Study Of Condition-based Maintenance Optimization Strategy For Transformer Based On Semi-Markov Decision Process

Transformer is large electric power equipment. Its stable operation concerns the stability of the grid. If the transformer has power failure, it will has a profound impact on the stability and safety of the grid. Therefore, it is necessary to pay attention to its real-time running state and fix it in time if there is deterioration. With the development and the promotion of condition-based maintenance, engineers gradually notice the importance of transformer's actual running state to the maintenance of transformer. The semi-markov decision process is an improvement and development of Markov decision process, which can describe the deterioration, inspection and the maintenance process and at the same time it takes the sojourn time into account so it has a good applicability.This paper, by considering the state inspection frequency changing with the deterioration degree, establishes an optimized maintenance strategy of transformer based on semi-markov decision process.Firstly, this paper has a push-over analysis of semi-markov decision process. The analysis includes the definition and solution of state transition probability and state transition probability density, state transition probability transition matrix, state transition probability density matrix and steady state duration time.Secondly, it has an analysis of the state diagram maintenance model. Based on the state diagram model, the paper introduces the detection and maintenance state to get the entire diagram model and then reform and simplify the state diagram model combining the actual condition based maintenance.Lastly, it has an analysis of transformer's maintenance strategy based on semi-markov decision process. In view of different state inspection frequency having different influence on transformer, it introduces the mathematical method of optimizing frequency of inspections and the maintenance strategy and policy iteration algorithm based on semi-markov decision process. It also test and verify the validity of the strategy by simulation analysis.