A Practical Method for Generating Day-Ahead Output Probabilistic Scenarios of Multiple Wind Farms Based on K-Shape Clustering

The volatility and uncertainty of wind power bring severe challenges to the power system, and the modeling of multi-wind farm output scenarios considering the uncertainty of wind power is of great significance in power system dispatching decision-making. Especially for area/ province with abundant wind power resources, where the reliability of the dispatching plan can be strongly affected by the volatility of large amount of wind power. In this paper, a practical method of generating probabilistic scenarios for day-ahead wind power output is proposed for provincial power grid with high proportion of wind power. Firstly, the K-Shape clustering algorithm is used to cluster the wind farm stations, and then the output scenario are generated by using the total wind power forecast value and forecast error probability distribution characteristics of each aggregated station. After arrangement and combination of the aggregated stations’ scenes, the scenario minimization based on probability distance was conducted for clustering reduction to obtain the typical output scenes, and then the typical output scenes were assigned to each station according to the forecast value of the station to obtain the probability output scenes of each station. It is verified by actual power grid data that the scenarios generated by this method can combine with wind power predictions to well depict the random fluctuation of multi-wind farms output and has a good application prospect in the practice of day-ahead dispatching decision making.