The load consumption always changes with time due to factors such as human behaviors

weather conditions and continuous on/off switches of different kinds of loads

thus the relationship parameters of the load model with external factors is complex and nonlinear

and the parameters are not constants. Therefore

the load voltage characteristics exponential model in which time-varying is taken into consideration is proposed to describe the load and voltage variation with time in distribution network. Moreover

the improved recursive least square（RLS） algorithm based on the cumulant and variable forgetting factor is proposed to estimate time-varying exponential model parameters. The identification results of the improved algorithm and the traditional RLS algorithm are compared based on the digital simulation example and the actual sampling data of the power grid.From the angles of all kinds of error evaluation indexes

the estimation accuracy of the improved algorithm is better than that of the traditional RLS algorithm

which proves that the improved algorithm has good tracking performance and noise suppression performance.