Prediction of Combination Probability Interval of Wind Power Based on Naive Bayes

In order to improve the performance of predicting the probability interval of wind power, a wind power interval prediction method combining normal exponential smoothing and mixed sliding kernel density estimation is proposed. Firstly, the point prediction model is established by the naive bayesian classifier. Then, the probability distribution of the prediction error is estimated by normal exponential smoothing and mixed sliding kernel density estimation, and the corresponding prediction interval under a certain confidence probability is obtained. Finally, the entropy method is used to reasonably combine the prediction interval of normal exponential smoothing and the prediction interval of mixed sliding kernel density estimation to generate the final wind power prediction interval. The results show that, compared with the normal exponential smoothing and mixed sliding kernel density estimation, the proposed combination prediction method combining with an entropy method can be employed to improve the interval coverage and reduce the average bandwidth of intervals, which proves that the method compromises the reliability and accuracy of the prediction. The research can provide a reference for the prediction of wind power. © 2020, High Voltage Engineering Editorial Department of CEPRI. All right reserved.