配电网空间负荷聚类及预测方法研究
发布时间：2018-07-22 16:01
【摘要】：空间负荷预测是配电网规划的前提和基础,负荷预测的精准性不仅影响电网的投资和后期运行,而且影响城市规划方案的合理性。本文在认真梳理和总结国内外先进理论和研究方法的特点的基础上,以提升空间负荷预测的精准性、实用性和适用性为目标,主要开展了以下三方面的研究:基于日负荷曲线的负荷分类模式研究。运用数据挖掘中的聚类技术对电力系统日负荷曲线进行分析,提出一种基于特性指标降维的日负荷曲线聚类方法——特性指标聚类(Pattern Index Clustering,PIC),通过负荷率、日峰谷差率等6个日负荷特性指标对日负荷曲线进行降维处理,利用基于聚类有效性修正的德尔菲方法配置各指标权重,以加权欧式距离作为相似性判据,对日负荷曲线进行聚类。算例分析结果表明所提方法运行时间短,鲁棒性好,提高了负荷曲线聚类质量,能直观反映典型负荷曲线的特点。利用该方法提取的典型日负荷曲线可作为空间负荷预测调研样本分类校验的依据,为调研样本提质奠定基础。考虑地域差异的空间负荷预测方法研究。针对基于智能算法的负荷密度指标法对样本依赖性强且在各地实际应用困难的不足,提出一种考虑地域差异的配电网空间负荷聚类及一体化预测方法。该方法首先通过大量调研得到分布不同地区、分属不同类型的负荷样本及所处地区信息;然后利用基于日负荷曲线的负荷分类校验及精选方法对所有调研样本进行分类精选;再根据区域分类、负荷分类对精选样本构成的全样本空间进行两级划分,得到分层级子样本空间;最后根据待测地块的属性信息对子样本空间进行匹配,选取与其最相似的子样本空间作为训练样本,构建支持向量机模型预测各地块的负荷密度,进而得到电力负荷的空间分布。工程实例分析表明了该方法的实用性和有效性。空间负荷预测指标体系优化研究。针对现有研究偏重对预测方法的理论创新和精度提升,缺乏对各地各类空间负荷分布规律研究的不足,提出一种基于聚类分析与非参数核密度估计的空间负荷分布规律研究方法。以浙江电网为例,对调研采集的空间负荷按城市发展类型、用地类型进行二级划分后,利用基于非参数核密度估计方法提取各类样本负荷密度的典型分布特征,结合实际对浙江11个城市的工业、商业、居住等多类空间负荷的分布规律进行分析研究,为配电网规划提供可靠支撑。
[Abstract]:Spatial load forecasting is the premise and foundation of distribution network planning. The accuracy of load forecasting not only affects the investment and later operation of power grid, but also affects the rationality of urban planning scheme. On the basis of combing and summarizing the characteristics of advanced theories and research methods at home and abroad, this paper aims at improving the accuracy, practicability and applicability of spatial load forecasting. This paper mainly studies the following three aspects: load classification model based on daily load curve. The daily load curve of power system is analyzed by using the clustering technology in data mining, and a new clustering method of daily load curve based on reducing dimension of characteristic index, pattern Index clustering (PIC), is proposed, which is based on load rate. Six daily load characteristic indexes, such as daily peak-valley difference ratio, are used to reduce the dimension of the daily load curve. The Delphi method based on clustering validity correction is used to configure the weights of each index, and the weighted Euclidean distance is used as the similarity criterion. The daily load curve is clustered. The results of example analysis show that the proposed method has the advantages of short running time and good robustness. It improves the clustering quality of load curves and can directly reflect the characteristics of typical load curves. The typical daily load curve extracted by this method can be used as the basis for the classification and verification of spatial load forecasting and investigation samples, which lays a foundation for improving the quality of the investigation samples. Research on spatial load forecasting method considering regional differences. In view of the shortage of intelligent algorithm based load density index method which is highly dependent on samples and difficult to be applied in various places, a method of spatial load clustering and integrated forecasting considering regional differences is proposed. In this method, firstly, a large number of load samples are obtained from different areas, which belong to different types of load samples, and then the load classification checking and selecting method based on daily load curve is used to classify and select all the investigation samples. Then according to regional classification and load classification, the whole sample space of selected samples is divided into two levels, and the sub-sample space is obtained. Finally, the sub-sample space is matched according to the attribute information of the plots to be measured. The most similar subsample space is chosen as the training sample, and the support vector machine model is constructed to predict the load density of each plot, and then the spatial distribution of power load is obtained. An engineering example shows the practicability and effectiveness of the method. Research on Optimization of Space load forecasting Index system. In view of the theoretical innovation and improvement of accuracy of the existing research on forecasting methods, there is a lack of research on the spatial load distribution law in various places. A research method of spatial load distribution based on clustering analysis and nonparametric kernel density estimation is proposed. Taking Zhejiang Power Grid as an example, the spatial load collected by investigation and acquisition is divided into two levels according to the type of urban development and the type of land, and the typical distribution characteristics of load density of all kinds of samples are extracted by using non-parametric kernel density estimation method. The distribution law of industrial, commercial, residential and other spatial loads in 11 cities of Zhejiang Province is analyzed and studied in order to provide reliable support for distribution network planning.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM715
[Abstract]:Spatial load forecasting is the premise and foundation of distribution network planning. The accuracy of load forecasting not only affects the investment and later operation of power grid, but also affects the rationality of urban planning scheme. On the basis of combing and summarizing the characteristics of advanced theories and research methods at home and abroad, this paper aims at improving the accuracy, practicability and applicability of spatial load forecasting. This paper mainly studies the following three aspects: load classification model based on daily load curve. The daily load curve of power system is analyzed by using the clustering technology in data mining, and a new clustering method of daily load curve based on reducing dimension of characteristic index, pattern Index clustering (PIC), is proposed, which is based on load rate. Six daily load characteristic indexes, such as daily peak-valley difference ratio, are used to reduce the dimension of the daily load curve. The Delphi method based on clustering validity correction is used to configure the weights of each index, and the weighted Euclidean distance is used as the similarity criterion. The daily load curve is clustered. The results of example analysis show that the proposed method has the advantages of short running time and good robustness. It improves the clustering quality of load curves and can directly reflect the characteristics of typical load curves. The typical daily load curve extracted by this method can be used as the basis for the classification and verification of spatial load forecasting and investigation samples, which lays a foundation for improving the quality of the investigation samples. Research on spatial load forecasting method considering regional differences. In view of the shortage of intelligent algorithm based load density index method which is highly dependent on samples and difficult to be applied in various places, a method of spatial load clustering and integrated forecasting considering regional differences is proposed. In this method, firstly, a large number of load samples are obtained from different areas, which belong to different types of load samples, and then the load classification checking and selecting method based on daily load curve is used to classify and select all the investigation samples. Then according to regional classification and load classification, the whole sample space of selected samples is divided into two levels, and the sub-sample space is obtained. Finally, the sub-sample space is matched according to the attribute information of the plots to be measured. The most similar subsample space is chosen as the training sample, and the support vector machine model is constructed to predict the load density of each plot, and then the spatial distribution of power load is obtained. An engineering example shows the practicability and effectiveness of the method. Research on Optimization of Space load forecasting Index system. In view of the theoretical innovation and improvement of accuracy of the existing research on forecasting methods, there is a lack of research on the spatial load distribution law in various places. A research method of spatial load distribution based on clustering analysis and nonparametric kernel density estimation is proposed. Taking Zhejiang Power Grid as an example, the spatial load collected by investigation and acquisition is divided into two levels according to the type of urban development and the type of land, and the typical distribution characteristics of load density of all kinds of samples are extracted by using non-parametric kernel density estimation method. The distribution law of industrial, commercial, residential and other spatial loads in 11 cities of Zhejiang Province is analyzed and studied in order to provide reliable support for distribution network planning.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM715
【相似文献】
相关期刊论文 前10条
1 符杨;朱兰;曹家麟;;基于模糊贴近度理论的负荷密度指标求取新方法[J];电力系统自动化;2007年19期
2 李双伟;范斗;王康元;邱家驹;鄢安河;;基于数据分区的负荷密度显示的实现[J];继电器;2008年03期
3 汤波;胡荣;符杨;宓佳t，
本文编号：2137988
本文链接：https://www.wllwen.com/kejilunwen/dianlidianqilunwen/2137988.html
