Aiming at the problem of ensemble learning having many parameters and lack of efficient and accurate parameter optimization methods, this paper proposes a bayesian optimized random forest (RF) transformer fault diagnosis method. The method uses multiple parallel CART decision trees to form an RF fault diagnosis model. Then, the Gaussian process is used as a probabilistic proxy model and a lifting strategy (PI) as an acquisition function to construct a bayesian optimization algorithm for RF model parameter optimization. At the same time, bayesian optimization is performed on support vector machine (SVM) and K nearest neighbor (KNN) respectively, and fault diagnosis is performed by three models. On the RF model, bayesian optimization and random search optimization are compared to optimize performance. The experimental results show that the RF integration strategy has higher fault diagnosis accuracy than the single classifier. When bayesian optimization method is applied to various algorithms, the diagnostic performance is significantly improved compared with the unoptimized. Compared with random search optimization, bayesian optimization method can find better model parameters with higher efficiency.