Abstract

This paper presents the development of an efficient security constrained economic dispatch using multi-thread parallel computations. One of the main obstacles in wide application of economic dispatch is the slowness of algorithm execution time, especially when a large system is utilized. Aside from the model development to overcome this, we have proposed a technique to perform parallel computing of the problem which can make solution converged in shorter time. In this sense, the serial economic dispatch is decomposed into four sub-process alternatives for multi-thread programming. Among them, the two processes of loss-coefficient and barrier-optimization are potentially qualified for multi-thread parallel computing. Results obtained shows that the multi-threaded loss-coefficient process has improved the speed-up for the serial computations in all the cases at where the multi-threaded barrier-optimization process can improve the speed-up only if the execution time for the process at single iteration can take more than one second, which mostly happens in the systems with a large number of units. The performance results of the multi-thread economic dispatch exhibit improved speed-up and performance characteristics compared to the serial execution time in all cases. A variety of case studies of standard IEEE 118, 300 bus and practical utility 664, 4995 bus are employed to validate the promising performance characteristics.

Key words: Security constrained economic dispatch, multi-thread parallel computing, barrier optimization, exact loss.