CN-121984129-A - Power grid dispatching method, control device and storage medium

CN121984129ACN 121984129 ACN121984129 ACN 121984129ACN-121984129-A

Abstract

The invention provides a power grid dispatching method, a control device and a storage medium, and belongs to the technical field of power grids. The method comprises the steps of inputting acquired operation data into a multi-target optimization model to output an optimization scheduling variable, checking constraint conditions of the optimization scheduling variable, generating a scheduling scheme of a target power grid when the optimization scheduling variable meets preset constraint conditions, issuing a scheduling scheme instruction, monitoring operation indexes of the target power grid in the process of executing the scheduling scheme instruction by the target power grid, calculating deviation between the monitored operation indexes and an optimization target, and adjusting the scheduling scheme of the target power grid when the deviation exceeds a preset threshold. According to the invention, the generation time of the scheduling scheme is greatly shortened through the technical path of offline training and online reasoning of the neural network, and the real-time response of the dynamic change of the power grid is realized.

Inventors

SU WEN
MA YABIN
CHEN QINGTAO
SHAO JUNWEI
LI BO

Assignees

电力规划总院有限公司
国网安徽省电力有限公司电力科学研究院

Dates

Publication Date: 20260505
Application Date: 20260203

Claims (10)

1. The power grid dispatching method is characterized by comprising the following steps of: Acquiring operation data of a target power grid, and inputting the acquired operation data into a trained multi-target optimization model for power grid dispatching so as to output an optimization dispatching variable; Performing constraint condition verification on the output optimized dispatching variable, generating a dispatching scheme of the target power grid when the output optimized dispatching variable meets a preset constraint condition, and issuing a dispatching scheme instruction; Monitoring the operation index of the target power grid in the process of executing the dispatching scheme instruction by the target power grid, and And calculating the deviation between the monitored operation index and the optimization target, and adjusting the scheduling scheme of the target power grid when the calculated deviation exceeds a preset threshold.
2. The grid dispatching method of claim 1, wherein prior to the obtaining the operational data of the target grid, the grid dispatching method further comprises: based on the improved BP neural network, constructing a multi-objective optimization model for power grid dispatching, and constructing a training set and a verification set; Configuring an objective function of the multi-objective optimization model and embedding the constraint into a loss function, and And training the multi-objective optimization model based on the constructed training set and verification set, the configured objective function and the loss function to obtain a trained multi-objective optimization model.
3. The power grid dispatching method according to claim 2, wherein the constructing a multi-objective optimization model for power grid dispatching based on the improved BP neural network comprises: Determining an input layer, a hidden layer and an output layer of the BP neural network, and And improving a weight updating mode of the BP neural network by utilizing the momentum factor to construct a multi-objective optimization model for power grid dispatching based on the improved BP neural network.
4. The power grid scheduling method of claim 2, wherein the configuring the objective function of the multi-objective optimization model comprises: the objective function of the multi-objective optimization model is configured with the aim of minimizing the network loss and the total running cost of the power grid, Wherein, based on a node voltage method, the power loss of each branch of the power grid is summed to obtain the power grid loss, And obtaining the total running cost of the power grid based on the fuel cost of the generator, the new energy power-discarding cost and the running cost of the energy storage system.
5. The power grid scheduling method of claim 2, wherein the training the multi-objective optimization model comprises: In the process of training the multi-objective optimization model, updating the weight and the threshold of the multi-objective optimization model through a back propagation algorithm to obtain the trained multi-objective optimization model.
6. The power grid dispatching method of claim 1 or 2, wherein the preset constraint conditions include one or more of a power balance constraint, a generator output constraint, a new energy grid-connected power constraint, an energy storage system constraint, a node voltage constraint and a branch power constraint.
7. The power grid dispatching method according to claim 1, wherein when the constraint check is performed on the outputted optimized dispatching variable, the power grid dispatching method further comprises: and when the output optimized dispatching variable does not meet the preset constraint condition, adjusting the boundary of the optimized dispatching variable.
8. The grid scheduling method according to claim 1, wherein adjusting the scheduling scheme of the target grid when the calculated deviation exceeds a preset threshold value comprises: Reconstructing a training set and a verification set of the multi-objective optimization model based on the acquired operation data; Retraining the multi-objective optimization model based on the reconstructed training set, verification set and preconfigured target function to obtain a retrained multi-objective optimization model; Outputting new optimized schedule variables using the retrained multi-objective optimization model, and And performing constraint condition verification on the output new optimized scheduling variable, and generating a new scheduling scheme when the output new optimized scheduling variable meets a preset constraint condition.
9. A control device, characterized in that it comprises a memory, a processor and a computer program stored on the memory and executable on the processor, the processor executing the computer program to implement the grid dispatching method according to any one of claims 1-8.
10. A machine-readable storage medium having stored thereon instructions that cause a machine to perform the grid dispatching method of any one of claims 1-8.

Description

Power grid dispatching method, control device and storage medium Technical Field The invention relates to the technical field of power grids, in particular to a power grid dispatching method, a control device and a storage medium. Background Aiming at multi-objective optimization requirements (including network loss minimization, cost minimization and the like, for example) in power grid operation, constructing a mathematical model comprising constraint conditions, simulating natural rules such as biological evolution, group cooperation and the like through a heuristic algorithm, searching an optimal scheduling scheme in a random search and iterative optimization mode, and finally outputting scheduling variables such as generator output, new energy grid-connected power and the like. The method is characterized by relying on manually preset search rules (such as cross variation of GA, particle flight of PSO and the like), and no data-driven self-learning capability is needed. The related technology has the defects of serious shortage of real-time performance and incapability of adapting to dynamic scheduling requirements of a power distribution network. Optimization techniques, mainly represented by Particle Swarm Optimization (PSO) and Genetic Algorithm (GA), require multiple iterative searches to find the optimal solution in the solution process. The method causes the dispatching response speed to be far behind the actual running requirement of the power grid, so that the real-time requirement of dynamic dispatching of the power grid cannot be met. Disclosure of Invention The embodiment of the invention aims to provide a power grid dispatching method which can solve the problems that the real-time performance is seriously insufficient and the power grid dispatching method cannot adapt to the dynamic dispatching requirement of a power distribution network. In order to achieve the above purpose, the embodiment of the invention provides a power grid dispatching method, which comprises the steps of obtaining operation data of a target power grid, inputting the obtained operation data into a trained multi-target optimization model for power grid dispatching to output an optimization dispatching variable, checking constraint conditions of the output optimization dispatching variable, generating a dispatching scheme of the target power grid when the output optimization dispatching variable meets preset constraint conditions, issuing a dispatching scheme instruction, monitoring operation indexes of the target power grid in the process of executing the dispatching scheme instruction by the target power grid, calculating deviation between the monitored operation indexes and the optimization target, and adjusting the dispatching scheme of the target power grid when the calculated deviation exceeds a preset threshold. Optionally, before the operation data of the target power grid is acquired, the power grid dispatching method further comprises the steps of constructing a multi-target optimization model for power grid dispatching based on an improved BP neural network, constructing a training set and a verification set, configuring an objective function of the multi-target optimization model, embedding constraint conditions into a loss function, and training the multi-target optimization model based on the constructed training set and the verification set, the configured objective function and the loss function to obtain a trained multi-target optimization model. Optionally, the multi-objective optimization model for power grid dispatching is constructed based on the improved BP neural network, and comprises the steps of determining an input layer, a hidden layer and an output layer of the BP neural network, and improving a weight updating mode of the BP neural network by utilizing momentum factors so as to construct the multi-objective optimization model for power grid dispatching based on the improved BP neural network. Optionally, the configuring the objective function of the multi-objective optimization model includes configuring the objective function of the multi-objective optimization model with the aim of minimizing the power grid loss and minimizing the total running cost of the power grid. The method is based on a node voltage method, and electricity is summed through the loss of each branch of the power grid, so that the power grid loss is obtained, and the total operation cost of the power grid is obtained based on the fuel cost of the generator, the electricity discarding cost of new energy and the operation cost of an energy storage system. Optionally, the training of the multi-objective optimization model comprises updating weights and thresholds of the multi-objective optimization model through a back propagation algorithm in the process of training the multi-objective optimization model so as to obtain the trained multi-objective optimization model. Optionally, the preset constraint condition includes