Abstract: To improve the accuracy of photovoltaic output forecast and provide sufficient forecast information for the scheduling decision makers, a short-term probabilistic forecast for power output of the photovoltaic station based on the high order Markov chain (HMC) and the Gaussian mixture model (GMM) is proposed. Firstly, the HMC model was carried out for the historical output data of the photovoltaic power stations. The orders of the Markov chain were determined by calculating the Pearson correlation coefficient of the photovoltaic output power in the adjacent periods, and the state transition probability matrix was obtained by analyzing the statistical historical data. Based on this, a probability prediction model in the form of GMM was established, and the mean and variance of each Gaussian distribution were modified based on the meteorological similarity, and finally the probability density function of photovoltaic output was obtained. Taking the actual photovoltaic station data as an example, the results show that the proposed probabilistic prediction method has high accuracy. Compared with the traditional point prediction method, the probabilistic prediction method can provide more information for the power grid operation decision-making.