DSpace Collection:
http://hdl.handle.net/123456789/115
2019-12-13T04:39:00ZPOWER QUALITY IMPROVEMENT USING MULTILEVEL INVERTER BASED HYBRID POWER FILTER
http://hdl.handle.net/123456789/14317
Title: POWER QUALITY IMPROVEMENT USING MULTILEVEL INVERTER BASED HYBRID POWER FILTER
Authors: Mahajan, Vasundhara
Abstract: Modern civilization has become reliant on the unremitting supply of clean electrical power. But the power supplied by the grid and transmission system may not be always clean and continuous. Therefore, it necessitates the elimination or minimization of disturbances causing impurity. The first step of providing the power quality solution is to understand the types of power quality problems in the received supply and the nature of the loads to be coupled. Recently, the sophisticated and advanced technology is being implemented more due to its immense benefits in terms of lifestyle, business, infrastructure, and health. Accepting these aids, however, intensifies reliance on electrical power and often that power has to be completely free from interruption and disruption of the system to function efficiently.
The ac distribution systems have experienced high harmonic pollution due to the wide use of power electronic loads that are nonlinear. These nonlinear loads generate harmonics which degrade the distribution systems and may affect the communication and control systems. The nonlinear load causes distortion and nonlinearity in current and/or voltage waveform for system. This deviation of current from the ideal sinusoidal waveform is described as harmonic and interharmonic current. Many of these new devices are more sensitive to the voltage quality than conventional linear loads. Nonlinear equipment such as induction machines, transformers, electric arc furnaces, welding machines, fluorescent lamps (with magnetic ballasts), ac/dc drives, and battery chargers are also accountable for the generation of harmonics in electric power systems. Besides, there are different types of power disturbances ranging from few micro seconds to few seconds duration problems which lead to waveform distortion. The harmonic distortion of voltage and current waveform are controlled by the harmonic regulations and guidelines such as IEEE-519-1992 and IEC 61000. These guidelines promote best practice in the design of both power systems and nonlinear equipment.
A harmonic survey is being conducted at different commercial and high power industrial loads to know the existing level of harmonics in the system and a brief report is presented. The harmonic distortion instigates various undesirable effects in power system especially reduces system efficiency, causes low power factor, deteriorates the performance of induction motors, increase losses in transformer and line, forces resonance phenomenon, neutral burning, mal-operation of relays, blowing of fuses, and interference with nearby communication networks.
Harmonic filters, in general, are designed to reduce the effects of harmonic penetration in power systems and they should be installed when it has been determined that the recommended harmonic content has been exceeded. There are two approaches to reduce the effect of the harmonic distortion, namely active filtering approach and passive filtering approach. The passive filter method is conventional, which have the drawbacks of large size, resonance, and fixed compensation. Moreover, to filter out manifold harmonics,
ii
multiple filters are required in parallel combination. The performance of passive filter is not much satisfactory due to the variable operating conditions. Active Power Filters (APFs) are feasible alternatives to passive filters and emerged as more efficient solution to the harmonics. The contemporary active filters have better performance, fast response, more adaptability, and flexibility as compared to passive elements. In the active filtering approach, the harmonic currents/voltages produced by the nonlinear loads are extracted, and their opposites are generated and injected into the power line using a power converter. Several active filtering approaches based on different circuit topologies and control theories have been suggested by researchers. The active filter mainly consists of a single high rating PWM power converter which takes care of all the harmonic components in the distorted signal.
The main advantage of active filters over passive ones is their fine response to changing loads and harmonic variations. In addition, a single active filter can compensate more than one harmonic, and improve or mitigate other power quality problems such as flicker. For applications where the system configuration and/or the harmonic spectra of nonlinear loads change, active elements may be used instead of the passive components to provide dynamic compensation. An APF is implemented when the order numbers of harmonic currents are varying. This may be due to the nature of nonlinear loads injecting time-dependent harmonic spectra (e.g. variable speed drives) or may be caused by a change in the system configuration.
The structure of an active filter may be that of series or parallel architectures. The proper structure for implementation depends on the types of harmonic sources in the power system and the effects that different filter solutions would cause to the overall system performance. The series topology is mainly applied for voltage harmonic elimination, whereas for current harmonic minimization the shunt topology is used. The shunt topology is more popular as compared to others due to its performance and ease of implementation for current harmonic elimination. In this thesis the current harmonics are required to be minimized, therefore, the shunt topology is identified for implementation and modifications.
APFs rely on active power conditioning to compensate undesirable harmonic currents replacing a portion of the distorted current wave stemming from the nonlinear load. This is achieved by producing harmonic components of equal amplitude but opposite phase angles, which cancel the injected harmonic components of the nonlinear loads. APFs are expensive compared with their passive counterparts and are not feasible for small facilities. The main drawback of active filters is that their rating is sometimes very close to the load (up to 80% in some typical applications), and thus it becomes a costly option for power quality improvement. Moreover, a single APF might not provide a complete solution in many practical applications due to the presence of both voltage and current quality problems. The combination of high power and high switching frequency results in excessive amounts of power losses. Furthermore, the reliability of the existing active filters is a major concern, as
iii
the failure of converter results in no compensation at all. In such cases, a more complicated filter design consisting of two or three passive and/or active filters (called a Hybrid Power Filter (HPF)) is recommended.
APF or HPF topology selection depends upon several factors such as harmonic distortion, kVA rating and cost of passive filter components, displacement factor, and losses in APF including switching ripple filters, ability to provide harmonic isolation between supply and load, and control complexity.
In addition to the advancement in power electronics and technology, the power and voltage requirement of the industrial applications also reached a higher level and better power quality in terms of minimum harmonic distortion with unity power factor. The conventional two-level inverters are found to be incapable of satisfying this demand; therefore multilevel inverters are being used. The Multilevel Inverter (MLI) allows the expansion of system rating with increased efficiency and better performance. The MLI can generate approximately sinusoidal voltages at low switching frequency with reduced switching stress and negligible electromagnetic interference and common-mode voltage. The rating of MLI is increased by adding more voltage levels without increasing individual device rating with reduced harmonics at the output voltage. Mostly, the active harmonic filter uses a standard two-level Voltage Source Inverter (VSI) for low power, low voltage applications. However, for medium/high voltage applications, MLI VSIs provide more advantages. The two-level inverter has many disadvantages such as it requires coupling transformer, large smoothing reactors, snubber components, complicated control and higher cost when applied at a medium or high voltage system. As compared to the two-level inverter, MLI provides many advantages such as no coupling transformer, small smoothing reactor, less switching stress, modular structure for medium or high voltage application. Furthermore, MLI has lesser harmonic distortion for more number of voltage levels to the reasonable increase in cost.
The MLIs are mainly classified in three topologies: a) Diode Clamped Multilevel Inverter (DCMLI) b) Flying Capacitor Multilevel Inverter (FCMLI) c) Cascade H-bridge Multilevel Inverter (CHBMLI). The selection of MLI topology and number of levels depends upon the system parameters, such as voltage/power rating, and cost. Among the available MLI topologies, the CHBMLI topology has comparatively simple control and modular layout for medium/high power applications with the least number of components for a given level.
The researchers have invariably utilized, modified, tested, and implemented the various MLI configurations for a large number of applications for medium/high power and medium/high voltage systems for power quality enhancement. The harmonic filtering and minimization, power factor improvement, reactive power compensation, static var compensation, and drives are among the main applications. The MLI mainly as the harmonic filter is focused and analysed for the thesis.
iv
The literature study indicates that the CHBMLI is the candidate topology for harmonic filtering and reactive power compensation at high-power, medium-voltage systems. As, the CHBMLI has modular units for each level, that allows the flexibility and simple extension of levels at the output. The CHBMLI requires the minimum number of components by eliminating the clamping components such as diodes and capacitors, which are necessary for DCMLI and FCMLI configurations respectively. These advantages have motivated the author of the thesis to exploit the CHBMLI scheme. However, the requirement of separate dc source for each unit of each phase is the main drawback of CHBMLI, as it complicates the voltage balancing with the increase in output levels.
Recently, Artificial Intelligence (AI) based controllers are used in harmonic filtering using MLI with higher efficiency and more dynamics; the researchers have reported various methods in the literature for various applications. The single-phase as well as three-phase MLIs have been implemented using a Fuzzy Logic Controller (FLC), Neural Network (NN), and/or Genetic Algorithm (GA) based controllers. These AI based controllers are employed to control the voltage, frequency, and/or current for power quality improvement by reducing harmonic distortions.
In most of the reported methods, the AI technique is used for dc voltage control for three or five-level MLI and uses traditional compensating techniques such as Instantaneous Power Theory (IPT) or Synchronous Reference Frame (SRF) and PI controllers with its inherent limitation of fine tuning. These drawbacks are subdued by using the AI for the controllers of the multilevel harmonic filter as described in this work. In which an AI controller based three-phase, five-level, multilevel Shunt Active Power Filter (SAPF) is designed developed to improve percentage Total Harmonic Distortions in the source current (%THDi
The described harmonic filter has an eleven-level, CHB as a filter and AI based control scheme. As compared with the two-level filter, the method achieves better compensation performance by using smaller filter inductors, with reduced switching stress at higher switching frequency and higher voltage. However, filters based on MLIs are generally more expensive and more complicated to control as it has more number of inverter switching states. Moreover, the application of these filters in medium and high voltage with high power level is justified in terms of cost and performance. The instantaneous power theory is modified and the reference/compensating component is extracted by using NN. Further, the scheme uses two separate FLCs, one is for regulating dc voltage (V) in high voltage system. The designed controller can be extended to any level of CHBMLI configuration and hence more versatile. The CHB topology using Level Shifted Pulse Width Modulation (LSPWM) is preferred over the other MLI topologies because of its comparatively simple control and modular layout. This scheme is advantageous as it uses AI, which is self-adaptive and self-adjusting and thereby less susceptible to change in system parameters.
dc) and the other is for generating gate pulses for the IGBTs in the inverter circuit. In CHBMLI, the clamping diodes
v
are not required, which reduces the control complexity, therefore has a simple modular structure for expansion. The proposed active harmonic filter is a shunt/parallel connected, current harmonic compensating device, which injects the harmonic cancellation component in the power circuit at the Point of Common Coupling (PCC) and thereby reduces the harmonic distortion in current and also improves the power factor. The scheme has AI based controllers that are robust, do not necessitate a mathematical model, and accommodates unpredictability.
In current controller scheme the reference signal is obtained by using the NN based Instantaneous Power Theory (IPT), which uses the basic IPT for its execution. The IPT uses Clarke/Inverse Clarke transformation and is based on time-domain. The IPT is more efficient and feasible as compared to other methods and can be applied to balanced/unbalanced systems under both steady state and/or transient condition. In the projected scheme, the NN is applied for harmonic component extraction. This controller requires only average/constant component and all others are undesirable and are to be eliminated. Traditionally a low pass filter is used for separating harmonic component, which requires tuning for changes in the system. The use of NN eliminates the repeated tuning and is self-adjusting according to the system parameter variation. The harmonic component with NN from active part is calculated. NN based controllers have self-adapting and high rated calculation characteristics that allow them to handle high nonlinearities and uncertainties to which systems are generally susceptible. A feed forward neural network is designed with three layers, the input layer, the hidden layer and the output layer respectively. The network is trained with large data of source current, reference dc voltage, power loss component and reference compensation current from a conventional PI method using tan sigmoidal and pure linear activation functions in the hidden and output layers respectively. The network is trained with Levenberg-Marquardt back-propagation (LMBP) algorithm. The NN based APF performance is evaluated through simulation results and it is found that the performance of APF improves with the use of NN as compared to that with conventional PI based controller. The fuzzy logic controlled dc voltage is calculated and active power loss is determined for reference estimation in coherence with harmonic components to be eliminated. This compensating component is then fed to the control circuit for harmonic filtering.
The IPT gives satisfactory results, when supply voltages are sinusoidal, but in case of distorted mains voltage, prevalent in the most industrial power system, this generates errors in reference currents and limits the compensation of harmonics. In such cases the average power method is applied, which uses supply voltages through Phase Locked Loop (PLL) and load currents to calculate the instantaneous power and average power over one sixth of a cycle. This in turn is used to obtain the peak current component of fundamental load current. The loss component of APF is obtained by comparing the actual capacitor voltage and reference capacitor voltage. The peak component of reference source current is obtained by
vi
adding the peak value of the load current component and the peak value of current providing the APF losses which is obtained from PI controller.
The MLIs are modulated by using multicarrier schemes, which are classified as Phase Shifted Pulse Width Modulation (PSPWM) multicarrier modulation or Level Shifted Pulse Width Modulation (LSPWM) multicarrier modulation. The PSPWM can be used for FCMLI and CHB, whereas LSPWM is topology independent. The technological growth in MLI configurations and increase in number of levels has forced the modification and expansion in modulation methods derived from conventional. These developments are aimed to utilize the additional switching state to compete with the added complexity.
There has been a lot of progress in MLI modulation schemes for diverse applications that have correspondingly agitated the growth in modulation methods to achieve the specific objectives. The modifications are intended to generate the best stepped output voltage at all modulation indices with adaptable parameters such as magnitude, frequency, and phase. Further, the switching frequency has to be minimized for maximum efficiency and minimum switching stresses. Although, the trade-off between the appropriate modulation method with ease of control and less complexity is tricky, the researchers have tried to modify the conventional schemes for their best.
The LSPWM is found to be a suitable modulation scheme for the presented work as it has the lowest harmonic distortion and lesser device switching frequency. To utilize these features it is required to overcome the basic drawback of uneven switching and unequal distribution of switching losses. The switching patterns are balanced by rotating the carriers. The basic LSWPM is described and modifications are proposed. There are several extended multicarrier LSPWM methods which have been discussed in literature by various researchers. In most of the methods as discussed in the literature for multicarrier modulation of MLI the switching sequence and repetition timings are variable and methods are applied and topology specific, which are difficult to extend to other types of MLI configuration.
This thesis demonstrates the modified modulation scheme for a multilevel inverter (MLI) to equalize the switching pattern at all (high and low) modulation indices. The LSPWM performance is better than PSPWM. But LSPWM has an unequal switching frequency and unequal conduction period as its inherent disadvantages, which are undesirable for high power and/or high frequency applications. The demonstrated scheme is a modified LSPWM scheme for m-level MLI with (m-1) multicarrier. In this method the carriers are rotated/swapped cyclically for equalizing switching patterns and thereby balancing the conduction period for all the devices of the MLI at all modulation indices in (m-1)/2 cycle of modulating signal. The scheme has the advantage of both PSWPM and LSPWM as it has better harmonic performance and identical switching pattern for each device at high as well as low modulation indices. The scheme can be straightforwardly extended to any level of the inverter. The simulation study is done for three-phase, five-level and eleven-level CHBMLI in
vii
MATLAB/SIMULINK using simpower system tool box. The results are compared with traditional multicarrier PSPWM and LSPWM and with proposed rotated LSPWM for full, 1/2, 1/3, and 1/4 cycle rotation. The four types of rotation schemes are detailed and compared to show the viability and its effect on CHBMLI performance. The simulation results demonstrate the matching of switching pattern among the devices of each phase as compared to conventional LSPWM.
Further, the CHBMLI is used as an active harmonic filter and requires dc voltage regulation. The five-level MLI has two series connected H-bridges per phase to be regulated. The batch control method is used for error adjustment, which reduces the complexity with less number of controllers. In batch control positive, negative, and zero errors are estimated for obtaining active power loss component corresponding to harmonics. The PI and fuzzy logic based controllers are used for comparison and simulation. Similarly, the eleven-level CHBMLI has five series connected H-bridges per phase. These voltages are controlled in a similar manner as that of five-level using batch control method.
The designed controller adjusts the voltage magnitude and current error according to the existing magnitude of the corresponding harmonic component in current/voltage. This results in optimum dc side voltage and minimal converter losses. The information available on the magnitude of each harmonic component allows us to select the active filter parameter accordingly. This results in higher efficiency and superior performance.
The performance of the method is investigated through simulation in MATLAB/Simulink, which is further verified by experimentation. The efficacy is demonstrated via exhaustive simulation and experimental results for different nonlinear loads and change in load. The simulation is carried out at a high voltage rating of the system; however the experimentation is carried out at low voltage due to practical limitations. The number of levels is also reduced to five for prototype preparation.
The AI controller based three-phase five-level CHBMLI based harmonic filter is simulated in MATLAB/Simulink for performance description and implementation in the Institute laboratory. The simulation results show that the designed scheme is better in performance, economical as well as modular. The laboratory prototype at reduced rating of 100 V, 2 kVA is developed for the corroboration of simulation performance of the AI controller based rotated multicarrier LSPWM for CHBMLI as harmonic filter.
The prototype arrangement consists of the following
a) Three-phase, five-level CHBMLI with conventional and rotated multicarrier LSPWM.
b) Three-phase, five-level CHBMLI with PI controller and conventional and rotated multicarrier LSPWM as shunt connected hybrid harmonic filter.
c) AI controller based three-phase, five-level CHBMLI conventional and rotated multicarrier LSPWM as shunt connected hybrid harmonic filter.
viii
The prototype module has 24 IGBTs as switching devices that are used for 6 H-bridges of three-phase, five-level CHBMLI. Each switch is galvanically isolated, floating dc capacitor. The IGBT (IRG4PH40KD) is used as switching device for designing this prototype. Each phase has two series connected H-bridges, each H-bridge has two legs and each leg has two series connected IGBTs. The complete experimental scheme has dSPACE DS1103 interface, Host PC, current sensor, voltage sensor, isolation circuit, dead-band circuit, and driver circuit, which are designed and developed in the laboratory. The conventional three-phase nonlinear load and three-phase ac input are used for experimentation. Although the simulation is carried out for eleven-level CHBMLI, the laboratory prototype is developed for the five-level because of practical limitation of fabrication and design. The robust control algorithm is designed and developed through exhaustive simulation, which is independent of the number of levels of CHBMLI. The developed prototype performance is validated through comprehensive experimentation in the laboratory.
The digital signal processor (DSP) DS1103 of dSPACE is used for the real-time simulation and implementation of control algorithm on hardware circuit. The control algorithm is designed in the MATLAB/Simulink. The Real-Time Workshop (RTW) of MATLAB is used to generate the optimized C-code for real-time implementation. The Real-time Interface (RTI) of dSPACE allows the interaction between virtual and real world for execution of the complete setup. The Multi-channel Analog to Digital Converter (ADC) is used to sense ac line currents, ac supply voltages, and dc capacitor voltages. An opto-isolated interface board is also used to isolate the entire DSP.2013-12-01T00:00:00ZMODEL REFERENCE ADAPTIVE CONTROL USING NEURAL NETWORKS
http://hdl.handle.net/123456789/14198
Title: MODEL REFERENCE ADAPTIVE CONTROL USING NEURAL NETWORKS
Authors: Shukla, Rohit
Abstract: The thesis presents a quick overview of Model Reference Adaptive Control
(MRAC). The thesis presents model reference based neural network structure that
can be used for adaptive control of linear and nonlinear processes. The proposed
MRAC neural network scheme is simulated to control the movement of a simple,
single-link robot arm. The simulation results show that neural network based
MRAC can give satisfactory performance requirements2016-07-01T00:00:00ZCONTROL OF SHIP'S ROLL BY ACTIVE FIN STABILIZERS
http://hdl.handle.net/123456789/14196
Title: CONTROL OF SHIP'S ROLL BY ACTIVE FIN STABILIZERS
Authors: Sunil, Thvvrn
Abstract: Stabilization of ship roll motion induced by wave disturbances is one of the important controls required in the modern day marine industry. Excessive roll motion makes the ship's crew uncomfortable and also causes damage to the cargoes and equipment on board. Active-fin stabilisers are the most widely used equipments out of the many other available stabilizers like anti-roll tanks, gyroscope and bilge keel in today's marine world to reduce the roll motion. It rotates about its stock to give a desired hydrodynamic list to stabilize the ship's roll.
Various control strategies have been used for the roll motion control of ships. However, most of the research includes stabilization of a ship with fixed mathematical model. The major changes in the ship's parameters like weight and meta-centric height due to practical reasons are generally not considered for controller design. This dissertation exploits this fact to study, analyze and design a controller for the ship with varied parameters including weight of the ship.
In this dissertation, three different conditions of a ship model are selected and PID controllers were separately designed for all the 3 conditions. These PID controllers are amalgamated to design a neural network controller. Thereafter, neural network controller is improved upon in various steps to arrive at an optimal controller to control the ship's roll in all the three possible conditions of the ship. The neural network controller could successfully control even the sinking ship.
There is a saturation limit for the stabilizer fin angle which is generally not considered by many researchers in their work or simulations. In this dissertation even the non-linearity in the form of fin angle's saturation is considered during the design of PID/NN/FL controllers. Performance of a controller may decrease by considering fin's angle saturation. However this is what is practical and designing controllers without considering saturation limit is meaningless and is of no practical use.
After the design of the NN controller, the creation of a fuzzy logic controller was envisaged. In the pursuit to make the best possible FLC even with non-availability of the system information for the rule creation, data from the previously designed 3 PID controller was used to frame the rules. Various FLCs with varied shapes, sizes and calculated parameters of the membership functions of the fuzzy sets were created, tested and analyzed. Some of the FLCs were even designed with 125 rules along with a separate weightage for each rule. Particle swamp optimization was later used to fine tune the gains added to the Fuzzy logic controller for better control of ship's roll. The finalised FLC also could save the sinking ship and give good performance. FLC also could reduce roll frequency to a great extent which is very important for some ships with specific roles.2016-05-01T00:00:00ZNOVEL MODEL ORDER REDUCTION AND CONTROLLER DESIGN TECHNIQUE USING BIG BANG BIG CRUNCH OPTIMIZATION ALGORITHM
http://hdl.handle.net/123456789/14195
Title: NOVEL MODEL ORDER REDUCTION AND CONTROLLER DESIGN TECHNIQUE USING BIG BANG BIG CRUNCH OPTIMIZATION ALGORITHM
Authors: Biradar, Shivanagouda
Abstract: This dissertation focuses on a meta-heuristic optimization algorithm i.e. big bang
big crunch optimization (BBBC) algorithm and major setbacks in BBBC algorithm
with respect to its conceptual and working structure. A modi ed BBBC optimization
algorithm is proposed, which works better than original BBBC. But, it is observed
that BBBC and modi ed BBBC like many other meta-heuristic optimization algorithm
su ers from the problem of getting trapped in local minima. Therefore, modi ed BBBC
is combined with chaos which e ectively enhances the searching e ciency and greatly
improves the searching quality. These algorithms validity is quanti ed using various
benchmark function.
Further, this thesis, contributes various results, techniques and focuses on application
of BBBC in areas of System and Control. Starting with Model Order Reduction
(MOR), which is an integral part of System Engineering. MOR techniques have proved
to be an important technique for accelerating time-domain simulation in a variety of
CAD tools for highly complex system and controller design. There are various reduction
techniques available in literature and most of them are either complex i.e. they are too
di cult to understand while other techniques work for particular class of problems. In
this report, a novel MOR technique has been proposed using BBBC and time moment
matching method, which works for many class of problems. Now, moving onto eld of
Control Engineering, utility of this algorithm for controller design has been elaborated.
In this method, a multi-objective function has been formulated and BBBC is used as
an optimization tool for ne tuning the PID controller. Above work i.e. MOR and
controller design have been validated on automatic voltage regulator system.
Another contribution of this thesis is study of utility of statistical methods in area
of Control System and Optimization. As it is known that BBBC is a relatively new
optimization technique, before which, many famous techniques like PSO and GA are
widely used in all eld of engineering and talked about in optimization society. But
question always raises which one of these algorithms are better in respect to solution
nding capability (e ectiveness) and computational e ciency. In this report we have
used inferential statistics as a tool to analyze this problem and bring out a concrete
conclusion in this regard.
At last we have used Taguchi method, a statistical technique, combined with BBBC to
ne tune the controller parameters.2016-05-01T00:00:00Z