Demand Response Uncertainty Considered Residential Strategic Bidding Optimization For Load Aggregators

Nowadays,the electricity demand is still growing rapidly,and constructing a new power system driven by renewable system has become the primary task of our country’s energy industry development,the proportion of distributed energy will appear a dramatical increase trend in power system,which brings greater challenges to its safe and stable operation.As one of the most significant applications in smart grid,demand response is applied to adjust electricity demand curves through ecnomic ways,which encourages user-side resources transfer from peak periods to other periods autonomously and relieve power strain.For this reason,local DR resources can more flexibly and economically realize the real-time balance requirements of large power grids in a certain region.In conclude,the full use of user-side resources can effectively alleviate power supply uncertainty problem.With the development of electricity market,load aggregators and virtual power plants,have entered into power market to participate in trade.On the one hand,as an agency for residential users,load aggregator needs to aggregate DR resources and provide electricity sales services to all resident users they managed,on the other hand,load aggregator has to join in the DR transaction competition as a market union entity.However,due to the complexity of describing DR behaviors,which involves predicting users’ electricity consumption habits,differentiating response subsidy sensitivities and designing the backup schemes in extreme cases,designing DR models and optimizing DR strategies have become a meaningful research topic.This paper takes maximizing the expected revenue of load aggregators participating in DR programs as the objective function and proposes a strategic bidding considering the uncertainty of demand response,then utilizes advanced soft actor-critic algorithm to optimize DR strategies.Under the circumstance that the response willingness response situation of lots of resident users is uncertain and the users’ electricity consumption behaviors also fluctuates greatly,this paper firstly models the internal charateristics of DR resources managed by load aggregators,and on the basis of probability prediction model of user group response ability,constructs the preliminary bidding strategy model,including DR declaration estimation model,incentive subsidy estimation model,load transfer income estimation model and the reliability evaluation model.Secondly,considering the competition factors among load agents in DR market,based on the preliminary bidding strategy,take the system operators’ evaluation model for DR bidding strategy as the measurement index and improve the strategy biddings for different styles of load aggregators respectively to guarantee the probability of winning DR bids,which take the internal and external factors into account at the same time.Furthermore,the backup schemes to relieve the shortage or excess response volume cases are also built in the amended strategic bidding.In order to overcome the limitations of some traditional algorithms in dealing with multi-dimensional continuous variable programming problem,such as low efficiency and high requirements on the number of data samples,this paper has utilized deep reinforcement learning to optimize the decision-making process of DR strategy,and verified the effectiveness and superiority of DRL algorithm in case studies.