大数据环境下变压器状态评估的关联集对分析方法研究

发布时间：2018-12-14 14:29

【摘要】：电力变压器是电力系统的关键枢纽设备，状态检修替代传统的定期检修是电力行业发展的必然趋势，而实现状态检修的前提是依据设备检测数据信息对变压器的运行状态进行准确的评估。随着国家智能电网建设的不断发展，电力设备状态监测所采集的状态量数据呈指数级增长，与之相关的区域电网内变电设备的大数据分析研究就显得越来越重要，因此，围绕故障诊断和状态评估所必须的数据处理和信息挖掘技术对智能电网设备侧业务而言具有重大意义。当前，电网公司针对大数据研究的工作重点之一就是对海量的数据做统计性的搜索、比较、分类、聚类等相关性分析。而针对设备侧的电力变压器状态评估，大多数方法要么侧重于方法理论的研究，忽视了变压器状态检测的数据信息的关联性分析及内在联系，要么没有注意变压器状态信息具有模糊、随机和信息不完全所导致的不确定性问题。有鉴于此，本文提出了智能电网大数据环境下融合集对分析和关联规则的变压器故障诊断方法。首先，本文依据关系型数据库理论和相关标准，结合实际运行经验，选取最具代表性且能够准确有效地反映变压器运行状况的状态量来作为故障类型和故障征兆；利用关联规则的支持度对故障类型与故障征兆进行分类管理，分析故障类型与故障征兆之间的相互耦合联系，并约简各故障类型的故障征兆集合，，最终建立状态评估模型；在权重系数的确定方法上，通过对比计算关联规则的置信度，得到故障征兆的常权重系数，解决了专家个人主观意见影响权重准确性的问题。同时在故障类型中使用因素空间变权理论，根据每个故障类型的评分情况确定其权重系数，能够有效避免常权重系数下单个故障类型不能准确反映变压器整体健康状况的缺陷。其次，对电力变压器的运行状态进行等级划分，确定联系度的元数，并构建联系度表达式，提高了变压器状态评估及故障诊断系统评价精确度；引入模糊理论中相对劣化度和隶属度来构建同异反评价矩阵，模糊理论的引入具有一定科学依据，可以有效避免专家意见或经验的主观性的问题；对于多元联系度表达式，采用均分法处理差异度系数，结合关联规则确定的权重系数得出变压器整体运行状态和各故障类型的联系数值，对比状态等级的划分从而对变压器健康状况进行评估和故障诊断。最后，给出算例和统计结果来验证本文所提方法的正确性与有效性。与关联规则和集对分析分别单独应用于变压器状态评估方法相比，本文方法具有较高的正判率，并且在多故障诊断方面也表现优异。
[Abstract]:Power transformer is the key equipment of power system. It is the inevitable trend of power industry to replace the traditional periodic maintenance with condition-based maintenance. The premise of condition maintenance is to accurately evaluate the operating state of transformer according to the information of equipment detection data. With the continuous development of national smart grid construction, the state data collected by power equipment condition monitoring is increasing exponentially. The analysis and research of big data, which is related to the transformer equipment in regional power grid, is becoming more and more important. The data processing and information mining techniques necessary for fault diagnosis and state assessment are of great significance to the smart grid equipment side services. At present, one of the focuses of research on big data is to do statistical search, comparison, classification, clustering and other correlation analysis of massive data. For the power transformer condition evaluation on the equipment side, most methods either focus on the research of method theory and ignore the correlation analysis and internal relation of the data information of transformer condition detection. Either the transformer state information is fuzzy, random and incomplete information caused by uncertainty. In this paper, a transformer fault diagnosis method combining set pair analysis and association rules under big data environment of smart grid is proposed in this paper. First, according to the relational database theory and related standards, combined with the actual operation experience, select the most representative and can accurately and effectively reflect the transformer operation status as the fault type and fault symptoms; The fault type and fault symptom are classified and managed by using the support degree of association rules, the coupling relation between fault type and fault symptom is analyzed, and the fault symptom set of each fault type is reduced to establish the state evaluation model. In the method of determining weight coefficient, by comparing and calculating the confidence degree of association rule, the constant weight coefficient of fault symptom is obtained, which solves the problem that the expert's subjective opinion affects the accuracy of weight. At the same time, the factor spatial variable weight theory is used in the fault type, and the weight coefficient is determined according to the grading of each fault type, which can effectively avoid the defect that the order type of the constant weight coefficient can not accurately reflect the overall health condition of the transformer. Secondly, the operation state of power transformer is classified, and the number of connection elements is determined, and the expression of connection degree is constructed, which improves the accuracy of the evaluation of transformer condition and fault diagnosis system. The relative deterioration degree and membership degree in fuzzy theory are introduced to construct the same difference and inverse evaluation matrix. The introduction of fuzzy theory has certain scientific basis and can effectively avoid the subjective problem of expert opinion or experience. For the expression of multivariate relation degree, the differential coefficient is treated by the method of average division, and combined with the weight coefficient determined by association rules, the overall operating state of transformer and the relation value of each fault type are obtained. Compare the level of state to assess and diagnose the transformer's health. Finally, an example and statistical results are given to verify the correctness and validity of the proposed method. Compared with the association rules and set pair analysis which are used in transformer condition evaluation separately, the proposed method has higher positive judgment rate, and it also performs well in multi-fault diagnosis.
【学位授予单位】：华中科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TM41

【相似文献】