Interval forecasting of solar PV power generation

Abstract

Solar photovoltaic power generation is more and more popular in recent year. However, when solar PV power stations connect to the grid, it is important to predict the solar PV power generation for the stability of the whole system. Based on such situation, this dissertation provides and validates a method by using a neural network algorithm which is an LSTM model to make prediction, and proposes some observation results from the simulation case. Briefly, in this case study, it is shown that different time steps can lead to different accuracy of the results, so it is important to choose suitable time steps to forecast. In addition, the results of this prediction model are also related to the correlation of input features, and it is necessary to select the input features when training the model in order to make it more effective.