Recent advancements in wireless power transfer technologies can solve several residual problems concerning the maintenance of wireless sensor networks. Among these, air-based acoustic systems are still less exploited, with considerable potential for powering sensor nodes. This thesis aims to understand the significant parameters for acoustic power transfer in air, comprehend the losses, and quantify the limitations in terms of distance, alignment, frequency, and power transfer efficiency. This research outlines the basic concepts and equations overlooking sound wave propagation, system losses, and safety regulations to understand the prospects and limitations of acoustic power transfer. First, a theoretical model was established to define the diffraction and attenuation losses in the system. Different off-the-shelf transducers were experimentally investigated, showing that the FUS-40E transducer is most appropriate for this work. Subsequently, different load-matching techniques are analysed to identify the optimum method to deliver power. The analytical results were experimentally validated, and complex impedance matching increased the bandwidth from 1.5 𝑘𝐻𝑧 to 4 𝑘𝐻𝑧 and the power transfer efficiency from 0.02% to 0.43%. Subsequently, a detailed 3D profiling of the acoustic system in the far-field region was provided, analysing the receiver sensitivity to disturbances in separation distance, receiver orientation and alignment. The measured effects of misalignment between the transducers are provided as a design graph, correlating the output power as a function of separation distance, offset, loading methods and operating frequency. Finally, a two-stage wireless power network is designed, where energy packets are inductively delivered to a cluster of nodes by a recharge vehicle and later acoustically distributed to devices within the cluster. A novel dynamic recharge scheduling algorithm that combines weighted genetic clustering with nearest neighbour search is developed to jointly minimise vehicle travel distance and power transfer losses. The efficacy and performance of the algorithm are evaluated in simulation using experimentally derived traces that presented 90% throughput for large, dense networks.