Fully GPU‐based electromagnetic transient simulation considering large‐scale control systems for system‐level studies

Shifted-frequency based electromagnetic transient simulation for power system dynamics using two-stage singly diagonally implicit Runge–Kutta method

Parallel-in-time methods are emerging to accelerate the solution of time-consuming problems in different research fields. However, the complexity of power system component models brings challenges to realize the parallel-in-time power system electromagnetic transient (EMT) simulation, including the traveling wave transmission lines. This paper proposes a system-level parallel-in-time EMT simulation method based on traditional nodal analysis and the Parareal algorithm. A new interpretation scheme is proposed to solve the transmission line convergence problem. To integrate different kinds of traditional EMT models, a component-based EMT system solver architecture is proposed to address the increasing model complexity. An object-oriented C++ implementation is proposed to realize the parallel-in-time Parareal algorithm based on the proposed architecture. The results on the IEEE-118 test system show 2.30x speedup compared to the sequential algorithm under the same accuracy with 6 CPU threads, and a high parallel efficiency around 40%. The performance comparison of various IEEE test cases shows that the system's time-domain characteristics determine the speed-up of Parareal algorithm, and the delays in transmission lines significantly affect the performance of parallel-in-time power system EMT simulations.

This study presents a method to interface an electromagnetic transient (EMT) simulator with a dynamic phasor (DP) simulation program. The hybrid DP–EMT simulator offers flexibility in deciding the harmonic contents to be preserved in the DP domain; additionally, it offers significant reduction in computing time of large networks compared with EMT simulation. The proposed interfacing method is based on a generic transmission line, which uses specialised numerical techniques to transfer discretised waveform samples between the two simulators. These numerical methods produce the required mapping between EMT simulation samples and their counterpart DPs. This study describes the mathematical foundations of the DP–EMT interface and demonstrates them using an illustrative example. Several large networks are then studied to assess the interface's accuracy and performance in reducing computing time.

Electromagnetic transient (EMT) simulation is one of the routine tasks being executed in modern power systems. Typically, the most common EMT simulation method is the trapezoidal integration method, however, sustained numerical oscillations are possible to appear using this method, for instance, getting the voltage waveform across an inductance during a sudden current interruption. Two-stage diagonally implicit Runge-Kutta (2S-DIRK) method is a L-stable numerical method, which is able to avoid numerical oscillation. The 2S-DIRK method has been applied to solve simple RLC circuit simulations, in this paper, it is extended to system level EMT simulations considering the electromagnetic transient characteristics and nonlinearity for generators, transformers and loads. Case studies are performed to verify the effectiveness of the 2S-DIRK method in power system level EMT simulation, by comparing the accuracy and numerical stability with traditional methods.

Electromagnetic transient (EMT) simulations become more time-consuming with the increasing complexity of power systems. Thus, it is necessary to exploit the feasibility of accelerated EMT simulations taking advantage of latest heterogeneous parallel computing technologies. A fine-grained parallel EMT algorithm is proposed in this work, which makes full uses of the advanced Single-Instruction-Multi-Threads (SIMT) and shared memory parallelisms supported by Graphic Processing Units (GPUs). Moreover, time-cost models of the proposed algorithm are also presented, which is valuable for optimizing the run-time configurations and enhancing the efficiency of the parallel simulations on GPUs. Test results of large-scale power systems not only verified the accuracy and efficiency of the proposed the algorithm and but also validated the time-cost models derived.

High Voltage Direct Current (HVDC) systems are increasingly popular in electric power grids. Control system of such devices can cause instabilities when interacting with other equipment in a power system. Depending on the control system dynamics, instability and resonance can occur in a wide range of frequencies (up to few kHz). Low and high frequency resonances have been registered in several power electronics based project. In this article, several realistic test cases where oscillations can occur between HVDC links and power system components are illustrated using electromagnetic transient (EMT) simulations. In order to analysis and identify the cause of such instabilities, this paper illustrates how to conduct a frequency domain analysis by means of EMT time domain simulation. As EMT-type models provided by the manufacturers are often inaccessible (or black-boxed), the method presented considers black-boxed HVDC link model.

three kinds of digital programs used in power system simulation are summarized in this paper. The electromagnetic and electromechanical transient simulation software ADPSS is introduced. Electromagnetic transient simulation, electromechanical transient simulation and electromagnetic-electromechanical transient hybrid simulation are performed respectively on Deyang-Baoji ±500KV HVDC system. The simulation result shows that the electromagnetic-electromechanical transient hybrid simulation has the advantage of electromagnetic transient simulation and electromechanical transient simulation. It can effectively analyze the process of the local electromagnetic transient in large-scale grid. The simulation speed is fast and the result is accurate and reliable.

An automatic grid partitioning technique of large grids based on CloudPSS

Electromagnetic transient (EMT) simulation is one of the most complex power system studies that requires detailed modeling of the study system including all frequency-dependent and nonlinear effects. Large-scale EMT simulation is becoming commonplace due to the increasing growth and interconnection of power grids, and the need to study the impact of system events of the wide area network. To cope with enormous computational burden, the massively parallel architecture of the graphics processing unit (GPU) is exploited in this paper for large-scale EMT simulation. A fine-grained network decomposition, called shattering network decomposition, is proposed to divide the power system network exploiting its topological and physical characteristics into linear and nonlinear networks, which adapt to the unique features of the GPU-based massive thread computing system. Large-scale systems, up to 240 000 nodes, with typical components, including synchronous machines, transformers, transmission lines, and nonlinear elements, and multiple levels modular multilevel converter with up to 6144 submodules, are tested and compared with mainstream simulation software to verify the accuracy and demonstrate the speed-up improvement with respect to sequential computation.

Faster calculation methods are desired for electromagnetic transient (EMT) simulation programs, since circuits simulated are becoming increasingly larger and more complex. If there are time-varying and non-linear elements in the circuit to be solved, the solution process requires repeated factorization of the coefficient matrix of the circuit equations. The repeated factorization, which is often called refactorization, significantly increases the computation time. Up to now, the bottom right arranging (BRA) method has been applied for refactorization in EMT simulation programs in order to reduce the computation load. In this paper, the author proposes the use of the varying elements tracing (VET) method, which is one of the refactorization methods proposed for power flow calculations, for an offline EMT simulation program. Using some test circuits, the computation time by the BRA method and that by the VET method are examined, and the following facts are found. In most test circuits, both methods greatly improve refactorization speed, but in some test circuits, the BRA method shows degraded refactorization speed due to increased fill-ins. As for the entire EMT simulation time, the VET method improves the refactorization speed by a factor of 1.6 times at most. On the other hand, the BRA method improves the refactorization speed by a factor of 1.5 times in the best case but degrades it in some test circuits.

With an increasing number of power electronics connected renewable energy sources, electromagnetic transient (EMT) simulations become increasingly important to accurately simulate the behavior of electric power grids. EMT simulations are computationally expensive, and often specialized hardware, such as Real-Time Digital Simulators (RTDS), are used to run simulations in real-time. As individual simulators are limited in their computational capabilities, they can be interconnected in order to fully leverage their potential and enable more complex simulations. However, since these simulators are expensive and often located near power hardware for hardware-in-the-loop simulation, they are regularly distributed across multiple locations of different research institutions or organizations. The main contribution of this paper is the assessment of different approaches to connect multiple RTDS systems to form a locally distributed simulation infrastructure. For this, we combine a realistic and comprehensive model of a campus grid with a microgrid model to investigate the benefits and limitations of the different connection approaches.

A multiscale power system modeling methodology for the integrative simulation of electromagnetic and electromechanical transients is introduced, implemented and validated. It makes use of frequency-adaptive simulation of transients (FAST) in which the shift frequency appears as a new parameter in addition to the time step size. For fast electromagnetic transients, tracking of instantaneous waveforms as in the Electromagnetic Transients Program (EMTP) is performed. When slower electromechanical transients involving power oscillations prevail, the Fourier spectra of the waveforms are shifted by typically either 50 or 60 Hz to eliminate the ac carrier and enable envelope following as in phasor-based simulation. An algorithm for the automatic setting of both shift frequency and time step size is proposed. Multiscale models of transformers, synchronous machinery, and transmission lines based on FAST are developed. The line model is distinguished in that it bridges both lumped and distributed parameters to efficiently emulate scale-bridging phenomena from steady state to traveling waves. The overall methodology is thoroughly validated: first against a staged field test involving transients of line energization and recovery on the network of the Bonneville power administration (BPA); then through comparison with an EMTP-type simulator in modeling diverse transients on a four-machine two-area power system.

Electromagnetic transients (EMT) simulation is the most accurate and intensive computation for power systems. Past research has shown the potential of accelerating such simulations using graphics processing units (GPUs). In this paper, an efficient GPU-based parallel EMT simulator is designed. Thread-oriented model transformations are first proposed for the electrical and control systems. Following the transformations, the electrical system is represented by connected networks of massive primitive electrical elements, the computations of which can be constructed as massive fused multiply-add operations and solutions to a linear equation. The control systems are represented by a layered directed acyclic graph with primitive control elements that can be dealt with using single-instruction-multiple-threads groups. Finally, code automation tools are designed to form the GPU kernels. Compared with past work, the proposed model transformations improve the degree of parallelism. Most importantly, the code automation tools improve computational efficiency by substantially reducing addressing and memory access, and render the implementation of the algorithm more general and convenient. Test systems of different sizes were created by connecting multiple IEEE 33-bus distribution systems and adding distributed generators. Simulations were performed on NVIDIA’s K20 $\times$ and P100 cards. The results indicate that the proposed method significantly accelerates EMT simulations compared with a CPU-based program. Real-time performance was also achieved under certain conditions.

In this paper electromagnetic transient (EMT) simulation of large scale power systems using graphics processing unit (GPU) based computing is demonstrated. As the size of power system networks increases, the simulation time using conventional central processing units (CPUs) based simulation increases drastically. This paper proposes a hybrid CPU-GPU environment for fast large scale power systems simulation. In this scheme the GPU is mainly deployed to perform the computationally intensive part of the simulation in parallel on its built-in multiple processing cores, and the CPU is assigned for other sequential jobs like flow control of the simulation and storing output data, etc. The GPU-based approach is used to simulate a network with 900 Buses, and it is shown that the CPU-GPU based implementation is 70 times faster than the conventional CPU-based implementations.

Modern power systems are experiencing high penetration of voltage-source converter (VSC)-interfaced distributed energy resources and loads. Design, analysis, and reliable operation of such systems require extensive offline and real-time electromagnetic transient (EMT) simulations. This paper proposes discretized impedance-based modeling (DIBM) of VSCs for efficient time-domain transient analysis in state-variable (SV)-based EMT simulators. Specifically, the VSC-based systems are first represented as admittance-based models in the Laplace domain, and then they are discretized and formulated to construct a Thévenin equivalent impedance matrix and history voltages that can be interfaced seamlessly with external systems in SV-based simulators. By replacing VSC subsystems with Thévenin equivalent circuits, the proposed DIBM technique significantly reduces the number of states and eliminates the need for fictitious snubbers that may be needed in SV-based EMT simulators for compatible interfacing. The effectiveness of the proposed DIBM approach over the conventional method that uses average value models of VSCs is demonstrated on a seven-bus VSC-based system in offline (MATLAB Simscape Electrical) and real-time (OPAL-RT) EMT simulators. It is verified that the proposed method enables high accuracy and larger simulation time steps while also significantly improving the overall computational performance.