Condition Comprehensive Assessment Of Power Transformers Based On Online Monitoring And Software Development

Condition assessment is an important basis of operation, repair and maintenance of transformer. The accuracy of condition assessment for transformer remains to be improved. On the one hand, the redundancy of evaluation parameter and its subjective selection criterion are the difficult problems. On the other hand, the stability and reliability of on-line monitoring data is often affected by factors such as sensors, environmental changes. So based on threshold decision method, it is difficult to discover the unusual situation in time, and identify noise data.Therefore, based on the in-depth analysis of demand and the present research for condition assessment, the paper constructs the multi-level comprehensive fuzzy evaluation model on the basis of key parameters system and D-S evidence theory. Firstly, an index system of key parameters is extracted by using factor analysis method. Then, analytic hierarchy process(AHP) is applied to obtain the optimal weights of each index in evaluation system, and results of each evaluation layer are computed based on their membership functions. Finally, taking results of project layers as inputs, the fusion evaluation decision is made based on D-S evidence theory. Example analysis shows the assessment model can effectively and reasonably to comprehensively evaluate transformer state.Moreover, a fast data detection method of on-line monitoring data is proposed, based on multivariate time series, in view of the multidimensional monitoring data and the angle of time sequence, innovatively record the time and type of abnormal points by sliding time window, and finally establish a judgment candidate model of a set of abnormal data. Then, make a multi- dimensional abnormal data detection on this set by using the clustering algorithm. Examples show that the method can detect the abnormal state of on-line monitoring data flows in realtime, which is of high practical value.At last, this thesis developed the software system for power transformer condition comprehensive assessment and online monitoring. This system offers functions of historical data of monitoring and management, embedded with the abnormal data processing algorithm, at the same time shows the state evaluation results based of online – monitoring data sets, and provides solid basis for transformer condition maintenance.