Research On Real-time Energy Optimization Management Method For Prosumer Considering Demand Response

As distributed energy resources(DERs)such as distributed photovoltaic generation and electric vehicles are connected to the distribution network on a large scale,more and more users are transforming from traditional electricity consumers to prosumers who can provide and receive electricity.Prosumers actively adjust energy consumption behavior of internal large-scale DERs in real time based on price or incentive demand response information,which not only helps to weaken the impact of the centralized access of a large number of DERs on the distribution network,but also promote the successful achievement of the power system’s carbon peak and carbon neutral goals.However,the current research on real-time energy management of prosumers has problems such as insufficient consideration of the requirements of different demand response types and the characteristics of different demand response signals,and the efficiency of real-time calculation and solution still needs to be improved.To takle the above problems,based on the implementation mechanism of different types of demand response,this paper separately researches the real-time energy optimization management models and methods of the prosumer in response to price and power demand signals,and explores the ability to quickly and accurately solve the energy optimization management of the DERs within the prosumer.First of all,this paper clarifies that the research object is commercial office building prosumer who are equipped with rooftop photovoltaic generation and multiple electric vehicle charging piles.The charging and discharging power of electric vehicles is used as a dispatchable resource for the prosumer energy management problem.Second,based on the mixed integer programming and rolling optimization methods,this paper establishes a basic model of prosumer energy optimization management that takes into account demand response,as the basis and comparison of follow-up research.Then,in order to achieve the real-time automatic response to the demand response price signal,a long short-term memory neural network(LSTM)based machine learning method is established to model the prosumer energy management problem,and the simulation verification of the accuracy and efficiency of the proposed method in real-time solution.The proposed LSTM network based machine learning method adopts an optimized scheduling mode of offline training and online execution,which can not only greatly reduce the pressure of prediction and calculation in real-time,but also achieve better performance than traditional rolling optimization method.The proposed LSTM network based method helps prosumer with multiple DERs realize real-time energy optimization management that quickly and automatically responds to demand response price signals.Finally,aiming at the fast and accurate tracking of demand response power signals,a convex optimization model is proposed,and the accuracy and efficiency of the proposed method in real-time solutions are analyzed through simulation.The established convex optimization energy management model of prosumer can still ensure that the optimal solution satisfies the constraints after relaxing the non-convex constraints in the basic energy management model,but greatly reduces the solution time,and has an advantage in application that the tracking error of the power signal is controllable when schedulable capacity is sufficient.The proposed convex model based method helps prosumer with multiple DERs to achieve real-time energy optimization management that responds to power signals quickly and accurately.