Please use this identifier to cite or link to this item: http://localhost:8081/xmlui/handle/123456789/490
Full metadata record
DC FieldValueLanguage
dc.contributor.authorAwchi, Taymoor Abdul-Majeed Ibrahem-
dc.date.accessioned2014-09-16T14:27:22Z-
dc.date.available2014-09-16T14:27:22Z-
dc.date.issued2004-
dc.identifierPh.Den_US
dc.identifier.urihttp://hdl.handle.net/123456789/490-
dc.guideSrivastava, D.K.-
dc.description.abstractMotivation to provide sustainable development for the benefit of present and future generations has led engineers to search for improved planning techniques which address all system impacts. Sustainability of reservoirs implies a need for the control and adjustment of reservoir planning and operation characteristics, resulting in optimal, or near optimal system performance throughout the life time of the reservoir. System analysis, which involves the use of optimization, simulation, and other decision-making techniques, is a set of powerful tools to solve reservoir planning and operation problems. Modeling the application of reservoir system analysis involves studies performed specifically to re-evaluate operating policies for existing reservoir systems. Periodic evaluations may be made to ensure system responsiveness to current conditions and objectives. The re-estimation of different water demands for the projects already existed, should be taken into consideration. These projects may fail to fulfill the actual water demands for different reasons. Special techniques for reservoir operation which take these aspects in the consideration should be adopted. In reservoir operation phase, more concentration is applied to determine more practical and user friendly initial policies for reservoir operation. In addition, some prospective realistic aspects should be incorporated in the reservoir operation in the optimization models. These aspects are usually considered in linear programming models but are not incorporated in the dynamic programming models. Recognizing most of the above mentioned aspects, the integrated re-planning and operation of the Mula project has been considered for the present study. The Mula project is a major irrigation project on the river Mula, a tributary of Pravara which is a sub-tributary of Godavari. The multipurpose project provides irrigation, and water supply to the nearby areas. Several system analysis techniques have been used in this study for the re-planning and operation of Mula reservoir, i.e., linear programming (LP), dynamic programming (DP), artificial neural networks (ANNs), and hedging rules (HRs). The reservoir yield model (YM) based on LP is used for reservoir planning to assist in re-estimation of different reservoir yields for the given actual reservoir storage capacity. Both the cases of single yield (irrigation demand) and multiple yields (water supply and irrigation demands) by incorporating a factor (failure fraction, dPii) which represents the percentage of yield which may be provided during failure years (4 failure years out of 18 years are considered for nearly 75% annual dependability/reliability). Previous studies used the in minimum food requirements to estimate the value of (6>A,), but due to the lack of economical information for Mula project, the reservoir yield model is tried for the determination of the optimal 6pj values. In addition, it was tried to use the reservoir yield model in estimation of different reservoir storage capacities for different demands and priorities, over-year and within year reservoir capacities, and identification of failure years. For operation phase, it was aimed to use the information obtained from YM to build a suitable continuous and discrete hedging rules. Dynamic programming (DP) models are used extensively in the field of reservoir planning and operation due to the fact that the nonlinear and stochastic features, which characterize a large number of water resources systems, can be translated into a DP formulation. For planning phase, an attempt was made to introduce reservoir yield reliability factor in both the DP models considered in this study, i.e., controlled output (CODP) and controlled input (CIDP) models. Simultaneously, the use of factor (9Pji) is also introduced. The DP models are used to estimate the optimal size of the reservoir. In addition, the effect of changing reservoir storage capacity on release targets (yield targets) is to be explored. An attempt was made to use both the CIDP and CODP models to obtain the monthly optimal releases for the reservoir. Due to the lack of economical information for Mula project, it is aimed to use different objective functions with both the DP models to choose a proper objective function which leads to the highest performance of the models. In the last decade, extensive attention is given to investigate the potential of the ANNs. The ability to identify a relationship from given patterns make it possible for ANNs to solve large-scale complex problems. In the field of reservoir planning and operation, the investigation and application of ANN models is still rare. Most of the studies in the literature consider the feedforward neural networks and less attention is given to other ANN models. In the present study, the radial basis networks (RBN) are considered beside the well known feedforward neural network models. For planning phase, it is aimed to use the ANN models in best fitting of the reservoir storage capacity-demand information obtained from sequent peak and DP models for different reliabilities and 9pj values. In addition, it is aimed to use the ANN models with storage capacity-demand plots to obtain the DP targets from sequent peak model results. For stochastic generation of monthly inflow to Mula reservoir which can be used later on for reservoir operation with hedging rules, it is aimed here to employ a hybrid model in which the ANN is a main component of it. Thomas -Fiering model is to be applied for comparison and verification of ANN based model results. ANN models are used IV for reservoir operation by developing a reservoir operation policy, through DP models functional relations between the inflow, water demand and reservoir storage as input, and reservoir release as output. To verify the ANN based functional relations, the reservoir releases determined from the ANN model are compared to the results of reservoir operation simulations using multiple linear and non-linear regression models. Several aspects are being investigated, e.g., the effect of changing different ANN structural and training parameters on model performance, and the potential of radial basis networks (RBN) for reservoir operation is to be scrutinized. In addition, the effect of differentRBNdesign and training parameterson the network performance are to be studied. The well known standard linear operating policy (SOP) is an example of a very simple reservoir operating rule. The main withdrawal of the SOP model is that it does not account for future conditions, i.e., the occurrence of low flow or drought periods especially in the late non-monsoon season. Therefore, the application of hedging rules (HRs) nowadays is becoming more accepted for the periods of anticipated droughts or low flows. Hedging rules curtail deliveries over some range of water supply to retain water in storage for use in later periods. This type of insurance is suitable for reservoirs with low refill potentials subject to variable annual inflows and operated for over-year storage. Hedging reduces the risk of large shortages but at the cost of having more frequent small shortages. Two types of hedging rules are considered in the present study, i.e., the continuous hedging rules (CHR) and the discrete hedging rules (DHR). Most of the studies in literature used the hedging rules for single demand cases (water supply), but using hedging with multipurpose reservoirs is very rare. In the present study, it is aimed to use the hedging rules for multiple yields (water supply and irrigation demands). Several reservoir operation performance indices to be used for the comparison of results obtained from hedging rules application to those of SOP model application and to select a proper operation policy. A study is to be conducted to compare the performances of CHR and DHR models as this comparison is not available in the literature. In the previous studies, the monthly hedging fractions (Kpt) values for CHR are obtained using the mixed integer linear programming optimization (MILP) models. In the present study, it is aimed to use the results of reservoir yield model and DP models to estimate initial (Kpt) values. For discrete hedging rules, two-phase hedging is to be used and the use of YM results to estimate the hedging trigger volumes is to be investigated. The selection of hedging fractions of each hedging phase is a difficult task as there are no defined guidelines. In the present study, an attempt was made to choose a proper hedging fraction combination and to study the effect of changing the hedging fraction values on various reservoir operation performance indices is to be investigated. The results obtained from reservoir yield model considering single yield showed that an annual yield of 720.0 MCM (674.6 MCM proposed value) can be provided from the Mula reservoir for the project's proposed capacity of 608.88MCMfor a near optimum value of 6pj equal to 0.5 at an annual project dependability of 75% against 697.0 MCM obtained by simulation. Considering multiple yields with water supply as the first priority (100% annual dependability), an annual yield of 65.9MCM(73.92MCMproposed value) for water supply and 659.0 MCM for irrigation can be achieved. If the annual inflow with 75% dependability (668.31 MCM) is considered as an annual yield target, then the reservoir can provide 430.2 MCM annually during failure years for a near optimal value of 6pi equal to 0.625. The model results showed that the annual reservoir yield can only fulfill the irrigation demand of the project. This leads to the conclusion that the reservoir was initially planned for irrigation purpose only and the water supply demands were considered later on. Employing the DP models for reservoir planning showed that the no deficit CODPnd and the CIDP models give almost identical annual release targets (yield targets). A release target of 667 MCM could be considered for a 6Pri value of 0.5. The ANN models showed very superior results in all the aspects considered for in the present study. In addition, it is revealed that the RBN is a competitive tool too besides the feedforward back-propagation ANN models and to derive a reservoir operation policy. Applying continuous hedging rules (CHR) for reservoir operation indicated that the using the monthly values of hedging fraction (Kpt) obtained from the results of YM model gave an acceptable performance especially during the months of late nonmonsoon season for both the single and multiple yields cases. The DHR showed a very strong influence in decreasing the number of reservoir empty conditions and the hazards during low flows and droughts. VIen_US
dc.language.isoen.en_US
dc.subjectRESERVOIR-PLANNINGen_US
dc.subjectRESERVOIR-OPERATIONen_US
dc.subjectSUSTAINABLE-DEVELOPMENTen_US
dc.subjectPLANNING-TECHNIQUESen_US
dc.titleOPTIMAL PLANNING AND OPERATION OF A RESERVOIRen_US
dc.typeDoctoral Thesisen_US
dc.accession.numberG12020en_US
Appears in Collections:DOCTORAL THESES (Hydrology)

Files in This Item:
File Description SizeFormat 
OPTIMAL PLANNING AND OPERATION OF A RESERVOIR.pdf39.6 MBAdobe PDFView/Open


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.