A Fundamental Study on the Potential of AI Prediction in Control Force Estimation to Maximize Power Generation of PA-WECs in Irregular Waves

-Second Report: Time Domain Analysis Including DNN-based Trained Models and Evaluation-

Abstract

　The Japanese government has established the target for the proportion of renewable energy in the electricity supply to 36–38 % by 2030. One of the ocean renewable energy sources is wave energy. This study focuses on a point-absorber wave energy converter (PA-WEC). Let irregular waves be expressed as the sum of regular wave components whose periodic components are “known”, then the control force required to maximize the average generated power can be solved theoretically. However, in real sea, the optimal control force cannot be identified instantaneously. This is because the periodic components of the irregular waves are “unknown”. Therefore, we have investigated that the optimal control force could be estimated instantaneously using artificial intelligence (AI) with only instantaneous input data. In this study, time-domain simulation is conducted while the control force estimated by AI is included in the motion equation each time step. We investigated the appropriate explanatory variables for training to increase the average generated power in regular and irregular waves. The average generated power was less likely to be negative using the heave velocity of the float as the explanatory variable, and the heave displacement made a large contribution in wave periods of 7.0–10.0 [s].