JavaScript is disabled on your browser. Please enable JavaScript to use all the features on this page.

Highlights

A new extremum-seeking control (ESC) with anticipative action (ESCaa) is proposed.

The anticipative action is performed using a neural-network model.

A theoretical analysis is done to compare ESC and ESCaa convergence time.

The performance of ESC and ESCaa is compared via simulation and experimentally.

ESCaa has very good tracking efficiency even for high frequency disturbances.

Abstract

This paper presents a fast and accurate real-time optimization (RTO) technique that can be applied to different types of renewable energy sources (RES). Two RES with very different dynamics in terms of complexity and convergence time towards the static regime have been chosen for this study: Photovoltaic panels (PV) and microbial fuel cell (MFC), a bioreactor that uses exoelectrogenic bacteria to produce electrochemical energy. The maximum power generated by these two RES is prone to vary when the system is subjected to various external disturbances. Extremum-seeking control (ESC) is a RTO method that has the ability to optimize the performance of a RES whatever its complexity. However, when the external disturbances affecting the RES result in fast variations of its optimal operating point, the slow convergence of ESC will induce a lack of precision. This paper proposes the addition of a neural network-based anticipative action to the existing ESC scheme to improve its performance in terms of speed and accuracy if the system is subject to the effect of measurable disturbances. The performance improvement of ESC is demonstrated theoretically for general systems, via simulation for an MFC and experimentally in the case of a PV.

Keywords

Extremum-seeking control

Neural networks

Real-time optimization

Photovoltaic system

Microbial fuel cell

© 2018 Elsevier Ltd. All rights reserved.