Parallel Computing Simulation in EMTP^®

As EMT power system models grow in complexity and are used more frequently for routine analysis, enhancing simulation speed becomes crucial. The EMTP^® development team addresses this challenge through several key advancements:

- Our innovative sparse matrix solver and ongoing enhancements to our code base have significantly boosted simulation speed, achieving performance levels unmatched by any other EMT tool. For example, simulating the IEEE 8500 system can now approach real-time speeds on a single CPU. 

- Additionally, our new Parallel Processing Toolbox provides unparalleled capabilities for distributing computations across multiple CPUs:

• Users can partition their system into multiple subsystems, each processed in a separate thread. This approach optimizes CPU utilization and reduces simulation execution time. 

• The toolbox supports multiple time steps (multi-time-step mesh), allowing users to select the most suitable time step for each subcircuit. This improves resource efficiency, as regions with rapid dynamics can be simulated with smaller time steps, while slower dynamics can use larger time steps. The EMTP^® solver maintains unique numerical stability and precision even with time steps of up to 250µs or more.

• Subsystems can be initialized independently and asynchronously, using load flow results obtained by the EMTP^® load-flow solver. Each subsystem is assigned its own initialization time, meaning that while one subsystems is being initialized, the rest of the network model is on pause, waiting for all subsystems to complete their initialization. This significantly reduces overall initialization time by reducing computational burden during initialization. 

• The network model remains unified; there is no need to manually create separate designs or go through a difficult database management process. 

With these techniques, extremely large transmission system models, such as those of Hydro-Quebec or Chile, can be simulated nearly in real time, with some cases achieving faster-than-real-time performance.