Thermostatically Controlled Load Aggregated Power Prediction Based on CLNet

Abstract

The user side thermostatically controlled load (TCL) scheduling is flexible and has little influence on the user comfort level. However, due to the decentralized distribution of TCL, it is difficult for the dispatching center to directly obtain its aggregated power and response potential. In order to guide TCL to participate in the grid regulation operation, the aggregation model of them is established by using a deep learning algorithm combining convolutional neural network and LightGBM, and the estimation value and upper and lower limit range of the aggregated power can be easily determined. Based on the model, a new evaluation method of TCL's aggregated response potential was proposed. The aggregated response potential and distribution characteristics of TCL were also evaluated.