基于特征提取相似日的ELM短期负荷预测研究

马立新,尹晶晶,郑晓栋

(上海理工大学 光电信息与计算机工程学院,上海200093)

摘要为解决短期电力负荷预测中预测精度差、计算时间长等问题,提出一种基于自组织特征映射网络进行特征提取相似日的极限学习机短期电力负荷预测方法。通过自组织特征映射网络找出与预测日同类型的历史数据作为训练样本;并采用预测能力强、计算时间短的ELM网络进行预测。以某市电力负荷数据进行仿真,并将上述方法与传统神经网络进行对比。仿真算例表明,基于特征提取相似日的ELM方法具有较高的预测精度,泛化性能好,且运算时间短。

关键词自组织特征映射;特征提取;相似日;极限学习机;短期负荷预测

收稿日期:2014-11-04

基金项目:国家自然科学基金资助项目(61205076);上海市研究生创新基金资助项目(JWCXSL1302)

作者简介:马立新(1960—),男,教授,博士。研究方向:电力系统分析与优化运行等。尹晶晶(1987—),女,硕士研究生。研究方向:电力系统负荷预测。E-mail:jing.jing.yin@163.com

doi:10.16180/j.cnki.issn1007-7820.2015.12.006

中图分类号TP183文献标识码A

Short-term Load Forecasting Based on Daily Feature Extraction of Similar Days and ELM

MA Lixin,YIN Jingjing,ZHENG Xiaodong

(School of Optical-Electrical and Computer Engineering,University of Shanghai for

Science and Technology,Shanghai 200093,China)

AbstractIn order to solve the problems of forecasting method,such as low forecasting accuracy,and long computation time in short-term electric power load forecasting,an approach to short-term load forecasting based on self-organizing feature mapping of similar days feature extraction and ELM (Extreme Learning Machine) combined method is proposed in this paper.Firstly,self-organizing neural network is used to the classification of related data.The data of the same type as that of the forecasting day are found out.Then these training samples are forecasted by ELM,which has strong ability to predict and short computing time.The power load data of one city were used for simulating.The proposed method is compared with ELM and back propagation (BP) neural network.The experimental results show that ELM method based on feature extraction of similar days has high prediction precision,good generalization performance and short running time.

Keywordsself-organizing feature map;feature extraction;similar days;extreme learning machine;short-term load forecasting

电力负荷预测是供电部门的重要工作之一,准确的负荷预测,可以经济合理地安排电网内部发电机组的启停,保持电网运行的安全稳定,减少不必要的旋转储备容量,合理安排机组检修计划,保证社会的正常生产和生活,有效降低发电成本,提高经济效益和社会效益[1]。

负荷预测的技术方法研究或者预测的数学模型研究是负荷预测研究的核心问题。一方面,电力负荷的变化过程较为复杂,许多不确定因素都会引起负荷的波动;而另一方面,负荷变化又具有规律性的周期变化特征。……