The basic error of electric energy metering equipment is greatly affected by environmental stress. And the relationship between multiple electrical stresses hard to be described under typical environmental stress. To address these issues, an improved particle swarm with long short-term memory network (IPSO-LSTM) is proposed to predict the basic error of electric energy metering equipment. Firstly, various stresses data in typical environment are normalized and data set allocation are preprocessed. To solve fluctuation trend of the error time series data, an extruded LSTM network architecture is established to analyze the variation trend characteristics of the error data. In this way, the nonlinear fitting ability of the model under multiple stress data is enhanced. Then, the improved PSO algorithm is used to optimize the model′s hyperparameters to reduce the influence of hyperparameters and improve the prediction performance of the model. In the experimental part, the proposed algorithm is evaluated and analyzed according to several electric energy metering equipment of one company. The environmental stress and error data are both considered by typical operating laboratories in Xinjiang region. The results show that the sample prediction accuracy indexes RMSE values reach 1. 08% and 1. 19% , respectively. And MAE values reach 0. 88% and 0. 96% , respectively.