Development of a day-ahead solar energy forecasting model using seasonal ARIMA for economic load dispatch

Forecasting the short-term solar irradiance is of paramount importance for efficiently planning the electricity distribution in the transmission network to ensure the smooth running of solar power plants. In this paper, a seasonal ARIMA model is implemented for predicting the global solar irradiance in Mauritius for a time horizon of 24 hours. The proposed technique uses three days historical irradiance measurements to predict the next day values in time steps of one hour. The model has been observed to work well in both cloudy and sunny sky regimes, with a higher predictive accuracy noted on average for the sunny sky conditions. A maximum coefficient of determination attaining 97.7% has been observed for one of the sunny day test-case while a maximum value of 98.0% for the coefficient of determination has been noted for one of the cloudy day scenarios. Coupling of the developed model to a solar farm has been discussed with the suggested approach for regularization of electricity distribution in the grid, despite occasional drops in the electricity generation curve arising from the movement of journeying clouds above the facility.