Study of Typical Scenarios of Wind Power Based on Data Dimension Reduction and Affinity Propagation Clustering

As more and more wind farms are connected to the grid, the high fluctuation of wind power has a great impact on the dispatch of the grid. Optimal dispatch model for power systems is a computationally-demanding task by use of wind power time series data with high dimensionality and redundancy, which leads to the time consuming and occupation of storage resources. To deal with this problem, the most common solution is to create typical scenarios for optimization through aggregating the original data set. However, there are no common ways to create accurate and stable scenarios. Therefore a novel approach is proposed based on wavelet entropy piecewise aggregation approximation and affinity propagation clustering in this paper. This method is proved to be valid by the actual output data of a wind farm in a northwestern province of China. The evaluation indexes show that the proposed method has a good aggregation effect and can be efficiently used for creating the typical scenarios for practice in optimal power dispatch.