Research On Reliability Evaluation Method Of Data Center Power Distribution System

In reliability assessment of power system,the traditional network method needs to be available based on accurate failure rate of components.However,the less sample capacity of data center power distribution system located at the end u.ser level,the difficulty of carrying out the test,and the lack of failure data lead to the impossibility of obtaining the accurate failure probability of the component.Also,the complicated UPS power supply device is used in the data center power distribution system to impro ve the power supply reliability of the end user equipment,which cannot be solved and analyzed by the network method.In this paper,by introducing multi-source evidence,this paper adopts improved evidence theory to synthesize and correct the statistical component failure rate,and to get the componen t failure rate including uncertainty.In addition,The Bayesian network method is u.sed to obtain the probability interval of the system top event.Firstly,this paper analyzes the fact that the Dempster evidence synthesis formula tends to be not reliable in the evidence synthesis of high-reliability equipment.Then,it proposes an improve.d evidence synthesis formula based on evidence distance,which can transform the evidence of insufficient conflict into more credible results.Secondly,this paper uses Markov theory to analyze the state transition process of UPS power equipment,and proposes a mathematical model of the failure probability by analyzing the fault path of UPS parallel system.Finally,this paper uses Bayesian network method to obtain the probability of top event occurrencs,and analyze the assessment result of reliability for data center power distribution system by means of component probability im portance and cognitive importance.