At present, photovoltaic plants observe a large dispersion in peak power values for their generators, with differences greater than 6% between these values. Firstly, this paper proposes a method, based on a specific distribution of modules given by R. Zilles [9], to generate a random series of PV modules from the parameters published by the manufacturer of a commercial module. Once obtained these modules has been carried out to select the generators needed to to form 9Sx4P, calculating its losses by mismatching. Later these results will be verified with real measurements made in a real photovoltaic plant, thereby, confirming that the average loss per mismatch is approximately 0.6%. Once verified, the simulation will be extended to different generators with different numbers of modules both in series and parallel (nSxmP). The conclusion reached was that for any structure mismatching losses do not exceed 0.5%. However, how much would losses be if a defective module appears within a generator? To answer this question, a classification, from an electrical point of view was made of the different defects that can occur in a module. The simulations were performed for nSxmP 9sx4P generators and incorporating in these generators, modules with different types of defects. In these cases the results demonstrated that the losses depend greatly on the type of defect, and these are greatest when the defect is a current fault. Similarly, it is also shown that the number of series and series modules only significantly affect the value of the losses when the default module is also current.