基于数学形态学聚类与果蝇优化算法的风电功率短期预测

WIND POWER SHORT-TERM PREDICTION BASED ON MATHEMATICAL MORPHOLOGY CLUSTER ANALYSIS AND FRUIT FLY OPTIMIZATION

针对预测模型训练数据的选择以及模型参数最优化的问题,提出一种基于数学形态学聚类与果蝇优化算法相结合的风电功率短期预测方法。数学形态学聚类方法通过膨胀腐蚀运算,自动将数值天气预报数据聚成类,然后寻找与预测日相似的类作为训练样本。果蝇优化算法能较快确定模型的最优参数。通过对依兰风电场的发电功率进行预测,证实该方法的有效性,其精度比基于K均值聚类方法和粒子群优化算法的预测模型要高,且训练数据对模型精度的影响会高于模型本身参数的优化。

In order to solve the problems of training data selection and parameter optimization of the prediction model,this paper introduces a method of the short-term prediction of wind power generating capacity based on mathematical morphology cluster analysis and fruit fly optimization algorithm.Mathematical morphology cluster analysis automatically divides numerical weather prediction data into several clusters according to dilation and erosion operation,and similar days with the predicted day are searched as training sample.Fruit fly optimization algorithm can determine quickly the optimization parameters of the prediction model.Simulation is performed to the wind power generation of Yilan wind farm.The results show that the method is effective and its precision is higher than that of the prediction model based on k-mean cluster analysis or particle swarm optimization algorithm.The effect of training data on model accuracy is higher than that of model optimization.