Controlo de um Conversor de Energia das Ondas usando um sistema multi-agente e métodos de aprendizagem automática

Nuno Miguel Antunes Pereira

Key information

Authors:

Nuno Miguel Antunes Pereira (Nuno Miguel Antunes Pereira)

Supervisors:

Duarte Pedro Mata de Oliveira Valério (Duarte Pedro Mata de Oliveira Valério)

Published in

06/25/2018

Abstract

Modern societies are characterized by an intense use of energy in transport, industry and electricity production. The transition to renewable energy sources presents a path to a sustainable development. In addition to the commonly known renewable energy sources, like solar and wind, wave energy from oceans and seas promises to be a valuable resource for electricity generation. This work aims to provide ISWEC (inertial sea wave energy converter) — a device designed to extract power from sea waves using gyroscoping motion to generate electricity — the capability to adjust its PD controller gains according to current sea wave condition, so its performance is maximized. The forecasting process is carried out by a set of machine learning methods, namely neural networks and a SVM (support vector machine) classifier. The distributed arrangement of ISWEC devices on “wave farms” is handled by a multi-agent platform, based on Java, ensuring all devices receive due forecast data from the central agent. It is believed this strategy, as opposed to relying on a exogenous forecast set on an hourly basis, will result in improved performance, since sudden transitions of sea wave conditions may be taken into account thanks to a shorter time basis of gain setting operation and to an autonomous forecasting agent.