A realistic linear decision rule for reservoir management

Abstract

A generalized linear decision rule is presented which takes into account the aspect of spill in a multi-lag LDR model. The proposed rule incorporates past inflow experience to determine the optimum release rules based on a stochastic (linear) programming optimization model. It also prescribes a procedure of determining spill, should it occur, and the method of adjusting the release policy, accordingly, for the subsequent periods, which are directly affected by the spill of the current period. The use of the rule also makes it possible to produce a specification for a reservoir with a smaller capacity by taking liberal constraints on the reservoir freeboard during the monsoon months. The problem is solved, for the purpose of illustration, using the historical data of a river located in central India. Two synthetic streamflow series of a duration of 50 years each are generated under lognormal flow assumption. The prescribed release rules are applied to a hypothetical reservoir with the optimum capacity determined by the linear programming method, and the generated series as the inflow. The results and findings are quite satisfactory.