Wave Height Prediction for Maximum Power Extraction Scheme of Air-Turbine of an OWC based Wave Energy Plant

Abstract

Ocean wave energy has not yet gained popularity as a renewable energy source, because of highly varying nature of available wave power. Despite the seasonal/daily variations in the input hydraulic power available, extracting the maximum power from the installation requires controlling the reflected electrical load on the generator dynamically guessing the wave height on short-time basis. Such an effort of maximizing the air- turbine's efficiency is not complete without a wave height prediction algorithm setting reference to the turbine speed. Considering wave data pertaining to the site of Indian Wave energy plant, prediction of Significant wave height on hourly basis is carried out using a Neural Network employing Levenberg Marquardt (LM) algorithm for back propagation/ supervised learning. Normal back-propagation based (Least Mean Square) algorithm is compared with LM method, and the latter one is found to yield better results. Based on predicted significant wave height, prediction of the actual height of the immediately following wave is obtained by employing Non-linear Auto-Regressive Neural Network (NARNN) using previous actual wave heights as input and this model is compared with NAR neural network with exogenous input (NARXNN). The results obtained for typical significant wave heights are provided. Improvement achieved in the efficiency of the impulse turbine of OWC based Wave Energy Plant is verified by employing the concept of prediction of wave height.