1 CHAPTER 1 INTRODUCTION 1.1 PREAMBLE Magnetic characteristics during saturation of synchronous machine are the main consideration for improving the design and modeling of the machines. In stability analysis of synchronous machines, magnetic saturation characteristics are taken into account, as they determine the initial conditions of the generator such as rotor angle, field current, rotor circuit flux linkages, etc. Two magnetic characteristics are present in a synchronous machine, one in direct axis (d-axis) and another in quadrature axis (q-axis). There are either no direct measurement methods available to simultaneously estimate both the d-axis and q- axis magnetic characteristics of a synchronous machine by simple experiments or to model the q-axis magnetic characteristics by uncomplicated methods. This thesis focuses on the measurement of d- and qaxis flux and modeling of q-axis flux of projected pole synchronous machines. 1.2 THE GENESIS OF THE THESIS The d-axis magnetic characteristics of synchronous machine can be determined easily from a conventional open-circuit test with machines excited from their field winding. On the other hand, the q-axis magnetic characteristics cannot be determined by any conventional method. Some experimental methods, such as maximum lagging current method, steady-state 2 on-load measurements method and back-to-back method for measuring the qaxis saturation characteristics may give the result and however the degree of complexity of test procedures, need of special, complex controllable excitation system and loading limitation of the machine are serious issues. Consequently, extensive research work is being carried out in order to measure and model the d-q axes flux characteristics of synchronous machines by simple, reliable and accurate heterogeneous techniques. 1.3 LITERATURE SURVEY A survey of existing literature reveals that various methodologies have been employed to determine and model the d-q axes magnetic characteristics of synchronous machines. The different methods of determination being dynamic and complex in nature, the literature survey of this problem are classified into the following sections: determination of q-axis saturation characteristics by analytical and experimental methods determination of intermediate axis saturation characteristics development of saturation models analysis of magnetic coupling between d-q axes estimation of synchronous machine parameters Neuro-Fuzzy techniques for control of machines and analysis of various flux measurement techniques. 3 1.3.1 Determination of q-axis saturation characteristics by analytical and experimental methods Kar et al (2000) discussed a new method to calculate q-axis saturation characteristics of cylindrical-rotor synchronous generator from the existing d-axis saturation characteristics. The saturated values of the d- and qaxis self magnetizing reactances and mutual cross-magnetizing reactances between the two orthogonal axes under the steady-state operation have been calculated by them. The effect of q-axis saturation, on the power system transient stability has also been presented. Four different analytical methods for calculating the q-axis saturation characteristics of the salient pole synchronous machines from the measured d-axis saturation characteristics were explored by Ahmed El-Serafi and Kar (2003). The q-axis saturation characteristics have been calculated from the measured d-axis saturation characteristics and the d- and q-axis unsaturated magnetizing reactances. A comparison made between the above four methods reveal that two methods out of the four produce accurate results. The saturation characteristics in both the direct and quadrature axis are needed for the accurate prediction of the performance of saturated synchronous machines. Three possible experimental methods for determining the q-axis saturation characteristics of both cylindrical-rotor and salient pole synchronous machines were presented by Kar and Ahmed El-Serafi (2004 and 2006). The measured q-axis saturation characteristics obtained by these experimental methods were compared and the various sources of errors which affect the accuracy of characteristics determination investigated. The maximum lagging current proposed is limited by human error. This method requires a complex positive and negative excitation system. The on-load measurement method can be conducted only upto certain values of load angles because of the stability limits. The back-to-back test requires a special 4 experimental set-up. In all the three methods, the stator resistance and the variation in the armature leakage reactance due to saturation were neglected which introduces some error. 1.3.2 Determination of intermediate axis saturation characteristics An innovated model for the saturated synchronous machines using the intermediate axis saturation characteristics has been proposed Narayan Kar and Ahmed El-Serafi (2002). Two sets of parameters represent the saturation characteristics in the various intermediate axes. The first set gave the information about the saturation levels in the intermediate axes while the second set, the corresponding phase angles between the resultant magnetic flux and the resultant ampere-turns. This approach has been applied to three laboratory synchronous machines of various designs. The accuracy and the universality of the new model have been demonstrated. Ahmed El-Serafi and Kar (2005) presented four analytical methods for determining the intermediate axis saturation characteristics of salient pole synchronous machines from the measured d-axis saturation characteristics. The accuracies of these four methods have been verified by comparing the measured field currents and load angles of two laboratory salient pole synchronous machines of different designs with those calculated values. In this context, the field currents and load angles have been calculated using an innovated approach which uses the intermediate axis saturation characteristics directly in the modeling of the saturated synchronous machines. A synchronous machine model was proposed which applies the intermediateaxis saturation characteristics directly in machine modeling. 5 1.3.3 Development of saturation models Shackshaft and Henser (1979) have used parameters obtained from both no-load and on-load measurements to obtain a saturation model, of adequate simplicity for use in multi machine studies. Both no-load and onload methods have been used to obtain the saturation characteristic of the quadrature axis. The methods have produced good results but on-load method has been recommended for its advantages over no-load method. An efficient numerical algorithm for including the effects of magnetic saturation in a model of a synchronous generator has been proposed by Brandwajn (1980). It has been reported that one axis saturation model is not totally adequate in the simulation of magnetic saturation effects. Harley et al (1980) presented the results of an investigation into the effect of different saturation models on the transient performance of a synchronous machine. Various methods of calculating and using saturation factors in the equations of synchronous machines in order to represent saturation in both the direct and quadrature axes during three-phase fault studies have been proposed. The capability and suitability of a couple of simple saturation model structures to reproduce steady state operating characteristics of synchronous machines were explored by De Mello and Hannett (1986). The results of sensitivity to variations in parameters and model structures have been presented. Joseph Ojo and Thomas Lipo (1989) have presented an improved model for the transient analysis of saturated salient pole synchronous motors. The modeling approach proposed, combines the advantages of the finite element method for parameter determination and the simplicity of the 6 equivalent circuit approach. The importance of this work is the derivation of the d-q equations of the synchronous machine in the rotor reference frame accounting for the instantaneous spatial saturation effects. Chuanli Shi and Ming Li (1990) have developed a general model for the analysis of salient pole machine as well as non-salient pole machine under steady- state conditions. This method provides refinements over the classical model on the basis of „two-reaction theory‟. It has been concluded that the resolution of armature mmf is not necessary for the analysis of synchronous machines under steady-state conditions and the real saturation condition of machines can be taken into account in measuring magnetization curves in d- and q- axis. A new method to express the main flux saturation in synchronous machines by using auxiliary currents and unsaturated magnetizing inductances instead of the saturated inductances was presented by Tamura and Takeda (1994). The new model has only constant coefficients defined in terms of unsaturated values of the magnetizing inductances and other constant parameters. A consistent method for including magnetic saturation in the general two-axis model of the salient pole synchronous machine was explored by Tahan and Kamwa (1995). The general model covers both the main path magnetic saturation and frequency effects in the dynamic equations. Implementation of the model requires the easily available no-load saturation curve and the two unsaturated reactances together with the saturated saliency ratio. Tsai et al (1995) presented a comprehensive procedure to develop an artificial neural network (ANN) based saturation model for synchronous generator stability assessment. A feed-forward network with multi 7 dimensional mapping of the synchronous machine saturation and an error back-propagation learning scheme was used. This developed scheme was implemented in the generator dynamic transient stability study requiring only small computational alteration in saturation model representation. Levi (1997) has presented a general procedure for main flux saturation modeling in smooth air-gap synchronous machines. The method relies on the concept of „generalized flux‟ and „generalized inductance‟. A number of models that result from the application of the method for different selections of state-space variables were presented. The accuracy of saturation representation in all the novel models was the same and these models find application in synchronous machine simulation and control. The concept has been extended to salient pole synchronous machines by Levi (1999), where the saturation accounted for by means of single saturation factor approach. An improved technique for incorporating saturation into the q-d axis model of a synchronous machine was proposed by Corzine et al (1998). The model has been verified for steady-state and transient conditions using a laboratory synchronous machine-rectifier system similar to those commonly used for Naval and aerospace power generation. The saturation function was applicable for an infinite range of flux. The model has been shown to be adequate in predicting the performance of a synchronous machine / rectifier system. A general, six winding salient pole saturated machine was discussed by Levi (1998). It was shown that numerous alternative models, with different sets of state-space variables may be derived. The major advantage of numerous novel models was considered to be the simpler structure of the system matrix, together with relatively small number of saturation dependent coefficients, while stator d-q axis current components 8 were retained as state-space variables. These models found application in synchronous machine simulation. Nhut-Quang Dinh and Jos Arrillaga (2001) described a salient-pole generator harmonic model in sequence components that takes into account the effects of rotor angle, saliency and saturation. Generator harmonic impedances at the characteristic harmonics were obtained from the measured generator terminal harmonic voltages and currents. The values were compared with those calculated using the derived formulation. This analytical model was found to be a powerful tool for a fast and accurate prediction of the harmonic levels on the generator terminals in the presence of saliency and distortion. An enhanced saturation model for wound-rotor salient pole synchronous machines based on experimental evaluation and analysis has been proposed by Wang (2003). This new model used three nonlinear lumped-parameter impedances to fully characterize the magnetic saturation phenomenon in wide frequency and load range. They characterize respectively, the d-axis saturation, the q-axis saturation and the nonlinear coupling between the field and the stator windings due to the saturation. This model considered to be convenient for implementation and tuning in drive controllers based on the field-oriented algorithm. Improved drive control accuracy and dynamic performance have been shown by simulation and experimental results. Adequacy of the modeling approach was confirmed by Finite element analysis. A new flux-correction method to represent magnetic saturation in classical q-d machine model was presented by Liwei Wang et al (2007). This proposed method is simple and does not require calculation of the derivatives of the saturated mutual inductances. It seemed to be very fast and accurate as compared to other more complex models. 9 A linearized model of synchronous reluctance generator has been developed using a new saturation model to perform the steady-state stability analysis by Souvik Guha and Kar (2008). The effect of the saturation both in the direct and quadrature axes on the determination of the steady-state and transient stability performances of synchronous reluctance machine has also been investigated using the proposed linearized machine model. The magnetic saturation phenomena were characterized by static saturation inductances, based on which the saturated salient pole synchronous motor model was established by Wei Jing and Guojun Tan (2010). The changing discipline of the inductances in d-q reference frame align with current was analyzed in detail under the magnetic saturation condition. Simulation results indicate that main physical variables act in accordance with the operating principles in nonlinear magnetic saturation situation and the validity of the model was verified. A new approach for identification and modeling of the saturated synchronous reactances of synchronous machines was proposed by Martin Jadric et al (2010). The saturated synchronous reactances were estimated and modeled using fitting surfaces from steady-state operating data at a wide range of operating conditions. Simulation studies have been conducted to verify the developed steady-state saturation model. Stiebler and Campeanu (2010) proposed a model which makes use of an approximation with a single non-linear characteristic. The main field saturation curve was approximated by a trigonometric function which reflects the machine‟s no-load saturation curve. The result supported the common practice to use the conventional Park‟s equations for modeling transient behavior. 10 1.3.4 Analysis of magnetic coupling between d-q axes Cross-magnetizing effect (magnetic coupling between d- and q- axis) in the d- and q- axis has been described by simple equations which show that the change in flux in any axis of the synchronous machine due to the cross-magnetizing effect is directly proportional to the ampere-turns in this axis as well as the saturation degree in the other axis. A modified power/load angle relationship including the effect of the cross-magnetizing has been derived by El-Serafi et al (1988). El-Serafi and Abdallah (1991) presented a model of a synchronous machine which includes the effect of both the machine saturated reactances as well as the cross-magnetizing phenomenon. This model was used to study the steady-state stability of a synchronous machine connected to an infinite bus system. These studies show that the effect of saturation has been noticeably large and depends mainly upon cross-magnetizing effect. A model for the d- and q- axis synchronous reactances was developed by El-Serafi and Abdallah (1992) in which the effects of both the d- and q- axis saturation factors and the cross-magnetizing phenomenon have been included. The effect of saturation was found to reduce the values of both d- and q- axis mutual reactances. The magnitude of this reduction was found to be appreciable. Moreover, the reduction in the q-axis mutual reactance was found to be larger than that in the case of the d-axis mutual reactance. El-Serafi and Wu (1993) introduced the concept of the crossmagnetizing phenomenon in saturated synchronous machines. Techniques for determining the parameters, which represent the cross-magnetizing effect, were proposed. The inclusion of the cross- magnetizing effect in the two-axis frame model of a saturated synchronous machine depends on the methods of defining the saturated d- and q- axis fluxes in the model. When the two 11 saturation factors concept was applied to define these saturated fluxes, the cross-magnetizing effect has been shown to exist in both the non-salient and salient pole machines and demagnetizing in both the d- and q- axis. Emil Levi and Viktor Levi (2000) discussed the impact of dynamic cross-saturation on simulation accuracy of various saturated synchronous machine models. Dynamic cross-saturation appear in all the saturated machine models, which was the existence of non-zero elements in the system matrix that describe cross-coupling between d- and q- axes. It has been found that for one particular class of models, omission of dynamic cross-saturation has practically no impact on accuracy. An analytical method for determining the parameters that represent the effect of the magnetic coupling between the d-axis and q-axis of saturated cylindrical rotor synchronous machines has been proposed by El-Serafi and Demeter (2000). Experimental investigations were carried out to verify the accuracy of the proposed method. 1.3.5 Estimation of synchronous machine parameters A test procedure for derivation of synchronous machine d-axis and q-axis parameters has been described by De-Mello and Rebeiro (1977). This involves simple procedures which can be implemented in the course of normal operation of the machine as it is synchronized or taken off line. De Mello and Hannett (1981) presented the results of transient response tests on synchronous generators along with derivations of generator model parameters from data. The adequacy of d- and q- axis generator model structures was investigated through comparisons of model simulations against tests. Transient response tests include load rejections from zero power factor, under excited and over excited conditions for identification of d-axis 12 characteristics. Rejection test from load conditions was used for identifying qaxis characteristics. A simple modification of model to account for low frequency shielding effect, without changing the order of the model has been proposed. Junya Matsuki et al (1993) have presented a new approach to determine the saturated values of synchronous reactances, based on permeance distributions in the air-gap inside a synchronous generator. Effect of saturation was reflected in the air-gap flux distributions, hence the permeance distributions in the air-gap, directly. In order to evaluate the saturation level in the air-gap, the permeance distributions have been used in this method. When the on-load permeance distributions are determined, saturation effect can be estimated by them and easily related to the saturated values of synchronous reactances. This method was found to be useful to the analysis of saturation effect on machine constants, using a single search coil placed on the stator. An improved method for calculating the steady- and transient-state leakage flux distributions in salient pole synchronous machines through the leakage inductances has been reported by Shima et al (2003). The method provides information on leakage flux distributions that cannot be obtained by terminal quantities. The leakage inductances of a synchronous motor were calculated under various loads and saturation of the inductances analyzed. 1.3.6 Neuro-Fuzzy techniques for control of machines Mohanasundaram et al (2009) presented the design and simulation of a fuzzy logic controller for speed regulation of voltage controlled induction motor drive. The fuzzy algorithms in the proposed controller have been systematically obtained from the intuition and experience about the motor drive dynamics. 13 Abul R. Hasan et al (1994) presented the design and application of fuzzy control to the problem of Automatic Voltage Regulation of a synchronous generator. The method explored deals with the use of binary input-output fuzzy associative memories for control. Radwan and Gouda (2005) have proposed speed control system for permanent magnet synchronous motor (PMSM) drive, which newly incorporates adaptive neuro-fuzzy with two input variables and one control output variable. This control methodology solves the problem of non-linearity and parameter variations of PMSM drive. These references are related to Neuro-Fuzzy approach for speed control and automatic voltage regulation in various electrical machines. 1.3.7 Analysis of flux measurement techniques The V-I characteristics of a large variety of commercial diodes of different types of germanium and silicon have been studied both at room temperature and at liquid air temperature to investigate the effect of magnetic field on the diode current by Parshad and Mehta (1967). Germanium diodes show a large decrease in the forward current and a small decrease in the reverse current on the application of the magnetic field, while the effect of magnetic field on silicon diodes was found to be negligible. The effect of a magnetic field in decreasing the forward current of a germanium diode has been observed by Bansal and Parshad (1969). A bridge set up using two diodes, to one of which the magnetic field is applied, and the other used for temperature compensation, gave the unbalance voltage as a function of magnetic field. To operate a bridge, it was first balanced by adjusting the resistance arms without the application of the magnetic field. As the magnetic field applied, the unbalanced voltage was measured. The relation 14 observed between the magnetic field and the resulting unbalance voltage could yield a practical method of measuring the magnetic field. The observations show that by proper change of bridge input voltage, the overall bridge sensitivity could be kept constant for all working temperatures. In practice, for any set of particular diodes used in the bridge, the proper working voltages for the bridge should be experimentally calibrated as a function of working temperatures. Measurement of magnetic field strengths has been one of the most important measurement techniques in Electrical engineering. Such measurements always involve inherent inaccuracies to a considerable magnitude. Difficulties arise from the fact that the magnetic materials were not homogenous and also suffer from the lack of uniformity in magnetic properties between different batches of material. Hence a review of magnetic flux measurement techniques has been discussed along with their applications to various engineering fields by Karambelkar and Banerji (1973). It is essential to understand the effect of magnetic field on circuit parameters of semiconductor devices so as to apply them to various control systems and equipment. Karambelkar (1979) has studied the effect of magnetic field on bipolar and unipolar transistors. New applications for the measurement of AC and DC voltage, current and power were also proposed. Traditionally the magnetic field distribution throughout the electrical machine core was obtained by positioning miniature enameled copper wire search coils on test laminations stacked in an experimental core. However their physical presence, when inserted in core, changes the flux and field distribution, creates localized stress and an air gap between two adjacent laminations. Basak et al (1995) proposed a method to overcome this drawback by depositing sensors comprising a thin film structure, on the surface of several laminations by vacuum evaporation of metals. 15 Reininger and Hanisch (1997) have described possible tools for the successful and competitive realization of a proper working magnetic sensor for a certain field of application. The boundary conditions and interfaces to be matched during the development of a magnetic switch towards a final product for the industrial automation industry have been sketched. The difficulties which go beyond the study of parameters influencing the working of a physical sensor were mentioned. A direction sensitive micro-magnetic field sensor was developed using the giant magneto impedance effect by Steindl et al (2000). A quickresponse micro-sized magnetic sensor was investigated for high reliability measurements of the stray fields of magnetic bar codes. Chiriac and Chiriac (2003) have presented a new sensor for magnetic field detection and mechanical displacement. A magnetic amorphous wire was used as the active element. The magnetic field and the displacement range can be modified by using different wire lengths. A new method was proposed by George Loisos and Anthony Moses (2003) for magnetic flux measurement in localized regions inside a magnetic material using a novel surface probe sensor. It found to be more accurate than the conventional probe sensor. It was proved to have an induced voltage proportional to the flux density rate of change in a triangular crosssectional area instead of traditional orthogonal cross-sectional areas. Precise magnetic sensors were used by Michal Vopalensky et al (2003) for geophysical and space research magnetometers, for position trackers and bio-magnetic measurements. Anisotropic magneto resistors and Giant magneto resistors were the competitors for small size fluxgate sensors. Their stability could be improved by ac techniques but the electronics becomes complex and consumes more power. These offered higher sensitivity 16 but unfortunately they could not be used for precise applications due to their high non-linearity and hysteresis. Nikitin et al (2003) have deposited thin film multi-layer structures and optimized for various low noise magnetic field sensors. The developed double-layer magneto resistive sensors revealed much less pronounced hysteresis, lower noise when compared to single-layer sensors and have high frequency band. Marius Volmer and Jenica Neamtu (2007) have presented a method to improve the quality of the Planar Hall effect used for micro compass application. A linear isolator, which was able to measure the electrical currents relatively with a high sensitivity, has been built. The behavior of the magnetic sensors was simulated using a micro magnetic simulator. Wheatstone bridge configuration was used to minimize the thermal drift and to increase the magnetic sensitivity. Yabukami et al (2009) have developed a thin film magnetic sensor for magneto cardiogram measurement using the phase detection technique. The technique was also applied for measurement of the low frequency field, which resulted in achievement of a high magnetic resolution. The literature survey shows that the measurement methods presented by many researchers for the estimation of d-q axes flux are highly complicated. Sensors for measuring flux in both the axes of synchronous machines which give best results are required. No magnetic sensor is used by the researchers for flux measurement in d- and q-axes of synchronous machines. The literature survey on modeling of flux reveals that the estimation of flux in synchronous machines using Adaptive Network Fuzzy 17 Inference System (ANFIS) based soft computing technique and mathematical regression techniques are also not yet attempted by any researcher. 1.4 AIM AND CONTENT OF THE THESIS From the review of literature, it has been inferred that various researchers have proposed the estimation of d- axis, q- axis and intermediate axis characteristics and made their analysis on the different aspects of degree of complexity and accuracy. Further, these analyses were carried out using few experimental techniques and certain analytical methods which use machine models, derived in terms of machine parameters and have not considered the measurement of flux using flux sensors and modeling by soft computing methods. Hence, specific approach is undertaken in this research work, to propose new methodologies to alleviate estimation of d-q axes flux in projected pole synchronous machines and to enhance the accuracy. A methodology has been developed and presented for measurement of d- axis and q- axis flux in synchronous machines using germanium diode flux sensors. The laboratory machine that has been taken up for experimental investigation is the projected pole (stationary field type) synchronous machine of 415 V, 3000 VA. Flux in d- and q- axes in the machine were obtained by measurements, which uses Germanium diode flux sensors. The change in the property of Germanium diodes due to the effect of application of magnetic field is used for the magnetic flux measurement. If a forward biased diode is placed in the magnetic field, its forward current decreases. The decrease in the forward current is directly proportional to the strength of the magnetic field. The measurement of variation in this forward current, leads to the measurement of magnetic flux. 18 Mathematical models have also been developed employing Multivariate linear regression and Multivariate polynomial regression for the estimation of q- axis flux. Soft computing models have been developed which uses ANFIS for the estimation of q- axis flux. Various ANFIS models use different membership functions for the two inputs viz., field excitation and d- axis flux. Along with the experimental test results measured using Germanium diode flux sensors, available q-axis magnetic flux data calculated from the measured d-axis flux by Ahmed El-Serafi and Kar (2003), for various Salient pole synchronous machines namely, a) Microalternator, b) Machine No. 1 and c) Machine No. 2 form the basis of all the modeling schemes reported in this thesis. The proposed measurement method which uses germanium diode flux sensor has to be a simplest instrumentation kit for the measurement of d-q axis flux in projected pole synchronous machines. The models developed have to be provided with simple known inputs to output an accurate value of q- axis flux. The graphical representations of measured and modeled values of flux have to be provided and comparisons of these are to be made available with the reference values of flux to further enhance the analysis. The objective of the thesis is structured as a block diagram in Figure 1.1. 19 Figure 1.1 Aim and Content of the Thesis 1.5 ORGANIZATION OF THE THESIS The research work carried out are presented in six chapters and structured in a top down approach as shown in Figure 1.2. The first chapter briefly reports the review of literature, which necessitates the scope of the present work. 20 Figure 1.2 Organization of the Thesis 21 The second chapter introduces d-q axes flux measurement in projected pole synchronous machines. Regression models, which estimate the q- axis flux from the d- axis flux and field excitation, aim to make a comparison between the two regression models are presented in the third chapter. Chapter 4 reports the q- axis flux modeling using adaptive network fuzzy inference system. The design, development and comparison of q- axis flux obtained by three ANFIS models are discussed in this chapter. In chapter 5 all the proposed models are compared with the actual reference values. It also presents the estimation of residual for the models. A comprehensive summary of conclusions obtained from this research work and scope for future work are presented in the sixth chapter.