Intelligent plan generation system and method for power grid operation fault

Abstract

The invention discloses an intelligent plan generation system and method for a power grid operation fault, and the system comprises a power grid situation sensing module which is used for obtaining the real-time operation data of a power grid, constructing a digital structure diagram of a power grid diagram, inputting the digital structure diagram of the power grid diagram into a diagram digital sensing model for the situation sensing of the power grid, and obtaining the power grid operation fault; the power grid fault positioning module is used for positioning a power grid operation fault and predicting power grid operation stability to obtain fault positioning information and an operation stability result; the pre-arranged plan intelligent generation module is used for inputting the fault positioning information and the operation stability result into a pre-arranged plan generation model based on deep learning and outputting a preliminary pre-arranged plan result; and combining the power grid topological structure and the output of the generator set as regulation and control actions, and generating a final power grid plan based on a reinforcement learning algorithm by using the electric power big data and the preliminary plan result. According to the invention, automatic detection, rapid positioning and intelligent plan generation of power grid operation faults are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid operation fault processing, and in particular to a system and method for intelligently generating plans for power grid operation faults. Background Art

[0002] In the power industry, the safety and stability of power grid operation are of vital importance. However, due to various reasons, such as equipment aging, natural disasters, human operational errors, etc., power grid operation failures occur from time to time. These failures may not only lead to power supply interruptions, but also damage equipment and even endanger personnel safety. Therefore, quickly and accurately locating faults and generating effective emergency plans are of great significance to ensure the stable operation of the power grid.

[0003] The traditional emergency plan for power grid failure mainly relies on human experience and expert judgment. This method is often difficult to respond quickly and accurately when facing complex and changeable power grid failures, resulting in low efficiency in handling power grid operation failures. In addition, with the continuous expansion of the scale of the power grid and the improvement of the level of intelligence, the types and complexity of power grid failures are also increasing, and the requirements for plan formulation are becoming higher and higher.

[0004] In view of this, this application is hereby filed. Summary of the invention

[0005] The technical problem to be solved by the present invention is that the existing methods for generating emergency plans for power grid failures often fail to respond quickly and accurately when faced with complex and changeable power grid failures, resulting in low efficiency in handling power grid operation failures. The present invention aims to provide an intelligent generation system and method for plans for power grid operation failures, which is based on artificial intelligence technology and integrates multiple modules such as a power grid situation awareness module, a power grid fault location module, and an intelligent plan generation module, and utilizes advanced technologies such as deep learning and machine learning to achieve automatic detection, rapid location, and intelligent generation of plans for power grid operation failures.

[0006] The present invention is achieved through the following technical solutions:

[0007] In a first aspect, the present invention provides an intelligent generation system for emergency plans for power grid operation failures, the system comprising a power grid situation awareness module, a power grid fault location module and an intelligent generation module for emergency plans;

[0008] The power grid situation awareness module is used to obtain real-time power grid operation data, build a digital structure diagram of the power grid diagram based on the real-time power grid operation data, input the digital structure diagram of the power grid diagram into the digital perception model to perform power grid situation awareness, and obtain power grid operation faults;

[0009] The power grid fault location module is used to locate power grid operation faults and predict power grid operation stability, and obtain fault location information and operation stability results; the fault location information includes the location of the fault node, the fault type and the fault time information; the operation stability result refers to the stability index of the power grid under different operation states, and the stability index includes voltage, frequency and power flow;

[0010] The intelligent plan generation module is used to input fault location information and operation stability results into the deep learning-based plan generation model, and output preliminary plan results; it combines the grid topology structure and generator output as control actions, and uses power big data and preliminary plan results based on the reinforcement learning algorithm to generate the final grid plan.

[0011] Furthermore, the power grid situation awareness module includes:

[0012] An acquisition unit, used for acquiring real-time operation data of the power grid, the real-time operation data of the power grid including generator node information, bus node information, transmission line information between different component nodes and attribute type information of different component nodes;

[0013] A structure diagram construction unit is used to extract the grid characteristics from the generator node information, bus node information, transmission line information between different component nodes and attribute type information of different component nodes based on the real-time operation data of the grid and form a graph digital feature; based on the graph digital feature, a digital structure diagram of the grid diagram is generated;

[0014] The virtual representation unit is used to input the digital structure diagram of the power grid diagram into the digital perception model to obtain the virtual representation μ i (t), μ i (t) represents the virtual representation of the i-th node at time t;

[0015] The sensing unit is used to determine whether the digital structure diagram of the power grid diagram of the i-th node at time t has changed based on the virtual representation. If so, the power grid operation has a fault; if not, the power grid operation has no fault.

[0016] Furthermore, the representation of the graph digital features is GDT = {V, E, K}, where V = {v d1 ,...v dn ,v z1 ,...v zn} represents the set of generator node information and bus node information features, v d1 ,...v dn represents the information characteristics of the generator node, v z1 ,...v zn Represents the bus node information characteristics, E={e1,e1...e n} represents the state of the transmission line between different component nodes, K=

[0017] {k d1 ,...k dn ,k z1 ,...k zn} represents the attribute type of different component nodes, k d1 ,...k dn Indicates the attribute type of the generator node, k z1 ,...k zn Represents the property type of a bus node.

[0018] Furthermore, the virtual representation μ obtained by the digital perception model is i (t) = {s i (t), Δd i , M i , L i (t)}, where s i (t) represents the current operating status of the bus and generator, Δd i Indicates the data change value of the bus and generator, M i represents the historical failure status of the i-th node, L i (t) represents the current operating status of the component node adjacent to the i-th node;

[0019] Based on the virtual representation, it is determined whether the digital structure diagram of the power grid diagram of the i-th node at time t has changed, including:

[0020] The virtual representation μ i Any quantity s in (t) i (t), Δd i 、M i , L i (t) is compared with the corresponding quantity at the previous moment. If there is a change, it means that the digital structure diagram of the power grid diagram of the i-th node at time t has changed, otherwise there is no change.

[0021] Furthermore, the power grid fault location module includes:

[0022] The fault location unit is used to construct a fault location model F(Z,A), where Z represents the fault node feature and A represents the adjacency matrix of the relationship between the components; calculate the propagation rule of the fault location model, verify the adjacency matrix A of the relationship between the components of each fault node based on the verification matrix of the propagation rule, and determine whether the verified adjacency matrix A is consistent with the adjacency matrix A of the relationship between the components. If so, determine the number of layers of each fault node, and determine the position of the digital structure diagram of the power grid diagram where the fault is located based on the number of layers of each fault node, and obtain fault location information;

[0023] The operation stability prediction unit is used to obtain the characteristic matrix of voltage and current of each node at time t, and perform feedforward propagation calculation on the digital structure diagram of the power grid diagram based on the graph convolution network to obtain the node feature classification value Mp; based on the node feature classification value M p The global representation of the graph is obtained by using the adjacency matrix, and the average pooling method is used to classify the graph to obtain the classification result; the classification result is compared with the preset operation stability result to obtain the operation stability result.

[0024] Furthermore, the expression of the propagation rule is:

[0025]

[0026] in, represents the operation of adding the identity matrix to the adjacency matrix of an undirected graph, represents the operation of adding the reference matrix to the adjacency matrix of the undirected graph, Q j is a trainable weight matrix, σ[.] represents the activation function, N j represents the validation matrix of the jth layer.

[0027] Furthermore, the node feature classification value M p The expression is:

[0028]

[0029] Where X is the characteristic matrix of voltage and current of each node at time t; is the weight matrix from the input layer to the hidden layer, is the weight matrix from the hidden layer to the output layer, τ(.) represents the function that transforms the node vector into a graph vector representing the entire graph, and softmax(.) represents the classification function.

[0030] Furthermore, the plan intelligent generation module includes:

[0031] A preliminary plan generation unit is used to preprocess the fault location information and the operation stability results to obtain preprocessed data; the preprocessed data is input into a plan generation model based on deep learning, and a preliminary plan result is output; the plan generation model based on deep learning adopts any one of a recurrent neural network, a long short-term memory network, a Transformer or a convolutional neural network;

[0032] The final plan generation unit is used to combine the grid topology and the output of the generator set as the control action, and generate the final grid plan based on the reinforcement learning algorithm using the power big data and the preliminary plan results; the power big data includes the operating flow data of the transmission network at the current moment and the transmission network operating scenario simulation data.

[0033] Furthermore, the execution process of the final plan generation unit is:

[0034] The actual operation status of the transmission network is simulated in the power grid simulation environment. In the operation scenarios of renewable energy grid connection and load fluctuation, the simulation environment solves the operation power flow data of the transmission network at the current moment through power flow calculation according to the current operation status of the transmission network, obtains the information of the topological structure of the transmission network, and uses the program interface to obtain the corresponding state space of the reinforcement learning algorithm;

[0035] The simulation data of the power transmission network operation scenario with a duration of about 100,000 time steps is used as input and runs in the power transmission network simulation environment. The power flow solver performs power flow calculation for each input scenario and records each calculation result. When the power transmission line is overloaded in the power transmission network, a topology control action is selected in the initial state space based on the exhaustive method. If the selected topology control action can restore the power transmission network from the overload state to the safe operation state, the action is saved and continuously simulated until the iteration of the operation scenario is completed. The selected topology control actions are merged as the action space for reinforcement learning algorithm training.

[0036] The state quantity given by the transmission network simulation environment should be observed in real time to determine whether it meets the preset mandatory constraints. If not, the current scene simulation ends; if it does, the load rate of the transmission line is observed to obtain the load rate; among them, the preset mandatory constraints are: ① Ensure the power consumption on the load side of the transmission network, that is, the output of the generator set must be balanced with the load power consumption; ② At most one generator set can be shut down; ③ The changed topology structure cannot produce an island in the transmission network; ④ The power flow calculation of the transmission network always maintains convergence; If there is no transmission line with a load rate greater than 100%, no action is required; If there is a load rate greater than 100%, the intelligent agent selects the optimal topology structure control action according to the current strategy;

[0037] After selecting an action and returning to the interactive environment, the agent updates the neural network parameters based on the current state and reward function, and outputs the final power grid plan.

[0038] In a second aspect, the present invention further provides a method for intelligently generating a plan for power grid operation failures, which is applied to the above-mentioned system for intelligently generating a plan for power grid operation failures; the method comprises:

[0039] Acquire real-time operation data of the power grid, construct a digital structure diagram of the power grid diagram based on the real-time operation data of the power grid, input the digital structure diagram of the power grid diagram into the digital perception model to perceive the power grid situation, and obtain power grid operation faults;

[0040] Locate power grid operation faults and predict power grid operation stability to obtain fault location information and operation stability results; fault location information includes the location of the fault node, fault type and fault time information; operation stability results refer to the stability indicators of the power grid under different operating conditions, including voltage, frequency and power flow;

[0041] The fault location information and operation stability results are input into the deep learning-based plan generation model to output the preliminary plan results. The grid topology and generator output are combined as control actions, and the power big data and preliminary plan results are used based on the reinforcement learning algorithm to generate the final grid plan.

[0042] Furthermore, the real-time operation data of the power grid is obtained, a digital structure diagram of the power grid diagram is constructed based on the real-time operation data of the power grid, and the digital structure diagram of the power grid diagram is input into the digital perception model for power grid situation perception to obtain power grid operation faults, including:

[0043] Acquire real-time operation data of the power grid, which includes generator node information, bus node information, transmission line information between different component nodes, and attribute type information of different component nodes;

[0044] Based on the real-time operation data of the power grid, the power grid features are extracted from the generator node information, bus node information, transmission line information between different component nodes and attribute type information of different component nodes to form a graph digital feature; based on the graph digital feature, a digital structure diagram of the power grid diagram is generated;

[0045] Input the digital structure diagram of the power grid diagram into the digital perception model to obtain the virtual representation μ i (t), μ i (t) represents the virtual representation of the i-th node at time t;

[0046] Based on the virtual representation, it is determined whether the digital structure diagram of the power grid diagram of the i-th node at time t has changed. If so, the power grid operation has a fault; if not, the power grid operation has no fault.

[0047] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0048] The present invention discloses an intelligent generation system and method for plans for power grid operation failures. The present invention obtains real-time power grid operation data, constructs a digital structure diagram of a power grid diagram based on the real-time power grid operation data, inputs the digital structure diagram of the power grid diagram into a digital perception model for power grid situation perception, and discovers power grid operation failures; locates power grid operation failures and predicts power grid operation stability to obtain fault location information and operation stability results; substitutes the fault location information and operation stability results into a plan generation model based on deep learning, outputs preliminary plan results, combines power grid topology and generator set output as control actions, and obtains a final power grid plan based on a reinforcement learning algorithm using power big data and preliminary plan results, thereby realizing automatic detection, rapid location, and intelligent generation of plans for power grid operation failures, and improving the efficiency of handling power grid operation failures. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of this application, and do not constitute a limitation of the embodiments of the present invention. In the drawings:

[0050] Figure 1 This is a system block diagram of an intelligent generation system for emergency plans for power grid operation failures according to Embodiment 1 of the present invention;

[0051] Figure 2 This is a structural diagram of the power grid situation awareness module of the present invention;

[0052] Figure 3 This is a structural diagram of a power grid fault location module of the present invention;

[0053] Figure 4 This is a structural diagram of the intelligent generation module of the plan of the present invention;

[0054] Figure 5 This is a flow chart of a method for intelligently generating plans for power grid operation failures according to Embodiment 2 of the present invention. DETAILED DESCRIPTION

[0055] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with embodiments and drawings. The exemplary embodiments of the present invention and their description are only used to explain the present invention and are not intended to limit the present invention.

[0056] Example 1

[0057] like Figure 1 As shown, Figure 1 1 is a system block diagram of an intelligent generation system for emergency plans for power grid operation failures according to Embodiment 1 of the present invention; the intelligent generation system for emergency plans for power grid operation failures according to the present invention comprises a power grid situation awareness module, a power grid fault location module and an intelligent generation module for emergency plans;

[0058] The power grid situation awareness module is used to obtain real-time power grid operation data, build a digital structure diagram of the power grid diagram based on the real-time power grid operation data, input the digital structure diagram of the power grid diagram into the digital perception model to perform power grid situation awareness, and obtain power grid operation faults;

[0059] The power grid fault location module is used to locate power grid operation faults and predict power grid operation stability, and obtain fault location information and operation stability results; the fault location information includes the location of the fault node, the fault type and the fault time information; the operation stability result refers to the stability index of the power grid under different operation states, and the stability index includes voltage, frequency and power flow;

[0060] The intelligent plan generation module is used to input fault location information and operation stability results into the deep learning-based plan generation model, and output preliminary plan results; it combines the grid topology structure and generator output as control actions, and uses power big data and preliminary plan results based on the reinforcement learning algorithm to generate the final grid plan.

[0061] In some embodiments, Figure 2 As shown, the power grid situation awareness module includes:

[0062] An acquisition unit is used to acquire real-time operation data of the power grid, which includes generator node information, bus node information, transmission line information between different component nodes, and attribute type information of different component nodes; specifically, the real-time operation data of the power grid includes not only the above data, but also voltage data, current data, etc.;

[0063] A structure diagram construction unit is used to extract the grid characteristics from the generator node information, bus node information, transmission line information between different component nodes and attribute type information of different component nodes based on the real-time operation data of the grid and form a graph digital feature; based on the graph digital feature, a digital structure diagram of the grid diagram is generated;

[0064] The virtual representation unit is used to input the digital structure diagram of the power grid diagram into the digital perception model to obtain the virtual representation μ i (t), μ i (t) represents the virtual representation of the i-th node at time t;

[0065] The sensing unit is used to determine whether the digital structure diagram of the power grid diagram of the i-th node at time t has changed based on the virtual representation. If so, the power grid operation has a fault; if not, the power grid operation has no fault.

[0066] Specifically, the representation of the graph digital features is GDT = {V, E, K}, where V = {v d1 ,...v dn ,v z1,...v zn} represents the set of generator node information and bus node information features, v d1 ,...v dn represents the information characteristics of the generator node, v z1 ,...v zn Represents the bus node information characteristics, E={e1,e1...e n} represents the state of the transmission line between different component nodes, K={k d1 ,...k dn ,k z1 ,...k zn} represents the attribute type of different component nodes, k d1 ,...k dn Indicates the attribute type of the generator node, k z1 ,...k zn Represents the property type of a bus node.

[0067] Specifically, the virtual representation μ obtained by the figure digital perception model i (t) = {s i (t), Δd i , M i , L i (t)}, where s i (t) represents the current operating status of the bus and generator, Δd i Indicates the data change value of the bus and generator, M i represents the historical failure status of the i-th node, L i (t) represents the current operating status of the component node adjacent to the i-th node;

[0068] Specifically, judging whether the digital structure diagram of the power grid diagram of the i-th node at time t has changed based on the virtual representation includes:

[0069] The virtual representation μ i Any quantity s in (t) i (t), Δd i 、M i , L i (t) is compared with the corresponding quantity at the previous moment. If there is a change, it means that the digital structure diagram of the power grid diagram of the i-th node at time t has changed, otherwise there is no change.

[0070] In some embodiments, Figure 3 As shown, the power grid fault location module includes:

[0071] The fault location unit is used to construct a fault location model F(Z,A), where Z represents the fault node feature and A represents the adjacency matrix of the relationship between the components; calculate the propagation rule of the fault location model, verify the adjacency matrix A of the relationship between the components of each fault node based on the verification matrix of the propagation rule, and determine whether the verified adjacency matrix A is consistent with the adjacency matrix A of the relationship between the components. If so, determine the number of layers of each fault node, and determine the position of the digital structure diagram of the power grid diagram where the fault is located based on the number of layers of each fault node, and obtain fault location information;

[0072] The operation stability prediction unit is used to obtain the characteristic matrix X of the voltage and current of each node at time t, and perform feedforward propagation calculation on the digital structure diagram of the power grid diagram based on the graph convolution network to obtain the node feature classification value Mp; based on the node feature classification value M p The global representation of the graph is obtained by using the adjacency matrix, and the average pooling method is used to classify the graph to obtain the classification result; the classification result is compared with the preset operation stability result to obtain the operation stability result.

[0073] Specifically, the expression of the propagation rule is:

[0074]

[0075] in, represents the operation of adding the identity matrix to the adjacency matrix of an undirected graph, represents the operation of adding the reference matrix to the adjacency matrix of the undirected graph, Q j is a trainable weight matrix, σ[.] represents the activation function, N j represents the validation matrix of the jth layer.

[0076] Specifically, the node feature classification value M p The expression is:

[0077]

[0078] Where X is the characteristic matrix of voltage and current of each node at time t; is the weight matrix from the input layer to the hidden layer, is the weight matrix from the hidden layer to the output layer, τ(.) represents the function that transforms the node vector into a graph vector representing the entire graph, and softmax(.) represents the classification function.

[0079] Specifically, an average pooling method is used to classify the graph to obtain a classification result; the classification result is compared with a preset operation stability result to obtain an operation stability result; wherein:

[0080] The preset operation stability results include one or more criteria to evaluate the operation stability. For example, it can be the average accuracy and standard deviation of the model on different test sets, or the sensitivity of the model to small changes in the graph structure.

[0081] Comparison: Compare the actual performance indicators of the model with the preset stability results to evaluate whether the model meets the requirements of operational stability.

[0082] In some embodiments, Figure 4 As shown, the plan intelligent generation module includes:

[0083] A preliminary plan generation unit is used to preprocess the fault location information and the operation stability results to obtain preprocessed data; the preprocessed data is input into a plan generation model based on deep learning, and a preliminary plan result is output; the plan generation model based on deep learning adopts any one of a recurrent neural network, a long short-term memory network, a Transformer or a convolutional neural network;

[0084] The final plan generation unit is used to combine the grid topology and the output of the generator set as the control action, and generate the final grid plan based on the reinforcement learning algorithm using the power big data and the preliminary plan results; the power big data includes the operating flow data of the transmission network at the current moment and the transmission network operating scenario simulation data.

[0085] Specifically, the execution process of the preliminary plan generation unit includes:

[0086] Data preparation: Fault location data: location of fault node, fault type, and fault time information; Operation stability data: stability indicators of the power grid under different operating conditions, including voltage, frequency, and power flow;

[0087] Preprocessing: cleaning, standardizing or normalizing the raw data so that the model can process it more effectively;

[0088] Model selection: one of recurrent neural network, long short-term memory network, transformer or convolutional neural network;

[0089] Model training:

[0090] Prepare training data: including grid status data under known fault conditions and corresponding emergency plan labels, use the training data to train the deep learning model, and adjust the model parameters through the back propagation algorithm to minimize the prediction error;

[0091] Monitor the training process: changes in the loss function, performance on the validation set to ensure that the model is not overfitting or underfitting;

[0092] Model evaluation and optimization: Evaluate the performance of the model on an independent test set, including accuracy, recall, and F1 score. Adjust the model architecture, hyperparameters, or training strategies based on the evaluation results to improve model performance.

[0093] These data are input into the model to generate preliminary scenario results.

[0094] Specifically, the execution process of the final plan generation unit is:

[0095] Step A, simulating the actual operation state of the transmission network in the power grid simulation environment. In the operation scenarios of renewable energy grid connection and load fluctuation, the simulation environment solves the operation flow data of the transmission network at the current moment through power flow calculation according to the current operation state of the transmission network, obtains the information of the topological structure of the transmission network, and uses the program interface to obtain the corresponding state space of the reinforcement learning algorithm;

[0096] Step B, using the simulation data of the transmission network operation scenario with a duration of about 100,000 time steps as input, running in the transmission network simulation environment, the power flow solver performs power flow calculation for each input scenario, and records each calculation result; when the transmission line overload occurs in the transmission network, a topology control action is selected in the initial state space based on the exhaustive method, and if the selected topology control action can restore the transmission network from the transmission line overload state to a safe operation state, then the action is saved and continuously simulated until the operation scenario iteration is completed, and the above selected topology control actions are merged as the action space for reinforcement learning algorithm training;

[0097] Step C, the state quantity given by the transmission network simulation environment should be observed in real time to determine whether it meets the preset mandatory constraints. If not, the current scenario simulation ends; if it does, the load rate of the transmission line is observed to obtain the load rate; among them, the preset mandatory constraints are: ① Ensure the power consumption on the load side of the transmission network, that is, the output of the generator set must be balanced with the load power consumption; ② At most one generator set can be shut down; ③ The changed topology structure cannot produce an island in the transmission network; ④ The power flow calculation of the transmission network always maintains convergence; If there is no transmission line with a load rate greater than 100%, no action is required; If there is a load rate greater than 100%, the intelligent agent selects the optimal topology structure control action according to the current strategy;

[0098] In step D, after selecting the action and returning to the interactive environment, the agent updates the neural network parameters according to the current state and reward function, and outputs the final power grid plan.

[0099] It is worth noting that the final grid plan generated will go through a system optimization and evaluation process. The system will adjust and optimize the plan based on historical experience, expert knowledge and actual conditions to improve its feasibility and effectiveness. At the same time, the system will also evaluate the plan to ensure that it meets the needs and expectations of grid operators.

[0100] Plan execution and monitoring: The generated plan will be provided to the grid operator for approval and execution. During the execution process, the system will monitor and provide feedback on the execution of the plan in real time. If any abnormal situation occurs or the plan needs to be adjusted, the system will promptly notify the operator and provide corresponding suggestions and support.

[0101] Plan learning and improvement: The system will record and learn each plan generated and its execution. By analyzing and learning these data, the system can continuously improve and optimize its plan generation algorithms and rules to improve their accuracy and efficiency.

[0102] It is worth mentioning that power big data includes transmission network operation scenario simulation data with a duration of about 100,000 time steps.

[0103] The present invention acquires real-time operation data of the power grid, constructs a digital structure diagram of the power grid diagram based on the real-time operation data of the power grid, inputs the digital structure diagram of the power grid diagram into a digital perception model for power grid situation perception, and discovers power grid operation faults; locates power grid operation faults and predicts power grid operation stability to obtain fault location information and operation stability results; substitutes the fault location information and operation stability results into a plan generation model based on deep learning, outputs preliminary plan results, and combines the power grid topology and the output of the generator set as a control action, obtains the final power grid plan based on the reinforcement learning algorithm using the power big data and the preliminary plan results, realizes automatic detection, rapid location and intelligent generation of plans for power grid operation faults, and improves the efficiency of handling power grid operation faults.

[0104] Example 2

[0105] like Figure 5 As shown, the difference between this embodiment and embodiment 1 is that this embodiment provides a method for intelligently generating a plan for power grid operation failure, which is applied to the system for intelligently generating a plan for power grid operation failure in embodiment 1; the method includes:

[0106] Step 1: obtain real-time operation data of the power grid, construct a digital structure diagram of the power grid diagram based on the real-time operation data of the power grid, input the digital structure diagram of the power grid diagram into a digital perception model to perceive the power grid situation, and obtain power grid operation faults;

[0107] Step 2: locate the power grid operation fault and predict the power grid operation stability, and obtain fault location information and operation stability results; the fault location information includes the location of the fault node, the fault type and the fault time information; the operation stability results refer to the stability indicators of the power grid under different operation states, and the stability indicators include voltage, frequency and power flow;

[0108] Step 3: Input the fault location information and operation stability results into the deep learning-based plan generation model to output the preliminary plan results; combine the grid topology and generator output as the control action, and generate the final grid plan based on the reinforcement learning algorithm using the power big data and the preliminary plan results.

[0109] As a further implementation, step 1 includes:

[0110] Step 11, acquiring real-time operation data of the power grid, the real-time operation data of the power grid including generator node information, bus node information, transmission line information between different component nodes and attribute type information of different component nodes;

[0111] Step 12, based on the real-time operation data of the power grid, extract the power grid features from the generator node information, bus node information, transmission line information between different component nodes and attribute type information of different component nodes and form a graph digital feature; generate a digital structure diagram of the power grid graph based on the graph digital feature; specifically, the graph digital feature is represented as GDT = {V, E, K}, where V = {v d1 ,...v dn ,v z1 ,...v zn} represents the set of generator node information and bus node information features, v d1 ,...v dn represents the information characteristics of the generator node, v z1 ,...v zn Represents the bus node information characteristics, E={e1,e1...e n} represents the state of the transmission line between different component nodes, K={k d1 ,...k dn ,k z1 ,...k zn} represents the attribute type of different component nodes, k d1 ,...k dn Indicates the attribute type of the generator node, k z1 ,...k zn Indicates the attribute type of the bus node;

[0112] Step 13: Input the digital structure diagram of the power grid diagram into the digital perception model to obtain the virtual representation μ i (t), μ i(t) represents the virtual representation of the i-th node at time t; specifically, the virtual representation obtained by the graph digital perception model is

[0113] μ i (t) = {s i (t), Δd i , M i , L i (t)}, where s i (t) represents the current operating status of the bus and generator, Δd i Indicates the data change value of the bus and generator, M i represents the historical failure status of the i-th node, L i (t) represents the current operating status of the component node adjacent to the i-th node;

[0114] Step 14, judging whether the digital structure diagram of the power grid diagram of the i-th node at time t has changed based on the virtual representation, if so, the power grid operation has a fault; if not, the power grid operation has not a fault; specifically, judging whether the digital structure diagram of the power grid diagram of the i-th node at time t has changed based on the virtual representation includes:

[0115] The virtual representation μ i Any quantity s in (t) i (t), Δd i 、M i , L i (t) is compared with the corresponding quantity at the previous moment. If there is a change, it means that the digital structure diagram of the power grid diagram of the i-th node at time t has changed, otherwise there is no change.

[0116] As a further implementation, step 2 includes:

[0117] Step 21, construct a fault location model F(Z,A), where Z represents the fault node feature, and A represents the adjacency matrix of the relationship between each component; calculate the propagation rule of the fault location model, verify the adjacency matrix A of the relationship between each component of each fault node based on the verification matrix of the propagation rule, and determine whether the verified adjacency matrix A is consistent with the adjacency matrix A of the relationship between each component. If so, determine the number of layers of each fault node, and determine the position of the digital structure diagram of the power grid diagram where the fault is located based on the number of layers of each fault node, and obtain fault location information;

[0118] Specifically, the expression of the propagation rule is:

[0119]

[0120] in, represents the operation of adding the identity matrix to the adjacency matrix of an undirected graph, represents the operation of adding the reference matrix to the adjacency matrix of the undirected graph, Q j is a trainable weight matrix, σ[.] represents the activation function, N j represents the verification matrix of the jth layer;

[0121] Step 22, obtain the characteristic matrix of voltage and current of each node at time t, perform feedforward propagation calculation on the digital structure diagram of the power grid diagram based on the graph convolution network, and obtain the node characteristic classification value Mp; based on the node characteristic classification value M p and the adjacency matrix to obtain the global representation of the graph, and use the average pooling method to classify the graph to obtain the classification result; the classification result is compared with the preset running stability result to obtain the running stability result;

[0122] Specifically, the node feature classification value M p The expression is:

[0123]

[0124] Where X is the characteristic matrix of voltage and current of each node at time t; is the weight matrix from the input layer to the hidden layer, is the weight matrix from the hidden layer to the output layer, τ(.) represents the function that transforms the node vector into a graph vector representing the entire graph, and softmax(.) represents the classification function.

[0125] Among them, the other steps correspond one-to-one to the corresponding module functions in Example 1, and will not be repeated one by one in this embodiment.

[0126] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0127] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes 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 generate 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 1A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0128] 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 comprising 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.

[0129] 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.

[0130] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. An intelligent generation system for emergency plans for power grid operation failures, characterized in that: The system includes: A power grid situation awareness module is used to obtain real-time power grid operation data, construct a digital structure diagram of a power grid diagram based on the real-time power grid operation data, input the digital structure diagram of the power grid diagram into a digital perception model to perform power grid situation awareness, and obtain power grid operation faults; A power grid fault location module is used to locate the power grid operation fault and predict the power grid operation stability, and obtain fault location information and operation stability results; the fault location information includes the location of the fault node, the fault type and the fault time information; the operation stability result refers to the stability index of the power grid under different operation states, and the stability index includes voltage, frequency and power flow; The intelligent plan generation module is used to input the fault location information and operation stability results into the plan generation model based on deep learning, and output the preliminary plan results; and combine the power grid topology structure and the output of the generator set as the control action, and use the power big data and the preliminary plan results based on the reinforcement learning algorithm to generate the final power grid plan.

2. The intelligent generation system for emergency plans for power grid operation failures according to claim 1 is characterized in that: The power grid situation awareness module includes: An acquisition unit, used to acquire real-time operation data of a power grid, wherein the real-time operation data of the power grid includes generator node information, bus node information, transmission line information between different component nodes, and attribute type information of different component nodes; A structure diagram construction unit is used to extract the grid characteristics from the generator node information, bus node information, transmission line information between different component nodes and attribute type information of different component nodes based on the real-time operation data of the grid and compose a graph digital feature; and generate a grid graph digital structure diagram based on the graph digital feature; A virtual representation unit is used to input the digital structure diagram of the power grid diagram into the digital perception model to obtain a virtual representation μ i (t), μ i (t) represents the virtual representation of the i-th node at time t; The perception unit is used to determine whether the digital structure diagram of the power grid diagram of the i-th node at time t has changed based on the virtual representation. If so, the power grid operation has failed; if not, the power grid operation has not failed.

3. The intelligent generation system for emergency plans for power grid operation failures according to claim 2 is characterized in that: The digital features of the graph are represented by GDT={V,E,K}, where V={v d1 ,...v dn ,v z1 ,...v zn } represents the set of generator node information and bus node information features, v d1 ,...v dn represents the information characteristics of the generator node, v z1 ,...v zn Represents the bus node information characteristics, E={e1,e1...e n } represents the state of the transmission line between different component nodes, K={k d1 ,...k dn ,k z1 ,...k zn } represents the attribute type of different component nodes, k d1 ,...k dn Indicates the attribute type of the generator node, k z1 ,...k zn Represents the property type of a bus node.

4. The intelligent generation system for emergency plans for power grid operation failures according to claim 2 is characterized in that: The virtual representation μ obtained by the digital perception model of the graph i (t) = {s i (t), Δd i , M i , L i (t)}, where s i (t) represents the current operating status of the bus and generator, Δd i Indicates the data change value of the bus and generator, M i represents the historical failure status of the i-th node, L i (t) represents the current operating status of the component node adjacent to the i-th node; Judging whether the digital structure diagram of the power grid diagram of the i-th node at time t has changed based on the virtual representation includes: The virtual representation μ i Any quantity s in (t) i (t), Δd i 、M i , L i (t) is compared with the corresponding quantity at the previous moment. If there is a change, it means that the digital structure diagram of the power grid diagram of the i-th node at time t has changed, otherwise there is no change.

5. The intelligent generation system for emergency plans for power grid operation failures according to claim 1 is characterized in that: The power grid fault location module includes: A fault location unit is used to construct a fault location model F(Z,A), where Z represents the fault node feature and A represents the adjacency matrix of the relationship between the components; calculate the propagation rule of the fault location model, verify the adjacency matrix A of the relationship between the components of each fault node based on the verification matrix of the propagation rule, determine whether the verified adjacency matrix A is consistent with the adjacency matrix A of the relationship between the components, and if so, determine the number of layers of each fault node, determine the position of the digital structure diagram of the power grid diagram where the fault is located based on the number of layers of each fault node, and obtain fault location information; The operation stability prediction unit is used to obtain the characteristic matrix of voltage and current of each node at time t, and perform feedforward propagation calculation on the digital structure diagram of the power grid diagram based on the graph convolution network to obtain the node feature classification value Mp; based on the node feature classification value M p The global representation of the graph is obtained by using the adjacency matrix, and the graph is classified by using the average pooling method to obtain a classification result; the classification result is compared with a preset operation stability result to obtain an operation stability result.

6. The intelligent generation system for emergency plans for power grid operation failures according to claim 5, characterized in that: The expression of the propagation rule is: in, represents the operation of adding the identity matrix to the adjacency matrix of an undirected graph, represents the operation of adding the reference matrix to the adjacency matrix of the undirected graph, Q j is a trainable weight matrix, σ[.] represents the activation function, N j represents the verification matrix of the jth layer; The node feature classification value M p The expression is: Where X is the characteristic matrix of voltage and current of each node at time t; is the weight matrix from the input layer to the hidden layer, is the weight matrix from the hidden layer to the output layer, τ(.) represents the function that transforms the node vector into a graph vector representing the entire graph, and softmax(.) represents the classification function.

7. The intelligent generation system for emergency plans for power grid operation failures according to claim 1, characterized in that: The plan intelligent generation module includes: A preliminary plan generation unit is used to preprocess the fault location information and the operation stability result to obtain preprocessed data; input the preprocessed data into a plan generation model based on deep learning, and output a preliminary plan result; the plan generation model based on deep learning adopts any one of a recurrent neural network, a long short-term memory network, a Transformer or a convolutional neural network; The final plan generation unit is used to combine the power grid topology and the output of the generator set as a control action, and generate a final power grid plan based on the reinforcement learning algorithm using the power big data and the preliminary plan results; the power big data includes the operating flow data of the transmission network at the current moment and the transmission network operating scenario simulation data.

8. The intelligent generation system for emergency plans for power grid operation failures according to claim 7, characterized in that: The execution process of the final plan generation unit is as follows: The actual operation status of the transmission network is simulated in the power grid simulation environment. In the operation scenarios of renewable energy grid connection and load fluctuation, the simulation environment solves the operation power flow data of the transmission network at the current moment through power flow calculation according to the current operation status of the transmission network, obtains the information of the topological structure of the transmission network, and uses the program interface to obtain the corresponding state space of the reinforcement learning algorithm; The transmission network operation scenario simulation data is used as input and runs in the transmission network simulation environment. The power flow solver performs power flow calculation for each input scenario and records each calculation result. When a transmission line overload occurs in the transmission network, a topology control action is selected in the initial state space based on the exhaustive method. If the selected topology control action can restore the transmission network from the transmission line overload state to a safe operation state, the action is saved and continuously simulated until the operation scenario iteration is completed. The selected topology control actions are merged as the action space for reinforcement learning algorithm training. The state quantity given by the transmission network simulation environment should be observed in real time to determine whether it meets the preset mandatory constraints. If not, the current scenario simulation ends; if it does, the load rate of the transmission line is observed to obtain the load rate; if there is no transmission line with a load rate greater than 100%, no action is required; if there is a load rate greater than 100%, the intelligent agent selects the optimal topology structure control action according to the current strategy; After selecting an action and returning to the interactive environment, the agent updates the neural network parameters based on the current state and reward function, and outputs the final power grid plan.

9. A method for intelligently generating emergency plans for power grid operation failures, characterized in that: The method includes: Acquire real-time operation data of the power grid, construct a digital structure diagram of the power grid diagram based on the real-time operation data of the power grid, input the digital structure diagram of the power grid diagram into a digital perception model to perform power grid situation perception, and obtain power grid operation faults; Positioning the power grid operation fault and predicting the power grid operation stability, obtaining fault location information and operation stability results; the fault location information includes the location of the fault node, the fault type and the fault time information; the operation stability results refer to the stability indicators of the power grid under different operation states, and the stability indicators include voltage, frequency and power flow; The fault location information and operation stability results are input into a plan generation model based on deep learning, and a preliminary plan result is output; the grid topology structure and the output of the generator set are combined as control actions, and the power big data and the preliminary plan result are used based on the reinforcement learning algorithm to generate the final grid plan.

10. The method for intelligently generating a plan for power grid operation failure according to claim 9, characterized in that: Acquire real-time operation data of the power grid, construct a digital structure diagram of the power grid diagram based on the real-time operation data of the power grid, input the digital structure diagram of the power grid diagram into a digital perception model to perform power grid situation perception, and obtain power grid operation faults, including: Acquiring real-time operation data of the power grid, wherein the real-time operation data of the power grid includes generator node information, bus node information, transmission line information between different component nodes, and attribute type information of different component nodes; Based on the real-time operation data of the power grid, power grid features are extracted from the generator node information, bus node information, transmission line information between different component nodes and attribute type information of different component nodes to form a graph digital feature; a digital structure diagram of the power grid graph is generated based on the graph digital feature; The digital structure diagram of the power grid diagram is input into the digital perception model to obtain a virtual representation μ i (t), μ i (t) represents the virtual representation of the i-th node at time t; Based on the virtual representation, it is determined whether the digital structure diagram of the power grid diagram of the i-th node at time t has changed. If so, a fault occurs in the operation of the power grid; if not, no fault occurs in the operation of the power grid.

Images