Research Of Power Equipment State Prediction And Maintenance Decision Method Driven By Data

Condition based maintenance(CBM)increases the availability of the power delivery equipment and reduces the maintenance cost,which enhances the safety and economy of the power system.Data analysis enables further application of CBM by mining various equipment condition data that may influence the equipment working state.State prediction is an important part of CBM.However,the analysis are primarily based on single or a few state parameters,and hence the potential failures of equipment can hardly be found or predicted.In this thesis,a data-driven method of association rule mining for equipment state parameters has been proposed by combining the Apriori algorithm and probabilistic graphical model,through which the disadvantage of Apriori algorithm is overcome.The proposed algorithm first finds out all frequent two-item sets using Apriori algorithm.Then the association rules are generated based on probabilistic graph.The algorithm overcomes the disadvantage that whenever the frequent items are searched the whole data items have to be scanned cyclically.The mined association rules are used in modifying the prediction results of a single state using radical based function neural network.The results show that the average prediction error has been reduced after taking the association rules into account.The maintenance strategy is how to achieve reliability and economy at the meantime based on state of the equipment.However,the existing maintenance decision process depends mostly on the experience of maintenance stuff.The deterioration process of power delivery equipment is a stochastic process.Therefore applying the theory of stochastic processes and considering the statistical characteristic of equipment parameters meets the reality better.In order to improve this condition,markov decision process(MDP)is applied in this paper to provide quantitative decision basis.First of all,two maintenance models are built,depending on whether online-monitoring devices are equipped on power delivery equipment.The impact of different maintenance on state transition is considered in the model.The difference of the two model are analyzed.Then the state transition probability are obtained based on steady state probability using markov process.Finally the optimum maintenance policy is solved using policy iteration method.Data of transformer and circuit breaker are applied to the model,and by changing failure cost the optimum policy varies.Results show that the proposed model can balance between maintenance and failure cost and provide the optimum maintenance decision.