A risk-based interactive multi-stage stochastic programming approach for water resources planning under dual uncertainties. (August 2016)

• Record Type: Journal Article
• Authors: Wang, Y.Y.; Huang, G.H.; Wang, S.; Li, W.; Guan, P.B.
• Is Part Of: Advances in water resources. Volume 94(2016)
• Page Start: 217
• Page End: 230
• Publication Date: 2016-08
• Subjects: Multi-stage -- Stochastic programming -- Dual uncertainties -- Interactive -- Random interval -- Risk analysis -- Water resources; Hydrology -- Periodicals; Hydrodynamics -- Periodicals; Hydraulic engineering -- Periodicals; 551.48
• DOI: 10.1016/j.advwatres.2016.05.011
• Languages: English
• ISSNs: 0309-1708

Abstract

Highlights: We develop a risk-based interactive multi-stage stochastic programming (RIMSP) approach to reflect uncertainties existing in both objective function and constraints of a general water-allocation optimization system, which enhance the applicability of the traditional methods. We can obtain a number of alternatives by setting different combinations of risk levels applied to the stochastic objective function and constraints, which is meaningful for performing an in-depth analysis of trade-offs between total net benefits and associated risks, as well as policies making. RIMSP approach is capable of reflecting dynamics of uncertainties and decision processes by constructing a set of scenarios that are representative for all possible random outcomes within a multi-stage context. Abstract: In this study, a risk-based interactive multi-stage stochastic programming (RIMSP) approach is proposed through incorporating the fractile criterion method and chance-constrained programming within a multi-stage decision-making framework. RIMSP is able to deal with dual uncertainties expressed as random boundary intervals that exist in the objective function and constraints. Moreover, RIMSP is capable of reflecting dynamics of uncertainties, as well as the trade-off between the total net benefit and the associated risk. A water allocation problem is used to illustrate applicability of the proposed methodology. A set of decision alternatives with different combinations of risk levels applied to the objective function and constraints can be generated for planning the water resources allocation system. The results can help decision makers examine potential interactions between risks related to the stochastic objective function and constraints. Furthermore, a number of solutions can be obtained under different water policy scenarios, which are useful for decision makers to formulate an appropriate policy under uncertainty. The performance of RIMSP is analyzed and compared with an inexact multi-stage stochastic programming (IMSP) method. Results of comparison experiment indicate that RIMSP is able to provide more robust water management alternatives with less system risks in comparison with IMSP.