Maximization of energy absorption for a wave energy converter using the deep machine learning

Abstract

7A controller is usually used to maximize the energy absorption of wave energy converter. Despite the 8 development of various control strategies, the practical implementation of wave energy control is still 9 difficult since the control inputs are the future wave forces. In this work, the artificial intelligence 10 technique is adopted to tackle this problem. A multi-layer artificial neural network is developed and 11 trained by the deep machine learning algorithm to forecast the short-term wave forces. The model 12 predictive control strategy is used to implement real-time latching control action to a heaving point-13 absorber. Simulation results show that the average energy absorption is increased substantially with 14 the controller. Since the future wave forces are predicted, the controller is applicable to a full-scale 15 wave energy converter in practice. Further analysis indicates that the prediction error has a negative 16 effect on the control performance, leading to the reduction of energy absorption. 17 Keywords: wave energy converter; wave energy control; energy absorption; neural network; deep 18 machine learning; wave force prediction. 19 Henriques [6] presented a review on the oscillating-water-column WEC. Stansby et al. [7] examined 29 the dynamics of multi-float WEC concept M4. 30Although a set of WEC concepts have been developed, the energy harvesting efficiency is still not 31 satisfactory, especially in the off-resonance state. One of the solutions is the usage of a non-linear 32 power take-off (PTO) system. Zhang and Yang [8] showed that a PTO system with nonlinear spring 33 could harvest more energy in random waves. Xiao et al. [9] investigated the power capture of an 34Eq. (4) is a first-order, one-variable differential formula, which is easier to handle. Given the initial 112 condition x(0) = 0, it becomes a classical initial-value problem and the time series of floater 113

Adaptive Neuro-Surrogate-Based Optimisation Method for Wave Energy Converters Placement Optimisation

“…The neuro-surrogate is trained prior to each placement using sampled positions used for the previous buoy placement. 2) Use EvalSet = {2 nd , 3 rd , 6 th , 9 th , ..., 15 th } so the neuro-surrogate is used to place buoys: 4, 5, 7,8,10,11,13,14,16.…”

Control Strategies Applied to Wave Energy Converters: State of the Art

“…In a similar way for a heaving point-absorber, a neural network is employed to forecast the short-term wave height and period in [53] to implement real-time adaptative latching control. This work presents some results comparing the differences of absorbed power for a particular wave scenario with and without control.…”

Non-Causal Linear Optimal Control of Wave Energy Converters With Enhanced Robustness by Sliding Mode Control

“…One class of prediction approaches are based on statistical methods, such as the Auto-Regressive (AR) prediction method [10] and the extended Kalman Filters (EKF) [11]. Artificial neural network (ANN) has also been used to forecast the short-term wave forces [12], [13]. Another class of prediction is based on the measurements of sea wave elevations at multiple upstream locations with certain distances away from the WEC, such Y. Zhang is a postdoctoral researcher with Queen Mary University of London, UK, E1 4NS.…”