基于小波变换和改进萤火虫算法优化极限学习机的短期负荷预测

A Novel Load Forecasting Method Based on WD-IFA-ELM

提出了一种基于小波变换和改进萤火虫优化极限学习机的短期负荷预测方法.通过小波分解和重构,对原始负荷序列进行降噪;在模型训练阶段利用改进的萤火虫算法优化极限学习机参数,获得各序列的最优模型;针对各子序列分别预测叠加得到最终预测值.通过在两种时间尺度的数据序列上进行数值计算,与传统的ARMA、BP神经网络、支持向量机及LSSVM等多种经典预测模型相比,模型预测效果更优.

A novel short term load forecasting method based on WD-LFA-ELM was proposed in this paper. First of all, the noise of original load series is reduced by wavelet decomposition and reconstruction. Then the parameters of ELM are optimized by improved firefly algorithm in the training procedure, and an optimal forecasting model for each sub series is obtained. The final prediction value is the superposition of all sub series prediction result. The method is applied to two time scale load prediction, the result show the superiority of this method compared with the classical model like ARMA, BP, SVM, LSSVM and standard ELM.