基于改进SVR的微网短期负荷预测研究  

Short term load forecasting of microgrid based on improved SVR

作　　者：滕爱国 单新文 李萌 王沈亮 TENG Aiguo;SHAN Xinwen;LI Meng;WANG Shenliang(State Grid JIANGSU Electric Power Co.,lad Information and Telecommunication Branch,Nanjing 210009,China;Nari Group Corporation/State Grid Electric Pouer Research Institute,Nanjing 210000,China)

机构地区：[1]国网江苏省电力有限公司信息通信分公司,南京210009 [2]南瑞集团有限公司,南京210000

出　　处：《自动化与仪器仪表》2020年第12期194-197,共4页Automation & Instrumentation

基　　金：国网电力公司科技项目,国家电网基于设备指纹技术的无线接入安全管理平台研究与应用(No.J2019125)。

摘　　要：微网的短期负荷预测是实现微网智能调度、智能化微网的重要依据和有效方法。但传统的LS-SVR算法预测受微电网影响因素角度的影响,导致预测带来很大的不确定性等问题,分别从稀疏性和鲁棒性的角度对LS-SVR算法进行改进。在鲁棒性改进方面,针对数据异常带来的不确定性,引入加权因子,以降低残差平方和损失函数的敏感型;在稀疏性改进方面,通过φ(·)将数据集映射到希尔伯特空间空间中,然后通过得到的希尔伯特集A求解出极大无关解,进而得到新的输入向量ω。最后通过上述两方面的改进,引入拉格朗日乘子进行求解,得到线性回归方程。最后,运用MATLAB仿真软件对上述算法进行编程,并以Ausgrid集数据作为样本进行预测。实验结果表明,改进的LS-SVR算法在对微网短期负荷的预测误差较小,具有较高的预测精度。The short-term load forecasting of microgrid is an important basis and effective method to realize the intelligent dispatching and intelligent microgrid.However,the traditional LS-SVR algorithm is affected by the influence factors of microgrid,which leads to great uncertainty and other problems.The LS-SVR algorithm is improved from the perspective of sparsity and robustness.In the aspect of robustness improvement,the weighted factor is introduced to reduce the sensitivity of the residual sum of squares loss function due to the uncertainty caused by data anomalies.In the aspect of sparsity improvement,the data set is mapped into Hilbert space throughφ(·),and then the maximum independent solution is solved through the obtained Hilbert set,and then a new input vectorωis obtained.Finally,through the improvement of the above two aspects,the Lagrange multiplier is introduced to solve the linear regression equation.Finally,matlab simulation software is used to program the above algorithm,and ausgrid set data are used as samples for prediction.The experimental results show that the improved LS-SVR algorithm has smaller prediction error and higher prediction accuracy.

关 键 词：微网 短期负荷预测 鲁棒性 加权因子 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]