Fast multireservoir multiperiod linear programing models

Multireservoir, multiperiod linear programing models are typically computationally very expensive. Some of these models can be formulated as network linear programs (NLPs), for which computer codes about 100 times faster than general linear programing codes are available. A NLP formulation is presented for determining water assignments in a multireservoir system over some time horizon. It provides for demand zone shortfalls due to drought or transfer limitations, instream flow requirements, which can be violated during droughts, and seasonal reservoir target volumes. It also allows the trade‐off between reliability and demand shortfall severity to be explored. A case study illustrates computational performance. A NLP was formulated for a three‐reservoir, two‐demand zone water supply system modeled over 324 periods. It was solved in 26 CPU seconds on a VAX 8550 computer. Finally, it is shown that a NLP can be formulated to find the required minimum capacity of one reservoir in a multireservoir system so that system demand is just meet over some planning period.