Deep Learning-Based Real-Time Solar Irradiation Monitoring and Forecasting Application for PV System

Due to the high penetration of renewable energy sources (RESs), especially, solar power or photovoltaic (PV), the operation of the power system has coped with several severe technical issues such as supply-demand balance and low inertia. To mitigate these problems, PV power generation or solar irradiance forecast is of great importance. In the last decades, myriad forecasting methods have been introduced, yet these approaches do not concentrate on the real-time forecast. This paper proposes deep learning-based real-time solar irradiation monitoring and forecasting system with data collected from sensors before being stored in the database. Two predictive models, namely, Long Short-term Memory (LSTM) and Convolution Neural Network - Long Short-term Memory (CNNLSTM), are constructed for day-ahead solar irradiance forecasting. These predicted results are aggregated employing Linear Least Square (LLS) method into the selected forecast values and updated periodically. The forecasting results show that the selected predicting values improve significantly compared to individual ones.