Analysis Model Of Residential Users’ Adjustable Potential Considering User Behavior Mode

As the scale of new energy entering power grid gradually increases and the social electricity consumption increases year by year,the absorbing ability and flexible adjustment ability of the power system face greater challenges.It has become a reliable and effective power grid regulation measure to ensure the balance of power supply and demand and maintain the stability of power grid through the adjustment of demand-side resources.It also plays a role in guiding users to use electricity reasonably and optimizing the operation efficiency of power system.As the object of power grid regulation,resident users have considerable adjustment potential.Exploring their adjustable potential and identifying their potential types can effectively assist the power grid and load aggregators to integrate user-side resources,promote the consumption of new energy and reduce the waste of power resources.In this thesis,the research on the adjustable potential prediction and potential type identification of resident users is carried out as follows:Firstly,the behavior pattern of electricity consumption is analyzed for residential users.From the natural environment,time and date,consumer electricity price regulation,household situation and interpersonal influence,the important factors affecting the behavior of residents are analyzed.According to the time scale,the residential load characteristics are divided into daily load characteristics,monthly load characteristics and annual load characteristics.The daily load characteristics of residents are analyzed,and the electricity consumption behavior patterns of residential users are obtained based on the clustering of residents’ daily load curves.Then,an adjustable potential prediction method for resident users is proposed,which combines LGBM and BiLSTM,with the inverse error method to make a weighted combination prediction.This method corrects the large error in the predicted value of a single model,so as to reduce the prediction error of a single model.Parameters of BiLSTM are optimized through experiments,the optimized LGBM-BilstM model and a variety of comparison algorithms are used to predict the adjustable potential,and a comprehensive comparative analysis is carried out through multiple evaluation indicators.Meanwhile,the adjustable potential prediction experiments of single resident user and aggregate resident users are carried out.The simulation results show that the LGBM-BiLSTM model has good performance in predicting the adjustable potential of resident users.Finally,both the consumption behavior pattern type and daily load characteristics of residents are used as the characteristic basis for the prediction of residents’ potential type,and a recognition method of residents’ adjustable potential type based on VMD and GRU is proposed.The variational mode decomposition VMD algorithm is used to decompose residents’ daily load curve to obtain multiple modes and extract time-frequency domain features to expand the feature space.Then the gated recurrent unit GRU is used for classification to identify the potential types of resident users.The VMD-GRU model is comprehensively compared with a variety of algorithms,and the classification accuracy of VMD-GRU model can reach 99.58%,which is higher than the comparison model.Simulation results show that the proposed VMD-GRU model is more accurate in identifying potential types of residents.