The invention relates to a wind power generation power prediction method and device for optimizing an echo state network, and the method comprises the steps: employing the historical data of wind power generation power as a training sample, and carrying out the training, and obtaining an echo state network model for predicting the wind power generation power; Optimizing parameters of the echo state network model by taking the minimum prediction error of the echo state network model as a target; and predicting the wind power generation power by using the optimized echo state network model. According to the technical scheme provided by the invention, the key parameters of the echo state network are continuously optimized and updated according to the historical input samples by utilizing thepowerful nonlinear mapping capability of the echo state network and combining the rapid and accurate optimization capability of the firefly optimization algorithm, so that the prediction accuracy of the echo state network is improved.