Estudo comparativo da calibração de rugosidades em redes de distribuição de água submetidas a regimes de escoamento transiente rápido e lento

Abstract

The calibration of roughness in water distribution networks submitted to a transient flow was neglected for some time, but has been the object of recent studies. In the water distribution networks, where demand varies with each user, reservoir levels are adjusted according to climatic conditions and pumping stations are continuously turned on and off, the use of computer models developed for steady state becomes limited. The use of simulation for the hydraulic network behavior is very useful, when well executed, for the accidents prevention, breaks or even lack of water in some determined network points. In order to simulate the network behavior with good accuracy, it is necessary to know all the physical parameters involved in the system and this results in values of pressure and flow very close to the real ones. The estimating parameters process of the hydraulic network is called calibration. It was searched to calibrate the roughness, which is the most difficult variable to determine, in a hypothetical distribution network formed by 8 nodes of variable demand (one quota reservoir equal to 60 mca) and 10 tubes. The analyzes were performed using programs that apply the genetic algorithm optimization technique to find the optimal values of the hydraulic load in the observed node. The piping was subjected to a fast transient (5 seconds) and a slow transient (2 minutes). The comparison criteria used were relative error, relative mean error, standard deviation and objective function. Three rates of elitism were applied to the initial population in each transient to observe the influence of the operator on the calibration results. The results indicate the best scenario to obtain the calibration of the roughness of a water distribution network and the effects of elitism on calibration. The Genetic Algorithm optimization technique is compatible for the calibration of water networks. In fast transient, the higher the rate of elitism, smaller the error. In the slow transient, the error increases with the growth of the elitism. The objective function value in the slow transient regime is lower than the fast transient regime one. The parameters of the genetic algorithm should not tend to maximum or minimum extremes. The occurrence of transient influences the definition of the genetic algorithm parameters, and the calibration error varies the behavior according to the fast or slow transient flow.