Solving Real-Life Hydroinformatics Problems with Operations Research and Artificial Intelligence

Many real life problems in the hydraulic engineering literature can be modelled as constrained optimisation problems. Often, they are addressed in the literature through genetic algorithms, although other techniques have been proposed. In this thesis, we address two of these real life problems through a variety of techniques taken from the Artificial Intelligence and Operations Research areas, such as mixed-integer linear programming, logic programming, genetic algorithms and path relinking, together with hybridization amongst these technologies and with hydraulic simulators. For the first time, an Answer Set Programming formulation of hydroinformatics problems is proposed. The two real life problems addressed hereby are the optimisation of the response in case of contamination events, and the optimisation of the positioning of the isolation valves. The constraints of the former describe the feasible region of the Multiple Travelling Salesman Problem, while the objective function is computed by a hydraulic simulator. A simulation–optimisation approach based on Genetic Algorithms, mathematical programming, and Path Relinking, and a thorough experimental analysis are discussed hereby. The constraints of the latter problem describe a graph partitioning enriched with a maximum flow, and it is a new variant of the common graph partitioning. A new mathematical model plus a new formalization in logic programming are discussed in this work. In particular, the technologies adopted are mixed-integer linear programming and Answer Set Programming. Addressing these two real applications in hydraulic engineering as constrained optimisation problems has allowed for i) computing applicable solutions to the real case, ii) computing better solutions than the ones proposed in the hydraulic literature, iii) exploiting graph theory for modellization and solving purposes, iv) solving the problems by well suited technologies in Operations Research and Artificial Intelligence, and v) designing new integrated and hybrid architectures for a more effective solving.