Maximisation en temps réel de la puissance de sortie de piles à combustible microbiennes en utilisant l'optimisation par essaims particulaires

Abstract

The industry is turning more and more to alternative sources of energy, and microbial fuel cells (MFCs) have been getting a lot of attention for many years in that regard, as they use microbes that transform the energy contained in organic matter into electric energy. However, MFCs have a very low power density, as well as very slow internal dynamics, which makes it difficult to harvest their energy. Indeed, it is only possible to harvest the maximum amount of energy from an MFC if its external load matches its internal resistance. Moreover, the slow dynamics of MFCs imply that their internal resistance may change over time. Real-time optimization (RTO) techniques are thus needed to address these problems. Particle swarm optimization (PSO) and perturb and observe optimization (P&O) are algorithms which are often used for the RTO of the power of photovoltaic (PV) systems. P&O has also been used on MFCs, whereas PSO hasn’t, even though PSO has many advantages over P&O. Thereby, the first contribution of this work is a study of the usability of PSO-based techniques for the RTO of the power of MFCs. The second contribution is the conception of a new RTO algorithm called parallel particle swarm optimization with classification (PPSOC) applicable on dynamic systems. Simulations have shown that PSO-based algorithms have a lot of potential for the RTO of the power of MFCs and that they are better than P&O. Experimental results on MFCs have revealed that, if the power optima are slowly changing, then the PSO-based algorithms need to be modified to adapt them to this constraint. PPSOC with diversity (PPSOCD) has been proposed to address this issue, and simulations have shown that PPSOCD would have a better performance than P&O in this context. The PPSOC algorithm has also been validated experimentally on a system comprised of 15 PV cells. The experimental results have shown that PPSOC has better performances than PSO and P&O, two algorithms that are well-tested on PV systems. As such, PPSOC has been proven to be an effective algorithm that can be applied to different types of systems.

Date	11 Jun 2018
Original language	French
Awarding Institution	École de technologie supérieure
Supervisor	Lyne Woodward (Supervisor)