nodes; Converting the power parameter into an electrical signal, amplifying the electrical signal through an amplifying circuit, removing interference components from the electrical signal through filtering, and converting the amplified and interference-free electrical signal into a digital signal through an analog-to-digital converter; Preprocessing the digital signal to calculate the maximum value, minimum value, average value, effective value, variance and standard deviation as characteristic data, and obtaining voltage data and current data based on the characteristic data; A power grid dispatching diagram is constructed based on the voltage data and the current data, and the power grid dispatching diagram visually displays the power grid topology and real-time operating status.

3. The smart grid supervision method based on the power grid dispatching diagram according to claim 1, characterized in that: The method of constructing a power grid anomaly detection model based on the power grid dispatch map using an isolation forest algorithm includes: Using the voltage data and the current data in the power grid dispatching diagram as model inputs of a power grid anomaly detection model; Extract subsamples from the model input and construct an isolation tree for each subsample; the process of constructing the isolation tree is to select the voltage effective value and the current peak value as input features, select a split value between the minimum value and the maximum value of the voltage effective value and the current peak value respectively, divide the data of the current node into two left and right child nodes according to the split value, put the data points whose voltage effective value and the current peak value are less than the split value into the left child node, and put the data points whose voltage effective value and the current peak value are greater than or equal to the split value into the right child node, if the number of data points in the child node is less than a predetermined threshold and all available features have been used, then stop splitting, if the number of data points in the child node exceeds the predetermined threshold and there are available features that are not used, then continue splitting; After the isolation tree is constructed, the path length and average path length of each voltage data point and current data point in the isolation tree are calculated, and the average path length is used as an indicator to measure whether the data point is abnormal; thus constructing a power grid anomaly detection model; The path length h(x) is the number of edges from the root node to the leaf node of the data point x; The average path length ; T represents the total number of isolated trees, is the path length of the data point x in the i-th tree.

4. The smart grid supervision method based on the power grid dispatching diagram according to claim 3, characterized in that: The performing of anomaly detection and early warning based on the power grid anomaly detection model, and identifying and responding to early warning events, includes: Converting the average path length into an anomaly score based on the power grid anomaly detection model , The expression is: ； Among them, the represents the anomaly score of the data point x, and c is the average path length of the voltage data and the current data minus two standard deviations, representing a preset threshold; Input the newly collected new voltage data and new current data into the isolation forest of the power grid anomaly detection model to evaluate their isolation degree; and compare them with the c. If there is a first data point with an average path length not less than the c, the first data point is a normal point; if there is a second data point with an average path length less than the c, the second data point is an anomaly point; The abnormal area corresponding to the abnormal point is marked in the power grid dispatching diagram, and early warning information including the time, location and impact range of the abnormality is generated. The early warning information is sent to relevant management personnel, and the early warning event and response results are recorded.

5. The smart grid supervision method based on the power grid dispatching diagram according to claim 1, characterized in that: The LSTM neural network model is constructed based on the power grid dispatch diagram, including: The pre-configured LSTM neural network model has an input layer, a hidden layer, and an output layer; The input layer receives the voltage data and the current data of the power grid dispatch diagram, corresponds the time step including the four features of voltage effective value, current peak value, variance and standard deviation to the voltage and current feature vectors at each time point, and represents the input vector of each time step as , represents the effective value of the voltage at the t-th time step, Indicates the peak current, the represents the variance, represents the standard deviation; The hidden layer uses a gating mechanism in the internal long short-term memory unit state to capture its long-term dependencies over time, and the gating mechanism includes a forget gate, an input gate, a cell state update, and an output gate; The forget gate ignores outdated voltage fluctuations and abnormal conditions that have been resolved. The calculation process is: ； Among them, the Represents the input vector of the current time step, Represents the output state of the previous time step, Represents the weight matrix corresponding to the forget gate, represents the bias term, Represents the sigmoid activation function, which maps the result to between 0 and 1, indicating the degree of retention and discarding. If it is close to 1, the information is retained; if If it is close to 0, the information is discarded; The input gate introduces new voltage characteristics and new current characteristics. The new voltage characteristics and the new current characteristics include a sudden high current peak and a low voltage effective value, which determine the new information added to the cell state. The calculation process is: ； ； Among them, the Represents the input factor, using the sigmoid activation function to determine whether each element is added to the cell state. The Tanh activation function is used to generate new candidate cell states ranging from -1 to 1. and stated Determine the selection mechanism of the input gate, and stated Control the generation of candidate cell states; Combining the results of the forget gate and the input gate, the cell state at the previous moment is bitwise multiplied by the output of the forget gate, and then added to the candidate state generated by the input gate to update the cell state storing long-term information. The calculation process is: ； The output gate determines the output state of the current time step , as the final health score output, the calculation process is: ； ； Among them, the Represents the output factor, using the Sigmoid activation function to determine whether each element is output. Make the output value fall between -1 and 1 to reflect the change of the current cell state; The output layer receives the output state from the hidden layer .

6. The smart grid supervision method based on the power grid dispatching diagram according to claim 5, characterized in that: The method of using the LSTM neural network model to obtain a grid health score and predict a health result includes: The output layer of the LSTM neural network model is used to convert the output state into Maps to a scalar value , It represents the grid health score at the current time point, and the expression is: ； Among them, the represents the weight matrix of the output layer, represents the bias term; A health score threshold is set. When the grid health score is lower than the health score threshold, an early warning mechanism is triggered to obtain a predicted health result.

7. The smart grid supervision method based on the power grid dispatching diagram according to claim 6, characterized in that: The predicted health results provided by the LSTM model provide support for power grid dispatch decision-making, including: Using a continuous hidden state sequence as an input to the LSTM model, and mapping the hidden state sequence to a grid health score at a future time step through a fully connected layer within an output layer of the LSTM model; Based on the grid health score provided by the output layer, trend analysis is assisted to identify deterioration trends. If the health of the target transmission line node is predicted to deteriorate, power distribution is optimized in advance to reduce the load pressure of the target transmission line node.

8. The smart grid supervision method based on the power grid dispatching diagram according to claim 1, characterized in that: The method of using the grid health score and the predicted health result as optimization targets of a genetic algorithm based on the grid dispatch decision support to obtain an optimized grid operation plan includes: The grid health score and the predicted health result output by the LSTM model are used as optimization targets of a genetic algorithm to generate initial individuals comprising different operating parameter combinations to form an initial population. The different operating parameter combinations include output power setpoints of different power plants and load distribution ratios of each transmission line. Based on the grid health score and the predicted health result provided by the LSTM model, a fitness function is defined to evaluate the performance of each individual by integrating grid health status, power supply and demand balance, and cost-effectiveness factors. The calculation formula is: ； Wherein, the p represents an individual, the represents the grid health score under the individual, Indicates the balance degree of power supply and demand under the individual. represents the cost of the individual, 、 and stated represents the weight coefficient; Individuals with higher fitness are selected from the current population through roulette to enter the next generation. Single-point crossover and multi-point crossover are performed on the selected individuals. Information is exchanged between different operation strategies to explore potential optimization solutions, generate new offspring individuals, change the operation parameter values ​​of the newly generated offspring individuals to mutate, introduce new gene mutations, and add the new individuals after selection, crossover and mutation to the new generation population to replace the old individuals with lower fitness. The selection, crossover and mutation process is repeated until the fitness no longer increases significantly. The individual with the highest fitness is selected as the optimal solution, and an optimized power grid operation plan is proposed based on the optimal solution.

9. A smart grid supervision system based on a power grid dispatching diagram, used to implement the smart grid supervision method based on a power grid dispatching diagram according to any one of claims 1 to 8, characterized in that: include: The power parameter monitoring module is used to monitor the power parameters of substations, power plants and transmission line nodes in real time through voltage sensors and current sensors to obtain voltage data and current data; Obtaining a power grid dispatching diagram according to the voltage data and the current data; A power grid anomaly warning analysis module is used to construct a power grid anomaly detection model based on the power grid dispatch map using the isolation forest algorithm; perform anomaly detection and warning based on the power grid anomaly detection model, and identify and respond to warning events; Based on the power grid dispatch map, a long short-term memory (LSTM) neural network model is constructed; using the LSTM neural network model, a power grid health score and a predicted health result are obtained; and based on the predicted health result provided by the LSTM model, power grid dispatch decision support is provided; On the basis of the power grid dispatch decision support, the power grid health score and the predicted health result are used as optimization targets of a genetic algorithm to obtain an optimized power grid operation plan.

10. The smart grid monitoring system based on the grid dispatching diagram according to claim 9, characterized in that: The system further comprises: Network communication center and integrated management module; The network communication center is connected to the power parameter monitoring module, the power grid abnormality early warning analysis module and the comprehensive management module; The comprehensive management module is used to integrate the functions of the power parameter monitoring module and the power grid anomaly warning analysis module.

Images