Decision Rules in Chance-Constrained Programming: Some Experimental Comparisons

A stochastic programming model for evaluation of investment in irrigation is presented. The model reflects the complex hydrologic and economic interactions that arise from consideration of the conjunctive use of ground and surface water, the need for the provision of drainage to irrigated lands, and the use of some lands for rice cultivation. The implications of stochastic precipitation are discussed. Several decision rules for system operation are presented and discussed in terms of computation requirements and the possibility of field implementation.

Deterministic equivalents to the probabilistic water requirement constraints are derived for zero-order, linear, and two-piece decision rules, and the model is solved with alternate assumptions regarding the form of the rules to be used. An irrigation project proposed for Bangladesh is chosen to illustrate some possible consequences for investment planning of the choice of decision rules. Decision rules in chance-constrained programming are seen as tools for effecting good approximate solutions to stochastic programming problems when used with knowledge of the technical aspects of the investment decisions to be analyzed.