تخصیص پویا از شاخص های کیفیت آب در یک مدل بهینه سازی چند مخزن

In the southern regions of Mediterranean Europe, the greatest part of water resources for supply systems come from artificial reservoirs. Eutrophication is one of the most serious problems affecting the quality of water stored in reservoirs. A simplified approach that includes water quality aspects as water use limiting factors in a multi-reservoir optimization model can be achieved by adoption of the Trophic State Index (TSI). This paper makes some improvements on the optimization model already presented by Sechi and Sulis in the management of large systems. Particularly, it addresses the possibility of dynamic attribution of quality indexes in the LP model. The application of the optimization approach to different operating rules in a real multi-reservoir system in southern Sardinia highlights the need for joint consideration of quality and quantity aspects for effective water management.

Eutrophication is one of the most serious problems affecting the quality of water in multireservoir systems. The increase in nutrients leads to greater productivity of the water system which may lead to excessive increase in algal biomassor other primary producers such as macrophytes. Excessive algal biomass can seriously affect water quality, especially if it creates anaerobic conditions. Therefore, even when using a simplified approach in a mathematical optimization tool, there is a requirement to include water quality indexes associated with the trophic state of reservoirs [1].This paper can be considered as a development of the optimization approach presented by Sechi and Sulis [2]. It is a modelling tool intended to assist decision-makers in identifying and evaluating management alternatives when considering simplified forms of water quality classification. It is well known that mathematical optimization procedures for large water resource systems are still unable to deal with all real-world complexities even when they can be easily incorporated in a simulation model. Nevertheless, optimization results can be seen as a reference target for simulation since optimization results can be considered as obtained by an ideal system manager [3]. Many studies have been carried out on the development of optimization models for multiple reservoir systems. However, few of them have taken water quality as an objective. Dandy and Crawley [4] modified an existing linear programming model to identify policies minimizing average salinity in water supply. A network optimization model for water allocation to demands with different quality requirements was described by Mehrez et al. [5]. Hayes et al. [6] integrated water quantity and quality modelling in an operational model for use in multi-reservoir hydropower systems. An essential step in the construction of an advanced optimization model for addressing real problems is linked to the definition of water quality in water bodies and the identification of constraints so as to set out correct system management criteria. In Sechi and Sulis [2], a criterion for classifying reservoirs in multiple reservoir systems was defined using Carlson’s Trophic State Index [1]. Trophic State Index (TSI), which in recent years appears to have attained general acceptance by the limnological community, can be evaluated using chlorophyll-a, total phosphorus and Secchi disk transparency measurements. In Sechi and Sulis [2] the formulation of water quality constraints was addressed using chlorophyll-a measurements by means ofwhich it is possible to evaluate the TSI(Chl) trophic index. In this paper an improvement on that approach is presented. It considers the dynamic introduction of quality index values in the optimization model. Instead of static introduction of quality values, the proposed model defines TSI(Chl) values according to changes in stored volumes in reservoirs. Two different operating rules are tested to highlight significant increases in system performance when water quality is considered as well as quantity in a common system management strategy.

Water drawn from reservoirs can be located at different depths. This allows for better quality in reservoir releases: the so-called “selective withdrawal”. By enabling control of the quantity and quality of releases, selective withdrawal is a typical reservoir-operating rule that integrates these aspects in the management of a reservoir system. When the selective withdrawal operating rule [OP(1)] is used, water released from the reservoir is characterized by a QE value equal to QEmean. On the other hand, a non-selective withdrawal operating rule [OP(2)] does not select the best layer and only considers quantity aspects. In this case, the QE value can be considered equal to QEmin. These operating rules were implemented in the model proposed by Sechi and Sulis [2] by means of two different constraints. The split-storage optimization model described above was implemented to assess improvement in system performances when considering quality and quality aspects in a common management strategy [OP(1)], instead of focusing attention on water quantity alone [OP(2)]. Optimization analysis spanned a 54-year time horizon and two optimization phases in order to highlight the consequences of examining quantity and quality issues jointly. Time reliability (expressed as percent value of the months in which deficit is equal or less than predefined thresholds) andvolumetric reliability (expressed as deficit rate on demand value for the entire period of analysis) are used to quantify system performances. Using the non-selective operating rule [OP(2)], Table 4 summarizes the degree of reliability of system management. Poorest volumetric reliability is obtained for irrigation (58%) since urban and industrial demands are marked by the highest deficit penalization cost in the OF. Temporal reliability values are estimated considering deficit values of 0% and 25%. As was to be expected, applying a selective withdrawal operating rule led to modification of optimal flux configuration. Operation of a reservoir under quality criteria can achieve the most restrictive downstream use, called designed use [12]. System performances in Table (5) show that OP(1) guarantees designated uses better than OP(2). Industrial demand is met during the whole time horizon. Temporal and volumetric reliability in civil demand improve significantly. Persistence of a significant deficit in irrigation demand occurred. In this case, not even a selective withdrawal rule is able to ensure better performances.Considering both storage volumes and associated quality indexes as decision variables of the problem, the proposed split-storage graph seems a good way to overcome a major obstacle to the application of linear programming in integrated quantity-quality analysis of complex water systems. Moreover, difficulties in incorporating all the complexity of the water system into the linear model still limit the effectiveness of practical application to real-world systems. Nevertheless, this approach can be viewed as a robust preliminary screening tool for providing information on operating policies. These policies will be further evaluated with more detailed simulation models not limited by many of the optimization model assumptions.