Machine Learning and Uncertainty Quantification for Surrogate Models of Integrated Devices With a Large Number of Parameters

Trinchero, Riccardo; Larbi, Mourad; Torun, Hakki Mert; Canavero, Flavio; Swaminathan, Madhavan

doi:10.1109/access.2018.2888903

Cited by 81 publications

(49 citation statements)

References 45 publications

(45 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For given frequency response in frequency domain, the residues and poles can be calculated using Vector Fitting method [30][31][32]. In this approach, each element of the entire scattering matrix S is approximated as (15),…”

Section: Proposed Residue-pole Methodsmentioning

confidence: 99%

“…However, the technique needs a large number of training samples for the PC black-box model to be trained. In [15] Least Square Support Vector Machine method (LSSVM) for modeling variability analysis of a system with a large number of uncertain parameters has been adopted, while the complexity of the black-box model is generally lower than PC based methods, it still needs a large set of training samples. Usually for creating a surrogate model of microwave circuits the values of the device's scattering parameters at selected frequency points are considered [16], [17].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Residue-Pole Methods for Variability Analysis of S-parameters of Microwave Devices with 3D FEM and Mesh Deformation

Rufuie¹,

Lamęcki²,

Sypek³

et al. 2020

RADIOENGINEERING

View full text Add to dashboard Cite

This paper presents a new approach for variability analysis of microwave devices with a high dimension of uncertain parameters. The proposed technique is based on modeling an approximation of system by its poles and residues using several modeling methods, including ordinary kriging, Adaptive Polynomial Chaos (APCE), and Support Vector Machine Regression (SVM). The computational cost is compared with the traditional Monte-Carlo method. To improve the efficiency, mesh deformation is applied within 3D FEM framework.

show abstract

Section: Proposed Residue-pole Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Residue-Pole Methods for Variability Analysis of S-parameters of Microwave Devices with 3D FEM and Mesh Deformation

Rufuie¹,

Lamęcki²,

Sypek³

et al. 2020

RADIOENGINEERING

View full text Add to dashboard Cite

show abstract

“…This can help utility and stakeholders to understand the effects of renewables' variability and unhealthiness of the network. It is to state that Monte Carlo simulation is used to observe the stochastic behavior of the network because we keep only two or three parameters of DGs as varying so simple Monte Carlo method holds, however, if the number of parameters is large, other methods are preferable for analyzing the uncertain behavior [36][37][38][39].…”

Section: Application: Dgs and Stochastic Analysismentioning

confidence: 99%

An Iterative Scheme for the Power-Flow Analysis of Distribution Networks based on Decoupled Circuit Equivalents in the Phasor Domain

et al. 2020

Self Cite

View full text Add to dashboard Cite

This paper presents an alternative solution for the power-flow analysis of power systems with distributed generation provided by heterogeneous sources. The proposed simulation approach relies on a suitable interpretation of the power network in terms of a nonlinear circuit in the phasor domain. The above circuit interpretation can be solved directly in the frequency-domain via the combination of a standard tool for circuit analysis with an iterative numerical scheme, providing directly the steady-state solution of the power-flow of a generic distribution network. At each iteration, the resulting circuit turns out to be composed by two decoupled subnetworks, a large linear part and a set of smaller nonlinear pieces accounting for the load characteristics, with evident benefits in terms of the computational time. The feasibility and strength of the proposed simulation scheme have been verified on a large benchmark consisting of the IEEE 8500-node test feeder. Then it is applied to the statistical simulation of a power network accounting for the variability effects of renewable generators. According to the results, the proposed tool provides an effective alternative to the state-of-the-art approaches for power-flow analysis further highlighting the benefits of the application of well-established tools for circuit analysis to power-flow problems.

show abstract

“…It is important to remark that in this nonparametric dual space formulation, and thanks to the so-called "kernel trick", the number of unknowns α i in (9) is completely independent of the number of parameters and basis functions considered in the LS-SVM regression model, and it always equals the number of training samples L that are used for the regression. For this reason, the LS-SVM regression in the dual space is an attractive candidate for the UQ of high-dimensional problems [37].…”

Section: And •mentioning

confidence: 99%

“…Recently, advanced machine learning (ML) methods [26,27] have been also employed for the UQ of several realistic problems in electrical engineering [28,29]. Specifically, flexible and powerful ML regressions, such as support vector machine (SVM) [30,31], least-square support vector machine (LS-SVM) [32], and Gaussian processes [33], were effectively applied to build accurate surrogate models starting from a limited set of training samples [34][35][36][37]. The resulting surrogate model is able to predict both the deterministic and the stochastic behavior of the system output for any configuration of the uncertain input parameters.…”

Section: Introductionmentioning

confidence: 99%

Compressed Machine Learning Models for the Uncertainty Quantification of Power Distribution Networks

et al. 2020

Self Cite

View full text Add to dashboard Cite

Today’s spread of power distribution networks, with the installation of a significant number of renewable generators that depend on environmental conditions and on users’ consumption profiles, requires sophisticated models for monitoring the power flow, regulating the electricity market, and assessing the reliability of power grids. Such models cannot avoid taking into account the variability that is inherent to the electrical system and users’ behavior. In this paper, we present a solution for the generation of a compressed surrogate model of the electrical state of a realistic power network that is subject to a large number (on the order of a few hundreds) of uncertain parameters representing the power injected by distributed renewable sources or absorbed by users with different consumption profiles. Specifically, principal component analysis is combined with two state-of-the-art surrogate modeling strategies for uncertainty quantification, namely, the least-squares support vector machine, which is a nonparametric regression belonging to the class of machine learning methods, and the widely adopted polynomial chaos expansion. Such methods allow providing compact and efficient surrogate models capable of predicting the statistical behavior of all nodal voltages within the network as functions of its stochastic parameters. The IEEE 8500-node test feeder benchmark with 450 and 900 uncertain parameters is considered as a validation example in this study. The feasibility and strength of the proposed method are verified through a systematic assessment of its performance in terms of accuracy, efficiency, and convergence, based on reference simulations obtained via classical Monte Carlo analysis.

show abstract

Machine Learning and Uncertainty Quantification for Surrogate Models of Integrated Devices With a Large Number of Parameters

Cited by 81 publications

References 45 publications

Residue-Pole Methods for Variability Analysis of S-parameters of Microwave Devices with 3D FEM and Mesh Deformation

Residue-Pole Methods for Variability Analysis of S-parameters of Microwave Devices with 3D FEM and Mesh Deformation

An Iterative Scheme for the Power-Flow Analysis of Distribution Networks based on Decoupled Circuit Equivalents in the Phasor Domain

Compressed Machine Learning Models for the Uncertainty Quantification of Power Distribution Networks

Contact Info

Product

Resources

About