Study On Interactive Operation Strategies For Multiple Agents In Power Distribution Systems With Multi-microgrids

As an individual power unit,a microgrid(MG)can effectively integrate the renewable energy(e.g.,wind and solar),demand response(e.g.,electric vehicle),local load,and storage system.As a part of the distribution system,multi-MGs can be connected interactively to the distribution network to improve the operational quality and economic benefit,which are beyond the limit of regional distribution regulation.Therefore,under the advocacy of low-carbon green and sustainable development,the multi-MG-based active distribution system will become an effective solution to the problem of the large-scale integration of distributed generators and the diversified development of loads in modern power systems.When the concept of multi-MG-based active distribution system is proposed,here come the following hotspots of MG cluster systems: how to detail the concept and establish the system;how to deal with the operational interaction among the MGs,distribution network,microsources,and users;how to effectively manage and trade the demand response and distributed generators;how to leverage multi-MG-based active distribution systems to realize the effective operation and economic benifit.Nowadays,the artificial intelligence(AI)technology marked by game theory and machine learning has been applied to each area including power systems.According to the structure of distribution networks and its multi-intelligent agents(MG,distribution network,and user),this dissertation divides the multi-MG-based active distribution system into the distribution level,MG level,and user level.Based on the MG and three core interactive relationships(MG-distribution network,MG-MG,and MG-user),several types of AI technologies and optimization methods are leveraged to build the interactive operation strategies of multi-agents from the top level(distribution level)to the bottom level(user level).The detailed work and contributions are listed below.1.The model for the interactive operation between MGs and the distribution network is built through bi-level gaming and the optimization of multi-objectives.The interactive strategy is proposed under the circumstances of dispatch and market,where the backup capacity of MG cluster is proposed.In the distribution network,the operation index such as power loss,voltage level,and tie-line fluctuation are taken as the control variables,which are achieved by nonlinear programming-based self-adapted NSGA-II algorithm.The affect of the capacity of distributed generators and the operation strategy on the operation index is analyzed under different scales of distribution systems.The effects of the interactive operation between MGs and the distribution network are compared under different operation strategies.The effectiveness of the proposed model is validated.2.Based on the above model,the cooperative/non-cooperative operation strategy for inter-MGs is proposed with the foundation of informatic game matrix(multi-agent technology)and dynamic decision making.The extent energy storage system is proposed to interact with other MGs.The stochastic programming is used to model the uncertainty of Intermittent distributed generators.The renewable integration in different regional MGs is optimized through hierarchical genetic algorithm and fuzy theory to improve the distribution reliability,the energy utility,and the operational cost.The influence of different operation strategies between MGs is compared.3.Based on the above model,according to the interactive operation strategy between the multi-MG system and the demand response,the inverse optimization is utilized to pull out the economic characteristic information which the users are featured with.The characteristic information is further optimized by a noisy function to best represent the users' economic behavior.The computational inefficiency and overfit of training caused by the large amount of high-dimensional data are solved to increase the precise and wideness of demand response prediction.The interactive Stakelberg model for multi-MG demand response is proposed.The predicted information is integrated into Stakelberg demand side management of distribution market.The dynamic transaction and power utilization are handled by a distributed hybrid Lagrange decomposition-gradient descent algorithm.The proposed model effectively predicts the user consumption,optimizes the distribution power transaction,and provides the demand response with a new regulation mode.4.Under the circumstance of multi-MG distribution power market,the interaction strategy for utility users and multi-MGs is proposed.In terms of the users,the electricity pricing plan decision system is built based on deep Q network.By the real-time interactive training with the environment and the optimization decision model,according to different utility plans based on the market price,the decision is made for the users to improve their satisfaction and the economic and energy-saving consumption.Aiming at solving the problem of this field,a comprehensive interactive operation system and an optimization methodology of operation(energy management and economic operation)are proposed with the consideration of different interactive objects(distribution system,MG,and user)and present popular and effective AI technologies.Simulation results and partial practical results are provided.The research achievement of this dissertation can effectively improve the power supply reliability,the power quality,and the economic benefit.It can also provide the future distribution system and the Energy Internet with a theoretical reference.