Abstract
The prediction of near future wave excitation force is the fundamental problem associated with optimal control of a wave energy converter. This becomes clear when the latching technique is considered. In this case the central control problem is the choice of the moment at which to release the latched working surface, such that the resulting velocity is in phase with the wave exciting force. To calculate the ideal time to unlatch, the time until the next peak or trough in the excitation force is required. This paper describes the development of various techniques, including neural networks, for estimation of the time until the next peak in excitation force. Two sets of synthetic excitation time series were used: one that was close to sinusoidal, with little variation in period and height, and the other with a high variation in these parameters. Several neural networks were trained and the results compared to alternative methods. It was found that neural networks performed better than the alternative methods for the time series that contained more variation.