基于贝叶斯网络的智慧园区数据中心供配电系统设计

Design of Power Supply and Distribution System for Smart Park Data Center Based on Bayesian Network

为了提高供配电系统配置的合理性,保障数据中心可靠运行,设计了一种基于贝叶斯网络的智慧园区数据中心供配电系统。结合可扩展性、可控性与可维护性等原则,确定系统整体结构;利用模块化设计思想,将系统分为市配电系统、不间断电源、照明配电、防雷保护四大模块;针对不间断电源,综合分析其负载率、可用性与电池放电时间三种因素,选择最佳电源类型;利用贝叶斯网络算法,根据各模块的运行样本信息,建立故障率的密度函数,计算其先验概率,并将其变换为伽马分布形式,分别获取故障率的平均值和方差,通过对超参数的计算得出供配电系统故障预测结果,完成系统设计。仿真实验表明,该方法可以保证输出电压平稳,同时可有效避免雷击故障。

In order to improve the rationality of power supply and distribution system configuration and ensure the reliable operation of data center,a power supply and distribution system of smart park data center based on Bayesian network is designed.Combined with the principles of scalability,controllability and maintainability,the overall structure of the system is determined.Using the idea of modular design,the system is divided into four modules,i.e.,municipal power distribution system,uninterruptible power supply,lighting power distribution and lightning protection.Among them,the load rate,availability and battery discharge time of uninterruptible power supply are comprehensively analyzed to select the best type of power supply.Using Bayesian network algorithm,according to the operation sample information of each module,the density function of failure rate is established,the prior probability is calculated,and it is transformed into gamma distribution form to obtain the average value and variance of failure rate,respectively.Through the calculation of super parameters,the fault prediction results of power supply and distribution system are obtained,and the system design is completed.Simulation results show that this method can ensure the stability of output voltage and effectively avoid lightning fault.