【作者】 管春；

【导师】 周雒维；

【作者基本信息】 重庆大学 ， 电气工程， 2011， 博士

【摘要】 电能质量信息检测与分析是监督、改善电能质量的前提条件,对保证电力系统的安全经济运行以及用户用电安全具有重要的理论与实际意义。论文针对电能质量综合分析问题,研究新的电能质量数据压缩、扰动分类以及综合评估方法,搭建电能质量信息综合检测与分析系统平台,对包括微网在内的电力系统的电能质量信息进行全面检测、分析和评估。针对目前电能质量数据信息量越来越大,对数据压缩的要求越来越高,而目前数据压缩方法没有考虑三相电能质量数据之间相关性的问题,论文提出一种新的基于图像编码算法的三相电能质量数据压缩方法。电能质量数据压缩的本质就是去掉空间冗余、时间冗余等各种冗余。为了挖掘三相电能质量数据各相之间的冗余性,采用dq0变换转换三相电能质量数据;为了挖掘电能质量数据的循环间冗余性,按整数倍周期将一维电能质量数据转换为二维矩阵;为了有利于二维小波变换编码,应用数字图像平滑技术对二维转换后的图像矩阵进行平滑处理;同时结合提升格式的二维离散小波变换、图像多级树集合分裂编码以及DEFLATE编码算法对电能质量数据进行压缩。仿真实验结果表明,相对现有的电能质量数据压缩方法,所提出方案不仅可以有效压缩三相电能质量数据(同等压缩比的情况下,信噪比更高),而且还可以根据实际需求通过对压缩码率的控制灵活调节电能质量数据压缩性能。针对目前电能质量分类方法不能有效识别复合扰动的问题,论文提出一种新的基于多标签分类的电能质量复合扰动分类方法。在特征值提取方面,对电能质量复合扰动信号进行离散小波分解,提取各层分解系数的规范能量熵作为分类识别的特征向量。在分类器设计方面,基于k–近邻和贝叶斯准则,并在多标签排位分类法的基础上引入k–近邻贝叶斯多标签分类法;基于c-均值和径向基人工神经网络,并在多标签排位分类法的基础上引入c-均值径向基神经网络多标签分类法。同时基于多标签分类的评判指标进行仿真实验研究,实验结果表明,在不同的噪声条件下,相对现有的电能质量扰动分类方法,所提出方案可有效分类识别电压骤降、电压骤升、电压中断、脉冲暂态、谐波和电压波动等电能质量扰动及其组合而成的复合扰动,其平均精度值能达到95%以上。针对在电能质量综合评估中由于各种单一综合评价方法属性层次相异所导致的各方法评价结论存在差异的问题,论文提出一种新的基于组合评价法的电能质量综合评估方法。首先分别应用4种单一综合评价方法(包括层次分析法、模糊综合评判法、人工神经网络评价法以及灰色综合评价法)对电能质量进行单一综合评估;其次基于Kendall协同系数构造统计量??进行组合评价的事前一致性检验,排除不一致的单一综合评价方法结果以保证各方法的评价结果具有一致性;然后分别运用算术平均值法、Borda法和Copeland法这三种组合方法对通过事前检验的电能质量单一综合评价结果进行组合;最后,基于Spearman等级相关系数构造统计量t进行组合评价的事后一致性检验,并根据其大小选取其中的最佳组合结果作为电能质量组合评价的最终结论。通过对多个评估对象的电能质量综合评估实验验证了所提方法的有效性。基于所提出的电能质量数据压缩方法、扰动分类方法以及电能质量综合评估方法,论文采用虚拟仪器技术设计了电能质量信息综合检测与分析系统。该系统既能对电网甚至大型电力用户的电能质量信息进行全面的在线实时检测、分析,同时又能实现电能质量的离线分析,并对系统的电能质量进行合理的综合评估。该系统能实现电网系统化、智能化、网络化的电能计量,提高电能计量的质量与效率,并能为电能的准确计量提供参考。该系统能全程记录电网的电能质量信息数据,为事故分析提供丰富完整的实测数据记录,能对各种电能质量扰动进行有效识别,为采取相应的改善措施提供重要依据。更多还原

【Abstract】 Power quality information detection and analysis is the precondition to monitor and control the power quality. So it has significant theoretical and practical value for ensuring the safe operation of power systems and safe use of the consumers. Aiming at the problem of the power quality detection and analysis, new methods to classify the power quality disturbances, to compress the power quality data, to comprehensively evaluate the power quality are studied in the paper. Meanwhile, the comprehensive power quality information detection and analysis system is designed for detecting and analysing the power quality information of the power system including microgrid system.Nowadays, the demand for data compression has increased considerably due to the the huge amount of the power quality data detected. Aiming at the problem that the current power quality data compression methods do not consider the correlation of the three-phase power quality data, a new method of data compression of three-phase power quality data based on image coding algorithm is presented in the paper. The nature of power quality data compression is to remove the redundancy including the spatial redundancy, the time redundancy, and other redundancy. The three-phase power quality data are transformed by dq0 transform to eliminate the redundancy of the three-phase data. The obtained one-dimensional power quality data are transformed to two-dimensional matrix according to the integral multiples of period to eliminate the redundancy between cycles. The two-dimensional image is smoothed based on image smoothing algorithm in order to make it more suitable for two-dimensional wavelet transform coding. Meanwhile, data compression is processed based on two-dimension wavelet transform, image set partitioning in hierarchical tree coding and DEFLATE coding algorithms. The simulation results show that the proposed methods can not only get higher signal noise ratio relative to the existing methods of the power quality data compression under the same compression ratio, but also control the power quality compression performance according to the practical requirement flexibly.Aiming at the problem that it is difficult to classify the category of multiple power quality disturbances for the current power quality disturbance classification methods, a new multiple power quality disturbances classification method is presented based on multi-label classification in the paper. In terms of the feature extraction, the signal of multiple power quality disturbances is decomposed by discrete wavelet transform, and the norm energy entropy of the wavelet coefficients of each level are extracted as eigenvector for classfication. In terms of the classifier design, the methods of the k-nearest neighbor Bayesian rule and the C-means RBF neural network based on ranking approach of the multi-label classification are introduced respectively to solve the problem of the multiple power quality disturbances classification. At the same time the simulation experiment is conducted based on the evaluation metrics of the multi-label classification. The simulation results show that the proposed methods can effectively recognize the multiple power quality disturbances including voltage sag, voltage swell, interruption, impulsive transient, harmonics, voltage fluctuation and their compound ones effectively under different noise conditions relative to the existing methods and the average precision achieves 95%.Aiming at the problem of inconsistencies in the conclusions, which is caused by adopting many different single comprehensive evaluation methods to evaluate the same objective, a new method of power quality comprehensive evaluation based on the combined evaluation method is presented in the paper. Firstly, the power quality is evaluated respectively based on four single comprehensive evaluation methods including the analytic hierarchy process, fuzzy comprehensive evaluation, artificial neural network evaluation and gray comprehensive evaluation. Secondly, in order to ensure the consistency of different single comprehensive evaluation results, the statistics is constructed based on Kendall correlation to check up the coherence of different evaluation methods and eliminate the inconsistencies results of the single comprehensive evaluation methods. Then, the arithmetic mean, Borda and Copeland methods are used respectively to combine the single comprehensive evaluation results which have passed the pre-test. Finally, the final result of the power quality combined evaluation is obtained according to the size of the statistics t which is established based on the Spearman’s rank correlation coefficient after post-test. The effectiveness of the proposed method is proved through evaluating the power quality of several locations.The comprehensive power quality information detection and analysis system is designed using the virtual instrument technology based on the former proposed methods of power quality classification, power quality data compression and power quality comprehensive evaluation. This system can not only detect and analyse the power quality information of the power grid even large power customers on-line, but also evaluate the power quality of the power system reasonably off-line. This system enables the systematic, intelligent networked detection of the electric energy. So it can improve the efficiency of the electric energy measurement and provide reference for accurate electrical energy measurement. This system can not only record the whole power quality information data of the power grid to provide rich real information for accident analysis, but also classify the power quality disturbance to provide important evidence for the corresponding improvement.更多还原