Photovoltaic (PV) systems integration is increasingly being used to reduce fuel consumption in diesel-based remote microgrids. However, uncertainty and low correlation of PV power availability with load reduces the benefits of PV integration. These challenges can be handled by introducing reserve. However, this leads to increased operational cost. Solar irradiance forecasting helps to reduce reserve requirement, thereby improving the utilization of PV energy. This paper presents a new solar irradiance forecasting method for remote microgrids based on the Markov switching model. This method uses locally available data to predict one-day-ahead solar irradiance for scheduling energy resources in remote microgrids. The model considers past solar irradiance data, clear sky irradiance, and Fourier basis expansions to create linear models for three regimes or states: high, medium, and low energy regimes for days corresponding to sunny, mildly cloudy, and extremely cloudy days, respectively. The case study for Brookings, SD, USA, discussed in this paper, resulted in an average mean absolute percentage error of 31.8% for five years, from 2001 to 2005, with higher errors during summer months than during winter months.