In this paper, a scheme has been developed for detecting individual, as well as multiple partial discharge (PD) sources within a transformer tank through acoustic sensors. For that purpose, in the laboratory, a transformer tank has been emulated inside which, provisions are made to place artificially created partial discharge sources at different locations. An experimental setup has been prepared to record the acoustic PD signal through multiple acoustic sensors placed at the outside surfaces of the walls. Presented results show the signature of the acquired data pattern for both individual and multiple discharges. Acoustic PD detection method experiences difficulty in locating the PD source accurately due to interference or attenuation of signals from the environmental noise. It is seen that Cross-wavelet transform (XWT) has an ability to nullify the effect of random noises present in the acquired acoustic PD signal. Thus, cross-wavelet transform based feature extraction with ensemble binary support vector machine based classifier is used for locating the partial discharge sources within the tank. These methodologies have been chosen as they suit well to discriminate between the PD sources at different locations inside the tank. Results indicate that the average classification accuracy is 92.8%. The contribution of this research work is in discriminating individual as well as multiple partial discharge sources which produce acoustic emission simultaneously. Experimental results indicate that the proposed methodology can be used to localize partial discharge sources in high voltage apparatus where the acoustic signals due to partial discharges find a path to propagate towards the outer surface.