1.NARI Group Corporation (State Grid Electric Power Research Institute), Nanjing 211106, China;2.Beijing Kedong Power Control System Co., Ltd., Beijing 100192, China

With the large-scale integration of wind power into the power grid, the demand for probability prediction of wind power is becoming more and more urgent. In order to realize the short-term probability distribution prediction of wind power, a short-term wind power probability prediction method based on the improved gradient boosting machine (GBM) algorithm is proposed. Firstly, the problems of the GBM algorithm applied to short-term wind power probability prediction are analyzed. Secondly, the negative log-likelihood loss function is used as the loss function in the GBM algorithm, and the Fisher information matrix is used to modify the gradient of the loss function in the parameter space of probability distribution and convert the gradient into the natural gradient of the probability distribution space. Then, based on the natural gradient, an improved GBM algorithm suitable for the short-term wind power probability distribution prediction is proposed. Finally, the proposed algorithm is compared with the traditional GBM algorithm and other methods. The results show that the training process of the proposed algorithm converges faster and has better prediction performance, which verifies the practicability and effectiveness of the proposed algorithm.

This work is supported by State Grid Corporation of China (No. 5700-202055368A-0-0-00).