OCR Text |
Show Chapter 6 Reservoir Operation Models Klemes ( 1979) identifies two physical objectives that are typically used for reservoir regulation where storage- mass curve techniques are used. These are: ( 1) regulation based on a firm value of the target release (' safe draft') which is either a constant or a periodic function of time and ( 2) regulation aimed at the greatest possible equalization of reservoir outflow. Klemes ( 1979) states that the second objective is usually regarded as the basic goal of reservoir regulation, since it tends to mitigate losses from both low flow periods and floods. He asserts that ' rule curve' based reservoir regulation approach the satisfaction of the second objective above, since they are derived using the historical record and are cognizant of both the past and future inflows. He states that the two objectives stated above may be regarded as the lower and upper limits of reservoir operation optimality. In the stochastic reservoir operation optimization literature, the yield formulation ( Loucks et al, 1981, pp. 339 - 355), the dynamic programming ( SDP) reservoir operation models ( Loucks et al, 1981, pp. 321- 332) and the chance constrained ( CCLP) models ( Loucks et al, 1981, pp. 55- 373) also consider the reliability associated with the target outflow or release. Klemes ( 1979) discusses the utility of the stochastic dynamic programming and chance constrained formulations, relative to the operation rules derived by Varlet ( 1923) - the ' stretched thread method'. He argues convincingly, using a calculus of variations approach that the stochastic dynamic programming and the chance constrained formulations, at best, approach the solutions from the stretched thread method, where a single release from a reservoir is considered and economic values are not discounted. Stedinger et al ( 1983) also show that the yield formulation is superior to the chance constrained formulations, in terms of the desired yield reliabilities being actually realized. Typically, the chance constrained formulations are too conservative in their exploitation of the reservoir storage. Klemes ( 1979) argues that while the systems analytic techniques ( SDP, CCLP) can approach the results from the stretched thread method in the limit, they are computationally expensive and fail to provide direct insights into the reservoir operation problem. By comparison for the stretched thread method he states: The piecewise constancy, independence of current storage, and abrupt changes at corner points of the optimal release in any specific situation should be sufficient to undermine our belief in the possibility of approaching the truly optimal release rules in the absence of foreknowledge of future flows, since no systems- analytic or other legerdemain can produce the needed but unavailable information. Perhaps the main practical virtue of the stretched thread method is that it helps us to understand this unpleasant reality by displaying the mechanism of optimal reservoir operation in its most naked form stripped of the camouflage of spurious mathematics and technical jargon. It demonstrates in the clearest possible way the importance of flow forecasting as well as the value of the historical flow record. It shows why small reservoirs need forecasts with smaller lead times, demonstrates the operational robustness of a large reservoir in which the mean flow is the safest bet for optimal release, and makes it clear why rule curve based operation' policies can be quite reasonable for reservoirs for annual regulation, but have hardly anything to offer in the case of large reservoirs intended for long term regulation. It explains why it is not so much the inflow rates that are important for optimum operation but rather the total inflow volumes for the reservoir ' working cycles': it is these volumes which determine the location of the corner points which are crucial for the determination of optimum outflows. 155 |