This paper presents a control scheme for improving the dynamic characteristic of the thermodynamics of hydrogen based metal hydride (MH) reactor by heating/cooling, thermal management, with a thermoelectric module. Normally, the uncertainty of the parameters obtained by mathematical model are deviated from the nominal parameters due to such as unexpected outside temperature of the reactor, or uncertainty of the time-delay of the reactor process. Moreover, the complicated problem of the integrated system model, thermoelectric module and MH reactor, make the system to be nonlinear which makes it hard to use the conventional controllers like proportional-integral-derivative, PID, or conventional fuzzy control to improve dynamic response. To obtain a high performance of those thermodynamics of hydrogen gas within the MH reactor, the Neuro-fuzzy network is employed to be control scheme for that thermodynamics non-linear system in real-time implementation. A back-propagation technique is applied to finely adapt the parameters of the Neuro-fuzzy PID. The results show the effectiveness of the proposed technique compared to conventional control approaches.