Research On Demand Response Capability Evaluation Method Based On Data Driv

Demand response has received more and more attention from scholars in recent years as an effective means to weaken the peak-to-valley difference of power system,optimize the stability of power system operation,and improve the ability of power system to consume new energy.The implementation of demand response needs to be based on customer-side responsiveness,and then reasonably formulate demand response strategies and mobilize demand-side resources,so a reasonable responsiveness assessment method is particularly important.Demand response capability assessment is to evaluate the electricity consumption that can be cut or can be shifted to other periods by power users during demand response hours.This paper conducts a methodological study of demand responsiveness assessment around flexible loads and the relationship between regional electricity consumption and electricity prices.A residential demand responsiveness assessment method based on probabilistic load decomposition is proposed,aiming to quickly assess the transferable electricity consumption of flexible loads with simple operating states within customers,and to provide conditions for large-scale responsiveness assessment.A non-intrusive load decomposition model is first established using a Gaussian probabilistic model and a weighted combination of multiple power features;then the electricity consumption that can be transferred by residential customers during demand response hours is evaluated based on the operating states of the flexible loads at each moment obtained from the decomposition of the total electricity consumption information of the customers;finally,experiments are conducted on the data set,and the results show that this model is faster in decomposition and evaluation with guaranteed accuracy.A regional residential demand responsiveness assessment method based on XGBoost non-intrusive load decomposition is proposed,aiming to efficiently assess the power that can be cut from flexible loads with complex operating states within the customers.A non-intrusive load decomposition model based on the XGBoost regression algorithm is firstly developed,and the model is optimally tuned using a simulated annealing algorithm.The final experimental results on the REDD public dataset and a large number of residential customers’ data show that this method is more accurate and faster in load decomposition for loads with complex operating states,with less error in responsiveness assessment.A regional demand responsiveness assessment method based on tariff sensitivity and graded price elasticity is proposed,aiming to evaluate the overall regional demand responsiveness from a more macroscopic perspective by directly using the relationship between real-time tariff and electricity consumption changes in the electricity market.Firstly,the regional electricity price sensitivity index is proposed based on the cosine similarity,which is used to screen the regions with more positive response;then,considering the characteristics of regional electricity price and electricity consumption fluctuation,the k-means clustering algorithm is used to grade the regional electricity price and consumption;finally,the price elasticity of demand coefficients of each level are used to calculate the electricity consumption that the region can cut during the peak hours of electricity consumption.Using the PJM electricity market data for validation,the results show that this method can calculate the regional response electricity more accurately and reasonably assess the regional response capability.