基于ZINB层次贝叶斯智能电能表的可靠性预估

Reliability evaluation and prediction of smart meters based on ZINB hierarchical Bayesian

智能电能表可靠性影响着千家万户的用电安全。针对环境应力对运行智能电能表可靠性的影响,采用贝叶斯零膨胀负二项分布(ZINB)建立智能电能表可靠性预估模型,进行精确的可靠性短期评估与预测。首先,针对数据的零膨胀数据特征,引入零膨胀分析过程,建立广义线性ZINB层次贝叶斯模型,融合3省7地区温度、压强等环境因子与智能电能表故障数据;然后,采用马尔可夫链蒙特卡洛方法实现非共轭条件下的后验分布求解,得到智能电能表可靠性的置信区间分布和变化趋势;最后,结合3省区域的故障样本进行交叉验证,对比分析贝叶斯泊松分布等模型,验证模型的准确性,并得到智能电能表可靠性在95.7%。实验仿真显示,贝叶斯ZINB模型试验能够有效融合典型区域环境因素,准确预估智能电能表故障在特定环境因素下的变化特征。

The reliability of intelligent watt-hour meter affects the safety of electricity consumption of millions of households.Aiming at the influence of environmental stress on the reliability of smart meter,this paper uses zero-inflated negative binomial(ZINB)to establish the reliability prediction model,and to carry out accurate short-term reliability evaluation and prediction.Firstly,aiming at the characteristics of zero expansion data,a generalized linear ZINB hierarchical Bayesian model is introduced to integrate the environmental factors such as temperature and pressure in three provinces and seven regions.Then,the posterior distribution under non-conjugate conditions is solved by using Markov chain Monte Carlo method,and the confidence interval of the reliability of smart watt-hour meter is obtained.Finally,cross-validation is carried out with fault samples from three provinces and regions,and Bayesian Poisson distribution model is compared and analyzed to verify the accuracy of the model,and the reliability of smart watt-hour meter is 95.7%.The simulation results show that the Bayesian ZINB model test can integrate the typical regional environmental factors and accurately predict the change characteristics of smart watt-hour meter faults under specific environmental factors.