Reactive voltage control method based on multi-time-scale multi-agent deep reinforcement learning

Abstract

The invention relates to a power system operation and optimization technology, and aims to provide a reactive voltage control method based on multi-time-scale multi-agent deep reinforcement learning. According to the method, on-load voltage regulation taps, capacitor banks and energy storage in photovoltaic power, draught fans and loads are all defined as agents, a method based on reinforcement learning is applied to the reactive power optimization problem, and a controller is allowed to learn a control strategy through interaction with a simulation model of a similar system. The action variable of reactive power regulation equipment interacts with a power distribution network environment, and the agents can finally realize the optimal response to the external environment, so that the maximum return value is obtained. The neural network method is used to analyze and fit a strategic function and an action value function of the agents, and the training process does not depend on prediction data results and accurate power flow modeling. By using the two-time-scale reactive power optimization method, the network loss can be smaller, the voltage stabilizing effect is better, and a more remarkable effect on improving the safety and reliability of the distribution network is achieved.

Description

technical field [0001] The invention relates to the technical field of power system operation and optimization, in particular to a reactive power and voltage optimization method based on multi-time scale multi-agent deep reinforcement learning. Background technique [0002] As a large number of renewable distributed power sources are connected to the distribution network, random fluctuations in the output of wind power equipment and photovoltaic equipment, and uncertain load fluctuations will lead to large voltage fluctuations in the operation of the distribution network, voltage cross-line, and network loss. Improvement and other issues will affect the power quality. [0003] The goal of reactive power optimization in distribution network is to effectively ensure the voltage stability of each node, reduce voltage fluctuation and reduce network loss of the power grid under the constraints of safe operation of the power grid. The reactive power optimization of distribution n...

AI Technical Summary

Problems solved by technology

Although most of the existing model-based reactive power and voltage optimization methods can effectively suppress the problem of voltage exceeding the limit, they largely rely on accurate models and forecast data, and the calculation speed is slow and easy to fall into local optimum.

Images