Optimisation of both energy use and pumping costs in water distribution networks with several water sources using the setpoint curve.

Abstract

Usually, pumping optimisation in water distribution networks is carried out by means of the system head curves (SHCs), also known as the resistance curves (RCs). These curves are subjected to the resistance generated by the users, according with the flow and pressure head needs at the final points of demand. Such resistance is highly variable and hard to determine. Thus the calculation of the RCs and all the points that define them results impractical. As a RCs suitable calculation is not possible, real flow and pressure needs of the network are not known. Therefore, neither energy excess of the pumping regarding the real requirements nor the raising operating costs due to such excess, are estimated. However, there is another type of SHC defined as the setpoint curve (SC). It can be easily calculated, but has been poorly studied so far. Thus, this work aims the optimisation of the energy use and operating costs in pumping systems by using the SCs. To achieve the objective, first of all the SC calculation is studied for networks with several pumping stations, non-pressure driven demands, pressure-driven demands, and without tanks. Next, the optimisation is performed only from the energy point of view (i.e. flow and pumping head required). For that, a search of the optimum flow distribution among pumping stations to find the optimum SCs is performed. Two methods are proposed: the discrete (D-M) and the continuous (C-M). The D-M considers the flow distribution as a discrete variable. The optimum flow distribution is obtained from a set of solutions defined previously. In the C-M, the flow distribution is assumed as a continuous variable. The optimum solution comes from using optimisation algorithms. Two algorithms have been applied: Hooke-Jeeves and Nelder-Mead. Then, the cost optimisation (pumping cost and water production cost) is developed. For that purpose, the M-C is used as starting point. Then, energy tariffs, water production fares and the minimum expected efficiency at the pumping stations, are included. The last step consists in the energy and cost optimisation in networks with tanks. When tanks are included the SC calculation methodology changes. Hence, the optimisation process also does. In that sense, besides the costs of pumping and water production, the cost function also considers penalty costs for unaccomplished minimum pressures and minimum storage leves. Moreover, tanks inclusion also rises the number of decision variables. Thus, the use of more powerful algorithms is required. In that context, the Differential Evolution and the Hybrid Algorithm have been applied. The last one is an additional contribution of this work. The optimisation methodology is applied to five distribution networks: TF, Catinen, Coplaca, Anytown and Richmond. Tanks are only considered in the last two networks. In the case of TF network, demonstrative pumps selection (without going into great depth) by the optimum SCs application is done. However pumps sizing and selection study is out of the scope of this research. Neither multiple operation conditions nor reliability (i.e. in the case that tanks or pumping stations are removed), are considered. Nevertheless, the results obtained evidence that pumping systems operated by mean of the optimum SCs could reduce their operating costs up 12%. The methodology also gives information about which pumping stations represent major savings and which are less important or not needed. Besides, the method demonstrates that better pumping conditions (i.e. low energy tariffs and high efficiencies) not always mean lower operating costs. Finally, some results show that the method could be useful for the optimisation of both placement and use of storage tanks.