皮尔逊相关系数在风电功率组合预测中的应用

Application of Pearson Correlation Coefficient in Wind Power Combination Prediction

针对目前权重组合预测方法存在的问题,提出一种新的相似序列组合预测方法。该方法以预测点处最近的一段时间序列为基础,采用皮尔逊相关系数,找出在一段历史数据中与之相似度较高的序列,对相似序列之后的点进行预测,通过对预测误差来确定最终预测点处各方法预测值所占的的权重。应用该方法对西北某风电场的风电功率预测,仿真结果表明:该方法避免了传统方法中以最近历史数据预测结果为计算依据所带来的误差较大问题,在一定程度上提高了组合预测的精度。

Aiming at existing problems in the weight combination prediction method, a new similar sequence combination prediction method is proposed. The proposed method takes a period of time sequence nearest the prediction point as a basis and uses Pearson correlation coefficient to find sequences with higher similarity in a period of history data. Points after the similar sequence are predicted and prediction errors are used to determine weights of prediction value of each method at the final prediction point. The method is applied in the wind power prediction of a certain wind farm in northwest. Simulation results show that the method avoids the larger error problem in traditional method which is caused by taking the nearest history data prediction results as calculation basis, and improves the precision of combination prediction in a certain degree.