The development of an Active Reflection Compensation (ARC) algorithm for a wave generator is challenging. Water behaves in a strong nonlinear way and the time available for the computation of the compensating segment displacements is short. Especially for the case where the wave height sensors are mounted on the surface of the segments. The current ARC algorithm of Bosch Rexroth is based on an approximation of the Biésel transfer function. Although this method gives fair results, it lacks performance regarding the absorption of oblique waves. This report introduces a wave model identification method based on the Discrete Fourier Transform (DFT) and Extended Kalman Filter (EKF). With the EKF having the ability to include measurement noise and determine model uncertainty, this leads to reasonable estimates of complex second order waves. This report does not only cover the identification part of the algorithm but the full ARC algorithm from wave height measurements to a time series for the segment displacements. Although the results are based on simulations, the algorithm shows a much improved absorption performance for a wide range of propagation directions. It shows decent results for tests with several different sets of wave characteristics. The identification process needs time to mature, which leads to a decrease in absorption performance in the early stage of the free-surface elevation time series. A hybrid method is designed, in which the short-term absorption is based on the current algorithm and, after a certain time interval, the model based algorithm will take over.