摘要
在线运行的超短期电力负荷预测用于潮流估计和电网调度,是制订发电计划的基础,具有显著的经济和安全意义。针对传统负荷预测算法对噪声数据的鲁棒性较差,预测精度无法进一步提升的缺陷,建立了一种基于半指数支持向量回归(SVR)的电力负荷预测模型。该模型提出了一种非线性的半指数损失函数,以解决负荷数据噪声导致的预测面偏移问题,从数学上泛化原始的铰链损失和硬间隔损失,通过设置不同的模型参数,获得了优于原始模型的分类效果。此外,通过引入对历史信息的挖掘理念,在输入量中加入了对时间的一阶和二阶微分,进一步提高了预测精度。最后通过理想数字模型仿真和使用真实的湖北省电网电力负荷数据进行预测,实验结果表明,提出半指数支持向量回归模型在速度上达到了在线运行的要求,而预测精度比现有方法有明显提高。
Online operation of ultra-short-term power load forecasting for tide estimation and grid dispatching,which is the basis for developing generation plans,has significant economic and safety implications.A semi-exponential support vector regression(SVR)-based power load forecasting model is developed to address the shortcomings of traditional load forecasting algorithms,which are less robust to noisy data and cannot further improve the forecasting accuracy.The model proposes a non-linear semi-exponential loss function to solve the problem of forecast surface shift caused by load data noise,mathematically generalises the original hinge loss and hard interval loss,and obtains a better classification effect than the original model by setting different model parameters.In addition,the prediction accuracy is further improved by introducing the concept of mining historical information and incorporating first-and second-order differentiation over time in the input quantities.Finally,through ideal numerical model simulations and forecasting using real power load data from the Hubei Provincial Grid,experimental results show that the semi-exponential support vector regression model proposed in this paper meets the requirements for online operation in terms of speed,while the forecasting accuracy is significantly improved over existing methods.
作者
王亮
王一鸣
侯威
贺元帅
纪超
Wang Liang;Wang Yiming;Hou Wei;He Yuanshuai;Ji Chao(School of Electronic Information,Xi'an Polytechnic University,Xi'an 710048,China;College of Information and Engineering,Xi'an University of Technology and Business,Xi'an 710200,China)
出处
《国外电子测量技术》
北大核心
2022年第12期164-170,共7页
Foreign Electronic Measurement Technology
基金
国家自然科学基金(51707141)
西安市科技计划项目(2019217114GXRC007CG008)资助。
关键词
电力负荷
预测
支持向量回归
鲁棒性
广义优化问题
electricity load
forecasting
support vector regression
robustness
generalized optimization problems
