Research On Short-Term Wind Farm Output Power Combination Prediction Model Based On Numerical Weather Prediction

The prediction of wind farm output power is considered as an effective way to increase the wind power capacity and improve the safety and economy of power system. It is one of the hot research topics on wind power. The wind farm output power is related to many factors such as wind speed, etc., which is difficult to be described by some mathematical expression. Previous researchers used to perform the wind power prediction by history data. However, we usually lack of history data, which only have forecast data. Hence, the model established by the history is not suitable for the wind power prediction calculated by the forecast data. Consequently, in this paper, we use the NWP data from meteorological agency to research the wind power prediction.There are three important jobs in this paper:(1)Firstly, we use Back Propagation (BP) neural network to perform wind power prediction, and then according to the disadvantage of BP neural network which is unstable prediction result caused by random connection weight local minimum value. We use genetic algorithm (GA) and particle swarm optimization (PSO) to optimize BP neural network. Meanwhile, we establish GA-BP and PSO-BP short-term wind power prediction model. The experiment shows that the prediction result of optimization is better than that of simple BP neural network.(2)In order to progress the prediction accuracy of the wind power, and eliminate the influence by different weather and different climate we establish wind power prediction model by months, which use the theory of combination prediction based on the GA-BP and PSO-BP method. The experiment shows that the prediction accuracy of combination prediction is the best.(3)We often find missing and abnormal data from NWP data. Precious researches often use interpolation method and replacement method. However, it will lead to the inaccurate result if it has lots of missing and abnormal data. Hence, in this paper, we use the combination prediction method to handle the missing and abnormal data.