Dynamic Operation Regulation Of Transmission Network Based On Deep Reinforcement Learning

In the process of pursuing the " double carbon " development goal,the huge fluctuation of renewable energy on the power side and the rapid change on the load side have put forward new problems and challenges to the operation regulation of transmission network.Therefore,this paper takes the topology structure of transmission network as the regulation variable,combines the regulation of power generation output and the regulation of energy storage unit,and realizes the regulation of transmission network in a fast,economical,safe and effective way.However,considering the excessive regulation variables and nonlinear characteristics of transmission network dynamic regulation problem,the traditional optimization algorithm is difficult to complete the solution in a limited time.In recent years,deep reinforcement learning technology is closely combined with power system research,which provides more sufficient technical support for the dynamic regulation of transmission network.Based on this,this paper adopts deep reinforcement learning algorithm to explore the dynamic regulation scheme of transmission network.Firstly,the typical substation main wiring structure and its bus-splitting capability are analyzed.Based on the main wiring structure of the substation double-bus 3/2 circuit breaker which is widely used in the actual transmission network,a bus-splitting model is established.This paper introduces the concepts and algorithms of deep neural networks and deep reinforcement learning,which provides the theoretical basis for the following chapters.Secondly,the topology control strategy of transmission network based on deep reinforcement learning is explored.The operation constraint conditions and regulation cost of transmission network are specified.The incentive function is constructed based on the load rate of transmission lines.The combination of electrical information and topology structure information is taken as the state,and the topology structure of transmission network is taken as the regulation variable.Using A3C algorithm,a 36-node transmission network is taken as an example to verify the effectiveness of the regulation strategy.Then,the regulation scheme of transmission network under N-1 line fault scenario is considered.Besides topology regulation,generator set regulation is also added.Based on PPO algorithm and A3C algorithm,the regulation scheme is verified by a 36-node transmission network example,and the test results are compared with other schemes.Finally,a transmission network regulation scheme based on deep reinforcement learning is explored by combining topological structure regulation,generator set regulation and energy storage unit regulation.PPO algorithm and A3C algorithm are used to verify the effectiveness of the regulation scheme,taking the larger 118-node transmission network with energy storage units as an example.In this paper,the structural characteristics,operation characteristics and safety criteria of transmission network are fully considered,and the dynamic control of transmission network is realized based on a variety of control means.The example results show that the regulation strategy explored in this paper effectively improves the safety of transmission network,and has good regulation efficiency and economy.Relevant research work can provide technical support for real-time regulation of intelligent power grid.