Research On Residential Customer Baseline Load Estimation Method For Demand Response Based On Spatio-Temporal Correlation Of Load

As an important form of demand response(DR),incentive-based demand response plays a significant role in improving the flexibility of power system.As the economic compensation calculation basis,the accuracy of customer baseline load(CBL)estimation directly affects the interests of the provider and participants of the incentive-based DR program.The research on CBL estimation is helpful to improve the fairness of incentive-based DR implementation and promote the application and promotion of incentive-based DR.The research of this paper is focus on the error generation reason analysis,the spatio-temporal correlation based CBL estimation method,the precision and accuracy co-optimization based baseline load estimation model.The main contents are as follows:1.The error generation reason analysis of current estimation methods.The error generation mechanism of baseline load estimation method based on temporal correlation and spatial correlation is analyzed respectively.The estimation method based on temporal correlation uses the historical load data of customer for baseline estimation.When the customer load pattern changes suddenly on the DR day,such method shows low accuracy;The estimation method based on spatial correlation uses the load data of customers in the CONTROL group for baseline estimation.When the number of customers in the CONTROL group is insufficient or the DR customer shares low load similarity with customers in the CONTROL group,the accuracy of such method is low.It is concluded that CBL estimation using load data of only one dimension(temporal dimension or spatial dimension)is the cause of error.2.The spatio-temporal correlation based CBL estimation method.The spatio-temporal correlation based method contains three main steps: firstly,all CONTROL customers on DR event day are grouped into several clusters,each DR customer is matched to the most similar CONTROL cluster according to the load pattern similarity.Secondly,features consist of temporal features and spatial features are extracted from load data in two different aspects,the temporal fea tures are extracted from DR customer’s historical load data and the spatial features are extracted from the load data of the CONTROL group members on the DR event day.Thirdly,the variables selection and the linear function between variables and predictors are achieved by the LASSO algorithm,then the baseline can be estimated by the model.Simulation results indicate that this method has better comprehensive performance than traditional estimation methods.3.The precision and accuracy co-optimization model.A framework for baseline load estimation based on co-optimization of precision and accuracy is proposed.The framework uses an optimization model to solve the optimal input feature combina tion,and both precision and accuracy are considered in the objective function.At the same time,the input information for estimation is expanded from one-directional(from pre-DR event time to DR event period)to bi-directional time scales(from pre-and post DR event time to DR event period),so that more useful information can be utilized to further improve the estimation performance.Simulation results indicate that the estimation performance of this method is significantly improved than traditional estimation methods.