An algorithm to measure and detect potential thermo-accumulation customers among the clients of a distributing utility was proposed in this work. Such algorithm uses statistical methods for the treatment of missing data and is applied to the measured hourly demands of medium and high voltage customers of the utility. Basically, the thermo-accumulation technique can be defined as a strategy to shift load from peak hours to off-peak hours. The algorithm allows the construction of load curves from the observed load series after the removal of outliers and/or missing values. From the profile of such curves, the algorithm produces the classification of the clients into three possible categories: higly potential, just potential and non potential (i.e. an already modulated client). Also important to mention this dissertation, is part of an R&D project developed for a brazilian southern utility. A friendly computer system named SIMCAR, developed in Visual Basic, implemented the proposed algorithm. As a final word, in surveys conducted with users of this technology, it was mentioned that the savings on energy consumption can be as high as 40 per cent.