The existing state evaluation experiments for online error estimation algorithms for energy meters have not considered the impact of uncertainty

making it difficult to achieve an effective balance between out-of-tolerance recognition rate and false positive rate. To solve the above problems

this paper proposes an uncertainty-based evaluation method for assessing an algorithm’s state discrimination capability and introduces the α- out-of-tolerance recognition rate metric to systematically evaluate this capability. Firstly

the uncertainty of the algorithm is evaluated under actual noise disturbance conditions in the substation area. Then

a probability distribution model for algorithm outputs is derived

analytical expressions for the α-out-of-tolerance recognition rate metric and false positive rate are established

and the relationship between state evaluation metrics and state determination threshold is revealed. Based on this

the error level α for accurate out-of-tolerance meter recognition and the optimal state determination threshold T are determined under target constraints. Finally

the sliding window algorithm is taken as an example for experimental verification

and the results show that the deviation between the calculated false positive rate of the evaluation method proposed in this paper and the measured value is 0.08%. For out-of-tolerance meters with an absolute error of 3%

the deviation between the calculated out-of-tolerance recognition rate and the measured value is 0.09%. These deviations fall within an acceptable range for practical applications

fully validating the effectiveness and reliability of the proposed method.