LSTM-based Distributed Photovoltaic Output Interval Prediction Method for Source-load Collaboration

With the continuous development of distributed energy, photovoltaic (PV) power emerges as a significant construction to tradeoff the balance of power supply and demand. Since the prediction of PV output is unavoidable, interval prediction becomes a powerful method to describe the output curve accurately. To this end, this paper proposes a long short term memory (LSTM)-based distributed PV output interval prediction method for source-load collaboration. It can correct the prediction error according to the spatiotemporal characteristic of PV power stations, and then calculates the interval prediction results under different confidence levels. Simulation results show that the proposed algorithm has the superior performance in the average width of the forecast interval, and the coverage width.