Sensor swarms can be a cost-effectieve and more privacy-friendly alternative for location based service systems in building automation and health-care. To increase the battery lifetime of such swarm networks, the energy consumption should be scaled to the required localization accuracy. In this paper we described the first steps in developing an energy model that couples localization accuracy to energy-related sensor parameters such as sample frequency and ADC resolution. The goal is to use the model for the localization of undetermined environmental sounds, by means of wireless acoustic sensors. We show that for TDOA-based localization, the signal sample rate can be under the Nyquist frequency, provided that enough frequency components remain present in the undersampled signal. The resulting localization error is comparable with that of similar localization systems.