A Comparison Of FPGA Implementation Of Latency-Based Solvers For Power Electronic System Real-Time Simulation

In the design of power systems, real-time simulation is a powerful tool to evaluate and validate designs before dedicating resources to develop such systems. Through use of real-time simulation, one can study the behavior of a power system in interaction with real, physical elements, all while avoiding the cost and risk in constructing and testing such systems before a design is finalized. In recent decades, effort has been made in the industry and academia to apply real-time simulation to that of switching power converters through use of high-speed digital signal processors (DSPs) and field programmable gate array (FPGA) devices. With these existing approaches, individual converters have been successfully simulated with relatively high switching frequency. At the advent of smart power grids of ever growing size, with switching converters operating at over 100kHz switching frequency, demand has increased to move high-fidelity, real-time simulation from converter-level modeling to system-level to simulate these power systems completely. However, many of the current simulation methods have difficulty to scale to system-level model size while maintaining capability to run in real-time for systems deploying high frequency converters. As such, efforts have been made to explore new simulation approaches that can meet these requirements. Simulation methods oriented towards high parallelism in their computations are perfect candidates for scalable, system-level real-time simulation. Two such methods include Latency-Based Linear Multi-step Compound Method (LB-LMC) and Latency Insertion Method (LIM). These methods exploit sources of latency in modeled systems to divide up computations into operations that can be performed simultaneously. Originally developed for software-based execution on traditional processors and DSPs, these methods are implementable on FPGA devices to take advantage of these devices hardware-based, low-latency execution and native parallelism. In this work, FPGA implementations of the LIM and LB-LMC methods for system-level real-time simulation of power systems are developed and compared for scalability and implementation challenges. FPGA-based simulation engines are developed for both methods on a Xilinx Virtex-7 FPGA evaluation platform. The LB-LMC simulation engine was realized to operate in single or multiple pass execution per simulation time step and apply mixed integration methods for model computations of various electrical components and converters. This engine applies use of subsystem decomposition allowable by LB-LMC to reduce computation time costs and FPGA resources needed for larger power system models. The LIM simulation engine was implemented to operate in a two pass, leapfrog approach to compute simulation solutions every time step. A novel method to handle converter switching action in LIM was developed to enable LIM to model power switching converters. For both simulation engines, various power systems were simulated in real-time with 50ns and below time steps, from a single threephase converter to an eight converter dual-bus shipboard power system. Simulation accuracy of both methods’ FPGA implementation are compared to high precision software-based simulators. The scalability of each method in real-time was analyzed and evaluated in terms of achievable time step, determined by computational delay, and resource usage on an FPGA. Finally, the FPGA implementation of each method is compared for implementation challenges in modeling power systems with these implementations