System identification of oscillating surge wave energy converter using physics-informed neural network

Optimizing the geometry and increasing the efficiency through PTO control of wave energy converters require the development of effective reduced-order models that predict their hydrodynamic response. We implement a multi-step approach to identify the coefficients of the equation governing the response of an oscillating surge wave energy converter. Data from quasi-static, free response and torque-forced experiments are successively used to respectively identify the hydrostatic stiffness, radiation damping, added mass, and nonlinear damping coefficients. The data sets were generated from experiments performed on a model of an oscillating wave energy converter. The stiffness coefficient was determined from quasi-static experiments. Physics-informed neural network was then applied to the free response data to identify the coefficients of a state-space model that represents the radiation damping. The same approach was applied to torque-forced response data to identify the added mass and nonlinear damping coefficients. Details of the implemented physics-informed neural network are provided. Validation of the identified coefficients and representative model of the response is performed through comparisons with experimental measurements. An analytical representation of the admittance function is derived using the identified coefficients. This representation is validated against experimentally determined values at discrete frequencies.