The calculation of high-frequency model parameters of a winding is crucial for the integrity assessment of transformer winding damages and its severity. Bat algorithm (BA) is one of the modern and efficient methods used in computational intelligence due to its efficiency to obtain better results for complicated non-linear problems. In this study, an iterative method to ensure the mutually coupled equivalent model synthesis of an actual isolated transformer winding is proposed via bat search algorithm based on frequency response analysis data measurement (resonance frequencies, equivalent inductance and effective capacitance). The equivalent ladder network synthesis of transformer winding using the BA is based on the construction of the search space of the model parameters, thus each model element to be identified must be limited between two bounds (the upper and lower bounds) based on constraints and inequalities defining the model parameters. The effectiveness of the BA was tested by simulation and experimentally on two case studies of two isolated transformer windings with an air core. Furthermore, a comparative study of BA, particle swarm optimisation and the magnitude vector fitting was made in terms of accuracy and convergence speed.