Search

The existing models of the Metal-Oxide arresters (MOA) have difficulty covering the behavior prediction under surges with front time ranging from microseconds to nanoseconds. This brief presents an equation-based dynamic nonlinear model and its circuit implementation, which can describe the behavior of the MOA in a wide range of frequencies and amplitudes with the consideration of the piecewise nonlinearity and the dynamic nonlinearity characteristics. The model can be easily constructed based on the manufacturer’s data and a multiplication factor table for the dynamic characteristic. Two types of MOA used in the power distribution systems are tested to validate the proposed model. The simulated responses of MOAs are in good agreement with the measured results when the MOAs are applied to impulses with front time ranging from 20 ns to $30~\mu \text{s}$ . Both the errors of the peak voltage and the absorbed energy between simulated and measured results are less than 6%.

Thanks to their convex formulation, kernel regressions have shown an improved accuracy with respect to artificial neural network (ANN) structures in regression problems where a reduced set of training samples are available. However, despite the above interesting features, kernel regressions are inherently less flexible than ANN structures since their implementations are usually limited to scalar-output regression problems. This article presents a vector-valued (multioutput) formulation of the kernel ridge regression (KRR) aimed at bridging the gap between multioutput ANN structures and scalar kernel-based approaches. The proposed vector-valued KRR relies on a generalized definition of the reproducing kernel Hilbert space (RKHS) and on a new multioutput kernel structure. The mathematical background of the proposed vector-valued formulation is extensively discussed together with different matrix kernel functions and training schemes. Moreover, a compression strategy based on the Nystrom approximation is presented to reduce the computational complexity of the model training. The effectiveness and the performance of the proposed vector-valued KRR are discussed on an illustrative example consisting of a high-speed link and on the optimization of a Doherty amplifier.

In this article, machine learning (ML) metamodels have been developed in order to predict the per-unit-length parameters of hybrid copper–graphene on-chip interconnects based on their structural geometry and layout. ML metamodels within the context of this article include artificial neural networks, support vector machines (SVMs), and least-square SVMs. The salient feature of all these ML metamodels is that they exploit the prior knowledge of the p.u.l. parameters of the interconnects obtained from cheap empirical models to reduce the number of expensive full-wave electromagnetic (EM) simulations required to extract the training data. Thus, the proposed ML metamodels are referred to as prior knowledge-based machine learning (PKBML) metamodels. The PKBML metamodels offer the same accuracy as conventional ML metamodels trained exclusively by full-wave EM solver data, but at the expense of far smaller training time costs. In this article, detailed comparative analysis of the proposed PKBML metamodels have been performed using multiple numerical examples.

This paper presents an automatic procedure for the optimization of the hyperparameters of a delay rational model approximating the frequency-domain behavior of highspeed interconnects. The proposed model is built via a kernel-based regression, such as the Least-Square Support Vector Machine (LS-SVM), by considering an ad-hoc kernel with two hyperparameters related to the propagation delays introduced by the system. Such hyperparameters, along with the Tikhonov regularizer used by the LS-SVM regression, are carefully tuned via an automatic approach based on a k-fold cross-validation and Bayesian optimization. The feasibility of the effectiveness of the proposed modeling approach are investigated on a high-speed link.

The aim of this article is to provide an overview of polynomial chaos (PC) based methods for the statistical analysis of transmission lines. The underlying idea of PC is to represent stochastic line voltages and currents as expansions of predefined orthogonal polynomials. The determination of the expansion coefficients allows obtaining pertinent statistical information and is generally much faster than running, e.g., a Monte Carlo (MC) analysis. There exist several approaches to calculate the PC expansion coefficients. The article briefly reviews virtually all existing methods, whilst focusing on the popular and accurate stochastic Galerkin (SG) method as well as on the recent, more efficient and non-intrusive formulation of the so-called stochastic testing (ST) method. These two techniques are introduced by way of a simple illustrative example, i.e., a single-wire line running above a ground plane. Numerical comparisons in terms of accuracy and efficiency are also provided for a four-wire line with nonlinear terminations.