Reliability evaluation based on the failure rate data is an important basis for the health status management and maintenance of smart meters. However, the small sample characteristics of outliers and failure rates limit the evaluation performance of traditional smart energy meter reliability prediction models. Therefore, a prediction model of smart meter failure rate under multi?environment stress based on weighted local outlier factor and Gaussian process regression is proposed in this paper. Firstly, a weighted local outlier factor is employed with the model to identify and then delete potential outliers in failure rate data sets; then, different kernel functions are selected to match the characteristics of multiple stress inputs in typical environments, and choose the best one. Finally, the interval change of the 95% confidence level of the failure rate is predicted by the posterior distribution of the Gaussian process, and the interval reliability is obtained based on this. Case analysis of fault samples of smart meters in two typical environmental areas shows that the proposed model could effectively predict the trend of failure rate of smart meters under multi?environmental stress, and could accurately solve its reliability.