摘要：As one indispensable part of power systems, the reliable-operated power transformers are vital for energy transmission, whereas they are remarkably threatened by potential fault events. To achieve the satisfying and valid operation of power transformers, any fault events that may impact their health ought to be evaluated and early warned. With such motivations, this paper presents original insights on the assessment of power transmission health states via their internal dissolved gas, and an enhanced Association Rule Mining (ARM) model incorporating the analysis of High-Impact-Low-Probability (HILP) components, as well as a dynamic fault event risk evaluation approach, is proposed. The first step is to differentiate the risky components. Unlike the standard ARM, the rarely occurred components in each feature can also be assessed explicitly as the common components to explore the underlying HILP components in the proposed model, rather than just being viewed as trivial data and directly omitted. The second step is to rate the risk level of each risky component. A component importance measure-based evaluation approach is deployed to assess the corresponding risk weights of distinguished risky components. In this approach, the risk weight is determined straightforwardly via the impacts of each component on the variation level of total risks in the system, rather than simply by its frequency of occurrence or data share. Finally, the parameters of the risk weight evaluation approach can be dynamically adapted in an adjustment framework as well. This model is testified through an empirical case study, and the leading results can demonstrate its flexibility and robustness during real applications.
