Optimal Management Strategy For Home Microgrids Considering Demand Charge Tariff

The introduction of the Energy Internet concept provides a new perspective for energy analysis and energy management.Home microgrid energy management is the application and extension of energy internet energy in residential customer side.Based on the above background,this paper designs a multi-level network collaborative decision making framework for home energy management considering load classification and demand rate suppression based on the principle of multi-intelligent deep reinforcement learning algorithm to achieve the reduction of electricity cost and peak consumption reduction and smoothing for residential customers.First,based on the analysis of the home electricity consumption environment,a power model is established for home electricity equipment in residential customers and load classification is performed according to the operating characteristics of home electricity equipment.From the perspective of community demand,the soft limitation method of demand rate is designed to regulate the demand rate billing method.From the household demand perspective,variable allocation coefficients are set according to the load classification,where the allocation coefficients of power variable loads are designed based on the indoor temperature change rate,coupled with the residential customer comfort without dynamics modeling.Secondly,the home energy management strategies are designed based on the proximal policy optimization and multi-intelligent deep deterministic policy gradient algorithms,respectively.A multilayer perceptron neural network is designed to predict real-time electricity price changes in a rolling time window,and with real-time data on demand rates and PV generation as model state quantities,so that the training process incorporates multilevel network key decision information.In the single-intelligence optimization algorithm,the reward function is combined with a soft-limit approach to demand rates to reduce the cost of electricity for residential customers.In the multi-intelligence optimization algorithm,four types of loads are set as four collaborating intelligences in the multi-intelligence Markov decision process,and the demand rate is flexibly assigned to the reward function to achieve the peak reduction and cost reduction of residential electricity consumption.Finally,the effectiveness of this paper’s method is verified through case design and simulation analysis.It is shown that the introduction of demand rate can effectively suppress the peak electricity consumption;the flexible allocation of demand rate coefficients is compared with the fixed allocation,and it is shown that the flexible allocation of demand rate coefficients based on the multi-intelligent deep deterministic policy gradient algorithm can effectively smooth the peak electricity consumption of residential customers and reduce the demand rate;the soft restriction of demand rate is compared with the hard restriction strategy.The soft restriction of demand rate is compared with the hard restriction strategy,which shows that the soft restriction of demand rate based on the multi-intelligent deep deterministic strategy gradient algorithm can effectively reduce the electricity consumption at each moment in the household to meet the diversity of residential customers’ electricity demand,and thus achieve the smoothing of community electricity peak.