Abstract: A self-adaptive monitoring method based on dynamic weight ensemble transfer learning is proposed to address the problem of negative transfer in traditional transfer learning caused by the difference in data distribution between the source and target domains in intelligent substation equipment status monitoring. Build multiple heterogeneous joint distribution adaptation base learners to capture domain shared information from different feature subspaces. Design a dynamic weighting strategy based on local maximum mean difference, adaptively adjust the integrated weights according to the degree of distribution deviation of the target sample in the local neighborhood of the source domain, and improve the model"s response ability to changes in operating conditions. The experimental results show that the proposed method achieves a macro average Matthews correlation coefficient and balance accuracy of 87.70% and 93.05%, respectively, under four types of device states, verifying its robustness and monitoring accuracy under distribution offset.