<p>Correct and timely detection of voltage sag events is of the utmost importance in any power network due to its disastrous nature. This paper addresses the same concern and proposes a novel methodology for identification of eight voltage sag events. The proposed methodology designs matched biorthogonal wavelets for eight types of voltage sag events using lifting scheme. Out of all matched wavelets, the wavelet which works best for the specific event signal is then looked for by using energy-to-Shannon entropy criterion. The selected matched wavelet extracts more distinguishing characteristic features out of that event signal, which further helps in enhancing the classification accuracy. On top of that, misclassifications possible due to one classifier in some cases are also removed by using three good classifiers (i.e., naïve Bayes classifier, extreme learning machine and modified probabilistic neural network) in form of an ensemble and aggregating their results by using majority voting. Proposed methodology has achieved 99.94% accuracy in identifying eight different voltage sag events. This method has also shown robustness towards noise as proved by 99.5% accuracy obtained with noisy data which proves the robustness of proposed method for real-time implementation.</p>