Multi-layer Perceptron based Photovoltaic Forecasting for Rooftop PV Applications in Smart Grid

In the smart grid, a consumer can choose either to expend energy from the grid or vend energy back to the grid. On this principle, for profit maximization based on the current day's electricity selling price, a smart home with a rooftop photo voltaic (PV) system can determine whether the yielded energy during the day should be expended by the consumer at night or stored in a storage cell for the purpose of selling on the following days. For this, the smart home system needs to predict the electricity generation of the PV system for making a better decision. In this paper, we propose a Multi-layer Perceptron based PV forecasting method for the PV systems. The system is trained with the historical data of irradiance, temperature, solar zenith angle, wind speed, and relative humidity. The performance is evaluated by the cross validation and testing with the help of real world PV data. Being light weight and high accuracy, our proposed algorithm can be a potential method for the roof top PV forecasting in the smart home applications.