Journal of Water Resources Planning and Management

This paper compares reservoir operating policies obtained from fuzzy and nonfuzzy explicit stochastic dynamic programming. The reservoir operation problem for the Mansour Eddahbi dam in Morocco can be formulated as either a classical stochastic dynamic programming (SDP) problem, where the objective function stresses energy maximization with particular volumes being released for irrigation, or a fuzzy stochastic dynamic programming (FSDP) problem, in which both hydropower generation and irrigation are considered as fuzzy constraints and aggregated by the weighting method. System performance is estimated from simulation based on continuous reoptimization models using either the cost-to-go function generated by the SDP algorithm, or the membership function generated by the FSDP algorithm. Despite major differences in the mathematical representation of operating objectives and/or constraints, we show that both formulations yield similar measures of system performance.

This research quantifies the impact of lake evaporation and rainfall on optimal reservoir capacity and water yield. A reservoir design and operation model was developed and applied to the Santa Ynez River basin of central California, which endures large evapotranspiration and extreme climatic variability. Reservoir design and average annual water yield were obtained in two cases. First, lake evaporation and rainfall fluxes were taken into account in the water balance of the reservoir system. Second, those same fluxes were ignored. The optimization-model results indicate that in-lake hydrology plays a considerable role on estimates of optimal reservoir capacity and yield. Furthermore, results indicate that the lack of proper consideration of in-lake hydrology leads us to err on the side of greater risk. Specifically, reservoir capacity and average water release are under and overestimated, respectively. The optimization model is particularly well-suited for modeling reservoir systems with active in-lake hydrologic fluxes and allows a variety of objective functions to be considered, thus, providing flexibility in the optimization of reservoir design and operation.

Yield models produce reasonable reservoir designs with release reliabilities near targets. This paper extends the basic yield model and presents a multiple-yield model for a multiple-reservoir system consisting of single-purpose and multipurpose reservoirs. The objective is to achieve prespecified reliabilities for irrigation and energy generation and to incorporate an allowable deficit in the annual irrigation target. A single-reservoir yield model illustrates how a single-yield problem can be converted to a multiple-yield problem that represents the same irrigation deficit criterion while maintaining the desired reliability. The yield model is applied to a system of eight reservoirs in the upper basin of the Narmada River in India. The results are analyzed for four cases. As the desired reliabilities for yields can be prespecified for different purposes by consideration of an allowable deficit in annual yield in a multiple-reservoir system, this model can act as a better screening tool in planning by providing outputs that can be very useful in improving the efficiency and accuracy of detailed analysis methods such as simulation.

An improved artificial neural network-based algorithm for short-term water inflow forecasting (STWIF) into run-of-river hydropower plants is presented. Deregulation of the electrical power industry and introduction of competition between power producers have prompted redefinition of their daily operational tasks. In a competitive market environment, accurate short-term production planning and profitable bidding strategies become an important issue, requiring water inflow forecasts for up to 36 h ahead. As a result, short forecasting horizons have been found to be a main drawback of first-generation STWIF. An additional module using forecast precipitation data is developed for enlarging the forecast horizon up to two days ahead. The water inflow forecaster is further enhanced by inclusion of new input variables. With a large number of potential input variables, a new algorithm for selection of input variables using average mutual information and nonparametric density estimation is applied to the specific problem of water inflow forecasting. The performance of the enhanced STWIF is applied to the Soca River cascade hydropower system in Slovenia, and results are presented along with some comparisons with the previous-generation STWIF.

Flow duration curves (FDCs) are useful tools for many water resource development and management purposes. Many water resource projects are located at ungauged sites, however, for which FDCs are required but unavailable due to the lack of observed flow records. Both the polynomial and area-index methods are introduced herein to synthesize FDCs at ungauged sites and estimate their uncertainties for the upper catchments of the Cho-Shuei Creek in Taiwan. The major difference between these two methods is that the area-index method only uses drainage area to explain regional variation, whereas the polynomial method uses annual rainfall, altitude, and drainage area. Results obtained by using cross validation at 15 stations indicate that the polynomial method produces better estimates of synthetic FDCs at ungauged sites than the area-index method. Furthermore, uncertainty analysis through bootstrap resampling reveals that the polynomial method has less uncertainty in synthesizing FDCs than the area-index method.

The remediation of groundwater contamination by pumping and injection is generally a long-term and costly strategy. Aquifer cleanup time is a highly nonlinear and nonconvex function of pumping rates. The cleanup objective often involves minimizing or constraining the cleanup time or cleanup cost. Linear programming and nonlinear optimization cannot guarantee the global solution. In this study, four global optimization (GO) algorithms, including a popular genetic algorithm, are used to minimize both cleanup time and cleanup cost taking pumping rates and/or well locations as decision variables. Groundwater flow and particle-tracking models (MODFLOW and MODPATH) and a GO tool (GLOBE) are used. Real and hypothetical contaminated aquifers are considered for application. The results are satisfactory and show that GO techniques can be widely applied in groundwater remediation strategy and planning. The comparison of the performance of algorithms did not reveal a clear winner. The results also show that in the particle-tracking method, excluding few particles from removal can significantly reduce the cleanup time.

Leaky distribution systems are costly in terms of lost water, potentially adverse water quality effects, and the energy consumed in supplying the leaks. To characterize the energy effectiveness of a leaky segment in a single pipe, several dimensionless parameters are analytically derived, which relate the leak size and location to its associated energy burden and water loss. The computer program EPANET is used to simulate the energy costs of leaks on representative distribution networks. In particular, analysis is performed to illustrate the influence of total system demand, leak location, and topological complexity. Furthermore, the connection between water loss and energy costs illustrates the potential importance of energy costs when pipes are leaky. The impact of leaks on water age is also evaluated through simulation and via a dimensionless expression relating leak size and location to residence time.

Hydrologic and sediment transport modeling were used to examine the effectiveness of typical stormwater management policies in reducing the potential for stream-channel erosion. Two bedload functions and three total-load transport relationships were applied to 8 mm gravel and 0.5 mm sand bed materials to compare the performance of the relationships in estimating detention requirements across modes of sediment transport. The various sediment-transport relationships yielded widely diverging estimates of sediment-transport capacity and yet suggested detention volume requirements that agreed within 20%. Detention design for control of cumulative sediment load required a detention storage volume 61% greater than a peak control detention facility and resulted in an altered temporal distribution of sub-bank-full shear stress. Design of stormwater facilities based on time-integrated sediment-transport capacity may inadvertently result in channel instability and substrate changes unless the approach accounts for the frequency distribution of sub-bank-full flows, the capacity to transport heterogeneous bed and bank materials, and potential shifts in inflowing sediment loads.

A recently developed multiple-reservoir version of the modified sequent peak algorithm is used to assess the benefits of integrated reservoir storage-yield-performance planning using data from Iran and England. The results showed that for a given yield, integrated reservoir planning could result in a saving of at least 6% in required reservoir capacity when compared with the aggregated storage from analyzing reservoirs as stand-alone systems. Similar synergistic gains of integrated reservoir planning were achieved in systems performance criteria and in reservoir surface evaporation fluxes.