Journal of Computer Applications ›› 2015, Vol. 35 ›› Issue (7): 2083-2087.DOI: 10.11772/j.issn.1001-9081.2015.07.2083

Previous Articles     Next Articles

Application of extreme learning machine with kernels model based on iterative error correction in short term electricity load forecasting

LANG Kun, ZHANG Mingyuan, YUAN Yongbo   

  1. Faculty of Infrastructure Engineering, Dalian University of Technology, Dalian Liaoning 116024, China
  • Received:2015-02-10 Revised:2015-03-26 Online:2015-07-10 Published:2015-07-17

基于迭代误差补偿的核极端学习机模型在短期电力负荷预测中的应用

郎坤, 张明媛, 袁永博   

  1. 大连理工大学 建设工程学部, 辽宁 大连 116024
  • 通讯作者: 张明媛(1981-),女,辽宁抚顺人,副教授,博士,主要研究方向:生命线工程、工程项目管理,myzhang@dlut.edu.cn
  • 作者简介:郎坤(1988-),女,陕西汉中人,博士研究生,主要研究方向:生命线工程、智能电网、负荷预测; 袁永博(1957-),男,辽宁大连人,教授,博士,主要研究方向:生命线工程、工程项目管理。
  • 基金资助:

    国家自然科学基金资助项目(51208081);辽宁省教育厅科学研究一般项目(L2014034)。

Abstract:

Focusing on the issue that the method of Back Propagation (BP) neural network limits the prediction accuracy of the short term electricity load, a prediction model based on Extreme Learning Machine with Kernels and Iterative Error Correction (KELM-IEC) was proposed. Firstly, an input index system was built, in which 7 factors were selected as the input of the prediction model, namely, month of the year, day of the month, day of the week, week number, holiday, daily average temperature, and maximum electricity load for the day before. Secondly, a load prediction model was built. It was based on a new kind of neural network called Extreme Learning Machine with Kernels (KELM). KELM introduced the kernel function mapping of Support Vector Machine (SVM) as the hidden layer nodes mapping of Extreme Learning Machine (ELM). It combined the advantages of ELM with simple structure and SVM with good generalization ability effectively, which could improve the prediction accuracy. Finally, an Iterative Error Correction (IEC) model was built based on the method of IEC in the field of time series prediction. The prediction errors of the load prediction model were trained by KELM and the prediction results could be corrected and revised. Thus, the prediction errors could be decreased and the predictive performance could be improved. In simulation experiments of two actual electricity load data sets, the KELM-IEC model was compared with the BP neural network model, and Mean Absolute Percentage Error (MAPE) respectively decreased by 74.39% and 34.73%, while Maximum Error (ME) decreased by 58.34% and 39.58%, respectively. At the same time, the KELM-IEC model was compared with the KELM model, and MAPE decreased by 18.60% and 4.29% respectively, while ME decreased by 0.08% and 11.21%, respectively, which verified the necessity of the IEC strategy. The simulation experiment results show that the KELM-IEC model can improve the prediction accuracy of the short term electricity load. It can benefit the plan, operation and management of the power system. It can guarantee the demand for production and living electricity. And it can improve both the economic and social benefits.

Key words: Extreme Learning Machine with Kernels (KELM), iterative error correction, input variable, short term, electricity load forecasting

摘要:

针对BP神经网络方法制约短期电力负荷预测精度的问题,提出一种基于迭代误差补偿的核极端学习机(KELM-IEC)预测模型。首先,建立短期电力负荷预测模型的输入指标体系,选择月份、日期、星期、周数、是否为节假日、日平均气温、前一日的最大负荷量等影响电力负荷的7个因素作为预测模型的输入;其次,基于新型神经网络模型——核极端学习机(KELM),建立负荷预测模型,引入支持向量机(SVM)的核函数映射作为极端学习机(ELM)的隐含层节点映射,有效结合ELM结构简单、训练简便与SVM泛化能力强的优势,提高负荷预测精度;最后,基于时间序列预测中迭代误差补偿(IEC)技术,建立IEC模型,再次利用KELM对负荷预测模型的预测误差进行学习,从而对预测结果进行补偿和修正,进一步减小模型预测误差,提高预测性能。采用两组实际电力负荷数据进行仿真实验,其中,KELM-IEC模型与BP神经网络模型相比,平均绝对百分误差(MAPE)分别降低了74.39%和34.73%,最大绝对误差(ME)分别降低了58.34%和39.58%;同时与KELM模型相比,平均绝对百分误差分别降低了18.60%和4.29%,最大绝对误差分别降低了0.08%和11.21%,说明误差补偿策略的必要性。实验结果表明,KELM-IEC预测模型能够有效地提高短期电力负荷预测的精度,有利于改善电力系统的计划、运营和管理,保障生产和生活用电,提高经济效益和社会效益。

关键词: 核极端学习机, 迭代误差补偿, 输入变量, 短期, 电力负荷预测

CLC Number: