Solar Power Forecasting Incorporated Energy Management Scheduling in a Microgrid

Abstract

Solar power is the key to a clean energy future. In a microgrid, powered by solar along with other power sources such as Diesel/Thermal/Hydropower plant, the solar power generated is uncertain. Since solar generation is variable with the unpredictability of sunlight in different places on earth, the ability to predict solar output will provide the solar energy industry with leverage. Forecasting solar power can help to predetermine the photovoltaic generation and avoid the imbalance between deficit generation and demand by planning for reserve supply ahead of time and hence to achieve increased efficiency in dispatching of the energy harvested. Predicting the solar power/irradiance using an artificial intelligence-based forecasting model could provide superior results and help secure an economic operation of the microgrid. This paper proposes a method of forecasting using afeed-forward neural network to train and predict solar power generation over the ThingSpeak IoT platform. This is very efficient and inexpensive when compared to other implementations which may require higher hardware complexity. Based on this forecasted value, scheduling of generation is carried out to meet the demand.