Wind Power Scenarios Analysis Method Based on Prediction Box and Improved k-means Algorithm

Abstract

The scenario analysis technology can generate time-series scenarios with probabilistic characteristics, and the deterministic scenarios describe the uncertainty information of wind power output, which is suitable for power system scheduling problems involving large-scale wind power grid-connected. The scenario analysis technology is divided into scenarios generation part and scenarios reduction part. In the scenarios generation part, the prediction box interval is first divided according to the distribution of a large amount of historical data, and the appropriate error fitting function is selected by calculating the skewness of each prediction box error data; then a random number sequence conforming to the multivariate standard normal distribution is generated. And performing two inverse transformations on the random number sequence according to the standard normal distribution and the prediction box fitting function; finally, superimposing the prediction error sequence obtained by the inverse transformation with the predicted value sequence to obtain a large number of initial scenarios sets; In terms of reduction, based on the traditional K-means algorithm, the optimal cluster number and initial cluster center are determined by using the CH index and the maximum and minimum principle, and the clustering level and clustering stability of the clustering algorithm are improved. The case study shows that the scenario analysis technology proposed in this paper has high accuracy for the description of wind power probability information.