Informations et ressources scientifiques
sur le développement des zones arides et semi-arides

Accueil du site → Doctorat → Japon → STOCHASTIC RESERVOIR OPERATION FOR WATER QUANTITY AND QUALITY USING ARTIFICIAL INTELLIGENCE TECHNOLOGIES

Kyoto University (2003)

STOCHASTIC RESERVOIR OPERATION FOR WATER QUANTITY AND QUALITY USING ARTIFICIAL INTELLIGENCE TECHNOLOGIES

dos Santos Chaves Paulo Henrique

Titre : STOCHASTIC RESERVOIR OPERATION FOR WATER QUANTITY AND QUALITY USING ARTIFICIAL INTELLIGENCE TECHNOLOGIES

Auteur : dos Santos Chaves Paulo Henrique

Université de soutenance : Kyoto University

Grade : Doctoral Dissertation (Science) 2003

Résumé partiel
In the last few decades most of the water storage systems, such as lakes and reservoirs, have suffered from poor water quality, mainly caused by human activities within their watersheds and inefficient operation of these systems. Even though there are a variety of techniques to improve water quality in storage, most are either very costly or not sustainable. Therefore, efforts to find new methods to achieve better water quality are extremely important. By taking advantage of the close relationship between the quality (limnological) and quantity (hydrological and climatological characteristics) of water, a better way to manage stored volumes is proposed to yield optimal benefits for water utilization while at the same time improving its quality. The aim of this research is to integrate the management of water resources, that is to say storage reservoirs, considering quantity and quality issues. The goal is to generate a formulation for the integration of quantity and quality within multipurpose storage reservoirs, which can serve as a basis for practical improvements of environmental and water quality and future developments in the field by using state-of-the-art techniques of artificial intelligence (AI), such as artificial neural networks (ANN), fuzzy theory and genetic algorithm (GA). Long-term planning operation is addressed considering multiple objectives, such as flow stabilization, power generation and improvement of water quality. The analysis focuses on the planning operation of the reservoir for monthly time-step considering optimization and simulation models. The relationship between storage volume, release from a single outlet, and quality in the reservoir are considered as the main driving forces of the system. Many works using dynamic programming (DP) in the optimization of water resources systems can be found in the literature, but they rarely address the water quality issues. The few works that attempted to operate reservoirs by stochastic DP (SDP) considering water quality, rarely integrate simultaneously water quality simulation and optimization models ; instead they consider quality through transition matrix developed based on a priori simulated quality results. Moreover, in the water quality optimization models, usually only a few, one or two, quality parameters are considered as the increased number of quality parameters or state variables can result in a problem computationally intractable by DP. Besides the problem related to the curse of dimensionality, SDP models face other drawbacks, such as uncertainty due to the discretization of variables and the calculation direction, being the last responsible for the difficulties on integrating simulation and optimization models simultaneously. With the use of artificial vii Abstract intelligence techniques, such as ANN, the fuzzy sets theory and genetic algorithm (GA), a methodology to overcome the problems of DP models related to the operation of storage reservoir optimization, when water quality is also of concern, is proposed.

Présentation

Version intégrale (3,94 Mb)

Page publiée le 12 décembre 2013, mise à jour le 20 avril 2020