Using Markov Switching Model for solar irradiance forecasting in remote microgrids

摘要

In recent years, there has been rapid growth of Photovoltaic (PV) system integration into diesel-based remote microgrids to reduce the diesel fuel consumption. However, due to low correlation of PV power availability with the load as well as uncertainty and variability of the PV power, the benefits of the integration have not been achieved properly. A large energy reserve is required to compensate the fluctuation and improve reliability, which leads to increased operational cost. Solar irradiance forecasting helps to reduce the reserve requirement and improve the PV energy utilization. In this paper, a novel solar irradiance forecasting using Markov Switching Model is proposed for remote microgrids. This forecasting method uses locally available historical irradiance data of the microgrid location to predict day-ahead irradiance. The case study for validating this method for Brookings, SD resulted in Root Mean Square Error (RMSE) of 99.6 W/m2 for 2008 and 106.8 W/m2 for 2011.