Use of artificial neural networks for real-time prediction of heave displacement in ocean buoys

Abstract

Many advanced control systems for wave energy converters (WEC's) require knowledge of incoming wave profiles to be implemented. This is due to the non-causal relationship between water elevation and force exerted on a floating body. This study focuses on the use of cascade feedforward neural networks to predict short-term incoming water surface displacements based on recently observed data in real time. Prediction networks are trained with time series data reconstructed from spectral data and recorded time series data from a data buoy deployed off the West Irish Coast. Both training methods are shown to have predictive capabilities with regression coefficients between 0.8-0.9 for a small range of sea states. Both networks prediction accuracies are tested on a large range of sea states as well. For sea states dramatically different from training data prediction accuracies decrease, but less so for the network trained on observed data. The need for accurate wave predictions in the field of WEC control design is also discussed.

Performance of artificial neural network in prediction of heave displacement for non‐buoyant type wave energy converter

“…The ANN becomes more knowledgeable about its environment as the iteration proceeds in the learning process. ANNs have been used by the researchers in various fields of renewable energy research [12–15]. In this work, a specific type of ANN called a non‐linear autoregressive network with exogenous inputs (NARX) consisting of neurons layer‐by‐layer (input, hidden and output) is implemented.…”

Rapid phase-resolved prediction of nonlinear dispersive waves using machine learning

“…However, ML for phase-resolve analysis of ocean surface waves i.e. finding the details of surface elevation as a function of space and time, is still not in hand: limitations include calm to moderate sea states [51][52][53], too short prediction periods [16], requirement for a timeconsuming preprocessing [16], and unrealistic simplifications such as the assumption of a fully developed sea [51]. To address these issues, this work proposes a novel approach toward the short-term rapid forecasting of ocean waves.…”