Decomposition of linear programs using concurrent processing on multicomputers
- Tennessee Univ., Knoxville, TN (USA)
This project is concerned with the distributed optimization of block-angular linear programming (LP) models using concurrent processing on multicomputers. The approach we use is that of Dantzig-Wolfe decomposition (DWD). The decomposition of an LP with R blocks is viewed as a coordination of a price-generating process (the Master problem) and R independent proposal-generating processes (the subproblems) towards a global objective. Each of these R + 1 processes is represented by one processor (node) in the multicomputer. The concurrent and asynchronous nature of a distributed system opens up the possibility of a variety of schemes in controlling the flow and utilization of information in the price-proposal coordination process. This provides not only the opportunity for speedup in Dantzig-Wolfe decomposition-based algorithms (DWDA), but also the prospect of truly distributed optimization models where the data for each subsystem is kept and maintained locally and optimization of the entire system is achieved by solely exchanging price and proposal information. An information scheme in distributed DWDA controls the timing of the availability and utilization of the coordinating information. This project investigates four information control schemes: Basic Information Scheme (BIS), Early Termination Information Scheme (ETIS), Early Start Information Scheme (ESIS) and Intermediate Prices Information Scheme (IPIS). A selection of problems are used to test such schemes in the environment of concurrent processing. The emphasis is on the underlying factors that determine the behavior of the information schemes so that guidelines can be developed to aid in the implementation of distributed DWDA for application on optimization problems arising from power systems planning and operations. 46 refs., 17 figs., 23 tabs.
- Research Organization:
- Electric Power Research Inst., Palo Alto, CA (USA); Tennessee Univ., Knoxville, TN (USA)
- Sponsoring Organization:
- EPRI
- OSTI ID:
- 6975879
- Report Number(s):
- EPRI-EL-6769
- Country of Publication:
- United States
- Language:
- English
Similar Records
Decomposition of linear programs using parallel computation. Technical report
Application of fuzzy sets to optimal reactive power planning with security constraints