Stochastic Operation Management Model for a Multi-Resevoir Inter-Basin Water Resource System

Abdolkarim Khajehmogahi, McMaster University


The discrepancy between the quantity and regime of water consumption and natural river flow usually gives rise to the need to create a multi-reservoir inter-basin water resource system to redistribute the river flow temporally and spatially. As the economic value of water increases and expensive sources of water supply diminish, the development of an optimal operation management model becomes more and more important. In this thesis, a stochastic operation management model composing of two integrated models of stochastic mutli-site flow generation model and deterministic Dynamic Programming optimization model is developed to determine the optimal operation of a multi-reservoir inter-basin water resource system. The stochastic multi-site flow generation model is used to generate synthetic flow series as input to the optimization model. In this model, the stochastic nature of historical flows to reservoirs, i.e. auto-correlation and cross-correlation, is explicitly considered. The deterministic Dynamic Programming optimization model is developed to determine the optimum operation policies for each of the many synthetic flow series through application of the optimization model to the multi-reservoir inter-basin water resource system. The approach of separating these two models will overcome the curse of dimensionality encountered in existing DP optimization models while allowing the stochastic nature of inflows to be incorporated into the optimization process and resulting optimum operation policies. The real case study approach, selection criteria and description of case study area, Lar-Kalan-Latian water resource system in Tehran, Iran as a multi-reservoir inter-basin water resource system is discussed. The superiority of a real case study in comparison with a hypothetical or abstract one is demonstrated. A comprehensive review and identification of stochastic multi-flow generation models and discussion of available flow generation computer programs are presented. The statistical analysis of historical monthly and annual flow data of the case study, setting up of HEC-4 program as direct method and SPIGOT program as disaggregation technique, and generation and verification of synthetic monthly and annual flow series for the case study are discussed. The comparison of HEC-4 and SPIGOT synthetic flow series with historical data shows the effectiveness of SPIGOT program against HEC-4, even for the short historical flow input data. A detailed review and discussion of the Dynamic Programming techniques and development of deterministic DP optimization model for the multi-reservoir inter-basin water resource system is carried out. The generation, comparison, statistical analysis, and reliability characteristics of optimum monthly operation policies determine by applying the developed DP optimization model is discussed. The comparison of optimum operation with historical operation demonstrates the usefulness and improvement of optimum operation upon historical operation. It is further concluded that the DP optimization model is not sensitive to the type of stochastic flow generation model use to generate synthetic flows. Finally, the development of optimum operation reliability characteristics demonstrates the application of stochastic operation management model in planning and operation of multi-reservoir inter-basin water resource system.