Reservoir operating rules for urban water supply systems

In recent times, a significant shift has emerged throughout the world from planning and\ud construction of water resource projects to the efficient operation of existing systems due to\ud many reasons. The reasons may include the non-availability of water resources for further\ud development, the limited availability of funds for capital works and the spirited lobbying of\ud environmental groups against construction of major water resource projects. Therefore, it is\ud necessary to determine the optimum operating rules for water resource systems.\ud The generalised water supply simulation model (REALM) is been currently used by many\ud water authorities around Australia for planning and operation of urban water supply\ud systems. The operating rules used in REALM are the restriction rules and target storage\ud curves. These operating rules currently in use do not produce the optimum operation and\ud are based on the subjective operator experience. Therefore, the need to determine optimum\ud operating rules in the form of restriction rules and target reservoir curves was important.\ud A general approach was used to produce the 'optimum' operating rules for an urban water\ud supply system. Melbourne water supply system is considered as the case study. An\ud objective methodology and computer software (namely Restrictions and Targets) were\ud developed to derive the operating rules in terms of restriction rules and target storage\ud curves.\ud The restriction rules were derived using a direct search method known as the Hookes and\ud Jeeves method. The objective function used was the maximisation of releases to demand\ud zones. The constraints of current security criteria were considered. A lumped single\ud reservoir and single demand centre approach was used in the study, however, the effects of\ud multi-reservoir interactions such as reservoir evaporation losses, spills from the system,\ud effect of carrier capacity on releases and demand shortfalls were considered implicitly in\ud the approach.\ud The target storage curves were derived using Discrete Differential Dynamic Programming\ud (DDDP), with the objective function of maximisation of releases to demand zones.\ud REALM system data of the Melbourne system was used and therefore all system details\ud incorporated in a planning study of the system were included in DDDP. Water allocation\ud among various parts of the water supply system was done through network linear\ud programming (NLP). REALM is used by Melbourne Water in their planning studies. The\ud Restrictions and Targets software were developed for the Melbourne water system in this\ud study.\ud Both Restrictions and Targets software were used for the Melbourne system using system,\ud streamflow and demand data provided by Melbourne Water (MW) in early 1994. The\ud restriction rules and target storage curves were derived for both static and dynamic\ud demands. The behaviour of the Melbourne system was analysed under the derived and\ud current MW rules using a REALM simulation model of system for the planning period of\ud 1994 to 2026, and a comparison study was performed. The restriction rules derived under\ud both static and dynamic demand analysis performed better than the current MW restriction\ud rules. It was also found that the restriction rules derived from the static demand analysis\ud were consistently better than those of the dynamic demand analysis while the target storage\ud curves derived from DDDP slightly under-performed the MW current target storage curves.\ud Finally, the target curves derived were fine-tuned using simulation results and expert\ud knowledge, and the system behaviour improved significantly.\ud It is recommended that the operating rules (both restriction rule curves and target storage\ud curves) derived from the static demand analysis be used for the Melbourne system for long\ud term operation as Melbourne system cannot be augmented by constructing reservoirs in the\ud MW catchments due to lack of suitable hydrologic sites