Traditional methods have limitations such as difficulty in adapting to complex and changeable environments, poor sensitivity and high complexity when dealing with practical problems. To achieve the goal of remotely monitoring the status of electricity meters and timely and accurately detecting abnormalities of electricity meters, the ITSO-XGBoost algorithm is innovatively introduced in this paper. By combining the improved teaching optimization algorithm (ITSO) and the extreme gradient boosting algorithm (XGBoost), a new error estimation method for electricity meters is proposed. A method for predicting the error of the calibrator under complex field conditions by using XGBoost is proposed. In order to balance the exploration and development capabilities of the algorithm, an Aquila Eagle hybrid tuna swarm optimization algorithm is proposed to improve the XGBoost algorithm and enhance its competitiveness. The results of multiple sets of experiments show that the ITSO-XGBoost algorithm performs well in the error estimation of electricity meters, not only significantly higher than the traditional machine learning algorithms in terms of prediction accuracy. Furthermore, this algorithm has a relatively low sensitivity to feature selection and can maintain stable performance when dealing with high-dimensional data, demonstrating strong robustness and adaptability.