Achieving accurate localization of partial discharge (PD) sources is of great importance to ensure the long-term safe operation of electrical equipment. To improve the accuracy of PD source localization, the adaptive inertial weight bacterial foraging localization algorithm (AIW-BFA) is proposed based on the principle of time difference of arrival (TDOA) method, which adaptively adjusts the search step size and performs global optimization seeking through repulsion, replication and dispersion. Since the electromagnetic wave propagation inside the transformer will be reflected, refracted and diffracted, which makes the localization of PD source more complicated, based on the principle of shortest optical range, the computational model of electromagnetic wave transmission path in the presence of obstacles inside the transformer box is proposed to reduce the time delay difference. Through simulation and field experiments, the AIW-BFA localization accuracy can meet the localization requirements and is better than the BFA and PSO methods, which proves the effectiveness of the proposed model and the localization algorithm.