Volume 31: Clean Energy Technologies towards Carbon Neutrality

Long-short Term Full-process Forecasting of Solar Power and Inelastic Load Bingtao Zhang, Hongchuan Qin, Xi Li, Zhe Cheng, Renjie Zhou, Jian Li , Jianhua Jiang



The forecasting of photovoltaic (PV) power generation and inelastic load is of great significance for the stable and efficient power supply of a microgrid power system. However, most of the PV prediction research in literature is based on known solar radiation which is difficult to obtained. In order to relieve the uncertainty in a microgrid, this work proposes full-process forecasting methods based on solar energy and load periodic characteristics analysis. For solar energy, a combination of Gaussian process regression (GPR) and physical model methods is utilized for the short-term accurate forecasting. The long-term trend forecasting is realized based on a cascade online TS fuzzy model. For inelastic load, an improved online long-short term memory (LSTM) rolling forecasting method is proposed. Simulation results show that the GPR & physical model methods can reach or even exceed the existing accuracy, while the cascade online TS fuzzy model method can achieve 5.5% higher accuracy than existing algorithms. Compared with the current offline LSTM method, the accuracy of the online method can be improved by up to 4.92%.

Keywords PV power generation, inelastic load, GPR, online TS fuzzy model, LSTM

Copyright ©
Energy Proceedings