For nearly a century, a core group of leading GE technical experts has focused on solving the electric power industry’s most pressing challenges—evolving electric power systems for greater affordability, reliability, and efficiency. Today, GE Energy Consulting continues this proud tradition by providing innovative solutions across the entire spectrum of power generation, delivery, and utilization. With our cross-company resources, GE Energy Consulting is able to serve a diverse global client base with a strong local presence while spearheading technology advances at our centers of excellence for these areas.
Improving PSLF Simulation Performance and Capability
Today’s power system planning tools simulate systems that are far larger and more complex than even those just a few years ago. Advances in renewable technologies and more pervasive control technology are driving planning engineers to analyze an increasing number of scenarios and system models with much more detailed network representations. By developing and customizing GE’s Concorda Positive Sequence Load Flow (PSLF) Software, our software experts provide clients a comprehensive set of state-of-the-art tools to assess the economic and technical performance of interconnected power systems.
PSLF is a software tool that enables power system planners to perform AC and DC steady-state power flow and dynamics analysis. Planners use PSLF to develop a detailed bus–branch model of the power system similar to the representation below and calculate the voltages and angles at every bus—up to 80,000—in the network. Dynamics analysis by PSLF represents dynamic performance under system disturbances. PSLF includes comprehensive dynamic modeling of generators, excitation controls, turbines, governors, driven loads, relays, and other system components. The computational demands of today’s advanced models, larger systems, and greater sensitivities have outstripped processor performance used to run these software tools.
This computational dilemma has reached a critical point, and technology must be developed to accurately model the systems of the future. Using the extraordinary high-performance computing resources available through LLNL’s HPC Innovation Center and the power systems domain expertise of GE Energy, the GE Concorda PSLF software will be demonstrated to perform contingency analysis in a massively parallel environment, significantly increasing performance. In this project, the modeling techniques to expand the computational capability of PSLF to efficiently deal with systems of greater than 150,000 buses – nearly eight times the size of the current Western Electricity Coordinating Council PSLF representation – will also be developed and tested.
One of the many things PSLF is used for is running contingency analyses on electricity grid systems. A contingency analysis is used to determine the stability of an electricity grid if electrical equipment such as generators, transformers, and power lines is removed from the system as depicted below. The electrical equipment is interconnected to efficiently generate and deliver power to consumers. This redundancy is designed into the interconnected system so that it continues to operate with the outage of some electrical equipment. Contingency analysis simulates the effects from equipment outage and informs planners and system operators how to operate under abnormal system conditions.
The collaborative GE Energy Consulting and LLNL team parallelized the PSLF code to run on High Performance Computing machines. In parallelizing PSLF, the amount of time required to conduct a complete set of contingency analyses on a data set decreases dramatically. In the depiction above, all 100 contingency analyses can be run concurrently. The time required to run all contingencies is reduced to the contingency that requires the most time to solve.
When applied to a study with 4,217 contingencies, the total time required to complete all analyses was 23 minutes.
Prior to the hpc4energy incubator
When calculated on a single desktop computer, contingency analyses occur consecutively. For the system above with 100 buses, the calculations would start with the first analysis and run through the 100th, analyzing the effects of an outage at one point. In direct comparison, the time required to run on one computer the complete analysis of 4,217 contingencies, calculating each contingency consecutively, is estimated to take 23.5 days.