To better handle nonlinearities in the dynamic modeling of wireless power transfer systems, a linear parameter-varying (LPV) Hammerstein model is derived, along with a method to identify the structure and estimate the parameters of the model from the experimental data. Precisely, the proposed modeling method identifies the model structure by reference to the specified identification criteria and the circuit topology of the system, and estimates the unknown model parameters using sampled input-output data. Compared with other data-driven models, such as linear models that are only accurate at an operation point, and Hammerstein models that only cope with the inverter nonlinearity, the proposed LPV Hammerstein model is able to handle both inverter and rectifier nonlinearities, as well as load and coupling variation, thus providing a more accurate description into the system dynamics. In addition, the LPV Hammerstein model has several merits compared with the conventional circuit-theory models: it does not necessarily require any circuit component parameter, and the resulting model is simple yet accurate - a first- or second-order model can provide a fairly good prediction of the system output in a quite wide operation range. The simulation and experimental results are included to confirm the effectiveness and superiority of the proposed model.