Abstract:
With the ever increasing demand for scarce water resources, it is imperative that water resource utilization is optimized in the short and the long-term. A major component of this endeavour is the optimization of reservoir system operation which until recently had been dominated by Linear Programming (LP) and Dynamic Programming (DP). An alternative to LP and DP is direct simulation-optimization within which the operating rule curves can be defined by decision variables and therefore obtained directly. The simulation—optimization approach which is now easily enabled by the use of evolutionary optimization techniques, is easier to understand and enables the incorporation of required system performance more easily and comprehensively than LP or DP. One of the main challenges experienced in applying simulation-optimization approaches is high computation intensity and reducing computation intensity whilst maintaining effectiveness leads to greater acceptability of the approach.
This paper investigates the applicability of simple trigonometric functions as a means of parsimoniously defining operating rule curves and thereby reducing the number of decision variables in reservoir system optimization. The procedure is applied to a system of two reservoirs whose total yield needs to be maximized while meeting multiple reliability constraints of supply and reservoir storage state. The results obtained do not indicate the need for trigonometric functions and suggest that an even simpler model that defines rule curves as straight horizontal lines with no monthly variation is adequate. While this may be the case for the system studied here, it is proposed that the trigonometric function or its variants be studied with more complex systems or other objectives. The shuffled complex evolution (SCE-UA) method is applied in the optimization and is found to be effective and efficient through the use of multiple randomly initialized runs.
