Novel parallel computing decoupling methods for power system distribution networks

This work presents and describes three strategies for decoupling distribution power networks among parallel processing cores in a real-time (RT) multi-core environment. The proposed techniques help remove computational limitations and speedup real-time simulation of networks. Prior to this work, the solutions employed include approaches such as decoupling with Stublines or delay-free State Space Nodal (SSN) solvers. Further, these approaches are limited to high-end proprietary hardware and software. The main issues addressed in this work are reducing computational complexity, storage complexity, and employing general-purpose computers to allow simulation of large power networks using RT-systems. The Compensated Distributed Line Decoupling (CDLD) method enhances the existing Stubline decoupling by improving its accuracy and transient response. In addition, CDLD combined with an SSN solver improves computational performance and removes bottleneck issues. The CDLD method was tested on three IEEE benchmark systems and resulted in significant improvement in the network response and computational performance compared to the Stubline method. When compared to SSN, mean computation time improved and overrun issues are removed. The combined SSN-CDLD method proved the most promising approach for network decoupling using dedicated high-end RT hardware and software. The second part of this work involved decoupling the power network into arbitrarily sized clusters, where each cluster was discretized using state-space equations. For each time-step, an approach following the published SSN method combined the individual clusters into a single nodal admittance matrix to resolve interfacing voltages at the nodes and use them for updating the clusters state-space equations. This method was programed and implemented on real-time general-purpose computers. The method was tested on two IEEE benchmark systems. The objective of this effort was to develop the SSN methodology for further investigations and make the capability available on none-dedicated hardware and software. Finally, a novel technique for decoupling the SSN nodal matrices that requires an approximation approach to resolve the complexity of the nodal admittance was developed. This technique allows distribution of computation effort to multiple cores and isolates the communications to interfaced cores only. The method was tested on an IEEE benchmark system and found to improve computational runtime complexity with acceptable accuracy.

Grantor dc:publisher

University of Tennessee at Chattanooga

Year dc:date.available

2025

Author dc:creator

• Saied, Noman

Contributors dc:contributor

• Karrar, Abdelrahman A.
• Liang, Yu; Eltom, Ahmed H.; Ahmed, Raga
• College of Engineering and Computer Science

• Decoupling (Mathematics)
• Parallel processing (Electronic computers)

Licence

• http://rightsstatements.org/vocab/InC/1.0/

Language dc:language

English, eng

Identifier

xxii, 143 leaves

OAI identifier oai:identifier

oai:scholar.utc.edu:theses-1935