基于信号分解和核极限学习机的风电功率预测

Wind Power Prediction Based on Signal Decomposition and Kernel Extreme Learning Machine

针对传统短期风功率预测模型在功率变化较大情况下的预测精度不高问题,提出了一种基于信号分解和量子粒子群算法优化核极限学习机的短期风功率预测模型。首先利用经验小波变换将原始风功率序列分解成为若干个模态分量,再利用核极限学习机建立每个模态分量的预测模型,为了提高模型预测精度,采用量子粒子群算法优化核极限学习机参数,最后将每个模态分量预测值相加得到最终的功率预测结果。以实际风电场发电功率为例,并与其他预测模型进行比较,结果表明所提模型具有较高的预测精度。

In order to solve the problem that the prediction accuracy of traditional short⁃term wind power prediction model is not high when the power changes greatly,a short⁃term wind power prediction model based on signal decomposition and quantum particle swarm optimization(QPSO)was proposed.Firstly,the original wind power sequence was decomposed into several modal components by using empirical wavelet transform.Then,the kernel extreme learning machine was used to establish the prediction model of each modal component.In order to improve the prediction accuracy of the model,the parameters of kernel extreme learning machine were optimized by using quantum particle swarm optimization algorithm.Finally,the predicted values of each modal component were added to obtain the final power prediction value.Taking the actual wind farm power generation as an example,and comparing with other prediction models,the results show that the proposed model has higher prediction accuracy.