Niewiadomski, Karol
(2024)
Statistical analysis in electromagnetic compatibility.
PhD thesis, University of Nottingham.
Abstract
Power electronic (PE) converters are electronic devices designed to process electrical power, transforming the input voltage into an output voltage with desired characteristics. Their main function, which is power conversion, is enabled through the usage of switching devices, such as transistors. Due to the rapid development of the Smart Grid (SG) power devices are more often used alongside or close to communication systems, in particular serial communication devices. The operation of the power converters might induce electromagnetic interference (EMI) and bring about malfunction of these communication devices, which poses a challenge in achieving electromagnetic compatibility (EMC) of the system. The range of switching frequencies used in modern power converters corresponds to the conducted EMI range (9 kHz – 30 MHz), and the impact of conducted EMI is highly influenced by the presence of parasitic elements, values of which exhibit uncertain behaviour due to environmental conditions, geometrical layouts of the circuits, and manufacturing tolerances of the circuit elements.
This thesis presents a statistical approach to identify EMI issues that might occur due to the coexistence of power converters and serial communication systems and proposes the use of stochastic simulation, surrogate modelling, and statistical analysis to compare methodologies for EMI reduction, identify the most influential parasitic elements and enable optimization of PE circuits.
Firstly, it presents two mathematical models for analysing EMI induced by PE converters when the switching function is realized with Random Modulation (RanM) and Deterministic Modulation (DetM) techniques, along with a novel averaging scheme to facilitate their comparison, thereby overcoming the existing limitations of the models and enhancing its ability to directly calculate the probability of communication error. Contrary to ex- pectations posed by many researchers, the research found that RanM does not improve EMC compliance in the context of PE converter’s impact on serial communication devices, with RanM and DetM demonstrating similar influences on average. Additionally, the thesis determined that a Polynomial Chaos (PC)-based surrogate model outperforms Support Vector Machines (SVM) in modelling complex PE converters as judged by prediction error and stability against Monte Carlo simulations. Finally, it presents a methodology using Sobol’ indices derived from the PC model to identify and optimize the most sensitive parasitic elements in PE converters, aiming to mitigate the influence of EMI on communication error probability.
This thesis provides the EMC engineers and system integrators methods to validate their approaches for achieving EMC compliance. Accompanying the thesis is an open source toolbox ‘UQSpice’ that allows to use the methods presented in this thesis for performing PC-based simulations and sensitivity analysis using an LTSpice circuit simulation software.
Actions (Archive Staff Only)
 |
Edit View |
Tools
Tools
