Accurate evaluation of photovoltaic power is of great significance to ensure safety and stability and economic operation of power systems. In this paper

a short-term photovoltaic power prediction model based on least square support vector machine(LSSVM) optimized by improved bare-bones differential evolution algorithm(IBBDE) is proposed. Initialization of population using generalized opposition based learning and self-adaption of crossover factor are introduced to enhance the global convergence of algorithm. The IBBDE algorithm is adopted to optimize the regularization parameters and kernel parameters of LSSVM

thus to improve the short-term photovoltaic power forecasting accuracy. Simulations are performed on a realistic photovoltaic power station in Tibet

and the results shows that

the proposed method has high prediction accuracy. In sunny weather

the mean absolute prediction error(MAPE) and root-mean-square error(RMSE) are 5.39% and 3.98%

respectively. In rainy weather

the MAPE and RMSE are 10.69% and 7.86%

respectively.