Font Size: a A A

Research On Coordinated Voltage Regulation Methods Of Active Distribution Networks Based On Deep Reinforcement Learning

Posted on:2023-06-16Degree:MasterType:Thesis
Country:ChinaCandidate:G BiFull Text:PDF
GTID:2532306836975339Subject:Logistics engineering
Abstract/Summary:PDF Full Text Request
To promote the green and low-carbon transformation of energy systems,China government has drawn up many guidelines on developing photovoltaic power generation.By the end of 2021,the total installed capacity of photovoltaic power generation has reached 306 million KW,of which the installed capacity of distributed photovoltaic power generation has reached 93.991 million KW,an increase of 85.72% over 50.61 million KW at the end of 2018.With the increasing penetration of distributed generation in distribution networks,distribution network systems gradually form active distribution networks(ADNs).However,the grid connection of large-scale distributed generation has changed the power flow distribution of traditional distribution networks and then distributed generation supplies power to this node or nearby nodes,resulting in safety problems such as frequent nodal voltage violation.Therefore,it is necessary to study the voltage regulation problem in ADNs.First,we studied the participation of distributed solar PV inverters in the cooperative voltage regulation of the active distribution network.We propose an attention-based Multi-Agent Proximal Policy Optimization algorithm(AMAPPO)combined with expert knowledge for collaborative voltage regulation in high percentage renewable energy distribution networks.The method has the stability brought by the Multi-Agent Proximal Policy Optimization algorithm with expert knowledge and the multi-resource coordination brought by the attention mechanism.The simulation results show that the proposed method can reduce the average voltage deviation by 0.404%-2.937%while guaranteeing the active output of the solar generator compared with the existing schemes.Second,the participation of source-load-storage multiple resources in the cooperative voltage regulation of the active distribution network is studied.Specifically,by taking static var compensator,distributed generation inverter,energy storage system and flexible load into consideration,we formulate the problem of minimizing voltage safety violation and active power reduction of distributed generation.Due to the interaction among voltage regulation resources and the unknown network model,it is very challenging to solve the above problem.To overcome the challenge,the above problem is reformulated as a Markov game.Then,we propose a multi-resource cooperative voltage regulation algorithm based on AMAPPO.The simulation results show that the proposed method can reduce the average voltage deviation by 0.239%-2.937% while guaranteeing the active output of the solar generator compared with the existing schemes.Finally,the work of this paper is summarized,and the future research work is prospected.
Keywords/Search Tags:distribution networks, multi-agent deep reinforcement learning, high-proportion distributed power, voltage regulation
PDF Full Text Request
Related items