Abstract: The paper aims to develop a model for forecasting energy production from a photovoltaic (PV) system given a short range of data. Real data for a recently installed PV system is available. An analytical model based on time series analysis has been developed. A feature of modeling is the insufficient amount of available data needed to make predictions. This limitation necessitates the use of a relatively simple forecasting model, such as the linear autoregressive model AR(1). To achieve better forecasting accuracy, a set of technologies, such as optimization, the least squares method, model predictive control, and a sliding procedure for shifting the beginning of historical data, was used. A comparison between actual and forecasted values has been made where possible. The proposed innovative approach is a good tool for planning energy production from photovoltaic systems. This can be beneficial for the declared power generation of an energy supplier, which has a cost-effective result. In addition, the forecast of photovoltaic system generation can help in the correct design of the inverter and battery parameters of the PV system.

Keywords: Analytical modeling, forecasting, linear regression models, optimization, photovoltaic energy production, time series

Pages: 503-510