Stochastic Reservoir Operation under Drought with Fuzzy Objectives

Parent, E.; Duckstein, L.; ,

Stochastic Reservoir Operation under Drought with Fuzzy Objectives

by E. Parent, ENGREF, Paris, France,
L. Duckstein, ENGREF, Paris, France,

Document Type: Proceeding Paper

Part of: Water Management in the '90s: A Time for Innovation

Abstract:

Biobjective reservoir operation under drought conditions is investigated using stochastic dynamic programming. As both objectives (irrigation water supply, water quality) can only be defined imprecisely, a fuzzy set approach is used to encode the decision maker (DM)'s preferences. The nature driven components are modeled by means of classical stage-state system analysis. The state is three dimensional (inflow memory, drought irrigation index, reservoir level); the decision vector elements are release and irrigation allocation. Stochasticity stems from the random nature of inflows and irrigation demands. The transition function includes a lag one inflow Markov model and mass balance equations. The human driven component is designed as a confluence of fuzzy objectives and constraints after Bellman and Zadeh (1970). Fuzzy numbers are assessed to represent the DM's objectives by two different techniques, the direct one (Bardossy and Duckstein, 1992) and indirect pairwise comparison (Saaty, 1980). The real case study of the Neste river system in southwestern France is used to illustrate the approach; the result are compared to a classical sequential decision theoretical model derived earlier (Parent et al., 1991) from the viewpoints of ease of modeling, computational efforts, plausibility and robustness of results.

Return to search