Wind power forecasting provides an important basis for power grid planning and design. Studying wind power forecasting methods is of great significance to ensure safe and stable operation of power grids. Aiming at the problem that the regularization coefficient C and the kernel parameter λ are model parameters that affect the accuracy of the kernel extreme learning machine prediction model, a PSO-KELM prediction method using PSO to optimize the parameters of the kernel extreme learning machine is proposed. Taking the regularization coefficient C and the kernel parameter λ as optimization objects, the parameters are jointly optimized by using the PSO method to establish a PSO-KELM wind power prediction model. Experimental research was performed on 3 sets of measured data. Root mean square error, mean absolute error, and relative standard deviation are introduced as evaluation indicators. The prediction error of this method is better than that of direct application of KELM method. The results are further compared with commonly used SVM and PSO-SVM method. The results show that the proposed method has high prediction accuracy and good stability, and it can provide a scientific and effective reference for wind farm power prediction and wind power grid safety.