基于改进极限学习机的电力市场实时电价预测方法  

Real time electricity price prediction method in the electricity market based on improved extreme learning machine

作　　者：王蕾[1] 李斌[1] 吴飞 张振明 徐绮 孙宇涵 WANG Lei;LI Bin;WU Fei;ZHANG Zhenming;XU Qi;SUN Yuhan(Northeast Electric Power University,Jilin 132012,China)

机构地区：[1]东北电力大学,吉林吉林132012

出　　处：《电子设计工程》2024年第20期21-25,30,共6页Electronic Design Engineering

基　　金：吉林省科技发展计划项目(20200401097GX)。

摘　　要：针对实时电价数据波动性强及其影响因素复杂,导致现有预测模型稳定性及预测精度低的问题,提出了一种基于改进极限学习机的电力市场实时电价预测方法。设计了基于经验小波变换的实时电价数据分解方法,将电价序列分解为接近价格的低频信号和噪声的高频信号。同时提出基于改进随机森林的实时电价特征提取算法,根据预测重要度获取最优的电价影响因素特征组合。以此为基础,将核函数替代极限学习机隐藏层构建了R-KELM预测模型,更好地反映了多因素影响下实时电价的不确定性和波动性。以PJM实时电价数据为例,结果表明,所提方法可以有效克服电价数据强波动性及高特征冗余的问题,预测模型准确性及稳定性得到显著提升。A real time electricity price prediction method based on improved extreme learning machine is proposed to address the issues of strong volatility and complex influencing factors in real time electricity price data,resulting in low stability and prediction accuracy of existing prediction models.A real time electricity price data decomposition method based on empirical wavelet transform is designed,which decomposes the electricity price sequence into low-frequency signals close to the price and high-frequency signals with noise.At the same time,a real time electricity price feature extraction algorithm based on improved random forest is proposed to obtain the optimal combination of electricity price influencing factors based on the predicted importance.Based on this,a multi-step recurrent R-KELM prediction model is constructed,which better reflects the uncertainty and volatility of real time electricity prices under the influence of multiple factors.Taking PJM real time electricity price data as an example,the results show that the proposed method can effectively overcome the problems of strong volatility and high feature redundancy in electricity price data,and significantly improve the accuracy and stability of the prediction model.

关 键 词：电力市场 实时电价预测 经验小波变换 特征提取 极限学习机 

分 类 号：TN-9[电子电信]