The short-term forecast of photovoltaic solar power is important for grid stable operation, economic dispatch, and renewable energy accommodation. However, photovoltaic power output is influenced by environmental factors such as radiation intensity and temperature, showing great volatility and randomness. In order to further improve the accuracy of prediction and the universality of different weather types, a hybrid photovoltaic power forecast (HPF) algorithm based on support vector regression combined with phase space reconstruction and similar day selection is proposed in this paper. Using the real data, we verify the prediction effectiveness of the proposed HPF algorithm. The results show that compared with the traditional support vector regression model, the prediction model in this paper can further improve the prediction accuracy. Additionally, the algorithm in this paper also shows good performance under both weak-fluctuation and strong fluctuation.