Microbial fuel cell (MFC) is a promising technology for wastewater treatment with simultaneousbioenergy production. To improve the power generation efficiency of MFCs, maximum power pointtracking control is a good choice. Three kinds of Q-Learning-based maximum power point trackingcontrol scheme based on ε-greedy exploration, Boltzmann exploration and greedy policy are proposedfor MFCs. The results show that the maximum power point tracking control based on Q-Learning hasbetter power tracking capabilities than perturbation and observation method. With the introduction of QLearning based on greedy policy, the time required for MFC to stabilize at the maximum power point isgreatly shortened by setting the action list of Q-Learning reasonably. In this case, the whole processfrom start-up to stabilization at the maximum power point was 42.9% faster than that of MFC using εgreedy exploration, and 50% faster than that of MFC using Boltzmann exploration. Q-Learningalgorithm based on greedy policy is an effective method to realize MPPT in MFC system.
展开▼
