During the use process of electricity meters, natural aging faults or human factors can cause abnormalities in the metering of electricity meters. Some existing detection algorithms for abnormal electricity meters cannot detect abnormal out-of-tolerance electricity meters quickly and timely due to their slow computing speed. Therefore, a rapid detection technology for abnormal out-of-tolerance electricity meters based on FIT model theory is proposed. Firstly, collect the power data in the power metering device to detect and fill in the missing power metering data. Secondly, according to the demand for power metering error detection, an objective function for error detection is constructed. Finally, genetic algorithms are used to search for the optimal configuration parameters of the model, and filtering strategies are formulated based on residual and other data to reduce the search space. The residual data is smoothed and denoised in the search space, and the power metering error detection is completed by solving the extreme value of the objective function. Compared to previous fit search algorithms, this algorithm can reduce the average detection time of a single substation area to about 30 seconds, and achieve a high hit rate of over 98% and a high-precision out-of-tolerance calculation range within an absolute difference of 0.005 for computable substation areas.