Emergency electric power supply vehicle dispatching system and method for recovering urban power supply

Abstract

The invention discloses an emergency electric power supply vehicle dispatching system and method for recovering urban power supply in the technical field of electric power systems. The method comprises the following steps: acquiring global state information in a current state; based on the global state information in the current state, through a pre-established power supply vehicle scheduling neural network model containing a DQN network and a shared attention module, obtaining a scheduling path and a power supply strategy of an individual power supply vehicle; after scheduling power supply is performed on the power supply vehicle based on the scheduling path and the power supply strategy, evaluating a reward value of the power supply vehicle; after training and updating the power supply vehicle scheduling neural network model based on the reward value, obtaining a scheduling path and a power supply strategy with the highest combined total value; and carrying out scheduling power supply on all power supply vehicles based on the scheduling path with the highest total value and the power supply strategy. According to the method, the dispatching paths and strategies of the power supply vehicles are optimized through the DQN and the power supply vehicle dispatching neural network model sharing the attention module, and the recovery speed and efficiency of the urban lifeline after the disaster are effectively improved.

Description

Technical Field

[0001] The invention relates to an emergency power supply vehicle dispatching system and method for restoring urban power supply, belonging to the technical field of power systems. Background Art

[0002] After a large-scale power outage caused by a major disaster in a city, the power supply vehicle becomes the core force for emergency power restoration in the entire city. Therefore, in order to improve the emergency response capability of the power system, it is necessary to conduct in-depth research on the power emergency material dispatch process and establish a scientific and practical resource dispatch mathematical model to promote the automation process of the power emergency mechanism.

[0003] The current emergency response system is not mature enough when facing a major power outage and needs to be improved. In power grid accidents caused by natural disasters, the emergency supplies stored in warehouses in the fault area are often not sufficient to meet the emergency needs of a major power outage. The existing emergency material dispatching technology mainly relies on experience and dispatches materials from nearby warehouses, lacking scientific and reasonable transportation route planning and optimized material allocation. Summary of the invention

[0004] The purpose of the present invention is to overcome the deficiencies in the prior art and provide an emergency power supply vehicle dispatching system and method for restoring urban power supply. DQN The power supply vehicle scheduling neural network model based on the network and shared attention module optimizes the scheduling paths and strategies of multiple power supply vehicles, improves the city's emergency power material scheduling capabilities, reduces power outage time and economic losses, and ensures the stability of urban operations and the safety of residents' lives.

[0005] To achieve the above object, the present invention is implemented by adopting the following technical solutions: In a first aspect, the present invention provides a method for dispatching an emergency power supply vehicle for restoring urban power supply, comprising: Get the global status information under the current state; Based on the global state information in the current state, through the pre-established DQN The power supply vehicle scheduling neural network model based on the network and shared attention module obtains the scheduling path and power supply strategy of individual power supply vehicles; After scheduling the power supply vehicle based on the scheduling path and power supply strategy, the reward value of the power supply vehicle is evaluated; After training and updating the power supply vehicle scheduling neural network model based on the reward value, the scheduling path and power supply strategy with the highest joint total value are obtained; All power supply vehicles are dispatched and supplied with power based on the dispatch path and power supply strategy with the highest combined total value.

[0006] Furthermore, the global status information includes user priority, power demand, real-time weather information, geographic information system data, traffic conditions, and other power supply vehicle locations. The power demand is obtained by power supply planning based on the power demand of users known before the disaster. The traffic conditions are obtained by monitoring and evaluating the real-time traffic conditions. LSTM The model is used to predict traffic conditions.

[0007] Furthermore, the traffic condition prediction step includes: Collect historical traffic data; Perform data cleaning and normalization on historical traffic data; Based on the cleaned data, features related to traffic conditions are extracted as input features; Create a time series dataset and organize the input features into a time series format; Build LSTM The model is trained using input features in time series format and uses the trained LSTM The model predicts traffic conditions and denormalizes the prediction results.

[0008] Furthermore, the user priorities are divided according to the national standard of urban lifeline load classification, including medical facilities, emergency service facilities, communication facilities, water supply and drainage facilities, energy facilities, transportation facilities, food supply facilities and other critical infrastructure.

[0009] Furthermore, the reward value is calculated by a reward function, and the expression of the reward function is:

[0010] in, represents the reward function, Indicates user The priority of The power supply vehicle meets the needs of users The electricity demand situation, Power supply vehicle The driving time, Indicates the traffic congestion situation of the road section where the power supply vehicle is currently traveling; is the weight coefficient.

[0011] Furthermore, after training and updating the power supply vehicle scheduling neural network model based on the reward value, the scheduling path and power supply strategy with the highest joint total value are obtained, including: Obtain the state transition of the power supply vehicle and store it in the experience playback buffer of the neural network, where the state transition refers to the current state, the action selected in the current state, the reward value obtained by whether the power supply can be provided to the user in the shortest time, and the next state to be transferred to; Each power supply vehicle passes through DQN Obtained from the network Q The value is passed into the hybrid network of the shared attention module to obtain the optimal overall joint total value. Q value, this Q The value is passed as the new state for the next step of training DQN In the network's experience replay buffer, the above process is iterated until a specified amount of data is accumulated in the experience replay buffer; Retrieve from the experience replay buffer batch The neural network parameters are trained and updated by state transfer until the overall loss function finally converges to close to zero, and the gradient of the overall loss function is updated at the same time DQN The parameters in the network and the parameters in the hybrid network in the shared attention module; The power supply vehicle scheduling path and power supply strategy are determined according to the updated neural network model.

[0012] Furthermore, the expression of the overall loss function is:

[0013] Where: is the overall loss function; B for batchsize ; yes t Always provide rewards for the electric vehicle agent; yes t Moment i The power supply vehicle agent chooses an action The value obtained afterwards; s Represents the state and environment information of the power supply vehicle intelligent body; represents the model parameters that control the decision-making of the power supply vehicle agent; G ( X )for Huber The loss function is expressed as:

[0014] The expression of the joint total value is:

[0015] Where: Represents the overall value, which is the common value generated by all agents after performing actions; Indicates status The bias term is part of the embedding vector of the global state and is used to integrate the information of all agents; H Indicates the total number of times, Represents the joint value of each power supply vehicle, expressed as:

[0016] Where: N Indicates the number of power supply vehicles; represents the individual value of each power supply vehicle; Represents the head coefficient of multi-head attention, expressed as:

[0017] Where: is the embedding vector of the cooperation feature, and are the global state and shared feature vector respectively; is the embedding vector of the global state.

[0018] In a second aspect, the present invention provides an emergency power supply vehicle dispatching system for restoring urban power supply, comprising: Information acquisition module: used to obtain global status information under the current state; Information processing module: used to process the global status information in the current state through pre-established DQN The power supply vehicle scheduling neural network model based on the network and shared attention module obtains the scheduling path and power supply strategy of individual power supply vehicles; Data evaluation module: used to evaluate the reward value of the power supply vehicle after scheduling the power supply vehicle based on the scheduling path and power supply strategy; Model training module: used to train and update the power supply vehicle scheduling neural network model based on the reward value to obtain the scheduling path and power supply strategy with the highest combined total value; Scheduling and power supply module: used to dispatch and supply power to all power supply vehicles based on the scheduling path and power supply strategy with the highest combined total value.

[0019] In a third aspect, the present invention provides an emergency power supply vehicle dispatching device for restoring urban power supply, including a processor and a storage medium; The storage medium is used to store instructions; The processor is used to operate according to the instructions to execute the steps of any of the methods described above.

[0020] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of any of the above methods when executed by a processor.

[0021] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention uses an intelligent and automated dispatching system to quickly respond and adjust after a disaster occurs, thereby improving the city's emergency power material dispatching capabilities, reducing power outage time and economic losses, and ensuring the stability of urban operations and the safety of residents' lives; Second, the present invention comprehensively considers user priority, power demand, traffic conditions, meteorological conditions, GIS The data and the location and power information of multiple power supply vehicle intelligent bodies are integrated and analyzed in multiple dimensions to improve the scientificity and effectiveness of scheduling decisions and avoid unreasonable scheduling problems under the influence of a single factor. 3. The present invention adopts experience playback and target network update strategy to make DQN The algorithm training process is more stable, avoiding shock and instability during training, and DQN Adding a shared attention module to the basis can dispatch multiple power supply vehicles at the same time. LSTM The model's time series data processing capabilities enable the system to maintain high prediction accuracy and decision-making capabilities in the face of complex and changing environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The accompanying drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the accompanying drawings: Figure 1 A schematic diagram of a flow chart of a method for dispatching an emergency power supply vehicle for restoring urban power supply provided in the first embodiment of the present invention; Figure 2 An overall framework diagram of a multi-agent reinforcement learning algorithm based on shared attention for a method of dispatching an emergency power supply vehicle for restoring urban power supply provided in the first embodiment of the present invention; Figure 3 A detailed diagram of a shared attention module of a method for dispatching an emergency electric power supply vehicle for restoring urban power supply provided in Embodiment 1 of the present invention. DETAILED DESCRIPTION

[0023] The present invention will be described in detail below with reference to the accompanying drawings and in combination with embodiments. It should be noted that the embodiments and features in the embodiments of the present invention can be combined with each other without conflict.

[0024] The following detailed description is an exemplary description, which is intended to provide further detailed description of the present invention. Unless otherwise specified, all technical terms used in the present invention have the same meaning as those generally understood by those skilled in the art to which the present invention belongs. The terms used in the present invention are only for describing specific embodiments, and are not intended to limit the exemplary embodiments according to the present invention.

[0025] Embodiment 1: This embodiment provides a method for dispatching emergency power supply vehicles to restore urban power supply. For a single power supply vehicle, the emergency power supply vehicle is dispatched according to user priority, power demand, real-time meteorological information, geographic information system ( GIS ) data, traffic conditions, and other power supply vehicle location initialization environmental states. It should be noted here that the power demand is obtained by power supply planning based on the power demand of users known before the disaster. The traffic conditions are obtained by monitoring and evaluating the real-time traffic conditions, that is, analyzing the traffic conditions through traffic monitoring cameras and video surveillance. GPS data, obtain real-time location information and driving speed and traffic application data from the vehicle navigation system, using Google Maps , BaiduMaps Traffic data provided by applications such as LSTM Traffic condition prediction is performed, and the traffic condition prediction specifically includes the following steps: step S 11: Collect historical traffic data, including: Traffic flow data: Collect vehicle flow data of each road section, which can be from traffic sensors, cameras, etc.; Road speed data: Collect the driving speed of vehicles on each road section; Road congestion data: Collect the congestion index of each road section, which is usually provided by the traffic management system; Timestamp data: Record the timestamp of the data, including time dimensions such as hours, days, and weeks; step S 12: Clean the acquired data, including removing outliers and missing values, ensuring the accuracy of the data and normalizing the data to make it conform to the LSTM The input requirements of the model; step S 13: Extract features from the cleaned data, including extracting features related to traffic conditions such as traffic flow, driving speed, congestion index, etc., or using time features (such as hours, days of the week, etc.) as input features; step S 14: Create a time series data set and organize the data into a time series format. LSTM The model requires continuous time step data as input and divides the data set; step S 15: Build LSTM Model and train, and use the trained LSTM The model predicts traffic conditions and denormalizes the prediction results.

[0026] In the current state, the depth Q network( DQN ) Algorithm based on QThe value function selects the optimal scheduling path and power supply strategy; the power supply vehicle supplies power according to the selected path and strategy and obtains the corresponding reward value; the current state, selected action, reward obtained and new state are stored in the experience playback buffer, and each power supply vehicle agent is replayed through DQN What you get Q The value is passed into the hybrid network of the shared attention module to obtain the overall optimal Q Value; randomly extract small batch samples from the experience replay buffer and update the target Q network; during the back propagation process, the gradient of the loss function is updated simultaneously DQN The parameters in the hybrid network in the shared attention module are repeated until the power supply strategy converges to ensure that the power supply vehicle supplies power to the highest priority user in the shortest time. It should be noted here that the depth Q network( DQN ) The scheduling and shared attention module gives the global state information given by the environment to each power supply vehicle agent and helps the agent make corresponding decisions. By adding shared vectors, vehicles can cooperate with each other to achieve the optimization of value combination. This embodiment refers to the national standard of urban lifeline load classification to divide the priority of users and determine the order of priority power supply. The priority of users includes but is not limited to medical facilities, emergency service facilities, communication facilities, water supply and drainage facilities, energy facilities, transportation facilities, food supply facilities and other key infrastructure. The power supply vehicle scheduling specifically includes the following steps: step S 21: First, construct a neural network and initialize the parameters. The initialization parameters include: initializing the learning rate of the neural network, TD The discount factor used when updating the algorithm and the batch size taken from the experience pool each time when training the network batchsize , state space and action changes; secondly, initialize the experience replay pool; step S 22: After obtaining the action space, the action is executed to select the driving direction and speed for the power supply vehicle. Then, according to the predicted traffic conditions and weather data, the new traffic conditions and the changed weather data are defined as the new state. During the operation of the algorithm, the new state will be given according to the action, and the predicted traffic conditions and weather data will be adjusted in real time. Q The value is passed into the shared attention module to obtain the overall optimal Q value, and this Q The value is passed as the new state for the next step of training DQN In the network, the specific value of the new state is unknown, and the above process is iterated until at least 200 data are accumulated in the experience replay pool; step S23: Train the neural network and take out the experience replay pool batch A state transition, wherein the state transition refers to the current state, the action selected in the current state, the reward obtained by supplying power to the user in the shortest time, and the next state to be transferred to. Then, the neural network parameters are trained and updated according to the state transition stored in the experience playback pool until the loss function of the algorithm finally converges to close to 0; it should be noted that the training basis in this embodiment is Strategy selection action.

[0027] In order to solve the problem of power supply vehicle scheduling, the present invention redefines the agent, state space, action space, reward function, transfer probability function and discount factor according to the reinforcement learning process, that is, , specifically: 1. Agent N :In the power supply vehicle scheduling problem, each idle power supply vehicle is regarded as an intelligent agent. Each intelligent agent has its own independent decision-making. Its goal is to supply power to higher priority and more demanding areas in the shortest time. The power supply vehicle intelligent agents are independent of each other and are only responsible for their own decisions.

[0028] 2. State Space :The state space represents the environmental information that the power supply vehicle can observe at each time step. The state space can be defined as a vector containing multi-dimensional information, including: 1. Current time step: represents the current time; 2. The location of the power supply vehicle: the geographical location of the power supply vehicle in the city; 3. The location and priority of each user: the geographical location and priority of each user; 4. The power demand of each user: the power demand of each user; 5. Real-time traffic conditions: traffic flow, speed and congestion on the route of the power supply vehicle; 6. Meteorological information: current meteorological data, such as rainfall, temperature, wind speed, etc.; 7. IS Data: terrain and road information. The state space is defined as follows:

[0029] in represents the current time step, , express t Moment i Status information of power supply vehicles, Indicates the location of the power supply vehicle. Indicates user location and priority, Indicates the user's electricity demand. Indicates real-time traffic conditions. Indicates weather information. express GIS data, Indicates the location information of other power supply vehicles.

[0030] Action Space A , : express t Moment i The actions of the power supply vehicle agent are all possible scheduling decisions that the power supply vehicle can take, including: 1. Selecting a driving route: the path selection from the current location of the power supply vehicle to a certain user location; 2. Selecting a service user: the power supply vehicle chooses which user to provide power for at the current time step.

[0031] 4. Reward Function : The power supply vehicle agents in the same time slice enjoy a common reward function, and the reward internal function needs to take into account the following four factors: 1. Reaching high-priority users: When the power supply vehicle successfully supplies power to high-priority users, a higher positive reward is given. 2. Meeting the user's electricity needs: When the power supply vehicle successfully meets the user's electricity needs, a positive reward is given. 3. Driving time and cost: The shorter the driving time of the power supply vehicle, the higher the reward given; the longer the driving time, the lower the reward given. 4. Traffic conditions: When the power supply vehicle is driving under smooth traffic conditions, a positive reward is given; when it is driving under traffic congestion, a negative reward is given. The definition of the reward function is as follows:

[0032] in, Indicates user The priority of The power supply vehicle meets the needs of users The electricity demand situation, Power supply vehicle The driving time, Indicates the traffic congestion situation of the road section where the power supply vehicle is currently traveling. is the weight coefficient, which is used to balance the influence of different factors.

[0033] 5. State transition probability function , which represents the probability of transitioning to the next state when the current state takes a joint action.

[0034] 6. Regarding It can be set to 0.99 because each power car agent tries to maximize its expected reward.

[0035] The overall framework of the multi-agent reinforcement learning algorithm based on shared attention is as follows Figure 2 As shown in the figure, the overall network framework is mainly divided into two parts. The first part is the intelligent agent network, which is mainly composed of DQNThe network consists of a network whose function is to deliver the global state information given by the environment to each power supply vehicle agent and help the power supply vehicle agent make corresponding decisions, and then evaluate the value of the action behavior selected by the power supply vehicle agent. The second part is mainly a hybrid network based on the shared attention module. By adding shared vectors, the power supply vehicles can cooperate with each other to achieve the optimal value combination.

[0036] The overall loss function is:

[0037] in B for batchsize ; yes t Always provide rewards for the electric vehicle agent; yes t Moment i The power supply vehicle agent chooses an action The value obtained afterwards; s Represents the state and environment information of the power supply vehicle intelligent body; represents the model parameters that control the decision-making of the power supply vehicle agent; G ( X )for Huber The loss function is as follows:

[0038] Shared attention module such as Figure 3 As shown in the figure, the shared attention module allows vehicles to understand each other's positions and take the positions of other vehicles into consideration when making decisions, thus transforming the local optimal solution into the overall optimal solution.

[0039] In the reinforcement learning process, it is necessary to measure the value of each action and the joint action of the vehicle: individual value : Indicated in t Moment i Agents in state Select action in case Value obtained; joint total value : Represents the overall value, which is the common value generated by all agents after performing actions. The size of indicates the quality of the overall behavior. The following is the joint total value The calculation formula is:

[0040] Where: Indicates status The bias term is part of the embedding vector of the global state and is used to integrate the information of all agents. HIndicates the total number of times, N Indicates the number of power supply vehicles, represents the head coefficient of multi-head attention, Represents the individual value of each agent.

[0041] This framework learns the self-attention of each agent by selectively focusing on information from other agents. Self-attention between each pair of agents is not performed, but self-attention is used to help the entire system model the influence of each individual on each agent. Specifically, the embedding vector of the global state is and the embedding vector of the cooperative feature The similarity between Softmax The head coefficient of multi-head attention is obtained , this process will be carried out H Second-rate:

[0042] in: is the embedding vector of the cooperation feature, and are the inputs of the shared attention module, which are the global state and the shared feature vector respectively. and It can be obtained by performing one or two layers of embedding transformation on the global state and shared feature vector. After that, And then with the individual value of each power supply vehicle Do the dot product to get the joint value :

[0043] Next, N The combined value of the power supply vehicles Summing, we get:

[0044] This solution uses the power supply vehicle warehouse to dispatch power supply vehicles to the power outage area in the city. The dispatch process comprehensively considers user priority, user power demand, meteorological conditions, geographic information system ( GIS ) data and traffic conditions to ensure that the power supply vehicle can provide power to users as quickly as possible. User priority adopts the Chinese standard lifeline load classification standard. User power demand has been predicted and known before the disaster occurs. Traffic conditions are determined by LSTM This scheme optimizes the dispatching paths and strategies of multiple power supply vehicles based on a multi-agent reinforcement learning algorithm with shared attention, thereby maximizing power supply efficiency and minimizing power outage time. This method can effectively improve the speed and efficiency of restoring the city's lifeline after a disaster.

[0045] Embodiment 2: An emergency power supply vehicle dispatching system for restoring urban power supply can implement an emergency power supply vehicle dispatching method for restoring urban power supply described in Embodiment 1, including: Information acquisition module: used to obtain global status information under the current state; Information processing module: used to process the global status information in the current state through pre-established DQN The power supply vehicle scheduling neural network model based on the network and shared attention module obtains the scheduling path and power supply strategy of individual power supply vehicles; Data evaluation module: used to evaluate the reward value of the power supply vehicle after scheduling the power supply vehicle based on the scheduling path and power supply strategy; Model training module: used to train and update the power supply vehicle scheduling neural network model based on the reward value to obtain the scheduling path and power supply strategy with the highest combined total value; Scheduling and power supply module: used to dispatch and supply power to all power supply vehicles based on the scheduling path and power supply strategy with the highest combined total value.

[0046] Embodiment three: The embodiment of the present invention further provides an emergency power supply vehicle dispatching device for restoring urban power supply, which can implement the emergency power supply vehicle dispatching method for restoring urban power supply described in the first embodiment, including a processor and a storage medium; The storage medium is used to store instructions; The processor is used to operate according to the instructions to perform the steps of the following method: Get the global status information under the current state; Based on the global state information in the current state, through the pre-established DQN The power supply vehicle scheduling neural network model based on the network and shared attention module obtains the scheduling path and power supply strategy of individual power supply vehicles; After scheduling the power supply vehicle based on the scheduling path and power supply strategy, the reward value of the power supply vehicle is evaluated; After training and updating the power supply vehicle scheduling neural network model based on the reward value, the scheduling path and power supply strategy with the highest joint total value are obtained; All power supply vehicles are dispatched and supplied with power based on the dispatch path and power supply strategy with the highest combined total value.

[0047] Embodiment 4: The embodiment of the present invention further provides a computer-readable storage medium, which can implement the method for dispatching an emergency power supply vehicle for restoring urban power supply as described in the first embodiment, and stores a computer program thereon, which implements the steps of the following method when executed by a processor: Get the global status information under the current state; Based on the global state information in the current state, through the pre-established DQN The power supply vehicle scheduling neural network model based on the network and shared attention module obtains the scheduling path and power supply strategy of individual power supply vehicles; After scheduling the power supply vehicle based on the scheduling path and power supply strategy, the reward value of the power supply vehicle is evaluated; After training and updating the power supply vehicle scheduling neural network model based on the reward value, the scheduling path and power supply strategy with the highest joint total value are obtained; All power supply vehicles are dispatched and supplied with power based on the dispatch path and power supply strategy with the highest combined total value.

[0048] It is known from common technical knowledge that the present invention can be implemented by other embodiments that do not deviate from its spirit or essential features. Therefore, the above disclosed embodiments are only illustrative in all respects and are not exclusive. All changes within the scope of the present invention or within the scope equivalent to the present invention are included in the present invention.

[0049] It should be understood by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of an all-hardware embodiment, an all-software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of one or more computer-usable storage media (including but not limited to disk storage, CD - ROM , optical storage, etc.).

[0050] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0051] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0052] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0053] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for dispatching an emergency power supply vehicle for restoring urban power supply, characterized in that: include: Get the global status information under the current state; Based on the global state information in the current state, through the pre-established DQN The power supply vehicle scheduling neural network model based on the network and shared attention module obtains the scheduling path and power supply strategy of individual power supply vehicles; After scheduling the power supply vehicle based on the scheduling path and power supply strategy, the reward value of the power supply vehicle is evaluated; After training and updating the power supply vehicle scheduling neural network model based on the reward value, the scheduling path and power supply strategy with the highest joint total value are obtained; All power supply vehicles are dispatched and supplied with power based on the dispatch path and power supply strategy with the highest combined total value.

2. The method for dispatching emergency power supply vehicles for restoring urban power supply according to claim 1, characterized in that: The global status information includes user priority, power demand, real-time weather information, geographic information system data, traffic conditions and other power supply vehicle locations. The power demand is obtained by power supply planning based on the power demand of users known before the disaster. The traffic conditions are obtained by monitoring and evaluating the real-time traffic conditions. LSTM The model is used to predict traffic conditions.

3. The method for dispatching emergency power supply vehicles for restoring urban power supply according to claim 2 is characterized in that: The traffic condition prediction step includes: Collect historical traffic data; Clean and normalize historical traffic data; Based on the cleaned and normalized data, features related to traffic conditions are extracted as input features; Create a time series dataset and organize the input features into a time series format; Build LSTM The model is trained using input features in time series format and uses the trained LSTM The model predicts traffic conditions and denormalizes the prediction results.

4. The method for dispatching emergency electric power supply vehicles for restoring urban power supply according to claim 2, characterized in that: The user priorities are divided according to the national standard for urban lifeline load classification, including medical facilities, emergency service facilities, communication facilities, water supply and drainage facilities, energy facilities, transportation facilities, food supply facilities and other critical infrastructure.

5. The method for dispatching emergency electric power supply vehicles for restoring urban power supply according to claim 1, characterized in that: The reward value is calculated by a reward function, and the expression of the reward function is: in, represents the reward function, Indicates user The priority of The power supply vehicle meets the needs of users The electricity demand situation, Power supply vehicle The driving time, Indicates the traffic congestion situation of the road section where the power supply vehicle is currently traveling; is the weight coefficient.

6. The method for dispatching emergency electric power supply vehicles for restoring urban power supply according to claim 1, characterized in that: After training and updating the power supply vehicle scheduling neural network model based on the reward value, the scheduling path and power supply strategy with the highest joint total value are obtained, including: Obtain the state transition of the power supply vehicle and store it in the experience playback buffer of the neural network, where the state transition refers to the current state, the action selected in the current state, the reward value obtained by whether the power supply can be provided to the user in the shortest time, and the next state to be transferred to; Each power supply vehicle passes through DQN Obtained from the network Q The value is passed into the hybrid network of the shared attention module to obtain the overall optimal Q value, this Q The value is passed as the new state for the next step of training DQN In the network's experience replay buffer, the above process is iterated until a specified amount of data is accumulated in the experience replay buffer; Retrieve from the experience replay buffer batch The neural network parameters are trained and updated by state transfer until the overall loss function finally converges to close to zero, and the gradient of the overall loss function is updated at the same time DQN The parameters in the network and the parameters in the hybrid network in the shared attention module; The power supply vehicle scheduling path and power supply strategy are determined according to the updated neural network model.

7. The method for dispatching emergency power supply vehicles for restoring urban power supply according to claim 6, characterized in that: The expression of the overall loss function is: Where: is the overall loss function; B for batchsize ; yes t Always provide rewards for the electric vehicle agent; yes t Moment i The power supply vehicle agent chooses an action The value obtained afterwards; s Represents the state and environment information of the power supply vehicle intelligent body; The model parameters that control the decision-making of the power supply vehicle agent need to be continuously learned to optimize the performance of the model; G ( X )for Huber The loss function is expressed as: The expression of the joint total value is: Where: Represents the overall value, which is the common value generated by all agents after performing actions; Indicates status The bias term is part of the embedding vector of the global state and is used to integrate the information of all agents; H Indicates the total number of times, Represents the joint value of each power supply vehicle, expressed as: Where: N Indicates the number of power supply vehicles; represents the individual value of each power supply vehicle; Represents the head coefficient of multi-head attention, expressed as: Where: is the embedding vector of the cooperation feature, and are the global state and shared feature vector respectively; is the embedding vector of the global state.

8. An emergency power supply vehicle dispatching system for restoring urban power supply, characterized in that: include: Information acquisition module: used to obtain global status information under the current state; Information processing module: used to process the global status information in the current state through pre-established DQN The power supply vehicle scheduling neural network model based on the network and shared attention module obtains the scheduling path and power supply strategy of individual power supply vehicles; Data evaluation module: used to evaluate the reward value of the power supply vehicle after scheduling the power supply vehicle based on the scheduling path and power supply strategy; Model training module: used to train and update the power supply vehicle scheduling neural network model based on the reward value to obtain the scheduling path and power supply strategy with the highest joint total value; Scheduling and power supply module: used to dispatch and supply power to all power supply vehicles based on the scheduling path and power supply strategy with the highest combined total value.

9. An emergency power supply vehicle dispatching device for restoring urban power supply, characterized in that: including processor and storage medium; The storage medium is used to store instructions; The processor is used to operate according to the instructions to execute the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method described in any one of claims 1 to 7 are implemented.

Images