Abstract:
To address the problems that traditional transformer state assessment methods are relatively simple and have difficulty handling ambiguity and conflicts in multi-source information, this paper proposes an intelligent assessment method based on multi-source data fusion. First, a multidimensional indicator system covering dissolved gas in oil, insulating oil characteristics, and electrical performance is constructed. Second, an improved membership function is used to characterize the fuzzy boundaries of operating states, and the optimal combined weights of the indicators are determined based on a sum of squared deviations maximization model, thereby completing the fuzzy comprehensive assessment of each indicator group. Finally, the assessment results are used as evidence, and D-S evidence theory is employed for multi-source fusion and decision making, outputting state assessment conclusions with quantified confidence degrees. Verification using measured data from a 110 kV transformer shows that the proposed method effectively identifies early comprehensive degradation of the equipment. After multi-source fusion, the confidence degree of the “attention” state increases from 0.401 to 0.503, making the method more forward looking and reliable than traditional methods. This provides a new approach for accurate transformer state assessment.