基于ELM和误差校正的风力发电区间预测方法

Wind Power Generation Interval Prediction Method Based on ELM and Error Correction

风力发电过程中具有不稳定性和随机性,传统的点预测获得的结果准确性欠佳且无法获得预测值的区间波动范围。提出了一种基于极限学习机(ELM)和误差校正的风力发电区间预测方法。首先,使用皮尔逊相关系数挖掘出数据集中重要特征;然后,建立ELM网络生成预测值,将生成的预测值与原始的风力发电功率值作对比得到风力发电功率的误差;再将误差和原始的数据结合成新的数据集输入到已经训练好的ELM网络模型中,得到校正过后的误差;最后,通过所提出的区间构造方法得到风电功率预测区间。仿真结果表明,校正过后的误差比原始误差小,构造出的区间具有较高的可靠性和较窄的区间带宽,能更准确地描述风力发电出力范围。

The process of wind power generation has instability and randomness,and traditional point prediction results have poor accuracy and cannot obtain the range of fluctuation of predicted values.This paper proposes an interval forecasting method for wind power generation based on extreme learning machine(ELM)and error correction.First,we used the Pearson correlation coefficient to excavate important features in the data set;then,established an ELM network to generate predicted values,and compared the generated predicted values with the original wind power value to obtain the error of wind power.Then the data were combined into a new data set and input into the trained ELM network model to obtain the corrected error.Finally,the wind power prediction interval was obtained by the proposed interval construction method.The experimental simulation shows that the corrected error is smaller than the original error,and the constructed interval has higher reliability and narrower interval bandwidth,which can describe the output range of wind power generation more accurately.