中长期电力负荷预测研究  

Study on Long-Term Power Load Forecasting

作　　者：宋晓茹[1] 李莉 张来青 

机构地区：[1]西安工业大学电子信息工程学院,陕西西安710032 [2]大唐陕西发电有限公司灞桥热电厂,陕西西安710038 [3]陕西德源府谷能源有限公司,陕西榆林719000

出　　处：《计算机仿真》2014年第9期132-135,共4页Computer Simulation

基　　金：陕西省科技厅专项基金;西安市科技厅专项基金的资助

摘　　要：为了有效地提高电力负荷的预测精度,针对影响中长期电力负荷多因素间的非线性和不确定性,提出了一种粗糙集(RS)和最小二乘支持向量机(LSSVM)相结合的中长期电力负荷预测方法。采用粗糙集理论把影响电力负荷的六个因素,属性约简为三个核心属性,减少了LSSVM的输入量,提高了电力负荷预测系统的快速性;粗糙最小二乘支持向量机回归建模,构造RS-LSSVM的电力负荷预测模型,提高预测的精度。最后进行仿真,改进模型应用于某地区的中长期电力负荷的拟合和预测中,采用RS-LSSVM模型,与BP神经网络的拟合预测结果相比,预测误差明显小于BP神经网络,具有更高的预测精度,为中长期的电力系统负荷预测提供了一种新的科学、有效的方法。In order to improve the accuracy of power load forecasting effectively, a combining method between RS and LSSVM is proposed for long-term power load nonlinearity and uncertainty factors. The six factors that influence the power load are attributed reduction to three core attributes by RS and LSSVM. And the LSSVM input variables are reduced. The rapidity of power load forecasting system is improved. The RS-LSSVM regression modeling is to con- struct the RS-LSSVM power load forecasting model. It is to improve the accuracy of prediction. The model is applied to the fitting and long term load forecasting in certain areas in an example. Compared with the fitting BP neural network prediction, the prediction error by using the RS-LSSVM model is significantly less than the BP neural network. It has higher prediction accuracy. It provides a new scientific and effective method for the long-term load forecasting of the power system.

关 键 词：电力负荷预测 粗糙集 最小二乘支持向量机 神经网络 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]