Laserfiche WebLink
underlying complexity of the flow network solution. It is inherently more difficult to grasp the <br />concepts related to solving the entire system network at once, given certain constraints and minimum <br />"cost" parameters. However, it is noted that the normal water user does not need to fully understand <br />the mathematical concepts of the MODSIM solutions, nor is it necessary to fully comprehend or <br />analyze the structure of the hidden "accounting" nodes and links that are automatically assigned and <br />solved by MODSIM. The steeper learning curve inherent with implementation of MODSIM is offset <br />by the flexibility available to address complex water resource issues and the general confidence in <br />the mass balance for the network. <br />The existing documentation for MODSIM is relatively complete and understandable. There is <br />considerable discussion of the underlying solution methodologies and a description of the data <br />requirements and input formatting. The documentation also includes a number of sample test <br />applications which are useful in gaining understanding of the model setup and operation. The source <br />code for MODSIM is reasonably well commented in regard to a description of the variables and the <br />subroutines. <br />Performance . It is difficult to assess the performance of one particular model in comparison to <br />another without an extremely complex test case of sufficient magnitude to really stress the <br />simulation process. It would be expected that a flow network model would take more computation <br />time than a strict allocation/accounting model because of the iterations necessary for the former to <br />achieve solution. There have been concerns expressed in regard to the computational time required <br />to run the Gunnison River basin model. It is our understanding that this model (CRAM) relies on the <br />out-of-kilter algorithm (OKA) as the solver. Recent versions of MODSIM have incorporated a <br />Lagrangian Relaxation algorithm as the network optimizer, in lieu of the OKA algorithm used in <br />1 <br />earlier versions. Studies by Bertsekas and Tseng indicate that the Relaxation algorithm can be <br />expected to improve the speed of computations, even for very large networks, by five to ten times. <br />Also, the code for the workstation version is being modified so that it will no longer be necessary to <br />open and close files during computations as is required by the earlier PC versions. This is expected <br />to greatly improve the efficiency. Finally, during the development stages of the model for the <br />CRDSS, the CRDSS Project Team will likely be making a higher number of iterations to ensure <br />proper solutions and higher confidence in the fundamental assumptions. After this development <br />period, it is expected that the number of iterations may be significantly reduced by establishing fixed <br />values to certain parameters (such as return flow patterns). This will further enhance the <br />performance of the model for the end-user. <br />Development for the CRDSS and Future Maintenance . MODSIM is supported by Colorado State <br />University and Dr. Labadie, and future modifications and enhancements can be expected. The State <br />and CSU are negotiating an agreement about use of the model and the source code, including the <br />right to change or modify the source code. The CRDSS Project Team is very familiar with the <br />application of the MODSIM program. <br />Level One Decision Criteria <br />Availability of Model Code . CSU and Dr. Labadie are negotiating an agreement with the State <br />concerning the State's use of the MODSIM program, including its source code. The State would be <br />1 <br />Bertsekas, D. and Tseng, P.; "Relaxation Methods for Minimum Cost Ordinary and Generalized Network Flow <br />Problems," Operations Research , Vol. 36, No. 1, (1988). <br />4 <br />A275 05.10.94 1.15-5 Fosha, Hyre <br />