论文部分内容阅读
针对单一预测模型误差波动较大和线性组合预测的局限性,提出了基于多属性决策和支持向量机(SVM)的风电功率非线性组合预测模型。首先基于多属性决策理论,在检验其预测有效的情况下选择3种最优模型作为单项预测模型,并分别建模预测得到3种不同的预测结果;然后将各单项的预测结果作为训练输入,将相应的实际值作为训练输出,建立SVM组合预测模型。为检验该模型预测的有效性,用2组不同的历史数据进行验证,结果表明:该组合模型综合了各单项模型的优点,其均方根误差和平均百分比误差均小于各单项模型及其他组合模型,有效地提高了预测精度。最后还研究了采样间隔对预测结果的影响,结论表明:当采样间隔为5～15min时,预测精度较高。
Aiming at the limitation of large error variance and linear combination forecasting of single forecasting model, a nonlinear forecasting model of wind power based on multiple attribute decision making and support vector machine (SVM) is proposed. Firstly, based on multi-attribute decision-making theory, three kinds of optimal models are selected as single prediction models to test their prediction effectiveness, and three different prediction results are obtained respectively by modeling and forecasting. Then, the individual prediction results are input as training inputs, The corresponding actual value is used as training output to establish SVM combined forecasting model. To verify the validity of the model prediction, two different sets of historical data were used to verify the results. The results show that the combination model combines the advantages of the individual models, the RMSE and the average percentage error are less than the individual models and other combinations Model, effectively improve the prediction accuracy. Finally, the influence of sampling interval on the prediction results is also studied. The conclusion shows that when the sampling interval is 5 ~ 15min, the prediction accuracy is high.