With the increasing proportion of the photovoltaic power system in the power grid , the influence of the photovoltaic power system on the power grid is becoming more and more high . The accurate forecasting of the photovoltaic power in advance is beneficial to the timely dispatching of the power grid and the guarantee of the power quality, so as to ensure the safe operation of the power grid. A power generation forecasting model based on regression analysis and Markov chain is proposed to predict the power of photovoltaic power generation. Considering the main influencing factors , e.g. season, weather type and meteorology , etc , we get the initial prediction value and relative residual sequence through regression model, and then combine the Markov chain theory to establish the state transition probability matrix, so as to correct the error sequence and improve the accuracy of the algorithm .According to the measured power data of a photovoltaic power station, the proposed model is tested and evaluated, which verifies the accuracy, simplicity and applicability of the regression analysis and Markov chain combination model.