Power output forecasting of solar photovoltaic plant using LSTM

Abstract

Renewable energy sources are gaining popularity, where solar photovolaics (PV) being the most preferred option due to its cleanliness, affordability, and abundance. The energy output of solar PV is primarily based on temperature & irradiance. Therefore, a weather-based intelligent model is needed for estimating solar energy output to fulfil energy demand and decision making. Predicting PV power output is essential for energy management, security, and operation. In addition to enhancing the output efficiency of PV power plants, the power grid's stability can be enhanced by enhancing the efficacy of PV power plants' electricity generation. This work focuses on LSTM and BPNN for forecasting solar plant power output and it is observed that their findings are virtually compatible with realistic power production in terms of MAE, MAPE, RMSPE, and R2 score. LSTM model comparisons with different layers for each weather season are also analysed. Comparing the extent of errors in the LSTM and BPNN models reveals that LSTM provides more accurate predictions.

Keywords

• Solar PV
• Machine learning
• RNN
• LSTM
• Mean absolute error (MAE)