Estimation method of meteorological sensitive load power based on correlation analysis and stacked auto-encoder
CHEN Yanxiang;QIN Chuan;JU Ping;ZHAO Jingbo;JIN Yuqing;SHI Jiajun;College of Energy and Electrical Engineering,Hohai University;Renewable Energy Power Generation Technology Engineering Research Center of Ministry of Education,Hohai University;State Grid Jiangsu Electric Power Company Research Institute;
The annual increase of meteorological sensitive load is an important reason for the increasing load of grid in summer and correct estimation of such load power is beneficial to the operation scheduling of power grid and the estimation of regional demand side response. An improved typical correlation analysis method is proposed,and a load-meteorology nonlinear correlation model is established,based on which,the meteorological sensitive load power in the historical load data can be calculated. An estimation model of meteorological sensitive load based on SAE( Stacked Auto-Encoder) is established. The unsupervised learning ability of SAE is utilized to extract the dimension reduction characteristics of daily load curve,and the calculative results of the correlation model is used as a label sample to train the full connection layers of the estimation model,thus the meteorological sensitive load power curve can be obtained directly by the daily load curve. Results of an example based on actual grid data verify the validity of the proposed methods.