In order to address the complex and diverse risk factors in substation operation and maintenance that are difficult to identify, a risk assessment model based on a dynamic Bayesian network (DBN) was proposed. Field research and literature review were used to investigate risk factors in substation operations, and rough set reduction theory was applied to screen and optimize risk indicators. This led to the establishment of a risk assessment system comprising 5 primary indicators and 17 secondary indicators. Subsequently, an interpretive structural model (ISM) was integrated with a temporal dimension to construct the DBN. Fuzzy set theory(FST) was used to investigate the quantification of expert evaluations, enabling the calculation of prior and transition probabilities for network nodes. The Leaky Noisy-or Gate extension model was introduced to refine conditional probabilities, and the bidirectional reasoning capability of the DBN was leveraged to achieve dynamic risk assessment. The results show that the probability of operational risk occurrence, derived from forward reasoning is 0.611, while reverse reasoning identified human, managerial, and equipment-related factors as the most significant primary risk indicators. Among secondary indicators, inadequate safety supervision, worker fatigue, lack of concentration, and equipment overheating are identified as critical contributors to operational risks. It is concluded that empirical case analysis validated the feasibility of the model by inputting real-world data into the framework. This research provides new theoretical and methodological support for the dynamic assessment of substation operation and maintenance risks.