This article presents a proper approach for the prediction of output power for photovoltaic generation system. Using the results of solar irradiation from the sunny-day-model as the base value, which are accurate when the sky is clear, and correcting the forecasting results with the assistance of Markov Chain in order to make them valid for other kinds of weathers. To be specific, by utilizing Markov Chain, the underlying stochastic effects of clouds coverage can be minimized and thus harvest more accurate results. The method is tested on the photovoltaic generation system of Electrical Engineering School, Wuhan University, P.R.China and rendered a satisfactory precision of forecasting. Finally, further measurements for enhancing accuracy are also discussed in this paper.