The levelized cost of energy (LCOE) from a parabolic trough collector power plant is influenced directly by its design parameters. In the present paper, the impact of five design parameters (solar multiple, direct normal irradiation, cycle efficiency, mirrors efficiency and absorber efficiency) is investigated in order to minimize the LCOE for a 1 MWe power plant. The artificial neural networks (ANN) and the response surface methodology (RSM) are used to model the LCOE as function of these five parameters. The methodology is applied to six cities of Morocco, and the results show that the ANN is the best model to predict the LCOE using the best topology of 5-7-1 neurons. This structure allows satisfactory prediction of LCOE obtained with Mean Absolute Error (MAE) of about 2% and Maximum Error (ME) of 9%. A minimal cost of 8.74 cents is obtained for Tata city with a direct normal irradiation of 7.09 Wh/m(2)/day, by coupling ANN model to desirability function. The optimal plant design point is 1.94 for the solar multiple, 20.88% for the cycle efficiency, 0.8673 for the mirror efficiency and 0.8537 for the absorber efficiency. (C) 2021 Elsevier Ltd. All rights reserved.