摘要
为了提高智能电网系统的负荷预测精度,基于支持向量回归(support vector regression,SVR)原理以及卡罗需库恩塔克(Karush-Kuhn-Tucker,KKT)条件,建立一种在线支持向量回归(online support vector regression,OSVR)预测模型。当数据库有新样本添加时,该模型利用递增算法及时调整参数,实时学习当前负荷曲线特征和气象变化状况,而无需结合旧样本重新训练模型;考虑到温积效应对电力负荷产生显著影响,基于Fisher信息(FI)原理,通过气温窗口数据计算当前气温的权值,并将气温加权积作为模型输入量;将温度影响的持续性引入建模过程,从而得到FI-OSVR负荷预测模型。实例仿真结果表明,FI-OSVR模型具有较高的预测精度,在智能电网负荷预测中具有优越的工程实用性。
To improve the load forecasting accuracy of the smart grid system,this paper establishes a prediction model based on online support vector regression(OSVR)according to the principle of support vector regression(SVR)and and Karush-Kuhn-Tucker(KKT)condition.When adding new samples to the database,this model adjusts the parameters in time using the incremental algorithm,learns the characteristics of current load curves and meteorological changes in real time,while without need retraining the model with old samples.Considering that the temperature accumulation effect has a significant influence on the power load,the paper calculates the weight of current temperature through the temperature window data,and takes the weighted product of temperature as the model input based on the Fisher(FI)information principle.It introduces the continuous influence of temperature into the modeling process,so as to obtain the FI-OSVR load forecasting model.The simulation results show that the FI-OSVR model has high prediction accuracy and superior engineering practicability in smart grid load prediction.
作者
任萌
高小征
REN Meng;GAO Xiaozheng(Meizhou Power Supply Bureau of Guangdong Power Grid Co.,Ltd.,Meizhou,Guangdong 514000,China)
出处
《广东电力》
2021年第4期78-84,共7页
Guangdong Electric Power
基金
广东电网有限责任公司科技项目(031400KK52180027)。
关键词
负荷预测
支持向量回归
卡罗需库恩塔克条件
温积效应
Fisher信息
load forecasting
support vector regression(SVR)
Karush-Kuhn-Tucke(KKT)
temperature accumulation effect
Fisher information
