Obtaining optimal operation rules by means of evolutionary computation and GRG algorithms: Cointzio Dam, Michoacán, Mexico

Abstract

In this study, operating policies of the Z curves type were obtained, using two optimization methods: the Generalized Reduced Gradient (GRG) Nonlinear and a simple Genetic Algorithm (GA), and they were applied to a reservoir located in the Cuitzeo lake basin, which plays an important role in supplying water to a part of the population of the city of Morelia and in the irrigation of several modules of the Irrigation District (DR) 020, known as Morelia-Queréndaro. With each policy obtained, the operation of the system was simulated, using the historical record of inflow volumes into the system, to compare with the reported by the National Water Commission (Conagua) and those that maintained the balance between compliance with demands and less presence of spills or deficits. Finally, to review the behavior of the system in the long term when using the selected policies, a series of synthetic monthly inflow volumes with 100 years of record were simulated, generated with the Svanidze method. The comparison of the deficit and spillage results at the monthly and annual level of the simulations carried out with the historical Z curve, GA and GRG, indicated that the GA had an acceptable behavior both in the historical period and in the long term, since, although the volumes spilled increased, the magnitude in which they did so is less than the magnitude in which the deficit was reduced and as the Cointzio dam has been used to supply water to the Vista Bella water treatment plant and for irrigation of 4 of the 5 modules of the irrigation district 020, the operating rule that best controlled the deficit event was chosen.