Abstract: In order to rectify the problem of measurement inaccuracies in secondary protection measurement circuits at substations, which undermine measurement precision, diminish protection sensitivity, and consequently impair the reliable operation of protective devices, this paper puts forth a multi-channel data verification and evaluation methodology founded on a bidirectional long short-term memory (BiLSTM) network that has been refined through a genetic algorithm (GA). This proposed methodology aims to guarantee measurement accuracy and to augment the usability of multi-channel information. An initial dataset is established based on relative deviations between multi-channel data groups, followed by the formulation of corresponding classification principles. Case study analysis demonstrates that compared to other network models, the GA-BiLSTM multi-channel evaluation model achieves an accuracy rate of 97%, representing a significant improvement.