The high proportion of outliers in photovoltaic power data has the characteristics of mutual overlap and multi-Gaussian distribution. How to accurately eliminate the high proportion of outliers from photovoltaic power data is a crucial problem for realizing precise photovoltaic power forecast. This paper analyzes the characteristics of high proportion of outliers in photovoltaic power data, and studies a high proportion of outlier detection method in photovoltaic power data based on Gaussian mixture model. Firstly, the algorithm model of Gaussian mixture model is established, and then the EM algorithm is used to estimate parameters of the Gaussian mixture model. At last, the proposed algorithm is used to detect and eliminate the high proportion of outliers in the photovoltaic power data. Experiments and comparative studies show that the proposed method is more suitable for high proportion of outlier detection in photovoltaic power than the classical outlier data detection method.