Análises e métodos de otimização para eficiência energética de estações elevatórias em sistemas de abastecimento de água.

Abstract

An efficient use of energy regarding a pump scheduling within water distribution companies is one of their major concerns. In practical problems of engineering, there are several situations in which researchers need to minimize or maximize certain numerical functions. Classical optimization techniques have been extensively documented but the difficulties in their implementation have given room for application of new techniques, sometimes with no verification of the applicability of the traditional methods. In this study, different tests of linear and nonlinear programming as well as genetic algorithms have been applied to the pumping scheduling of the water supply system of Campina Grande city, State of Paraíba, which has a population of 400,000 people. Changes in the pump scheduling solutions in addition to variation in processing time have been identified for different sets of initial data and maximum iteration stopping criterion. First, a performance comparison was made for the application of linear and nonlinear programming and genetic algorithm to a smaller pumping system, where the decision variables could take real values. Besides requiring quite different computational times, different initial conditions set to the decision variables resulted in different pumping scheduling solutions. Secondly, a comparison between two methods that results in binary integers for the decision variables has been performed. These methods are a two steps model, which applies a linear programming, and an integer linear programming optimization model, and a standard rule pumping scheduling simulation model. The results have shown a profit increase of 20,27% and a energy saving of 16,86% when applying a two steps model. Finally, in order to be able to solve large problems, another two steps model was applied to an extended number of pumping stations. This model is based on a linear programming, which generates a global optimum for cost minimization, and a quadratic programming, which minimizes maintenance costs as well as force real numbers, attained at the first step, to become binary integers. .