Global Solar Radiation Prediction. Using Artificial Neural NetworksCase study: the city of Er-Rachidia, Morocco

      Аннотация: This book aims to develop a new RSG preversion model based on deep learning. This approach will be able to boost the prediction accuracy of RSG data. Subsequently, the present proposed algorithm effectively handles the dynamics of our targeted weather component by integrating a recurrent and dynamic model named LSTM neural network with an autoregressive process. The raw data available for training this model is divided into two sets, the first is used for the training phase while the second is reserved for testing. The specific objective therefore is to generate accurate semi-hourly RSG forecasts at the level of the city of Er-Rachidia, MOROCCO (Latitude: 31°55′53″N; Longitude: 4°25′35″ W; Elevation: 1039 m), while adopting a powerful learning algorithm named Adam. The indices and results established in this study demonstrate the robustness and confidence that can be adopted to this model which can provide power system managers with reliable forecasts to ensure better management of solar energy and power service systems.

Ключевые слова: Artificial intelligence, Machine Learning, Deep Learning, Artificial Neural Networks, LSTM, Forecasting, Solar Radiation