Research On Transformer Condition Assessment Method Based On Feature Optimization And Parameter Fusion

With the continuous deepening of the digital transformation of the power system,more and more applications of sensor networks and intelligent measurement technologies have provided powerful data support for the implementation of accurate assessment of equipment status under panoramic monitoring data.Power transformer is a kind of core device in power system,and its operation status directly affects whether the power system can operate safely and stably."Compulsory repairs should be repaired" and "intelligent repairs" have become the inevitable trend of equipment repairs.When evaluating the status of power transformers,the indicator system is complicated and the data types are diverse,and due to problems such as communication equipment failures and data specifications,there are a large number of vacancies and abnormal values in the relevant monitoring indicators.Therefore,in the absence of information,research on power transformer feature selection,index weight distribution and equipment status evaluation has important theoretical and practical significance.This article first starts with the current status of power transformer maintenance,monitoring methods,data distribution characteristics,etc.,through the analysis of power transformer fault types and mechanisms,establishes a basic index system based on component-performance-fault types,and expounds from a multi-modal perspective.The power transformer holographic data perception framework is developed,and the power transformer information decision table based on equipment components and fault types is established according to the differentiated threshold screening of actual cases.The use of data set cutting reduces the difficulty of calculation while ensuring the difference of samples.A relative attribute reduction method based on the positive region of rough set is proposed to obtain the "core" feature of power transformers with missing information.In addition,in view of the problem of large data volume and data dimension,the above algorithm is optimized and improved by taking advantage of the faster convergence of PSO in obtaining the optimal solution.Secondly,considering the multi-modality and distribution differences of equipment "core" characteristics,based on the current CIM data modeling standards,the modeling methods of online data and offline data are explored.And based on the constraints of Task Flow Graph and REST service mechanism,a data integration mechanism based on task queue list mode is proposed.Experimental results show that compared with the traditional integration mechanism,it requires less resource space and stronger flexibility.Finally,in view of the existing "core" attribute characteristics,the complexity of power transformer evaluation and fault diagnosis is simplified through a progressive multi-level evaluation model.That is,a combination of subjective and objective weights based on the combination of analytic hierarchy process and association rules is used to avoid evaluation bias caused by excessive subjectivity and insufficient samples.When the actual evaluation process is due to technical level or production restrictions,the relevant indicators are missing.The weight is redistributed.Use DSm T to integrate the membership of various indicators,faults,performance,and components to complete the overall equipment evaluation.At last,through the analysis of the simulation results of the case,it is proved that the evaluation model described in this article has good application value and reference significance for equipment evaluation under the situation of lack of guidance information and integration of multiple system indicators.