Simplifying Real-Time Simulation of Large Power Networks
Power system equations represent a particular class of stiff systems, with dynamically very fast and highly damped components, difficult to solve numerically and challenging to simulate in real time. A new approach for real-time simulation of large, nonlinear power networks within the framework of the Resistive Companion Method (RCM), is the goal of this project. The method consists in a systematic substitution of the components of the network, in order to conveniently transform the nonlinear system into a network composed of only discretized sources and resistors, which can be solved with the RCM.
With the substitution of nonlinear components for current sources, and the subsequent discretization the simulation
Boils down to finding the DC operating point of a Linear Time Invariant (LTI) system.
The real-time execution forces the simulation time step to Be fixed. The numerical stability requirement of the simulation also imposes an upper bound to the step size to be used. In order to meet these constraints, the combination of an explicit integration method for the nonlinear components, and an implicit one for the linear components, makes it possible to satisfy both step size constraints of the simulation.
The RCM produces a natural partition of the network, where the equations of every discrete equivalent component can be solved separately and simultaneously, which allows a parallel solution scheme. The proposed method is to be implemented on a multicore machine based on Digital Signal Processors (DSPs) for computational purposes, and on Field Programmable Gate Arrays (FPGAs) for signal routing. The implementation is under development at the E.ON Energy Research Center. The software implementation done within the framework of the simulation tool Virtual Test Bed (VTB).
For the implementation, we used VTB´s code generation capability in order to produce the C-Code solver file, with the proposed simulation scheme, that runs on the DSP Cluster. Moreover, the user is able to run the proposed method also on the desktop version of VTB; the main reason behind the incorporation of this feature is not to perform desktop real time simulation, but instead to give the user the possibility to evaluate, case by case, the effects of solving explicitly some components, before deploying the code on the DSP cluster for the real-time execution.
From user’s point of view, this feature allows for reusability of the simulation schematic setup. In fact, once a simulation scenario is created using the schematic editor of VTB, the user will be able to simulate that schematic, using the traditional resistive companion solver or using the proposed new approach.
As the purpose of the desktop version is not to speed up the desktop simulation, its development was focused on the impact on stability and accuracy, reusing the current structure of the VTB solver as much as possible. In fact, we were able to implement such a capability by modifying only the individual component models, without changing the solver structure. A new parameter called “Integration”, and a new method named “Internal_Step”, were defined for nonlinear components. Linear components were left unchanged.