Day-ahead forecasting of solar irradiance using hybrid improved Cuckoo Search-LSTM approach

Solar energy has emerged as one of the popular alternative sources of energy as the demand for clean fuels gathers pace around the globe. However, solar irradiance is intermittent and stochastic by nature, directly affecting the performance of solar-based energy systems. Accurate solar irradiance forecasting can help in significant improvement in their design and operation. This work presents a novel forecasting approach for day-ahead Global Horizontal Irradiance using LSTM networks. For this purpose, historical hourly time-series data of meteorological variables including temperature, pressure, wind direction, relative humidity, dew point and, wind speed have been considered as input factors.The proposed model is a combination of LSTM and an improved variant of Cuckoo search algorithm. The model performance is evaluated using two statistical parameters: RMSE and, MAE. Low values of these parameters as compared with other models, indicate that the proposed model provides a better day-ahead forecasting of global horizontal irradiance.