Optimización en tiempo real del modo de operación de un abastecimiento de agua mediante técnicas metaheurísticas. Aplicación a la RED de suministro a Valencia y su área metropolitana.

Abstract

[EN] Due to the rise of energy prices, the energy cost of pumping represents most of the total operation cost in water distribution systems. In addition, energy price varies according to the tariff structure. For these reasons, optimal real-time control of the pumping stations and the valve operation settings is required in order to minimize energy costs while controlling pressure in the network. To achieve this objective it is necessary to define an optimization model that results in a reliable operation scenario. Moreover, simulation of hydraulic models allows tracking the daily operation schedule of the water supply network. Its integration into a SCADA system is a valuable tool to aid the making of decisions. For this aim, the mathematical model must be calibrated, and regularly updated. Finally, to perform optimal control in real-time, it is necessary to define a demand forecast model to know in advance the state of the network for the following 24 hours. This thesis exposes a methodology for optimal real-time operation of a hydraulic network based on the development of a platform that allows the integration of hydraulic simulation models, forecast models and optimization algorithms in SCADA systems. With the aim of performing the most adjusted model to actual scenarios, a demand forecast model has been developed based on historical patterns obtained from the SCADA system. Finally, an optimization module has been developed to minimize the energy costs of pumping systems, which basically consists of two blocks. In the first block, the goal is to optimize the operating regime of pumps using a linear programming model to minimize the operating costs while maintaining the water levels in tanks within a threshold range. In the second block, once optimal pump operation is set, setting of pressure reducing valves is optimized in order to satisfy the total expected demand within the system's restrictions of maximum and minimum pressure, according to the pump operation. An important contribution of this thesis has been the replacement of the hydraulic model by an equivalent model based on parallelized architecture neural network (ANN), which significantly reduces the computation time, a very important aspect when a system for real-time optimization is implemented. The proposed methodology has been applied to the water supply network of Valencia and its metropolitan area, taking into account current energy tariff structure. Results prove the effectiveness of the proposed model in the search for feasible operation strategies, to be applied to a real distribution system with two water pumping stations, several tanks, and various regulation valves remotely operated. Many of the physical components of the system have been replaced by using ANN architecture, without affecting the pressure constraints at critical points. The achieved energy saving highlights the importance of applying optimization models as a mean of improving system energy efficiency.; [ES] Con el aumento de precios de la energia, el coste de la energia…