| dc.description.abstract | Contemporary Power systems with renewable generators, power electronics possess new challenges to the system operators. The significant increase of variable energy resources in the power grid leads to a stressed grid with much higher variabilities at the operational stage. Such variabilities together with today’s lack of accurate simulation capabilities lead to significant uncertainties in predicting the dynamics across the grid.
Hence, Fast simulation tools are required for transient analysis such as the electro-magnetic transient program (EMTP). Electromagnetic transient (EMT) simulation is the most powerful tool which could handle various detailed device models and capture high-speed dynamic behavior in the power system. However, due to the enormous complexity of the power systems, these simulators tend to be slow in simulating real-time data. In order to accelerate these computations, we can take the advantage of GPUs and FPGAs. But still, the ‘real-time’ simulation is not fast enough to simulate the true real-time of a large system. To overcome this limitation, a custom chip (ASIC) for simulating power systems is necessary.
Matrix Multiplication is one of the most fundamental basic operations in the current complex digital circuits and is vastly used in the image, signal processing applications[1]. In EMTP simulation, Matrix-Vector Multiplication is used in solving network equations and the input to these are current(I) from different buses of a large system. These vectors tend to be periodic with a fundamental frequency of 60 Hz along with other higher-order harmonics. When sampled with a relatively higher frequency rate, the consecutive samples of the inputs tend to be close and it is inefficient to compute the matrix operation again and again at every time step. Instead, In this thesis, we aim to propose an architecture based on the higher-order difference which takes variation with respect to the previous sample to compute the actual result.
Park Transformation and Inverse Park Transformation are the interfaces between non-network and network part. They convert state variables from DQ0 coordinate to three-phase coordinate and vice versa. Linear interpolation is a method of curve fitting using polynomials to construct new data points within the range of a discrete set of known data points. In this thesis, we propose a ROM architecture combined with linear interpolation technique to compute the high precision sinusoid values. | en |
