To address the issues of strong subjectivity in weighting and insufficient multi-source information fusion in the insulation system evaluation of pumped storage units

a multi-dimensional data-driven method for comprehensive insulation health assessment was proposed. Based on the multi-physics stress coupling mechanism

a hierarchical evaluation framework was constructed

encompassing three dimensions(electrical

thermal

and mechanical)and 18 specific indicators. Fuzzy membership functions of the 鈥渟maller-the-better鈥?and 鈥渕edium-the-better鈥?types were designed to quantify the health status of each indicator. A hierarchical weighting strategy was proposed: the entropy weight method(EWM)was applied for objectiv

e weighting at the secondary indicator level

while a preference coefficient was utilized to integrate the weights from the analytic hierarchy process(AHP)and EWM at the primary indicator level. Validation using measured data from two operational cycles of a specific power station demonstrates that Samples X

1

and X

2

were evaluated as 鈥淕ood鈥?with Good membership degrees of 0.426 8 and 0.473 7

respectively). Sample X

3

exhibiting degradation in the thermal and mechanical dimensions(with a Qualified membership degree reaching 0.492 9)

was accurately identified as 鈥淨ualified鈥? Comparative analysis indicate that the proposed method can effectively identify samples with critical degradation

avoiding the misclassification of such samples as 鈥淗ealthy鈥?or the occurrence of ranking inversions common in traditional methods. The evaluation results are highly consistent with the actual operational status of the units.