A supplementary strategy is proposed to use operational errors as a reliable source of load point errors in traditional methods, in response to the problem of insufficient detection of load points in electric energy meters and the inability of some load point errors to fully represent the full range measurement characteristics caused by the limited load point calibration method. On the basis of the continuous accumulation of operating errors in the power system, generate full-scale error data of the electric energy meter, divide the processed full-scale error time series and key features such as timeliness, identify the change points by comparing the Pearson correlation coefficient matrix of multiple acquisition periods, and establish a data-driven error correlation prediction model that does not depend on specific power consumption environments and only relies on the intrinsic components of the electric energy meter error. Empirical analysis shows that the proposed method can dynamically analyze the fluctuation state of energy meter errors within a safe range, explore potential high-risk information in the periodic characteristics of energy meter errors, and achieve long-term trend prediction of energy meter errors in the substation area. It is of great significance for the evaluation of energy meter errors with time-varying characteristics and the analysis of energy meter rotation.