The paper deals with issues related to the optimization of pulse supply of thermoelectric cooling systems to obtain possibly low temperatures, below the minimum steady-state temperature. The pulse shape was optimized with a view to minimizing the period between its successive repetitions to achieve the set cyclic effect of supercooling. An analysis was conducted of the way of the current pulse modelling using three different methods. It was observed that its definition using 3 degrees of freedom was a good compromise between the achieved results and the computational cost. The calculations were carried out for a one- and a two-stage system taking account of the thermal capacity of the heat reservoirs. An analysis was also carried out of the impact of the thermal capacity of individual reservoirs on the optimization results. The notions of the limit temperature and of the limit period defining the rationality threshold of further reduction of the supercooling temperature were introduced. The optimization process was run for two variants of computations, i.e. firstly to minimize the period between subsequent supercooling temperature occurrence, and secondly to maximize the supplied cooling power at supercooling temperature range. It was demonstrated that in the case of the assumed two-stage system, it was possible to maximally reduce the cold reservoir temperature by over 6 K at the working cycle of about 13 min. Diagrams of supercooling temperature as a function of the minimum period were presented. Smaller supercooling values but with shorter periods are obtained for one-stage systems. Pulse supply also produces the benefit of obtaining cooling power at temperatures below the steady-state minimum and close to it.