Implementation of the unit commitment problem on supercomputers

This paper proposes the application of vectorization and parallelization techniques for solving the unit commitment problem on supercomputers. A modified dynamic programming method which takes full advantage of the vector processors present in a supercomputer is presented. Parallel processing is used to further speedup the computation time. Computer simulation of a 26-unit power system is performed on a CRAY Y-MP2/216 in both scalar and vector codes. The computation time results for the conventional dynamic programming algorithm and modified algorithm are compared to demonstrate the effectiveness of the proposed method.<
>