A Statistical Approach to Day Ahead Forecasting of Solar Irradiance in a Local Point

Abstract

Today, for the management of energy supply systems forecast information on load and the production of meteorology dependent (wind, solar, hydro) generation is ever rising. Solar irradiance forecasting is given a unique priority as it spans over major applications such as management of grids with a high share of photovoltaic generation and thermal power supply systems relying on solar heat generation. This research addresses the day ahead prediction of the local irradiances intended to be applied for the management of solar assistant systems for heat and hot water supply. The forecast method presented here is based on the statistical analysis of historical data in Kristiansand, Southern Norway. For this, satellite derived irradiance data covering seven years provided by Geomodel Solar, Slovacia (D. Heinemann, 2005) can be used. In this approach, it is assumed that the irradiance sum of today shows a dependence on the irradiance sum of yesterday (B.O. Ngoko, 2014). This day to day dependency is assessed by obtaining conditional probability distributions of irradiance sum on next day for a given status of weather, given here by the irradiance sum on previous day. Based on such probabilistic approach two schemes are introduced to obtain values for the forecasting. The first scheme is based on most probable expected irradiance sum of tomorrow and the second approach is based on the average expected irradiance sum, both extracted from the probability distributions. Having obtained forecasted values for the irradiance, the validity of prediction methods are investigated by

comparing with the actual measured data giving the statistical parameters, relative monthly Bias and relative monthly Root Mean Square Error (RMSE). The comparison reveals that the approach using the average expected irradiance sum, gives more accurate results showing low RMSE. Concerning the application, the irradiance data, both measured and forecasted can be used to analyze the daily energy gain of a solar thermal collector and its forecastability.