A recent trend for the development of these disciplines is the application of artificial intelligence (AI) tools, such as expert systems (ES), artificial neural networks (ANN), fuzzy logic systems (FLS), genetic algorithms (GA), and, more recently, multi-agent systems (MAS). These tools have been proven to be able to boost the performance of these systems in real-world and industrial applications thanks to features such as "learning," "self-organization," and "self-adaptation." With particular regard to ANNs for nonlinear function approximation, in power electronics and electrical drives applications, they are used for control and identification, such as the multilayer perceptron (MLP) or the radial basis function (RBF). Another kind of neuron that has also been applied recently is linear neurons (ADALINE), whose simplicity has given surprisingly good results. On the other hand, the detailed unified mathematical treatment of space-vectors has made it possible to embed the theory of linear neural networks, resulting in improvements, both theoretical and experimental, of classical approaches in electrical drives and power electronics. This standpoint is the goal of the book: to present in a systematic way the classical theory based on space-vectors in identification, control of electrical drives and of power converters, and the improvements that can be attained when using linear neural networks. With this outlook, this book is divided into four parts: o Part I deals specifically with voltage source inverters (VSI) and their control. o Part II deals with AC electrical drive control, with particular attention to induction and permanent magnet synchronous motor drives. o Part III deals with theoretical aspects of linear neural networks. o Part IV deals with specific applications of linear neural networks to electrical drives and power quality. Chapter 1 presents the theory of space-vectors and instantaneous power. This chapter is fundamental for understanding the rest of the book. Chapter 2 describes the open-loop and closed-loop control of voltage source inverters. With regard to open-loop techniques it also explains the different kinds of pulsewidth modulation (PWM) strategies, and with regard to closed-loop techniques it analyzes both current and power control of VSIs. Voltage-oriented control (VOC) and direct power control (DPC) are also presented. Chapter 3 explains the fundamentals of power quality; parallel active filters (PAFs) and series active filters (SAFs), with reference to their operating principle and control strategies, are investigated. Passive and hybrid filter configurations are also analyzed. Chapter 4 deals with induction machine (IM) static and dynamic space-vector models. The dynamic model of the IM, including saturation effects, is shown. Finally, the spacevector dynamic model of the IM, including rotor and stator slotting effects, is described. Chapter 5 describes, first, scalar control strategies of IM drives with impressed voltages and currents. It then derives field-oriented control (FOC) strategies, with reference to rotor, stator, and magnetizing flux linkage orientations. Related flux models are also presented. Finally, direct torque control (DTC) strategies are presented, particularly the classic switching table (ST) DTC, the space-vector modulation (SVM) DTC, and the electromagnetically compatible (DTC). The so-called direct self-control (DSC) is also described. Chapter 6 covers sensorless control of IM drives, with particular reference to both model-based and anisotropy-based techniques. With regard to model-based techniques, the following estimators/observers are described: open-loop speed estimators, model reference adaptive systems (MRAS), full-order Luenberger adaptive observer (FOLO), full-order sliding-mode observer, reduced-order adaptive observer (ROO), and, finally, the extended Kalman filter. With reference to anisotropy-based techniques, the following methodologies have been described: revolving Carrier techniques, pulsating carrier techniques, and high-frequency excitation techniques. Chapter 7 derives the permanent magnet synchronous motor space-vector model. Field-oriented control (FOC) with both impressed voltage and currents is described. Various control strategies are presented to maximize the electromagnetic torque production or the drive efficiency. The DTC of the PMSM is also presented. Finally, both anisotropy- and model-based sensorless techniques are explained. Chapter 8 deals with the theory of linear neural networks, particularly the neural EXIN family. Starting from the adaptive linear neuron (ADALINE) structure, it presents more recent and performing linear neural networks: the TLS EXIN neuron, the Ge-TLS EXIN neuron, the MCA EXIN neuron, and, finally, the MCA EXIN + neuron. Chapter 9 covers, first, the sensitivity analysis of the classic flux models of IM drives versus parameter variations. It then presents some on-line parameter estimation techniques of IMs by the least-squares (LS) technique, including both unconstrained and constrained estimations. Finally, it shows the neural self-commissioning of IM drives. Chapter 10 deals with the application of neural adaptive filtering to distributed generation (DG) and active power filter (APF) systems. The ADALINE design criteria for the fundamental frequency extraction and the harmonic load current compensation are presented. The stability issues of the entire system are included. Experimental verification of the neural approach is presented in comparison with classic approaches. Chapter 11 presents some applications of LS-based techniques to speed estimation of IMs. In particular, the following neural-based observers are presented and discussed: the MCA EXIN + MRAS observer, the TLS EXIN full-order Luenberger adaptive observer, and, finally, the reduced order adaptive observer. The general approach of this book is presenting initially the theoretical background of each subject immediately followed by a set of simulations of experimental results supporting the analytical part. It is the authors' opinion that the presence of many results should help the reader in understanding better the treated theoretical aspects.
Power Converters and AC Electrical Drives with Linear Neural Networks
M Pucci;G Vitale
2012
Abstract
A recent trend for the development of these disciplines is the application of artificial intelligence (AI) tools, such as expert systems (ES), artificial neural networks (ANN), fuzzy logic systems (FLS), genetic algorithms (GA), and, more recently, multi-agent systems (MAS). These tools have been proven to be able to boost the performance of these systems in real-world and industrial applications thanks to features such as "learning," "self-organization," and "self-adaptation." With particular regard to ANNs for nonlinear function approximation, in power electronics and electrical drives applications, they are used for control and identification, such as the multilayer perceptron (MLP) or the radial basis function (RBF). Another kind of neuron that has also been applied recently is linear neurons (ADALINE), whose simplicity has given surprisingly good results. On the other hand, the detailed unified mathematical treatment of space-vectors has made it possible to embed the theory of linear neural networks, resulting in improvements, both theoretical and experimental, of classical approaches in electrical drives and power electronics. This standpoint is the goal of the book: to present in a systematic way the classical theory based on space-vectors in identification, control of electrical drives and of power converters, and the improvements that can be attained when using linear neural networks. With this outlook, this book is divided into four parts: o Part I deals specifically with voltage source inverters (VSI) and their control. o Part II deals with AC electrical drive control, with particular attention to induction and permanent magnet synchronous motor drives. o Part III deals with theoretical aspects of linear neural networks. o Part IV deals with specific applications of linear neural networks to electrical drives and power quality. Chapter 1 presents the theory of space-vectors and instantaneous power. This chapter is fundamental for understanding the rest of the book. Chapter 2 describes the open-loop and closed-loop control of voltage source inverters. With regard to open-loop techniques it also explains the different kinds of pulsewidth modulation (PWM) strategies, and with regard to closed-loop techniques it analyzes both current and power control of VSIs. Voltage-oriented control (VOC) and direct power control (DPC) are also presented. Chapter 3 explains the fundamentals of power quality; parallel active filters (PAFs) and series active filters (SAFs), with reference to their operating principle and control strategies, are investigated. Passive and hybrid filter configurations are also analyzed. Chapter 4 deals with induction machine (IM) static and dynamic space-vector models. The dynamic model of the IM, including saturation effects, is shown. Finally, the spacevector dynamic model of the IM, including rotor and stator slotting effects, is described. Chapter 5 describes, first, scalar control strategies of IM drives with impressed voltages and currents. It then derives field-oriented control (FOC) strategies, with reference to rotor, stator, and magnetizing flux linkage orientations. Related flux models are also presented. Finally, direct torque control (DTC) strategies are presented, particularly the classic switching table (ST) DTC, the space-vector modulation (SVM) DTC, and the electromagnetically compatible (DTC). The so-called direct self-control (DSC) is also described. Chapter 6 covers sensorless control of IM drives, with particular reference to both model-based and anisotropy-based techniques. With regard to model-based techniques, the following estimators/observers are described: open-loop speed estimators, model reference adaptive systems (MRAS), full-order Luenberger adaptive observer (FOLO), full-order sliding-mode observer, reduced-order adaptive observer (ROO), and, finally, the extended Kalman filter. With reference to anisotropy-based techniques, the following methodologies have been described: revolving Carrier techniques, pulsating carrier techniques, and high-frequency excitation techniques. Chapter 7 derives the permanent magnet synchronous motor space-vector model. Field-oriented control (FOC) with both impressed voltage and currents is described. Various control strategies are presented to maximize the electromagnetic torque production or the drive efficiency. The DTC of the PMSM is also presented. Finally, both anisotropy- and model-based sensorless techniques are explained. Chapter 8 deals with the theory of linear neural networks, particularly the neural EXIN family. Starting from the adaptive linear neuron (ADALINE) structure, it presents more recent and performing linear neural networks: the TLS EXIN neuron, the Ge-TLS EXIN neuron, the MCA EXIN neuron, and, finally, the MCA EXIN + neuron. Chapter 9 covers, first, the sensitivity analysis of the classic flux models of IM drives versus parameter variations. It then presents some on-line parameter estimation techniques of IMs by the least-squares (LS) technique, including both unconstrained and constrained estimations. Finally, it shows the neural self-commissioning of IM drives. Chapter 10 deals with the application of neural adaptive filtering to distributed generation (DG) and active power filter (APF) systems. The ADALINE design criteria for the fundamental frequency extraction and the harmonic load current compensation are presented. The stability issues of the entire system are included. Experimental verification of the neural approach is presented in comparison with classic approaches. Chapter 11 presents some applications of LS-based techniques to speed estimation of IMs. In particular, the following neural-based observers are presented and discussed: the MCA EXIN + MRAS observer, the TLS EXIN full-order Luenberger adaptive observer, and, finally, the reduced order adaptive observer. The general approach of this book is presenting initially the theoretical background of each subject immediately followed by a set of simulations of experimental results supporting the analytical part. It is the authors' opinion that the presence of many results should help the reader in understanding better the treated theoretical aspects.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.