Search

Different water agencies use different modeling tools for water resources planning and management. For example, different jurisdictions in Australia use a variety of river system models and these models vary considerably in approach and assumptions, including different time steps (monthly and daily), flow routing (different types of routing and no routing), ordering solution (optimization and heuristic) and representation of management and operational rules. These fundamental differences in approach make it difficult to integrate existing models of connected river systems at a basin scale. A collaborative joint venture (the eWater CRC) between research, industry and government partners has recently developed an integrated river system modelling tool called “eWater Source” to improve on the existing river system models in Australia. One of the major advances in Source is the implementation of two decision algorithms, heuristic and NetLP approaches, for water distribution modelling in the same modelling platform. This paper describes the implementation of heuristic and NetLP approaches for water delivery under management and operations rules in Source, and compares the performance of the two approaches through a case study in the Goulburn-Broken-Campaspe-Loddon (GBCL) river system in south-eastern Australia. The key performance measures used to compare the approaches include the efficiency and equity of water delivery to water users, impacts on the reliability of supply, agreement with storage operating targets, and model application run time. The results demonstrate that the heuristic approach implemented in Source can replicate the performance of the NetLP approach for a model of reasonable complexity. This is important because the run times of models with heuristic approaches are shorter than models with NetLP approaches, so this will allow more complexity to be represented than was previously practical in equivalent NetLP applications. Agreement between jurisdictions to move to the single river system modelling platform will contribute to overcoming the problems faced by river managers in Australia in transboundary river basins.

Policy-makers face planning challenges such as deciding whether to build reservoirs or choosing how to distribute water to irrigators and the environment. They rely on advice from planners who compare scenarios with river models that they run over the recorded (and/or forecasted or generated) flows. Planners and modellers compose the river models in specialist software. This paper defines the ordering problem as the problem of determining how to release water from reservoirs to meet downstream demands. This paper compares algorithms for solving the ordering problem in a complex regulated environment. Linear programming, especially fast network flow mathematical solvers, have been extensively applied in water resource management software and tools for ordering such as REALM, which is used widely in Victoria. Rules-based approaches, which are essentially heuristic ordering systems, have also been widely applied. For example, most Queensland and New South Wales models are built in IQQM, which applies rules-based methods for ordering. The eWater CRC has recently developed eWater Source software, which provides both ordering systems. Consequently, it allows an easy comparison of the efficiency of water distribution solutions. In a case study in the Goulburn catchment of Victoria, using a simplified river system model, this paper compares the efficiency of water distribution from NetLP based ordering with a rules-based ordering method. The comparison of efficiency is based on overall shortfall percentages, and total water released but not extracted by demand centres, which reflects both the short-term (single time step) and long-term (multiple time step) efficiency. REALM's network linear programming algorithms are also compared with the results of those from NetLP ordering in eWater Source. Given the many limitations of the analysis, this paper found that eWater Source can achieve relatively higher efficiency than REALM in the case study. Refereed/Peer-reviewed

Refereed/Peer-reviewed