Life prediction of electric energy meters based on multi-source data fusion and stochastic degradation modeling

The widespread application and continuous improvement of intelligence of electric energy meters have made their reliability issues prominent. Therefore, this paper proposes an adaptive residual life prediction method for electric energy meters based on multi-source data fusion and random degradation modeling. Initially, a strong correlation is identified between the degradation features of electric energy meters and the dual stress factors of temperature and humidity through data mining during the acceleration degradation test. Then, a multivariate stochastic degradation model is developed, considering multiple degradation factors, to predict the lifespan of intelligent energy meters under normal operating conditions. This model allows for adaptive remaining life prediction in complex operating environments, reducing the time required for assessment. A three-phase temperature and humidity dual stress accelerated degradation test (ADT-DH) was conducted to validate the method. The results indicated that this approach's reliability prediction closely aligned with the degradation trends observed in the intelligent energy meters, outperforming the conventional Peck model. This methodology significantly reduces testing costs and time, demonstrating its practical value for engineering applications.