Skip to Main Content U.S. Department of Energy
EIOC at PNNL

Massive Contingency Analysis and Visualization Tool

Framework of parallel massive contingency analysis
Figure 1. Framework of parallel massive contingency analysis with single counter. Click for a larger image.

Contingency analysis is a key function in an energy management system (EMS) to assess the impact of various combinations of power system component failures based on system state estimates. Because of the heavy computation involved, today’s contingency analysis can only be updated every a few minutes for only a select set of “N-1” contingency cases. This limitation can result in an incomplete situational awareness of the power grid, leading to inefficient and unreliable grid operation. The trend of operating power grids closer to their capacity and integrating more and more intermittent renewable energy demands faster analysis of massive contingency cases to safely and reliably operate power grids, not to mention the need for analyzing high order (N-x) contingency events. These challenges require computational speed to be significantly improved, in a short time frame of seconds. High performance computing techniques hold the promise of accelerating power grid contingency analysis.

Framework of parallel massive contingency analysis
Figure 2. Evenness of execution time for WECC full N-1 contingency analysis with different computational load balancing schemes. Click for a larger image.

Contingency analysis is naturally a parallel process because the communication between different contingency cases is very minimal. Therefore, the challenge in parallel contingency analysis is not on the low-level algorithm parallelization but on the computational load balancing (task partitioning) to achieve the evenness of execution time for multiple processors. At Pacific Northwest National Laboratory (PNNL), we have investigated different load balancing schemes. Figure 1 shows the framework of parallel contingency analysis with one master processor that manages case allocation and load balancing. The performance of static and dynamic load balancing schemes are studied and compared. For example, the execution time with 32 processors for WECC full N-1contingency cases is shown in Figure 2. As can be seen, the dynamic load balancing scheme successfully improved speedups.

The massive contingency analysis results using a MPP2 machine with 512 processors are shown in Table 1. As shown in the table, massive contingency analysis can achieve over 500 times speedup, and full “N-1” WECC contingency analysis can be completed with half a minute.

Table 1. Summary results of the massive contingency analysis on the MPP2 machine

512 Processors Used
Wall Clock Time (Seconds)
Total Computation Time (Seconds)
Total I/O Time (Seconds)
Total Counter Time* (Seconds)
Speedup
Scenario 1 (20,094 N-1 cases) 31.0 14235.2 82.7 0.899 462
Scenario 2 (150,000 N-2 cases) 187.5 93115.5 489.1 5.550 499
Scenario 3 (300,000 N-2 cases) 447.9 226089.8 1087.1 9.984 507

*Includes waiting time.

Framework of a multi-counter based dynamic load balancing scheme
Figure 3. Framework of a multi-counter based dynamic load balancing scheme with task stealing. Click for a larger image.

To better manage counter congestion, we have studied a multi-counter based dynamic load balancing scheme with task stealing, as illustrated in Figure 3.

The performance of the dynamic load balancing scheme with 1 million contingency cases has been summarized in Table 2. It should be noted that this is the first time the contingency analysis has been scaled up to more than 10 thousand cores. The results show that when the number of cores is no more than 2048, their performance is close, with the two-counter scheme being slightly better. However, when the number of cores is larger than 4096, the two-counter scheme is much better. This result proves that the two-counter dynamic load balancing scheme is suitable with a large number of processors.

Table 2. Performance of the counter-based dynamic load balancing schemes with one million N-2 contingency cases

# of cores
Single Counter Time (s)
Two- Counter Time (s)
Single- Counter Speedup
Two- Counter Speedup
512 1023.4 1021.0 511 513
1024 517.4 512.7 1011 1021
2048 250.1 250.1 2093 2092
4096 163.4 132.0 3202 3964
10240 74.3 66.9 7040 7877

To make the contingency analysis results more useful, advanced visualization techniques are needed to present data in a meaningful way. To provide better situational awareness and decision support, a graphical contingency analysis (GCA) tool has been developed at PNNL. This tool not only utilizes advanced visualization techniques, but integrates analytical functions, which supports an operators’ decision-making process. The main features of the GCA tool, shown in Figure 4, include:

  • graphical representation of CA outputs to improve wide-area situational awareness,
  • contingency ranking to help focus operators’ attention on the most severe contingency and properly prioritize preventive actions, and
  • interactive assessment of operators’ actions to help them determine the effect of the actions.
Framework of a multi-counter based dynamic load balancing scheme
Figure 4. The GCA key features and user study: 1) GCA user interface, 2) operation action options, 3) interactive assessment of different actions, and 4) a picture from a WECC operator training class. Click for a larger image.

The GCA tool has been evaluated by North American Electric Reliability Council (NERC)-certified operators during several Western Electricity Coordinating Council (WECC) operator training classes. The evaluation assessment results demonstrate the validity of the developed GCA tool. Results suggest that use of the graphical tool can significantly reduce the number of steps that operators need to reach an acceptable contingency analysis solution. During the assessment, 30% savings were achieved by using the graphical tool instead of the tabular tool, even with a small 60-bus test system. If this advantage translates into time-savings, one may project that the elapsed time to address a “violation condition” will be significantly reduced using the graphical visualization tool. In practice, when a violation condition continues past 30 minutes, the balancing authority receives a sanction (monetary fine) that can be as much as $1million. Shaving off even 10% or 15% of the time to remove contingencies can save a significant amount of money.


Notable Publications

Papers

Z. Huang and J. Nieplocha. 2008. “Transforming power grid operations via high performance computing.” In Proceeding IEEE Power Energy Society General Meeting, Institute of Electrical and Electronics Engineers, Piscataway, NJ.

I. Gorton, Z. Huang, Y. Chen, B. Kalahar, S. Jin, D. Chavarria-Miranda, D. Baxter, and J. Feo, “A high-performance hybrid computing approach to massive contingency analysis in the power grid,” in Proceeding IEEE e-Science International Conference, 2009, pp. 277–283.

Z. Huang, Y. Chen, and J. Nieplocha, “Massive contingency analysis with high performance computing,” in Proceeding IEEE Power Energy Society General Meeting, 2009.

Y. Chen, Z. Huang, and D. Chavarria-Miranda, “Performance evaluation of counter-based dynamic load balancing schemes for massive contingency analysis with different computing environments,” in Proceeding IEEE Power Energy Society General Meeting, 2010.

S. Jin, Z. Huang, Y. Chen, D. Chavarrá-Miranda, J. Feo, and P. C. Wong, “A novel application of parallel betweenness centrality to power grid contingency analysis,” in Proceeding IEEE International Parallel and Distributed Processing Symposium, 2010, pp. 1–7.

Y. Chen, Z. Huang, and N. Zhou, “An advanced framework for improving situational awareness in electric power grid operation,” in Proceeding 18th IFAC World Congress, 2011, pp. 12162–12170.

Z. Huang, Y. Chen, F. L. Greitzer, and R. Eubank, “Contingency visualization for real-time decision support in grid operation,” in Proceeding IEEE Power Energy Society General Meeting, 2011, pp. 1–7.

Y. Chen, Z. Huang, and M. Elizondo, “Value of faster computation for power grid operation,” in Proceeding 8th Power Plants Power System Control Symposium, 2012, pp. 242–247.

Y. Chen, Z. Huang, Y. Liu, M. Rice, and S. Jin, “Computational challenges for power system operation,” in Proceeding 45th Hawaii International Conference on System Sciences (HICSS), Jan. 2012, pp. 2141–2150.

Y. Chen, Z. Huang, and M. Rice, “Evaluation of counter-based dynamic load balancing schemes for massive contingency analysis on over 10,000 cores,” in Proceeding 2nd International Workshop on High Performance Computing, Networking and Analytics Power Grid, 2012.

M. Rice, Y. Chen, Z. Huang, C. Allwardt, and P. Mackey, “Alleviating contingency violations through visual analytics and suggested actions,” in Proceeding IEEE Power Energy Society General Meeting, 2013.

Book Chapters

Z. Huang and Y. Chen. “High-performance computing for advanced smart grid applications”, in Book: Smart Grids: Infrastructure, Technology, and Solutions, Editor-in-Chief: Stuart Borlase, CRC Press Taylor & Francis, Jan. 2012.

About the EIOC

Collaborations

Resources

Contacts

Additional Information

High-Performance Computing Research Areas

Technical HPC Contact