Photovoltaic power prediction for solar micro-grid optimal control

Abstract

In a solar micro-grid, a hybrid renewable energy system generates electricity for a building’s onsite use. The battery storage and the main power grid connection are used to facilitate the matching between the demand and production. To control energy flows optimally, an accurate day-ahead prediction of the photovoltaic (PV) panels output is required. However, this is a challenging task due to the fluctuating nature of solar radiation availability. The accuracy of the prediction is influenced by the modelling method and input parameters. In this study, the measured power and weather data is gathered from an experimental installation of PV panels to predict PV output for a 24-hours horizon in 15 min intervals. The multiple linear regression (MLR) and artificial neural network (ANN) methods are considered in the prediction modelling and compared using performance indicators. The micro-inverter technology is used to gather the individual PV panel output in addition to the overall system output. The results show that the modelling methods have different accuracy performances and the ANN model built with the individual PV output data results in the highest accuracy. Utilizing the micro-inverter technology leads to an advantage of having more accurate PV prediction for the control purpose.

Keywords

• Solar energy
• Photovoltaics
• Prediction model
• Multiple linear regression
• Artificial neural network
• Machine learning