This paper proposes a behavioral model of power losses for multiphase converter that will be used to optimize the efficiency of the converter by dynamically modifying the number of working phases. The aim of this work is to develop a simple (in terms of simulation resources) and but accurate (in terms of energy conversion) model of one cell of the multiphase converter that can be used to deduce the model of the whole converter, depending on the number of active phases. This model could also be used for long-time range simulations of photovoltaic systems. To build the model, a large number of tests has been performed on a multiphase converter at different values of the input voltage, the duty-cycle of the switch drive signals and the output currents. Experimental power losses of the boost cells have been compared with a model proposed in the literature. As the accuracy of this model was not satisfactory, we proposed to fit experimental data with feed-forward artificial neural networks