Dynamic Wind Power Probabilistic Forecasting Based on Wind Scenario Recognition

Most traditional wind power forecasting methods are deterministic and static without considering external changing conditions as well as probabilistic information of forecasting error. To solve the above problem, a new wind power probabilistic forecasting method is presented based on wind scenario recognition in this paper. Firstly, a wind scenario recognition model is established based on K means clustering algorithm. The natural wind feature is extracted from wind speed and wind direction. Then, the probabilistic forecasting models based on relevance vector machine (RVM) are built for each wind scenario. During real time forecasting, the power generation process in different conditions can be recognized and mapped by adjusting the parameters of the pre established probabilistic forecasting model. Taking a wind farm in Northwest China as an example, the results show that the accuracy of deterministic forecasting, the reliability and skill score of probabilistic forecasting, and forecasting efficiency are improved by the proposed method, which provides a new solution for refined forecast modeling.