A Statistical Photovoltaic Power Forecast Model (SPF) based on Historical Power and Weather Data

Abstract

In this paper, we present an empirical statistical model, photovoltaic (PV) power forecast model (SPF) based on the historical power and weather type data. Firstly, the historical power data is grouped into 4 levels, according to the influence of different weather types on the penetration rate of solar radiation. Secondly, K-means and the parabolic least square fitting are used to model the PV power of each level. Thirdly, the variations to the initial fitted power are further investigated by analyzing the probability distribution of the residuals. The final power forecast is validated by a ten-day forecast for three stations, and the mean absolute percentage error (MAPE) value are 12.85%, 11.92%, 16.69%, respectively. Superior to the PV power forecast from the numerical weather data, this approach is computationally fast and needs less meteorological inputs, which can apply to the power forecast and energy yield estimation for the widely distributed PV plants.