Emergency command and dispatch system and method based on power business

Abstract

The invention relates to the technical field of electric power system maintenance, and discloses an emergency commanding and dispatching system and method based on electric power business, and the system constructs an electric power business emergency commanding and dispatching platform comprising a real-time monitoring module, a data analysis module, a decision support module and a dispatching and commanding module. The real-time monitoring module is responsible for collecting electric power system data, and the data analysis module identifies abnormity and potential risks. And the decision support module is used for defining an emergency state space, an action space and a response strategy and combining an emergency strategy model constructed by a reinforcement learning algorithm to realize dynamic optimization and updating of a decision algorithm and a rule so as to ensure that an emergency processing scheme is rapidly and accurately formulated. And the scheduling command module is responsible for coordinating resource allocation and scheduling according to a decision result, and ensuring rapid recovery and stable power supply of the power system. According to the method, reinforcement learning is introduced, so that the intelligence and the adaptability of emergency response are improved, and the efficiency and the accuracy of electric power emergency command and dispatch are effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power system maintenance, and in particular to an emergency command and dispatch system and method based on power business. Background Art

[0002] In the power business, it is crucial to ensure the stable operation of the power system and respond to sudden failures in a timely manner. At present, the power emergency command and dispatch system often relies on manual experience and preset fixed rules to make decisions. When faced with complex and changeable power system failures, it may not be possible to quickly and accurately formulate the best emergency response plan.

[0003] Existing power emergency command and dispatch systems usually include basic functional modules such as real-time monitoring, data analysis, decision support and dispatch command. The real-time monitoring module is responsible for collecting real-time operation data, fault alarm information and related weather condition data of the power system to provide a basis for subsequent decision-making. The data analysis module processes and analyzes these real-time data to identify abnormal conditions and potential risks in the power system. However, when making decisions, these systems often rely on preset decision-making algorithms and rules. These algorithms and rules are usually based on past experience and knowledge and may not be able to adapt to the ever-changing operating environment and failure modes of the power system.

[0004] Especially when facing large-scale power system failures or power emergency response under extreme weather conditions, traditional decision-making methods may not be able to quickly assess the scope, severity and duration of the failure, and it is difficult to select the best combination of multiple possible emergency actions. This may lead to delayed or improper emergency response, which in turn has an adverse impact on the restoration of the power system and the power supply to users. Summary of the invention

[0005] The purpose of the present invention is to provide an emergency command and dispatch system and method based on electric power business to solve the problems raised in the above background technology.

[0006] To achieve the above-mentioned purpose, the present invention provides the following technical solutions: an emergency command and dispatch system based on electric power business, the system comprising:

[0007] Build an emergency command and dispatch platform for power business, which includes real-time monitoring module, data analysis module, decision support module and dispatch and command module;

[0008] The real-time monitoring module is used to collect and integrate real-time operation data, fault alarm information and weather condition data of the power system;

[0009] The data analysis module is used to process and analyze real-time data to identify abnormal conditions and potential risks in the power system;

[0010] The decision support module generates emergency response plans and suggestions based on the data analysis results and using preset decision algorithms and rules;

[0011] The dispatching and commanding module is used to receive the output of the decision support module, coordinate and command the allocation and dispatch of power emergency resources, and communicate with relevant departments and personnel;

[0012] Wherein, the decision support module further includes:

[0013] Define the emergency state space, including the type, scope, severity and duration of power system faults;

[0014] Define emergency action space, including starting backup power supply, adjusting grid operation mode, dispatching emergency repair teams, implementing load control and notifying users;

[0015] Define the emergency response strategy evaluation function and select the optimal emergency action combination according to different situations in the emergency state space;

[0016] A reinforcement learning algorithm is used to construct an emergency strategy model, which is used to continuously optimize and update decision-making algorithms and rules.

[0017] Preferably, the power system fault types include line faults, equipment faults, power supply faults and grid instability; the impact range is determined based on the location of the fault point and the topological structure of the power system; the severity is evaluated based on the impact of the fault on the power supply capacity, voltage stability and equipment safety of the power system; and the duration is predicted based on the difficulty of repairing the fault and the resources required.

[0018] Preferably, the real-time monitoring module collects the operating parameters of the power system, including voltage, current, power factor and frequency, and status information of the fault indicator in real time through sensors, monitoring devices and remote terminal units (RTUs) of the power system;

[0019] The weather condition data includes wind speed, wind direction, temperature, humidity, rainfall and lightning activity, which are obtained through meteorological stations and are used to predict and assess the impact of weather on the operation of the power system.

[0020] Preferably, the emergency response strategy is composed of multiple sub-reward functions and penalty functions, specifically including:

[0021] Fault recovery speed reward function, whose calculation formula is: R_recovery = k1*(fault recovery time threshold - actual fault recovery time) / fault recovery time threshold; where k1 is the fault recovery speed reward coefficient, which is adjusted according to the actual operation of the power system and the emergency response target; the fault recovery time threshold is the expected value of the fault recovery time set according to historical data and experience;

[0022] Resource utilization efficiency reward function, whose calculation formula is: R_resource = k2*(actual number of emergency resources used / expected number of emergency resources used); k2 is the resource utilization efficiency reward coefficient, which is adjusted according to the resource reserves of the power system and the emergency response needs; the expected number of emergency resources used is estimated according to the fault type, severity and emergency response plan;

[0023] The coordination command efficiency reward function is calculated as follows: R_coordination = k3*(actual coordination command time / expected coordination command time); k3 is the coordination command efficiency reward coefficient, which is adjusted according to the organizational structure, communication capability and emergency response process of the power system; the expected coordination command time is the expected value of the coordination command time set according to the emergency response plan and historical data;

[0024] Information accuracy penalty function, whose calculation formula is: P_information = k4*(number of erroneous information / total number of information); where k4 is the information accuracy penalty coefficient, which is adjusted according to the information source, transmission method and verification mechanism of the power system; the number of erroneous information refers to the number of erroneous or inaccurate information transmitted during the emergency response process, and the total number of information refers to the number of all information transmitted during the emergency response process;

[0025] Response delay penalty function, whose calculation formula is: P_delay = k5*(actual response time-response time threshold); where k5 is the response delay penalty coefficient, which is adjusted according to the emergency response requirements and actual conditions of the power system; the response time threshold is the maximum value of the response time set according to the emergency response plan and historical data of the power system;

[0026] Through the above reward and penalty functions, the final reward function R is obtained: R = R_recovery+R_resource+R_coordination-P_information-P_delay; this reward function is used to evaluate the effectiveness of the emergency response strategy and guide the decision support module to select the optimal combination of emergency actions.

[0027] Preferably, the decision support module uses a deep Q network DQN algorithm in a reinforcement learning algorithm to construct an emergency strategy model.

[0028] Preferably, the step of constructing the emergency strategy model by the decision support module includes:

[0029] S1: Define the state space S and action space A: According to the actual scenario and requirements of emergency response, determine the environmental state s that the model needs to perceive and the executable emergency action a to form the state space S and action space A;

[0030] S2: Initialize the parameters of the reinforcement learning algorithm: Initialize the parameters of the DQN algorithm, including the learning rate, discount factor, and exploration rate;

[0031] S3: Build and initialize the policy network: Build one or more neural networks as the policy network, with the input being the state s and the output being the selection probability or Q value of action a; initialize the network parameters θ;

[0032] S4: Initialize the experience replay buffer D: Set the buffer size N to store the state transition samples (s, a, r, s') generated by the model during training, where s is the current state, a is the action performed, r is the reward obtained, and s' is the new state after the action is performed;

[0033] S5: Perform model training: In the emergency response simulation environment, let the model select actions according to the current strategy and execute them, observe the results and calculate the rewards; store the generated state transition samples in the experience replay buffer D; randomly extract a batch of samples from the buffer D, and use the reinforcement learning algorithm to update the parameters θ of the policy network to optimize the strategy;

[0034] S6: Regular evaluation and strategy update: During the training process, regularly use the validation set or test set to evaluate the performance of the current strategy; based on the evaluation results, adjust the parameters of the reinforcement learning algorithm, including the learning rate and exploration rate, as well as the weight coefficient of the reward function;

[0035] S7: Repeat steps S5 and S6 until the preset training conditions are met, including reaching the maximum number of training rounds, the strategy performance is stable, or specific performance indicators are met.

[0036] Preferably, in step S5, the model parameters are updated using a gradient descent algorithm, and the calculation formula is:

[0037]

[0038] Among them, θ represents the model parameters, including all weights and biases that the model needs to learn; ω represents the learning rate, which is used to control the step size of parameter update; represents the gradient of the loss function J(θ) with respect to the parameter θ, is a vector pointing to the direction where the loss function grows fastest; := represents the assignment operation, that is, updating the value of the parameter θ.

[0039] Preferably, in step S6, the mean square error (MSE) algorithm is used to evaluate the model performance, and the specific implementation steps include:

[0040] Step 1: After each training iteration, use the validation set data to evaluate the model performance;

[0041] Step 2: Calculate the mean square error (MSE) between the predicted output of the LSTM model and the actual output of the validation set. The calculation formula for MSE is:

[0042]

[0043] Among them, N is the number of samples in the validation set, y i is the actual output value of the i-th sample in the validation set, is the predicted output value of the LSTM model for the i-th sample;

[0044] Step 3: Evaluate the performance of the model based on the calculated MSE value. The smaller the MSE value, the higher the prediction accuracy of the model and the better the performance.

[0045] Step 4: Based on the changing trend of the MSE value and the preset performance threshold, decide whether to continue training the model, adjust the model parameters, or stop training early.

[0046] Preferably, the dispatching and commanding module adopts a distributed architecture and establishes communication connections with the control center, emergency repair team, users and other relevant departments of the power system.

[0047] Preferably, an emergency command and dispatch method based on power business, the method comprising:

[0048] A: Build an emergency command and dispatch platform for power business, which includes real-time monitoring module, data analysis module, decision support module and dispatch and command module;

[0049] B: Collect and integrate real-time operation data, fault alarm information and weather condition data of the power system through the real-time monitoring module;

[0050] C.: Use the data analysis module to process and analyze the collected real-time data to identify abnormal conditions and potential risks in the power system;

[0051] D: In the decision support module, the emergency state space is first defined, including the type of power system fault, impact range, severity and duration; then the emergency action space is defined, including emergency measures such as starting the backup power supply, adjusting the grid operation mode, dispatching the repair team, implementing load control and notifying users; then, based on the defined emergency state space and action space, the emergency response strategy is formulated;

[0052] E: The emergency strategy model is constructed in the decision support module using a reinforcement learning algorithm. The model optimizes and updates the decision-making algorithm and rules through continuous learning and iteration to select the optimal combination of emergency actions in an emergency state.

[0053] F: The dispatching and commanding module receives the emergency response plans and suggestions output by the decision support module, and coordinates and directs the allocation and dispatching of power emergency resources accordingly.

[0054] Compared with the prior art, the present invention has the following beneficial effects:

[0055] The system can collect and process the operation data, fault alarms and weather information of the power system in real time, and quickly identify abnormal situations through the data analysis module, providing immediate basis for emergency response. Combined with the reinforcement learning algorithm, the decision support module can quickly evaluate the fault status and automatically select the optimal combination of emergency actions, greatly shortening the emergency response time.

[0056] The existing decision-making methods that rely on manual experience and fixed rules are difficult to adapt to the dynamic changes of the power system. The present invention uses a reinforcement learning algorithm to continuously learn and optimize decision-making algorithms and rules, making the emergency response plan more in line with the actual fault situation, improving the effectiveness and pertinence of the plan, and helping to reduce the impact of faults on the operation of the power system and user power supply.

[0057] The defined emergency state space and action space cover various types of power system failures and various emergency response measures, providing a wide range of decision-making space for the system. The reinforcement learning algorithm can automatically adjust the emergency strategy based on historical data and real-time feedback, enabling the system to better adapt to power system failures of different scales and complexities, as well as emergency response needs under extreme weather conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 This is a working principle diagram of an emergency command and dispatch system based on power business described in the present invention;

[0059] Figure 2 A schematic diagram of the design of the emergency response strategy evaluation function;

[0060] Figure 3 Diagram of the training steps for the emergency strategy model. DETAILED DESCRIPTION

[0061] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0062] See also Figure 1-3 The present invention provides a technical solution: an emergency command and dispatch system based on power business, the system comprising:

[0063] Real-time monitoring module: The real-time monitoring module is responsible for collecting and integrating the real-time operation data, fault alarm information and weather condition data of the power system. The module is connected to the data interface of various monitoring points, sensors and meteorological departments of the power system to obtain the voltage, current, power and other operating parameters of the power system, as well as the time, location and type of fault alarm information in real time. At the same time, it collects relevant weather condition data, such as wind speed, rainfall, temperature, etc., to provide a basis for subsequent data analysis and decision-making.

[0064] Data analysis module: The data analysis module processes and analyzes the real-time data collected by the real-time monitoring module to identify abnormal conditions and potential risks in the power system. This module uses data processing algorithms and models to clean, integrate and mine real-time data, extract useful information, and identify abnormal changes, potential fault points and possible risk factors in the power system through comparative analysis, trend prediction and other means, providing accurate data support for the decision support module.

[0065] Decision support module: The decision support module is the core part of the system. It generates emergency response plans and suggestions based on the results of the data analysis module and uses preset decision algorithms and rules. The module first defines the emergency state space, including the type of power system fault, impact range, severity and duration, etc., to fully describe the emergency state. Then the emergency action space is defined, including emergency measures such as starting the backup power supply, adjusting the grid operation mode, dispatching the repair team, implementing load control and notifying users to deal with different emergency states. Then, according to the different situations in the emergency state space, an emergency response strategy is formulated and the optimal combination of emergency actions is selected. Finally, the reinforcement learning algorithm is used to construct the emergency strategy model, which can continuously learn and iterate, optimize and update the decision algorithm and rules according to the actual emergency response effect, and improve the accuracy and efficiency of emergency handling.

[0066] Dispatching and commanding module: The dispatching and commanding module is responsible for receiving the output of the decision support module, coordinating and commanding the allocation and dispatch of power emergency resources, and communicating with relevant departments and personnel. Based on the emergency handling plan and suggestions, this module formulates a specific dispatching plan, including the allocation of emergency resources, the dispatch of emergency repair teams, the implementation of load control, etc., and communicates information and instructions with relevant departments and personnel in real time through the communication system to ensure rapid and effective emergency response.

[0067] The present invention will be further described below in conjunction with Examples 1 to 3:

[0068] Embodiment 1:

[0069] In the power business emergency command and dispatch system, the power system fault types are subdivided into four types: line fault, equipment fault, power supply fault and grid instability. The definitions of these fault types are as follows:

[0070] Line failure: refers to a failure of power transmission lines (such as overhead lines, cables, etc.) caused by external force damage, aging, insulation damage, etc., such as broken wires, short circuits, etc.

[0071] Equipment failure: refers to the failure of electrical equipment such as transformers, switches, circuit breakers, mutual inductors, capacitors, etc. in the power system due to internal defects, external influences or improper operation.

[0072] Power failure: refers to the failure of a power plant's generator, excitation system, speed control system and other power equipment to work properly or shut down due to mechanical failure, electrical failure or control failure.

[0073] Grid instability: refers to the deviation of electrical quantities such as voltage and frequency from the normal range due to load changes, line failures, equipment failures or improper control of the power system, which may even lead to serious consequences such as system decoupling and collapse.

[0074] The real-time monitoring module collects the operating parameters of the power system and the status information of the fault indicators in real time through the sensors, monitoring devices and remote terminal units (RTU) of the power system. The specific implementation method is as follows:

[0075] Sensors and monitoring devices: Sensors and monitoring devices, such as voltage sensors, current sensors, power factor sensors, frequency sensors, etc., are installed on key nodes and equipment in the power system to monitor the operating parameters of the power system in real time.

[0076] Remote Terminal Unit (RTU): RTU is an important device in the power system. It can process and package the data collected by sensors and monitoring devices, and transmit them to the real-time monitoring module through the communication network. RTU also has a fault indication function, which can reflect the fault status of the equipment in real time.

[0077] Weather condition data: The real-time monitoring module also obtains weather condition data such as wind speed, wind direction, temperature, humidity, rainfall and lightning activity from the meteorological station to predict and evaluate the impact of weather on the operation of the power system. These data are transmitted to the real-time monitoring module in real time through the communication network and correlated with the operating parameters of the power system for analysis.

[0078] Embodiment 2:

[0079] The emergency response strategy evaluation function consists of multiple sub-reward functions and penalty functions, which are used to evaluate the effectiveness of the emergency response strategy and guide the decision support module to select the optimal combination of emergency actions. The following is the specific implementation of each function:

[0080] ① Fault recovery speed reward function:

[0081] The fault recovery speed reward function is used to evaluate the speed of fault recovery, and its calculation formula is: R_recovery = k1*(fault recovery time threshold - actual fault recovery time) / fault recovery time threshold; where k1 is the fault recovery speed reward coefficient, which is adjusted according to the actual operation of the power system and the emergency response goals. The fault recovery time threshold is the expected value of the fault recovery time set based on historical data and experience, which represents the expected fault recovery time under normal circumstances. The actual fault recovery time refers to the actual time required from the occurrence of the fault to the complete recovery of the fault. Through this reward function, the emergency response team can be encouraged to recover the fault as soon as possible and improve the recovery speed.

[0082] ② Resource utilization efficiency reward function:

[0083] The resource utilization efficiency reward function is used to evaluate the efficiency of emergency resource use, and its calculation formula is: R_resource = k2*(actual number of emergency resources used / expected number of emergency resources used); where k2 is the resource utilization efficiency reward coefficient, which is adjusted according to the resource reserves of the power system and the emergency response needs. The expected number of emergency resources used is estimated based on the fault type, severity and emergency response plan, indicating the number of resources expected to be used in the emergency response. The actual number of emergency resources used refers to the number of resources actually used during the actual emergency response process. Through this reward function, the emergency response team can be encouraged to save resources as much as possible while ensuring fault recovery.

[0084] ③ Collaborative command efficiency reward function:

[0085] The coordination and command efficiency reward function is used to evaluate the coordination efficiency between departments and personnel during the emergency response process. The calculation formula is: R_coordination = k3*(actual coordination and command time / expected coordination and command time); where k3 is the coordination and command efficiency reward coefficient, which is adjusted according to the organizational structure, communication capability and emergency response process of the power system. The expected coordination and command time is the expected value of the coordination and command time set according to the emergency response plan and historical data, indicating the expected coordination and command time under normal circumstances. The actual coordination and command time refers to the actual coordination and command time spent between departments and personnel during the actual emergency response process. Through this reward function, the emergency response team can be encouraged to improve the coordination efficiency and shorten the coordination and command time.

[0086] ④Information accuracy penalty function:

[0087] The information accuracy penalty function is used to evaluate the accuracy of information during the emergency response process. The calculation formula is: P_information = k4*(number of erroneous information / total number of information); where k4 is the information accuracy penalty coefficient, which is adjusted according to the information source, transmission method and verification mechanism of the power system. The number of erroneous information refers to the number of erroneous or inaccurate information transmitted during the emergency response process, and the total number of information refers to the total number of information transmitted during the emergency response process. Through this penalty function, the behavior of transmitting erroneous information can be punished to improve the accuracy of information.

[0088] ⑤Response delay penalty function:

[0089] The response delay penalty function is used to evaluate the timeliness of emergency response, and its calculation formula is: P_delay = k5*(actual response time-response time threshold); where k5 is the response delay penalty coefficient, which is adjusted according to the emergency response requirements and actual conditions of the power system. The response time threshold is the maximum value of the response time set according to the emergency response plan and historical data of the power system, indicating the maximum response time allowed under normal circumstances. The actual response time refers to the actual time required from the occurrence of the fault to the start of the emergency response team's response. Through this penalty function, the behavior of response delay can be punished to improve the timeliness of emergency response.

[0090] Final reward function:

[0091] Through the above reward and penalty functions, the final reward function R can be obtained, which is used to evaluate the effectiveness of the emergency response strategy and guide the decision support module to select the optimal combination of emergency actions. The calculation formula is:

[0092] R=R_recovery+R_resource+R_coordination-P_information-P_delay

[0093] This reward function comprehensively considers multiple aspects such as fault recovery speed, resource utilization efficiency, coordinated command efficiency, information accuracy and response delay. By adjusting the coefficients of each reward and penalty function, different emergency response strategies can be evaluated and compared according to the actual situation of the power system and the emergency response goals, so as to select the optimal emergency action combination.

[0094] Embodiment 3:

[0095] This embodiment describes in detail the steps and methods of building an emergency strategy model using the deep Q network (DQN) algorithm in the reinforcement learning algorithm. The following is the specific implementation process:

[0096] S1: Define the state space S and action space A

[0097] According to the actual scenarios and requirements of emergency response, determine the environmental state s that the model needs to perceive and the executable emergency action a. The state space S includes information such as the current operating state of the power system, fault type, fault location, and emergency measures taken. The action space A includes executable emergency actions, such as starting backup power, adjusting grid load, and dispatching maintenance teams. These states and actions form the basis for the model to make decisions.

[0098] S2: Initialize reinforcement learning algorithm parameters

[0099] Initialize the parameters of the DQN algorithm, including the learning rate α, discount factor γ, and exploration rate ε. The learning rate α determines the step size of the model when updating parameters, the discount factor γ is used to calculate the current value of future rewards, and the exploration rate ε is used to control the balance between exploration and exploitation of the model. The initial values ​​of these parameters are set based on experience or experiments and adjusted during the training process.

[0100] S3: Build and initialize the policy network

[0101] Construct one or more neural networks as policy networks, whose input is state s and output is the selection probability or Q value of action a. The architecture of the policy network can be designed according to the complexity and characteristics of the emergency response problem, and usually includes an input layer, a hidden layer, and an output layer. Initialize the network parameters θ, which will be updated through the back-propagation algorithm during the training process.

[0102] S4: Initialize the experience replay buffer D

[0103] Set the buffer size N to store the state transition samples (s, a, r, s') generated by the model during training. Among them, s is the current state, a is the action performed, r is the reward calculated according to the reward function, and s' is the new state after the action is performed. The experience replay buffer is used to break the time association between samples and improve the training efficiency of the model.

[0104] S5: Perform model training

[0105] S501: In the emergency response simulation environment, let the model select and execute actions according to the current strategy. The simulation environment generates a new state s' and reward r based on the executed action and the current state.

[0106] S502: Store the generated state transition samples (s, a, r, s') into the experience playback buffer D.

[0107] S503: Randomly extract a batch of samples from the buffer D, and use the DQN algorithm to update the parameters θ of the policy network. Specifically, for each sample, calculate the target Q value, and use the mean square error loss function to perform back propagation to update the network parameters.

[0108] S504: Repeat the above steps until a preset number of training rounds is reached or other stopping conditions are met.

[0109] S6: Regular evaluation and strategy update

[0110] S601: During the training process, the validation set or test set is used regularly to evaluate the performance of the current strategy. Evaluation indicators may include fault recovery time, resource utilization efficiency, collaborative command efficiency, etc.

[0111] S602: According to the evaluation results, adjust the parameters of the reinforcement learning algorithm, including the learning rate α, the exploration rate ε, and the weight coefficient of the reward function. These adjustments are intended to improve the training efficiency and policy performance of the model.

[0112] S603: If the evaluation result shows that the policy performance is stable or meets specific performance indicators, the training can be stopped and the final policy model can be saved.

[0113] S7: Repeat training and evaluation

[0114] Repeat steps S5 and S6 until the preset training conditions are met. These conditions may include reaching the maximum number of training rounds, the strategy performance is stable, or specific performance indicators are met. Through continuous training and evaluation, the model can gradually learn effective emergency strategies and play a guiding role in actual emergency response.

[0115] In step S5, the model parameters are updated using the gradient descent algorithm, and the calculation formula is:

[0116]

[0117] Among them, θ represents the model parameters, including all weights and biases that the model needs to learn; ω represents the learning rate, which is used to control the step size of parameter update; represents the gradient of the loss function J(θ) with respect to the parameter θ, is a vector pointing to the direction where the loss function grows fastest; := represents the assignment operation, that is, updating the value of the parameter θ.

[0118] In step S6, the mean square error (MSE) algorithm is used to evaluate the model performance. The specific implementation steps include:

[0119] Step 1: After each training iteration, use the validation set data to evaluate the model performance;

[0120] Step 2: Calculate the mean square error (MSE) between the predicted output of the LSTM model and the actual output of the validation set. The calculation formula for MSE is:

[0121]

[0122] Among them, N is the number of samples in the validation set, y i is the actual output value of the i-th sample in the validation set, is the predicted output value of the LSTM model for the i-th sample;

[0123] Step 3: Evaluate the performance of the model based on the calculated MSE value. The smaller the MSE value, the higher the prediction accuracy of the model and the better the performance.

[0124] Step 4: Based on the changing trend of the MSE value and the preset performance threshold, decide whether to continue training the model, adjust the model parameters, or stop training early.

[0125] Taking power system fault recovery as an example, the state space S can include the operating status information of the power system, such as voltage, current, power, as well as the fault type (such as short circuit, open circuit, etc.) and fault location (such as substation, transmission line, etc.). The action space A can include emergency actions such as starting the backup power supply, adjusting the grid load, and dispatching a maintenance team. Through the training of the DQN algorithm, the model can learn the strategy of selecting the optimal emergency action under different states to quickly restore the normal operation of the power system. During the training process, the performance of the strategy can be optimized by adjusting parameters such as the learning rate, discount factor, and exploration rate, as well as the weight coefficient of the reward function. Finally, the trained model can be used as a decision support module to provide intelligent decision support for the emergency response of the power system.

[0126] The dispatching and commanding module adopts a distributed architecture and establishes communication connections with the control center, emergency repair team, users and other relevant departments of the power system.

[0127] The present invention also includes an emergency command and dispatch method based on power business, the method comprising:

[0128] A: Build an emergency command and dispatch platform for power business, which includes real-time monitoring module, data analysis module, decision support module and dispatch and command module;

[0129] B: Collect and integrate real-time operation data, fault alarm information and weather condition data of the power system through the real-time monitoring module;

[0130] C.: Use the data analysis module to process and analyze the collected real-time data to identify abnormal conditions and potential risks in the power system;

[0131] D: In the decision support module, the emergency state space is first defined, including the type of power system fault, impact range, severity and duration; then the emergency action space is defined, including emergency measures such as starting the backup power supply, adjusting the grid operation mode, dispatching the repair team, implementing load control and notifying users; then, based on the defined emergency state space and action space, the emergency response strategy is formulated;

[0132] E: The emergency strategy model is constructed in the decision support module using a reinforcement learning algorithm. The model optimizes and updates the decision-making algorithm and rules through continuous learning and iteration to select the optimal combination of emergency actions in an emergency state.

[0133] F: The dispatching and commanding module receives the emergency response plans and suggestions output by the decision support module, and coordinates and directs the allocation and dispatching of power emergency resources accordingly.

[0134] The implementation of this method refers to the above embodiment and will not be described in detail in the specification.

[0135] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0136] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An emergency command and dispatch system based on power business, characterized in that: The system comprises: Build an emergency command and dispatch platform for power business, which includes real-time monitoring module, data analysis module, decision support module and dispatch and command module; The real-time monitoring module is used to collect and integrate real-time operation data, fault alarm information and weather condition data of the power system; The data analysis module is used to process and analyze real-time data to identify abnormal conditions and potential risks in the power system; The decision support module generates emergency response plans and suggestions based on the data analysis results and using preset decision algorithms and rules; The dispatching and commanding module is used to receive the output of the decision support module, coordinate and command the allocation and dispatch of power emergency resources, and communicate with relevant departments and personnel; Wherein, the decision support module further includes: Define the emergency state space, including the type, scope, severity and duration of power system faults; Define emergency action space, including starting backup power supply, adjusting grid operation mode, dispatching emergency repair teams, implementing load control and notifying users; Define the emergency response strategy evaluation function and select the optimal emergency action combination according to different situations in the emergency state space; A reinforcement learning algorithm is used to construct an emergency strategy model, which is used to continuously optimize and update decision-making algorithms and rules.

2. The emergency command and dispatch system based on power business according to claim 1 is characterized in that: The power system fault types include line faults, equipment faults, power supply faults and grid instability; the impact range is determined according to the location of the fault point and the topological structure of the power system; the severity is evaluated according to the impact of the fault on the power supply capacity, voltage stability and equipment safety of the power system; The duration is predicted based on the difficulty of repairing the fault and the resources required.

3. The emergency command and dispatch system based on power business according to claim 1 is characterized in that: The real-time monitoring module collects the operating parameters of the power system in real time, including voltage, current, power factor and frequency, as well as the status information of the fault indicator, through sensors, monitoring devices and remote terminal units (RTUs) of the power system; The weather condition data includes wind speed, wind direction, temperature, humidity, rainfall and lightning activity, which are obtained through meteorological stations and are used to predict and assess the impact of weather on the operation of the power system.

4. The emergency command and dispatch system based on power business according to claim 1 is characterized in that: The emergency response strategy evaluation function is composed of multiple sub-reward functions and penalty functions, specifically including: Fault recovery speed reward function, whose calculation formula is: R_recovery = k1*(fault recovery time threshold - actual fault recovery time) / fault recovery time threshold; where k1 is the fault recovery speed reward coefficient, which is adjusted according to the actual operation of the power system and the emergency response target; the fault recovery time threshold is the expected value of the fault recovery time set according to historical data and experience; Resource utilization efficiency reward function, whose calculation formula is: R_resource = k2*(actual number of emergency resources used / expected number of emergency resources used); k2 is the resource utilization efficiency reward coefficient, which is adjusted according to the resource reserves of the power system and the emergency response needs; the expected number of emergency resources used is estimated according to the fault type, severity and emergency response plan; The coordination command efficiency reward function is calculated as follows: R_coordination = k3*(actual coordination command time / expected coordination command time); k3 is the coordination command efficiency reward coefficient, which is adjusted according to the organizational structure, communication capability and emergency response process of the power system; the expected coordination command time is the expected value of the coordination command time set according to the emergency response plan and historical data; Information accuracy penalty function, whose calculation formula is: P_information = k4*(number of erroneous information / total number of information); where k4 is the information accuracy penalty coefficient, which is adjusted according to the information source, transmission method and verification mechanism of the power system; the number of erroneous information refers to the number of erroneous or inaccurate information transmitted during the emergency response process, and the total number of information refers to the number of all information transmitted during the emergency response process; Response delay penalty function, whose calculation formula is: P_delay = k5*(actual response time-response time threshold); where k5 is the response delay penalty coefficient, which is adjusted according to the emergency response requirements and actual conditions of the power system; the response time threshold is the maximum value of the response time set according to the emergency response plan and historical data of the power system; Through the above reward and penalty functions, the final reward function R is obtained: R = R_recovery+R_resource+R_coordination-P_information-P_delay; this reward function is used to evaluate the effectiveness of the emergency response strategy and guide the decision support module to select the optimal combination of emergency actions.

5. The emergency command and dispatch system based on power business according to claim 1 is characterized in that: The decision support module uses the deep Q network DQN algorithm in the reinforcement learning algorithm to build an emergency strategy model.

6. The emergency command and dispatch system based on power business according to claim 5 is characterized in that: The steps of constructing the emergency strategy model by the decision support module include: S1: Define the state space S and action space A: According to the actual scenario and requirements of emergency response, determine the environmental state s that the model needs to perceive and the executable emergency action a to form the state space S and action space A; S2: Initialize the parameters of the reinforcement learning algorithm: Initialize the parameters of the DQN algorithm, including the learning rate, discount factor, and exploration rate; S3: Build and initialize the policy network: Build one or more neural networks as the policy network, with the input being the state s and the output being the selection probability or Q value of action a; initialize the network parameters θ; S4: Initialize the experience replay buffer D: Set the buffer size N to store the state transition samples (s, a, r, s') generated by the model during training, where s is the current state, a is the action performed, r is the reward obtained, and s' is the new state after the action is performed; S5: Perform model training: In the emergency response simulation environment, let the model select actions according to the current strategy and execute them, observe the results and calculate the rewards; store the generated state transition samples in the experience replay buffer D; randomly extract a batch of samples from the buffer D, and use the reinforcement learning algorithm to update the parameters θ of the policy network to optimize the strategy; S6: Regular evaluation and strategy update: During the training process, regularly use the validation set or test set to evaluate the performance of the current strategy; based on the evaluation results, adjust the parameters of the reinforcement learning algorithm, including the learning rate and exploration rate, as well as the weight coefficient of the reward function; S7: Repeat step S5 and step S6 until the preset training conditions are met, including reaching the maximum number of training rounds, the strategy performance is stable, or the preset performance indicators are met.

7. The emergency command and dispatch system based on power business according to claim 6, characterized in that: In step S5, the model parameters are updated using the gradient descent algorithm, and the calculation formula is: Among them, θ represents the model parameters, including all weights and biases that the model needs to learn; ω represents the learning rate, which is used to control the step size of parameter update; represents the gradient of the loss function J(θ) with respect to the parameter θ, is a vector pointing to the direction where the loss function grows fastest; := represents the assignment operation, that is, updating the value of the parameter θ.

8. The emergency command and dispatch system based on power business according to claim 6 is characterized in that: In step S6, the mean square error (MSE) algorithm is used to evaluate the model performance. The specific implementation steps include: Step 1: After each training iteration, use the validation set data to evaluate the model performance; Step 2: Calculate the mean square error (MSE) between the predicted output of the LSTM model and the actual output of the validation set. The calculation formula for MSE is: Among them, N is the number of samples in the validation set, y i is the actual output value of the i-th sample in the validation set, is the predicted output value of the LSTM model for the i-th sample; Step 3: Evaluate the performance of the model based on the calculated MSE value. The smaller the MSE value, the higher the prediction accuracy of the model and the better the performance. Step 4: Based on the changing trend of the MSE value and the preset performance threshold, decide whether to continue training the model, adjust the model parameters, or stop training early.

9. The emergency command and dispatch system based on power business according to claim 1 is characterized in that: The dispatching and commanding module adopts a distributed architecture and establishes communication connections with the control center, emergency repair team, users and other relevant departments of the power system.

10. An emergency command and dispatch method based on power business, characterized in that: The method comprises: A: Build an emergency command and dispatch platform for power business, which includes real-time monitoring module, data analysis module, decision support module and dispatch and command module; B: Collect and integrate real-time operation data, fault alarm information and weather condition data of the power system through the real-time monitoring module; C.: Use the data analysis module to process and analyze the collected real-time data to identify abnormal conditions and potential risks in the power system; D: In the decision support module, the emergency state space is first defined, including the type of power system fault, impact range, severity and duration; then the emergency action space is defined, including emergency measures such as starting the backup power supply, adjusting the grid operation mode, dispatching the repair team, implementing load control and notifying users; then, based on the defined emergency state space and action space, the emergency response strategy is formulated; E: The emergency strategy model is constructed in the decision support module using a reinforcement learning algorithm. The model optimizes and updates the decision-making algorithm and rules through continuous learning and iteration to select the optimal combination of emergency actions in an emergency state. F: The dispatching and commanding module receives the emergency response plans and suggestions output by the decision support module, and coordinates and directs the allocation and dispatching of power emergency resources accordingly.

Images