This study sought to investigate the effect of the number of input variables on both the accuracy and the robustness of the artificial neural network (ANN) method for predicting the performance parameters of a solar energy system. Tests were conducted on a solar energy system in Ottawa, Canada during summer under different weather conditions. Three different ANN models, i.e., one each with nine, eight and seven input variables, were developed and compared to a baseline ANN model previously developed by the authors . The experimental data were used for constructing the ANN models in order to assess their reliability. Each of the models was applied in an effort to predict several performance parameters of the system. The data revealed that the optimal algorithms and topologies were the Levenberg-Marquardt algorithm and the structure with 9/8/7 inputs, 20 hidden and 8 outputs, respectively. The simulation results demonstrated the efficiency of this approach and delivered a good measure of precision, even with models employing reduced input variables. However, it is likely true that the degree of model accuracy would gradually decrease with reduced inputs. Overall, the results of this contribution reveal that the ANN technique provides both high precision and strong robustness for predicting the performance of highly nonlinear energy systems.