Forecast of Photovoltaic Plant Output Using Long Short-Term Memory (LSTM) Network: North Morocco Case Study

Abstract

AbstractDue to the volatile and intermittent nature of solar energy, the outputs of photovoltaic (PV) power plants present an uncertainty that can disrupt the security and stability of the electricity grid. Therefore, accurate forecasting of PV output power is of great importance to solve this problem and maintain the distribution network integrity. In this paper, the long short-term memory (LSTM) network, which is the memory of recurrent neural networks (RNNs), is proposed and applied to the data collected from the PV installation of the Faculty of Sciences and Technology of Tangier (FSTT). The LSTM architecture of this study has been improved according to different network simulations, all based on the root mean square error (RMSE), mean absolute error (MAE) and mean square error (MSE) performance evaluation indicators. The results showed that the LSTM network used gives good PV output power forecasting accuracy.KeywordsPV output power forecastingSolar energyRecurrent neural networks (RNNs)Long short-term memory (LSTM)Solar radiation