GPU-accelerated N-1 static security analysis based on fine-grained parallelism HELM

With the ever-increasing demand for power, ensuring the reliability, robustness and security of existing power systems has become a big challenge. In particular, power system security has drawn significant research attention owing to its high vulnerability to internal and external disturbances. Power system security comprises system monitoring, contingency analysis, and corrective action. Thus, for a power system to be described as N-1 secure, it should be able to maintain normal operations even during single contingency events like unprecedented loss of a transmission line.

N-1 static security analysis (SSA) is a common method for analyzing the steady-state security of power systems. It requires solving N power flows and is often used for real-time analysis and decision-making. Unfortunately, there are no fast, accurate and robust methods for solving the resulting N-1 SSA problems, especially for large-scale power systems. Researchers have developed several methods to solve the N-1 SSA problems. These methods, however, have various degrees of accuracy and efficiency that may not be suitable for some circumstances. Additionally, their practical applications are limited by several drawbacks like complexity, as they involve a lot of iteration.

Recently, holomorphic embedding load flow method (HELM) has emerged as a promising method for solving the steady-state equations of power systems owing to its fast calculation speeds, enhanced accuracy as well as non-iterative and deterministic nature. In order to enhance the acceleration effect of HELM and make it more suitable for N-1SSA, parallelization is required. This can be realized using parallel accelerated algorithms capable of performing connectivity tests of power systems. Nevertheless, since the available algorisms fail to meet desired outcomes, it is necessary to develop more reliable ones.

On this account, Dr. Weiran Chen, Dr. Jin Xu and Dr. Keyou Wang from Shanghai Jiao Tong University developed a Graphics Processing Units (GPU)-accelerated N-1 SSA method for steady-state security analysis of power systems. In this method, alternating current power flow (ACPF) was performed by HELM. Fine-grained parallelism Padé approximation of HELM was performed to improve the computational efficiency of ACPF. Additionally, block-level parallelism algorithm was designed to ensure the accuracy of the connectivity tests. Their work is currently published in the peer-reviewed International Journal of Electrical Power and Energy Systems.

The research team showed that the proposed method proved effective in satisfying the increasing efficiency and accuracy requirements of N-1 static security analysis of power systems. In addition to reducing the computational burden, this method was also capable of improving the accuracy of ACPF and connectivity tests simultaneously. These remarkable improvements were mainly attributed to the significant improvement in the HELM computation capability to the use of Padé approximation to solve sparse linear systems.

Compared to conventional graph theory algorithms, the proposed method effectively used shared memory to achieve block-level parallelism, which also contributed to improved connectivity test speed and accuracy. Its other advantages included lower time complexity, ease of use, suitability for parallelization and improved performance. The feasibility of this method was successfully validated with large-scale power systems of up to 2869 buses.

In summary, Shanghai Jiao Tong University scientists reported the optimization of the two main steps of N-1 SSA to realize accelerated N-1 SSA. The feasibility of the proposed combination of the parallel algorithm design and GPU architecture was validated. Overall, the study results revealed the effectiveness and robustness of the proposed method in reducing the computation burden and enhancing the accuracy of ACPF and connectivity tests. In a statement to Advances in Engineering, corresponding author, Dr. Jin Xu stated that their study provides valuable insights that would contribute to advanced static security analysis of large-scale power systems.