Machine Learning Based Maximum Power Point Tracking in Tidal/Ocean Energy Conversion System
In this paper, an efficient and supervised algorithm is presented to estimate a maximum power point (MPP) in
Tidal/Ocean energy conversion systems by the implementation of machine learning (ML). Tidal energy extracted from the
turbines depends on ocean current speed, tidal height, sea temperature, up cross period which most artificially intelligent
algorithms such as neural networks ignore. Hill climb search is the most common and accurate methodology to track the
maximum achievable power (MAP). However, the convergence speed to the maximum power point varies immensely and is
slow. The proposed method uses machine learning to estimate a MPP using ocean current speed, sea temperature, up cross
period and tidal height as input variables at every iteration. This MPP is then passed to the conventional hill climb search
algorithm (HCS) to retrieve the MAP thereby reducing the perturbation time by a significant amount. At the end of each
iteration, the machine learning algorithm is updated with the correct MAP thus avoiding overfitting which is predominant in
artificial neural networks (ANN) and deep learning systems. The accuracy of the estimation increases after every iteration.
Thus, for every tracked power point, the system is being trained recursively to predict an accurate MPP in the subsequent
iterations. The simulation performed yielded an efficiency of 99.99% in estimating the MPP after 2500 iterations which
corresponds to 9 hours of data.
Index Terms - Artificial Intelligence, Hill climb search, Machine, Learning Maximum Power Point Tracking, Tidal energy