Parallel SDP for Multi-Reservoir Management

by Elizabeth A. Eschenbach, (M.ASCE), Univ of Colorado, Boulder, CO, USA,
Christine A. Shoemaker, (M.ASCE), Univ of Colorado, Boulder, CO, USA,
Hugh M. Caffey, Univ of Colorado, Boulder, CO, USA,



Document Type: Proceeding Paper

Part of: Water Policy and Management: Solving the Problems

Abstract:

The use of stochastic dynamic programming (SDP) for solving water resources problems has been restricted to problems with low dimension due to the 'curse of dimensionality'. Algorithm improvements and hardware advancements have reduced and will continue to reduce the curse of dimensionality of SDP. This paper reports results of parallel algorithms for continuous state and continuous control SDP applied to the solution of a 4 reservoir problem. We present parallel results from a supercomputer with 6 vector processors and results from 6 RS/6000 workstations. Efficiencies up to 93% are achieved on the shared memory supercomputer, while efficiencies from the message-passing workstations are less than 60% due to communication overhead.

Subject Headings: Stochastic processes | Computer programming | Water resources | Reservoirs | Algorithms | Water management | Systems management

Services: Buy this book/Buy this article

 

Return to search