Abstract: This paper extends the point value prediction to interval prediction. The interval midpoint and interval radius of the similar day sample are used to forecast the PV output using the conventional BP neural network algorithm,the GM（1,1）algorithm and the support vector machine（SVM）algorithm respectively. The combination weights of various interval forecasting are optimized by using seeker optimization algorithm（SOA）. Then the multi-objective optimization is converted to the single-objective optimization by specifying the willingness coefficient. The simulation results showed that the proposed interval forecasting method has high accuracy and practical value.