Reliability Assessment and Maintenance Optimization of Oil-Immersed Transformers in Windblown Sand and Alpine Regions: A Bayesian Copula–Mixture Weibull Framework

High-voltage oil-immersed transformers operating in windblown sand and alpine regions are exposed to multiple environmental and operational stresses that may affect their degradation and reliability. To address limitations associated with single-distribution lifetime modeling, independent component assumptions, static operating-condition corrections, and small-sample estimation, this study proposes a Bayesian Copula–mixture Weibull framework for reliability assessment and maintenance optimization. A three-component mixture Weibull model is used to represent multi-stage lifetime behavior, while a Gumbel Copula describes statistical dependence among component lifetime variables. A four-dimensional scenario-based operating-condition correction incorporates altitude, windblown sand, temperature, and load effects. Bayesian MCMC estimation is applied to small-sample, right-censored lifetime data, and an imperfect-maintenance age-reduction model is integrated with a reliability-constrained maintenance optimization formulation. On the common estimation dataset, the complete integrated model yields the highest in-sample R2 and the lowest in-sample MAE and MSE among the compared model configurations; these metrics quantify goodness of fit and are not interpreted as evidence of out-of-sample predictive superiority. Under the prescribed severe operating-condition scenario, the corrected model exhibits substantially faster reliability decay than the uncorrected reference. Under the prescribed operating-condition, maintenance, reliability, and economic assumptions of the case study, the optimized differentiated maintenance strategy yields an approximately 24.5% lower accumulated cost over the specified 8-year reporting horizon than the fixed-interval strategy. This value represents a conditional scenario result rather than a population-level or full-lifecycle cost-saving estimate. These results demonstrate the applicability of the proposed framework for scenario-based reliability assessment and maintenance decision support under difficult operating conditions.