Research On Online Optimal Scheduling Of Residential Energy Consumption Considering The Operational Probability State Distribution Of Equipment

At present,the smart grid has become a major trend in the development of the world's power network.The home energy management system is an important part of the smart grid.It participates in demand response through real-time interaction with the power flow,information flow and business flow on the grid side.In order to ensure a comfortable living environment for users,the home energy management system encourages users to actively participate in the operation of the power system and respond to demand response signals.Users can use the home energy management system to schedule household power loads to reduce load peaks,reduce residential energy consumption and electricity costs.In this paper,the following research is performed on the load dispatching of household electrical equipment for residential users:This paper takes the user-side resource optimization scheduling as the core,and studies the operating characteristics and optimization scheduling problems of household electrical equipment.First,the composition and technical characteristics of the home energy management system are introduced.According to the time distribution characteristics and technical characteristics of the operating state of the consumer electrical equipment,the electrical equipment is classified and the operable actions and probabilistic operating states of the equipment are described.Secondly,based on the user's historical power consumption data,a resident power task scheduling optimization model based on the probability distribution of the operating status of the equipment was established in the context of time-of-use electricity prices.This paper proposes a probabilistic selection-based asynchronous advantage actor-critic algorithm that includes the ability to sense data.By combining the probability distribution of device operating states,the model is optimized and solved in an offline learning and online assessment mode.For a large number of user-side devices,multi-agents are embedded in the CPU,and the multi-threading function of the CPU is used to realize the online decision-making optimization of the residential multi-devices;the convergence ability,stability and timeliness of the method are analyzed.The results of the calculation examples show the effectiveness and correctness of the method for real-time residential load scheduling online.Finally,considering the uncertainty of grid-side energy prices and user demand.The long-term load scheduling problem for users is studied.The changes in electricity price information and load demand are modeled as Markov decision processes,and the interaction between users is described as a partially considerable random game.In order to make the problem easy to handle,this paper uses Markov perfect equilibrium with incomplete information to approximate the user's optimal scheduling strategy.An online load scheduling learning algorithm based on actor-critic is proposed to determine the optimal scheduling strategy for users.The simulation results show that compared with the benchmark without demand response,the peak-to-average ratios of the user's expected cost and total load are reduced by 28%and 13%,respectively.Compared with short-term scheduling strategies,long-term scheduling strategies can reduce users' expected costs by 11%.