基于卡尔曼滤波和支持向量机的风电功率实时预测研究  

Real-time Prediction for Wind Power Based on Kalman Filter and Suport Vector Mahines

作　　者：杨茂[1] 黄宾阳 江博[1] 林思思 

机构地区：[1]东北电力大学电气工程学院,吉林吉林132012 [2]国网福建南安市供电有限公司,福建南安362300

出　　处：《东北电力大学学报》2017年第2期45-51,共7页Journal of Northeast Electric Power University

基　　金：国家重点基础研究发展计划项目(973计划)(2013CB228201);国家自然科学基金(51307017);吉林省产业技术研究与开发专项项目(2014Y124);吉林省科技发展计划(20140520129JH)

摘　　要：为了提高风电场风电功率实时预测精度,并为风电场输出功率的合理调度提供参考依据,提出了一种基于滚动的自回归积分滑动平均模型(ARIMA)和支持向量机(SVM)相结合的卡尔曼融合预测模型。通过对风电功率序列进行分析得出ARIMA模型,用其作为卡尔曼滤波的状态方程。再用SVM预测得出观测方程,用卡尔曼滤波将二者结合起来实现融合多步预测。具体的实例分析中采用了国家能源局的评价指标对预测精度进行评价。通过预测结果可以看出,融合预测算法中可以实现预测误差相互抵消的状况,减少了误差累积,提高了预测的精度。In order to improve wind farm wind power real-time predictive accuracy,provide a reference for rational management of wind farm output power,this paper presents a combination based on rolling autoregressive integrated moving average model（ ARIMA） and support vector machine（ SVM） of Kalman fusion prediction model.By analyzing the sequence of wind power,obtaine ARIMA model,using it as a Kalman filter equations of state. Then SVM prediction derives observations equations. Kalman filtering combines the two methods to achieve integration of multi-step prediction.Finally,the paper gives specific examples of analysis used in the analysis of the evaluation of the National Energy Board to evaluate the prediction accuracy.By predicting the results we can see that the fusion algorithm can predict the prediction error cancel each situation,reducing the accumulation of errors and improve the accuracy of prediction.

关 键 词：ARIMA 卡尔曼滤波 SVM 实时预测 

分 类 号：TM614[电气工程—电力系统及自动化]