基于改进支持向量机的短期电力负荷预测研究
发布时间：2018-10-12 11:14
【摘要】：有效准确的电力负荷预测既是使电网安全、经济运行的有力保障，也为切实解决人民群众最关心、最直接、最现实的用电问题提供了先决服务。因此，对该领域的研究一直是学术界的热点问题。 支持向量机（Support Vector Machine，简称SVM）是一种新兴的学习机器，具有较为完备的理论基础和较好的学习性能，成功解决了神经网络难以克服的诸多问题，被称为神经网络的替代算法。因此，本论文将其引入到电力系统的短期负荷预测中来。在研究中本文发现，负荷预测的影响因素有很多，有些因素是可以在特定情况下被去除的。在进行预测时，如果不对众多因素（属性)进行处理，势必会提高预测模型的复杂程度并影响其实现效果，从而导致预测失准等问题。若仅凭经验来对各属性进行约减与提取，则又会因为缺乏依据，导致一些有用的信息被去除，同样会致使预测失准。 针对上述问题，本文进行了进一步研究。首先，采用粗糙集的有关理论与方法，对基于支持向量机的电力负荷预测技术进行改进，通过属性约减与特征提取等工作，使得有用的信息被完整保留，，无用的信息被基本剔除，在最大限度上减少了外界不良因素对负荷预测系统的干扰。其次，进行算例分析与效果比较，对照改进前后的负荷预测技术在预测效果上的差别，从而验证改进方案的有效性与可行性。通过验证发现，上述改进所得到的新技术确实取得了更加精确的预测效果。通过分析认为，其对解决电力负荷预测这一与企业管理者的决策息息相关的热点问题又提供了一套更加合理的方案。
[Abstract]:Effective and accurate power load forecasting not only ensures the security and economic operation of the power grid, but also provides a preliminary service for solving the most concerned, direct and realistic problems of electricity consumption among the people. Therefore, the research in this field has been a hot topic in academic circles. Support Vector Machine (Support Vector Machine,) is a new learning machine with relatively complete theoretical foundation and better learning performance. It has successfully solved many problems that can not be overcome by neural network and is called the substitute algorithm of neural network. Therefore, this paper introduces it into short-term load forecasting of power system. In this paper, it is found that there are many factors affecting load forecasting, and some factors can be removed under certain circumstances. In forecasting, if many factors (attributes) are not dealt with, the complexity of the prediction model will be increased and the effect of its implementation will be affected, which will lead to the misalignment of prediction and other problems. If each attribute is reduced and extracted only by experience, some useful information will be removed because of lack of basis, and the prediction will also be inaccurate. In view of the above problems, this paper has carried on the further research. Firstly, the theory and method of rough set are used to improve the power load forecasting technology based on support vector machine. Through attribute reduction and feature extraction, the useful information is preserved completely. Useless information is basically eliminated, which minimizes the interference of external adverse factors to the load forecasting system. Secondly, an example analysis and effect comparison are carried out to verify the effectiveness and feasibility of the improved method by comparing the difference of forecasting effect between before and after the improved load forecasting technology. Through verification, it is found that the new technique obtained by the above improvements has achieved a more accurate prediction effect. Through the analysis, it provides a more reasonable scheme for solving the hot problem of power load forecasting, which is closely related to the decision of enterprise managers.
【学位授予单位】：华北电力大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP181;F426.61
本文编号：2265914
[Abstract]:Effective and accurate power load forecasting not only ensures the security and economic operation of the power grid, but also provides a preliminary service for solving the most concerned, direct and realistic problems of electricity consumption among the people. Therefore, the research in this field has been a hot topic in academic circles. Support Vector Machine (Support Vector Machine,) is a new learning machine with relatively complete theoretical foundation and better learning performance. It has successfully solved many problems that can not be overcome by neural network and is called the substitute algorithm of neural network. Therefore, this paper introduces it into short-term load forecasting of power system. In this paper, it is found that there are many factors affecting load forecasting, and some factors can be removed under certain circumstances. In forecasting, if many factors (attributes) are not dealt with, the complexity of the prediction model will be increased and the effect of its implementation will be affected, which will lead to the misalignment of prediction and other problems. If each attribute is reduced and extracted only by experience, some useful information will be removed because of lack of basis, and the prediction will also be inaccurate. In view of the above problems, this paper has carried on the further research. Firstly, the theory and method of rough set are used to improve the power load forecasting technology based on support vector machine. Through attribute reduction and feature extraction, the useful information is preserved completely. Useless information is basically eliminated, which minimizes the interference of external adverse factors to the load forecasting system. Secondly, an example analysis and effect comparison are carried out to verify the effectiveness and feasibility of the improved method by comparing the difference of forecasting effect between before and after the improved load forecasting technology. Through verification, it is found that the new technique obtained by the above improvements has achieved a more accurate prediction effect. Through the analysis, it provides a more reasonable scheme for solving the hot problem of power load forecasting, which is closely related to the decision of enterprise managers.
【学位授予单位】：华北电力大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP181;F426.61
【参考文献】
中国期刊全文数据库 前10条
1 康重庆,夏清,张伯明;电力系统负荷预测研究综述与发展方向的探讨[J];电力系统自动化;2004年17期
2 张林,刘先珊,阴和俊;基于时间序列的支持向量机在负荷预测中的应用[J];电网技术;2004年19期
3 吴宏晓,侯志俭;基于免疫支持向量机方法的电力系统短期负荷预测[J];电网技术;2004年23期
4 杨延西,刘丁;基于小波变换和最小二乘支持向量机的短期电力负荷预测[J];电网技术;2005年13期
5 陆建宇;王亮;王强;吴江;刘涌;;华东电网气象负荷特性分析[J];华东电力;2006年11期
6 杨镜非,谢宏,程浩忠;SVM与Fourier算法在电网短期负荷预测中的应用[J];继电器;2004年04期
7 祝志慧;孙云莲;季宇;;基于经验模式分解和最小二乘支持向量机的短期负荷预测[J];继电器;2007年08期
8 吴军基,倪黔东,孟绍良,刘皓明;基于人工神经网络的日负荷预测方法的研究[J];继电器;1999年03期
9 张学工;关于统计学习理论与支持向量机[J];自动化学报;2000年01期
10 杜京义;侯媛彬;;基于遗传算法的支持向量回归机参数选取[J];系统工程与电子技术;2006年09期
本文编号：2265914
本文链接：https://www.wllwen.com/guanlilunwen/shengchanguanlilunwen/2265914.html
