论文部分内容阅读
基于风速在日时间尺度下的变化周期,提出一种风电场功率分类组合预测模型.该模型采用Morlet小波变换,分析数值天气预报中的风速在日时间尺度下的变化周期及特征;结合主成分分析和谱聚类方法对具有不同周期特征的风速变化过程进行分类;针对不同的风速变化类型分别建立遗传优化BP神经网络、RBF神经网络和支持向量机的预测模型,并选取每类对应的最优算法进行组合,预测功率时根据未来风速过程动态切换相应模型.以中国某风电场为例进行验证,结果表明,按8h的变化周期对风速变化类型进行分类,可得到较好的分类组合预测结果,其精度较单一预测模型提高0.87％,合格率提高1.05％,验证了所提模型的有效性,为风电场功率预测提供了新思路.,Wind power prediction is one of the necessary means to ensure the safe,stable and economic operation of power system.However,it’s difficult to predict the output power accurately with one single prediction method.Based on the change period of wind speed in the day time scale,the classification and combination prediction model of wind power was proposed.The Morlet wavelet transformation was used to analyze the change period and characteristics of wind speed in day time scale of NWP.Combining principal component analysis and spectrum clustering method,the change process of wind speed with different period features were classified.Different types of prediction models were established according to the corresponding wind speed type,such as genetic optimization BP neural network,radical basis function neural network and support vector machines etc.algorithms.Choosing the best corresponding algorithm for each class wind speed to combine,the corresponding model was dynamically switched according to future wind speed process in predicting the wind power.To take one wind farm in China as example,the results show that the prediction accuracy of the model based on 8 hours wind speed period is increased by 0.87％ than single prediction model,qualified rate is increased by 1.05％.The proposed model is effective to improve prediction accuracy,providing a new idea for wind power prediction.