Proceedings ECT 2012

8/10/2019 Proceedings ECT 2012

1/277


2/277

ISSN 1822-5934

ELECTRICAL AND CONTROL

TECHNOLOGIES

PROCEEDINGS

OF THE 7THINTERNATIONAL CONFERENCE ON

ELECTRICAL AND CONTROL TECHNOLOGIES

Sponsored by:

Research Council of Lithuania

ABB UAB

KAUNAS, TECHNOLOGIJA, 2012


3/277

UDK 621.3(474.5)(06); 681.5(474.5)(06)

The 7th

International Conference on Electrical and Control Technologies,

ECT2012Selected papers of the International Conference, May 3-4, 2012, Kaunas, Lithuania

(Kaunas University of Technology), edited: A. Navickas (Editor-in-Chief), A. Sauhats,

A. Virbalis, M. Aubalis, V. Galvanauskas, K. Brazauskas, A. Jonaitis

The 7th

International Conference on Electrical and Control Technologies, ECT2012 is

the continuation of series of annual conferences on electrical, power engineering,electrical drives and control technologies held earlier in Kaunas Polytechnic Institute

and from 1990 in Kaunas University of Technology.

The aim of the Conference is to create the forum for presentations and discussions toconsider the newest research and application results of investigations on electrical and

power engineering, electric drives, power converters, automation and control

technologies.

All papers were reviewed by Conference Editorial Committee, Conference Program

Advisory Committee, independent paper reviewers.

For information write to:

Kaunas University of Technology

Department of Electric Power SystemsStudent Str. 48, LT-51367 Kaunas

Lithuania

Kaunas University of Technology, 2012


4/277


5/277

CONTENTS

Multimodal Automation Controller for Smart Homes .......................................................... ........................................ 9Rytis Maskeliunas*, Vidas Raudonis*, Grigor Stambolov**, Rimvydas Simutis** Kaunas University of Technology, Lithuania; ** Technical University-Sofia, Bulgaria

Identification of Interconnected Systems by Instrumental Variables Method .......................................................... 13Grzegorz MzykWroclaw University of Technology, Institute of Computer Engineering, Control and Robotics, Wroclaw, Poland

Investigation of Automatic Control System in Wood Drying Facilities ............................................................... ..... 17

Bronius Karalinas*, Dalia Lukoien*, Liudmila Andriuien***Vilnius Gediminas Technical University, Lithuania

**Vilnius College of Technologies and Design, Lithuania

Artificial Intelligence Approach to SoC Estimation for Smart BMS ......................................................................... 21K.L. Man

1,2,3, C. Chen

5, T.O. Ting

1, T. Krilaviius3,4, J. Chang6, S.H. Poon71Xi'an Jiaotong-Liverpool University, China; 2Myongji University, South Korea; 3Baltic Institute of Advanced Technology,Lithuania; 4Vytautas Magnus University, Lithuania; 5Global Institute of Software Technology, China;6Texas Instruments, Inc, USA; 7National Tsing Hua University, Taiwan

Closed-Loop Optimization Algorithms in Digital Self-Tuning PID Control of Pressure Process ........................... 25

Gediminas Liauius, Vytautas KaminskasDepartment of Systems Analysis, Vytautas Magnus University, Lithuania

Toolbox for Automatic Control of Fermentation Processes for the Production of Recombinant Therapeutic

Protein ............................................................................................................................................................................. 30

Artur Kuprijanov*, Sebastian Schaepe**, Mathias Aehle**, Andreas Lbbert**, Rimvydas Simutis* * Institute of Automation and Control Systems, Kaunas University of Technology, Lithuania** Institute of Biochemistry and Biotechnology, Martin-Luther-University Halle-Wittenberg, Germany

An Automatic System of Lithuanian Speech Formant Synthesizer Parameter Estimation ................................... 36

Graina Py*, Virginija imonyt**, Vytautas Slivinskas***Vilnius University Institute of Mathematics and Informatics, Lithuania

**Lithuanian University of Educational Sciences, Lithuania

Wireless Sensor Networks Application at Water Distribution Networks in Latvia ................................................. 40Anatolijs Zabasta*, Viesturs Selmanovs-Pless**, Nadezda Kunicina**Riga Technical University, Institute of Industrial Electronics and Electrical Engineering, Latvia**Micro Dators ltd, Latvia

IEEE802.11n Standards Capability to Support Wireless Device in Vehicular Environment ........................... ..... 44Janis Jansons, Ernests Petersons, Nikolajs BogdanovsDepartment of Transport Electronics and Telematics, Riga Technical University, Latvia

Predicting Trends of Financial Attributes by an Adaptive Committee of Models .................................................... 48

Zivile Kalsyte*, Antanas Verikas*,**, Asta Vasiliauskaite***

*Department of Electrical & Control Equipment, Kaunas University of Technology, Lithuania

**Intelligent Systems Laboratory, Halmstad University, Sweden***Faculty of Economics and Management, Department of Finance, Kaunas University of Technology, Lithuania

Forecast of Healthcare Industry Distress Ratio According to the Companies Sectors............................................. 54Audrius Kabasinskas*, Zivile Kalsyte**, Asta Vasiliauskaite****Deptartment of Mathematical Research in Systems. Kaunas University of Technology. Lithuania** Department of Electrical & Control Equipment.Kaunas University of Technology, Lithuania

***Faculty of Economics and Management, Department of Finance, Kaunas University of Technology, Lithuania

Modeling Results of Vehicular Network for File Transfer Protocol ......................................................... ................ 60

Nikolajs Bogdanovs, Ernests Petersons, Janis JansonsDepartment of Transport Electronics and Telematics, Riga Technical University, Latvia

Temperature Control Prototype for Education and Training Purposes ................................................................... 64Gediminas Valiulis, Edvardas BielskisElectrical Engineering Department, iauliai University, Lithuania

5


6/277

Face Geometric Normalization Based on Eye Position ........................................................ ...................................... 70

Audrius Bukis, Paulius Lengvenis, Rimvydas SimutisDepartment of Process Control, Kaunas University of Technology, Lithuania

Orientation Invariant Objects Recognition in Infrared and Visible Lights on Hierarchical Temporal Memory . 74

Saulius Sinkeviius*, Vytenis Sinkeviius***Department of Information Systems and Communications, Panevezys College, Lithuania

**Department of Electrical Engineering, Kaunas University of Technology, Lithuania

Microcontroller Based Model of Fitzhugh-Nagumo System ............................................................ ........................... 78

Andrius Petrovas*, Saulius Lisauskas*, Alvydas lepikas* *Vilnius Gediminas Technical University, Lithuania

Investigation of the Multidimensional Automatic Control System, Having Structure of the Chain, Applying

Lambert W Function ...................................................................................................................................................... 82

Irma Ivanoviene, Jonas RimasDepartment of Applied Mathematics, Kaunas University of Technology, Lithuania

Research of Temperature Regulation at Variable Air Volume ....................................................... ........................... 87Ansis Klga*, Romualds Beinskis*, Andris Krmi** *Department of Transport Electronics and Telematics, Riga Technical University, Latvia**Institute of Heat, Gas and water technology, Riga Technical University, Latvia

Building Classification Models Using Adaptive Basis Function Construction .......................................................... 91Gints JekabsonsInstitute of Applied Computer Systems, Riga Technical University, Latvia

Technical Characteristics of TI DM3730 Embedded Platform When Processing Video Information ................... 95

Paulius Lengvenis, Rimvydas SimutisProcess control department, Kaunas University of Technology, Lithuania

Investigation of Avatars Face Features Influence on Users Emotional State Using Neuro-Feedback ................. 99

Egidijus Vakeviius, Edvinas Bazeviius, Aura Vidugirien, Vytautas KaminskasDepartment of System Analysis, Vytautas Magnus University, Lithuania

Business Process Dimension in Design of Control Systems .................. .............................................................. ...... 105Edvinas Pakalnickas, Dalius MisinasLietuvos energija AB, Lithuania

Exploration of Color Histograms Applicability for Amber Surface Analysis ......................................................... 109

Saulius Sinkeviius, Arnas LipnickasDepartment of Control Technology, Kaunas University of Technology, Lithuania

Hexapod Kinematic Model Optimization for Control System.................................................................................. 113Tomas Luneckas, Dainius UdrisVilnius Gediminas Technical University, Lithuania

Investigation into Frequency Controlled Electric Drive ........................................................................................ ... 117

Jonas Kriauinas, Roma Rinkeviien

Vilnius Gediminas Technical University, Lithuania

Optimization of Spare Parts Warehouse Costs ........................................................... ............................................... 121Rytis Briedis, Vygandas Vaitkus, Saulius iukasKaunas University of Technology, Lithuania

Mobile Robot Prediction of Potential Conflicts ......................................................................................................... 127

Stanislovas Bartkeviius, Kastytis arkauskas, ydrnas JakasDepartment of Electrical Engineering, Kaunas University of Technology, Lithuania

Mobile Robot Coordinates Detection Problems Using Centre of Gravity of Visible Environment ....................... 131

Stanislovas Bartkeviius, Alma Dervinien, Olga Fiodorova, Kastytis arkauskasKaunas University of Technology, Lithuania

Application of Smart Grid Technologies in Emergency Automation ............................................. ......................... 135Vladimir Chuvychin, Roman PetrichenkoRiga Technical University, Latvia

6


7/277

Impact on Reliability of Electrical Power Supply by Feeding of Offshore Wind Parks......................................... 139

Marco Fleckenstein, Gerd BalzerTU Darmstadt, Germany

The Means of Magnetic and Electric Fields Indication ......................................................... .................................... 143

Gabija Nedzinskait*, Juozapas Arvydas Virbalis*, Algis Vegys***Kaunas University of Technology, Department of Electrical Engineering, Lithuania

**Kaunas University of Technology, Department of Ergonomics, Lithuania

Stability Analysis of Electrical Power System Trough Characteristic Polynomial Roots Locality ....................... 147Georgi Georgiev*, Inga Zicmane**, Sergey Kovalenko**

,***

* Energy Institute JSC Sofia, Bulgaria; ** Riga Technical University, Institute of Power Engineering, Latvia;*** JSC Siltumelektroprojekts, Latvia

Mathematical Model for 110/10 kV Transformer Substations Optimum Power Choice...................................... 151

Svetlana Guseva, Olegs BorscevskisRiga Technical University, Latvia

Lightning Performance on Transmission Line Towers ......................................................... .................................... 157Olegs Sliskis, Ilja Dvornikovs, Karlis KetnersRiga Technical University, Latvia

Dynamic Mathematical Model and Single-Phase Short Circuit Analysis of Variable Speed PumpStorage Unit ................................................................ ............................................................... .................................... 161

Rimantas Pranas Deksnys, Darius AliauskasKaunas University of Technology, Lithuania

Operating Condition Processes Research of InverterThree Phase Dual System Converter .............................. 166

Povilas Balinas, Povilas NorkeviiusKaunas University of Technology, Lithuania

Assessment of the Reactive Power Flows Impact on the Nodal Prices of Active Power ......................................... 170

Anatolij Mahnitko, Inga UmbrashkoInstitute of Power Engineering, Riga Technical University, Latvia

Measurements of Electrical Drive ........................................................... ................................................................. ... 175Robertas Janickas, Zita SavickienVilnius Gediminas Technical University, Lithuania

Modeling and Analysis of Voltage Unbalance and its Influence on Transmission Network and Power Quality . 179Tanel Sarnet, Jako KilterDepartment of Electrical Power Engineering, Tallinn University of Technology, Estonia

Wind Park and Transmission Network Cooperation Considering Grid Code Requirements ............................. 183Jako Kilter, Edgar Dubbelman, Ivo Palu, Jaan NiitsooDepartment of Electrical Power Engineering, Tallinn University of Technology, Estonia

Transmission Network Development Planning by Using a Cooperative Game Theory Approach ....................... 189

Igor Moshkin, Svetlana Berjozkina, Antans Sauhats

Riga Technical University Riga, Latvia

The Power Distribution in the Electrical Power System Taking into Account Reservation ................................. 195Renata Varfolomejeva, Anatolij MahnitkoRiga Technical University, Latvia

High Temperature Low Sag Conductors as Method for the Improvement of Electrical Transmission Lines ..... 200

Svetlana Berjozkina*, **, Antans Sauhats *, **, Vladimirs Bargels **, Edvins Vanzovichs ** Riga Technical University, Institute of Power Engineering, Latvia; ** JSC Siltumelektroprojekts, Latvia

Supply Voltage Level Optimisation in Low Voltage Networks Using Shunt Capacitors ....................................... 206

Toomas Vinnal, Kuno Janson, Heljut Kalda, Lauri KttDepartment of Fundamentals of Electrical Engineering and Electrical Machines

Tallinn University of Technology, Estonia

Design of High-Frequency Transient Current Sensor for Powerline on-Line Measurement ................................ 212

Lauri Ktt*, Jaan Jrvik*, Jako Kilter*, Muhammad Shafiq**, Matti Lehtonen***Tallinn University of Technology, Estonia; **Aalto University, Finland

7


8/277


9/277

MULTIMODAL AUTOMATION CONTROLLER FOR SMART HOMES

Rytis MASKELIUNAS*, Vidas RAUDONIS*, Grigor STAMBOLOV**, Rimvydas SIMUTIS** Kaunas University of Technology, Lithuania; ** Technical University-Sofia, Bulgaria

Abstract:Authors suggest an approach of a multimodalcontroller based on a client-server architecture running

on a x86 Windows server platform connected to theUSB I/O hardware interface and capable of remotecontrol via web application using speech commands,touch gestures or gaze tracking. The paper presents the

state-of-art situation, the proposed architecture, theassociative control algorithm as well as the initial

experimental evaluation on the recognition accuracy ofthe input modalities analyzed and the effectiveness ofcontrol tasks.

Keywords:HCI, remote, speech, touch, automation.

1. Introduction

Wireless communication of various machines anddevices in mobile communication networks is a fast

growing business and application area. Modern homeautomation systems need to make use of moderntechnological components available, implementingvarious communication capabilities such as GSM,Internet, etc. together with speech recognition and othermodalities. All these techniques can successfully be

merged in a single wireless home automation system [1]offering a low cost, powerful and user friendly way of

real-time monitoring and remote control of a homeenvironment. Another sample design [2] integrates the

device to be controlled, the microcontroller, and theGSM module so that it can be used for a wide range of

applications and is illustrated by two experimentallytested home applications using PC-based equipment. A

home server platform can effectively integrate thefunctions of communication, digital broadcastingreception, and home automation applications providinga client-server model [3].Speech in information services and reading support isconsidered as a most natural modality. In [4] authors

present a case study concerning the design of atelephone-based interface concept for a modular homeautomation system. The evaluation revealed some minor

usability problems indicating a potential to improve thehedonic quality (may be induced by the addition of non-speech sounds and enriching user experience). Other

researchers have explored the possibilities of speech and

gesture based interfaces in desktop scenarios, but it wasalso necessary to explore these technologies in the

context of an intelligent environment, in which theusers focus moves off of the desktop and into aphysically disparate set of networked smart devices [5].

The results of such tests indicate that speech is thepreferred interface for controlling the devices. Moreimportantly, the study indicated that the locationawareness and gaze tracking are vital when building a

usable intelligent environment system for the home,even though users did not perceive added value from

these technologies.Other input modalities such as gesture detection systemcan provide the additional control capabilities. Forexample a solution [6] of a wearable device for control

of home automation systems via hand gestures solutionhas many advantages over traditional home automation

interfaces because it can be used by those with a loss ofvision, motor skills, and mobility achieving 95%accuracy of control gestures and 97% accuracy of user-defined gestures. Authors of [7] present multi-interfacesensor network gateway architecture for homeautomation and other distributed monitoring

applications providing multiple interfaces for supportingvarious application scenarios in home environments,

ranging from on-site configuration to mobile access.Results of the experiment showed that the proposedgateway gives good support to managing the home

network from different user terminals and allows theusers to better interact with the ambient environment.Alternatively in [8] paper intuitive home environment

was proposed and image processing software wasconnected with eye tracking system in order towirelessly control electric devices.Authors of this work present the architecture of theirvery own low-cost remote home automation controlsolution by utilizing a proprietary multimodal control

algorithm on industry standard hardware and softwaresolutions.

2. The concept, architecture and implementation

The inspiration and the concept of a practicalimplementation of a multimodal automation controllercame from a W3C Multimodal Interaction Working

9


10/277

Group Charter [9]. The upcoming standard envisions thefollowing: extending the Web to allow multiple modesof interaction (GUI, Speech, Vision, Pen, Gestures,Haptic interfaces); Anyone, Anywhere, Any device,

Any time (Accessible through the user's preferredmodes of interaction with services that adapt to thedevice, user and environmental conditions).Basically we consider a model of an interactionillustrated in figure 1.

Fig. 1.Multimodal interaction concept based on a MultimodalInteraction Working Group Charter [9]

A user utilizing modern, web capable computing deviceshould interact with the application using various built-in I/Os and associated modalities, for example speechrecognition, image recognition, touch interactions, etc.

Fig. 2.Our implementation built on a client-server model

The application itself should be built on a multimodalarchitecture thus capable of parsing the semantic

meaning of various input modalities and issuing acontrol action to a connected automation device, such as

a PLC or chip based programmable controllers (e.g.Arduino). Since it is hard to overcome and unify thedifferences in many modern hardware and softwarearchitectures, such as x86 (Windows, Linux), ARM

(Android and iOS) we targeted the client-server modelas the best approach for our implementation (figure 2).

Basically a user uses his/her device only to interact witha system.Our system implementation was based on the above

concept and further built on the basis of our previousexperimentation on multimodal interactions [10].Basically a prototype used was a web interface running

on windows PC based display with built-in webcam,

microphones and speakers. A system was able to gather

a user input via either of affirmed inputs (webcam wasused only for the confirmation (nodding)) and displaythe graphical feedback on screen. With this approach wewent further by moving into an eyes-free interface forhealthy users, by not providing a GUI for the user(disabling the screen on our test device). Additionally a

gaze tracking system was provided for the paralyzedpersons (looking at a designated area divided in to

button like squares). The concept of the proposedalgorithm is shown in figure 3.

Fig. 3.The concept of our algorithm

All input was sent using an internet connection to theserver. Since we had no GUI, naturally all feedback hadto be done via speech. So a Microsoft Speech API [11]was chosen for this purpose. An advantage of thissystem was that it allowed us to insert a custom code(i.e. touch and eye processing) within the Speechapplication framework while still maintaining

compatibility with Microsoft Windows Server IIS basedweb services allowing web interaction (gathering a userinput and outputting a feedback).For the automation tasks we have utilized the PhidgetsI/O hardware with a compatible relay block and theArduino controller for some of the autonomous control

processing. Traditionally the home automation isdivided into four categories by purpose. The firstcategory includes lightening control aimed at turningthe lights on and off remotely. The next categoryinvolves home security integrating a range of security

mechanisms to protect one from intruders by not lettingthem enter premises and issuing alerts. The third

category involves the climate control playing a crucialrole in maintaining the temperature of the roomautomatically and also conserving the energy byswitching the system off when there is no one at home.The last category involves the automation of homeentertainment system by managing music system andtelevision with the use of the multimodal interface. Our

current smart home implementation covers only threetypes of automation categories, i.e., the control oflightening, heating and entertainment system.

3. Control algorithm

The device application was built on three main modules:

gaze (), touch () and voice control () modules10


11/277

and the usage of each in the application is controlled bycoefficients a, b, and c (see formula (1)). The prioritiesof control modalities are identified by the user. Forexample, when interface user chooses to use a speech

modality all other modalities is switched off by settingcorresponding coefficients to zeros, i.e., a, b=0. In thiscase, the control signal is switched from a one modalityto a second modality when control command C is notrecognized.

; (1) 1, 0, 0 1; 0, 1, 0 2; 0, 0, 1 3; (2)Here - is the formal representation of the userselection of the prepared modality.

The simplified algorithm of the touch modality () isshown below, where the user must collect the sequence

of the key numbers which encodes the certain command(). The collecting process is stopped when the userpushes the escape button (ESC).

{ 0;whil TRUE 1; ;if ESC, FALSE;n ;

Here K is the vector of numbers which corresponds to

key number of the button on the interface device, n is

the amount of numbers (or buttons) which user ispressing, is the variable that controls the executionof the while cycle. The cycle is stopped, when userpushes the escape (ESC) button.The simplified principle of the gaze tracking modality

() is shown in the algorithm below, where the usercollects the sequence of the key numbers by gazing. Theuser confirms a selected key button by closing the eye.The collecting process is stopped when a user selectsthe escape button (ESC).

{

0;whil

TRUE

1; ; ;if 0.9, min ;

if ESC, FALSE;n ;

Here - is the normalized luminosity function of thehuman eye image,

- is the luminosity function of the

eye image, - is the number of column of theluminosity function, is the Euclidian distancebetween center coordinates of the gaze point () and thecoordinates of the button area (), is the number of

key buttons, is a correlation coefficient betweenluminosity functions of the real time eye image () andtemplate of the closed eye image ().The simplified algorithm of the speech modality () isshown in the paragraph below, where the user collectsthe sequence coded in key numbers by pronouncing the

required name of the command (or the key sequence).

The collecting process is stopped when a userpronounces the name of the escape function (ESC).The speech recognition of the certain command ornumber is based on a computation of cepstral analysisand correlation used as similarity measure.

{

0;whil TRUE 1; ln ; 1,2, , ; ;

;

; ; max ;if ESC, FALSE;n ;

Here - is the coefficient of the cepstral function, - isthe spectrum of the source signal S(t), - is thecoefficient of the linear propagation model (LPG), -the similarity value between real time cepstral

coefficients () and template commands ().In the next chapter we present the initial experimentalinvestigation of the proposed multimodal controller forsmart home environment control.

4. Experimental setup and investigation

The experimental smart home hardware consisted of aUSB controller with a relay block that commutates table

and ceiling lamps, heating radiator, according theambient sensor and user needs. The automation programis running on the server that receives the commandsfrom the user. The entertainment system consists of two

powered speakers and music software with a playlistcontrol interface (winamp). 10 people have participatedin the experimental investigation. Each user was askedto test the proposed control interface by executing four

tasks (10 times each, presented randomly): (1st) turning

on the table lamp, (2nd) turning on the ceiling lamp, (3rd)increasing the room temperature (setting dial on the

heater to 25 degrees) and (4th) selecting a different song.

The control of the smart environment was achievedusing all three modalities. Acquired results werecompared with a traditional way of the control.Only speech input modality was evaluated at this stage.The relation between speech recognition accuracy and a

specific user is shown in figure 4. Three different curvesrepresent the speech recognition performance in thethree different environments, i.e., noisy, normal andthe controlled environment in the laboratory. As it was

11


12/277

expected the best performance and accuracy (on average940.2 %) was reached when the experiments weredone in the controlled environment (vs 840,3 % innormal environment). Unfortunately the recognition

accuracy dropped to ~ 500,8 % in the noisyenvironment.

Fig. 4. Speech recognition accuracy in all environmentsanalyzed

The performance of the multimodal control interfacewas evaluated by measuring the task completion time in

seconds (see figure 5). Each of the tasks were selectedto illustrate a different time frame necessary tocomplete. However, the multimodal controller achieved~3,6 x faster performance for the second task and ~2 x

for the third task (~12,75 s by manually turning thelamp and ~23,61 s presetting the regulator on a heater

versus ~3,5 s and ~11,28 s). Another interesting aspectwas that the tasks requiring decision (e.g. which musictrack to select (4

th)) were done in similar time-frame.

Fig. 5.The performance (in time) of a multimodal controller

Each participant was asked to compare and subjectively

rank the satisfactory level of two different controlapproaches on the scale from 0 (worst) to 10 (best). The

results of this survey are illustrated in the figure 6.

Fig. 6. The satisfactory level using two types of interaction,i.e., normal and interface based

All participants have found the multimodal control

interface more convenient than the traditional way of

manual control. The multimodal control interface wasassessed with ~ 8.3 rank values versus ~ 6.2 rank valuesfor the traditional way of control.

5. Conclusions and future works

The results of the experimental investigation showedthat every participant found alternative multimodalspeech based control scheme more convenient (~23%more) and up to ~50 % more efficient way to control the

devices of the smart home.The functionality of the proposed system allowedachieving even better performance in quieter (up to 3times faster in time) environments due to a lowernumber of recognition errors, i.e. ~ 6 %.We plan to update the results in near future with theexperimentation on the effectiveness of remaining

modalities and the adaptation of multimodal interface torecognize the gesture commands using a stereo visiontechnique.

6. Acknowledgements

Parts of this research were funded by the LithuanianScience Council under the Postdoctoral FellowshipImplementation in Lithuania project No.: 20101216-90.

7. References

1. Yuksekkaya B. et al. A GSM, internet and speech

controlled wireless interactive home automation system.IEEE Transactions on Consumer Electronics, Vol. 52

Iss. 3, 2006, p. 8378432. Alheraish A. Design and implementation of home

automation system. IEEE Transactions on ConsumerElectronics, Vol.50, Iss. 4, 2004, p. 1087109

3. Intark Han, Hong-Shik Park, Youn-Kwae Jeong, Kwang-Roh Park. An integrated home server for communication,

broadcast reception, and home automation. IEEETransactions on Consumer Electronics, Vol. 52, Iss. 1,

2006, p. 104-1094. Sandweg N., Hassenzahl M., Kuhn K. Designing a

Telephone-Based Interface for a Home AutomationSystem. International Journal of Human-Computer

Interaction, Vol. 12, Iss. 3-4, 2000, p. 401-4145. Brumitt B., Cadiz JJ. "Let There Be Light!" Comparing

Interfaces for Homes of the Future. Technical Report,

MSR-TR-2000-92, 2000, 10 p.6. Starner T., Auxier J., Ashbrook D., Gandy M. The gesture

pendant: a self-illuminating, wearable, infrared computervision system for home automation control and medicalmonitoring. Wearable Computers, 2000, p. 87-94

7. Guangming Song; Yaoxin Zhou, Weijuan Zhang, AiguoSong. A multi-interface gateway architecture for homeautomation networks. IEEE Transactions on Consumer

Electronics, Vol. 54, Iss. 3, 2008, p. 1110-11138. Kairys A., Raudonis V., Simutis R. iHouse for advanced

environment control. Electronics and ElectricalEngineering, nr. 4(100), 2010, p. 37-42

9. Multimodal Interaction Working Group Charterhttp://www.w3.org/2011/03/mmi-charter.html

10.Maskeliunas R. Modeling aspects of multimodalLithuanian human - machine interface. Multimodal

Signal: Cognitive and Algorithmic Issues. LCNS Vol.5398, 2009, p. 75-82.

11.Dunn M. Pro Microsoft Speech Server 2007: DevelopingSpeech Enabled Applications with .NET, 2007, 275 p.

1 2 3 4 5 6 7 8 9 10

0,00 %

10,00 %

20,00 %

30,00 %

40,00 %

50,00 %

60,00 %

70,00 %

80,00 %

90,00 %

100,00 %

Noisy

Normal

Controlled

Number of the user

Recognitionaccuracy,

[%]

1 2 3 4

0,00

5,00

10,00

15,00

20,00

25,00Normal

Interface based

Task number

Performance,

[sec

]

1 2 3 4 5 6 7 8 9 10

0

2

4

6

8

10

12

Normal

Interface based

Number of participant

Levelofsatisfaction

12


13/277

IDENTIFICATION OF INTERCONNECTED SYSTEMS

BY INSTRUMENTAL VARIABLES METHOD

Grzegorz MZYK

Wroclaw University of Technology, Institute of Computer Engineering, Control and Robotics, Wroclaw, Poland

Abstract:The paper addresses the problem of parameter

estimation of elements in complex, interconnectedsystems. Similarity between causes of biases in the least

squares estimates for a simple SISO linear dynamic

object, and for a MIMO linear static system with

composite structure, was noticed in the paper. For linear

complex static system, the instrumental variable

estimate was proposed and compared with the least

squares approach. The strong consistency of the

presented parameter estimate was proved. Also the

optimal values of instrumental variables were

established, and the method of their suboptimal

generation was presented. The conclusions were

verified in numerical experiments.

Keywords: System identification, complex systems,

parameter estimation, instrumental variables.

1. Introduction

We consider the problem of parameter estimation in

complex, interconnected systems with the presence of

random noises. In a lot of commonly met hierarchical

control problems, the accurate mathematical models of

the particular system components are needed. Under the

term 'complex' we understand the fact that the system is

built of a number of interconnected components(subsystems), e.g., in the typical production system each

element is excited by the outputs of other blocks (see

[1]). In consequence of mutual interconnections, the

components are dependent and their separation may be

impossible or too expensive. In general, excitations of

particular element cannot be freely generated in the

experiment. It leads to the problem of structural

identifiability (i.e. identifiability of separate elements

does not imply identifiability of the whole

interconnected system [3]) and usually badly

conditioned numerical tasks. Moreover, some

interaction signals are hidden, and cannot be directly

measured. For these reasons, the algorithms dedicatedfor single element cannot be directly applied in complex

system analysis.

Identifiability of the element, which operates in

complex system, depends additionally on the systemstructure and the values of parameters of other

elements. Particularly, the components preceding

identified object must guarantee persistency of the input

excitation. In the paper we apply and compare two

methods least squares (l.s.) and instrumental variables

(i.v.) approach.

It is commonly known from the linear system theory,

that the least squares approach applied for the simple

SISO linear dynamic object leads to biased estimate.

The reason of the bias results from the property of

autoregression, i.e. the correlation between the noise

and the values of previous outputs of the identifiedobject (see the Appendix). Analogously, for the

complex, interconnected systems with random noises,

the least squares estimate has the non-zero systematic

error even if the number of measurement data tends to

infinity. The reason of the bias is that the output noises

are transferred to the inputs through the structural

feedback.

In the paper, the formal similarity of these problems is

shown and the instrumental variables technique, used so

far for the linear dynamics identification, was

successfully generalized for the systems with complex

structure. It is shown that the proposed i.v.estimate is

strongly consistent independently of the systemstructure and the color of the noise. Moreover, the

computational complexity of the method is comparable

with the ..sl algorithm. In Section 2 the identification

problem and the purpose is formulated in detail. Next,

in Section 3, the properties of the least squares based

algorithm proposed in [3] are reminded. In particular,

the reason of its bias is shown, and finally, in Section 4

the new i.v. estimate is introduced and analyzed.

Finally, in Section 5, the performance of the method is

demonstrated by the simulation example.

13


14/277

2. Statement of the problem

Consider the system shown in Fig. 1. It consists of n

linear elements described as follows

,),...,2,1( niubxayiiiiii

=++= (1)

where

( )

( )

( )Tn

T

n

T

n

yyyy

xxxx

uuuu

,...,,

,...,,

,...,,

21

21

21

=

=

=

(2)

are the external inputs, interaction inputs, and system

outputs, respectively. The processes

( )

( )Tn

T

n

,...,,

,...,,

21

21

=

= (3)

are random disturbances. The block H determines the

system structure in the following way

,iii

yHx += (4)

whereHi is the i-th row of the binary matrixH (i.e.

0, =jiH 'no connection', 1, =jiH 'is connection').

1u

1y

nu

iu

11,ba

nn ba ,

ii ba ,

H

ny

iy

1x

nx

ix

1

i

n

1

i

n

Fig. 1.The complex n-element linear static system

The goal is to estimate parametersn

iiiba

1)},{( = of the

particular elements using the set of data Nk

kkyu

1

)()( )},{( =

collected in the experiment. We emphasize that the

internal excitations )(k

x cannot be measured.

We assume that:

(A1)The structure of the system (i.e. the matrix H ) is

known.

(A2)The system is well defined, i.e. for any ),,( u it

exists unique y (see [3]).

(A3) The noises , and are zero-mean, mutually

independent, and independent of .u

(A4) In the noise-free case ( 0= , and 0= ) the

system would be identifiable (see [3]).

(A5)The excitations are rich enough, i.e. the matrix

( ),,...,,21 NN

eeeE = (5)

where

,and),( +== Aue TTT (6)

is of full rank with probability 1 .

Introducing the matrices

( )

TT

n

TT

n

n

HHHH

bbbB

aaaA

,...,,

),...,,(

),...,,(

21

21

21

=

=

=

(7)

the whole system can be described in the following

compact form

.

+=

++=

Hyx

BuAxy (8)

Inserting x to the first equation in (8) we obtain

( )( ) ,

,

++=

+++=

ABuyAHI

BuHyAy (9)

which leads to

GKuy += (10)where

.)(

,)(1

1

GBBAHIK

AHIG

==

=

(11)

The equation (10) resembles description of the object

with the input u , the output y , the transfer matrix K,

and the noise G . Invertibility of )( AHI in (11) is

equivalent to assumption (A2).

3. Least squares approach

Introducing the vectors of input-output data of i -thelement

,),(where

,],...,,[

,],...,,[

)()2()1(

)()2()1(

T

iii

N

iiiiN

N

iiiiN

uxw

wwwW

yyyY

=

=

=

(12)

we obtain the measurement equation

.),(iiNiiiN

WbaY += (13)

Since the inputi

x included ini

w is unknown (cannot

be measured), the least squares estimate cannot be

derived directly from (13). Owing to (4) the natural

substitution is

( , ) , where .T

i i ii i i iw x u x H y x = = =

(14)It leads to the following least squares estimate

( )1

. . . .

( , ) ,T Tl s l s

i i iN iN iN iNa b Y W W W

=

(15)

where

(1) (2) ( )

[ , ,..., ].N

iN i i iW w w w= (16)

Remark:The estimate (15) originates from the modified

version of measurement equation (13), in whichiN

W

was substituted with WiN( , ) .iNiN i i iN Y a b W = + (17)

Consequently, in (17) the disturbance

],...,,[ )()1()1( NiiiiN

= (18)

appears instead ofiN

. The situation is similar to the

problem of identification of the simple linear dynamics

with autoregression (see example in the Appendix). It

was shown in [3], that because of correlation between

the elements ofiN

and WiN, the estimation error

( )1

. . . .

( , ) ( , )Tl s l s

i i iN iN iN i i iN a b a b W W W

=

(19)

does not tend to zero, as N .

14


15/277

4. Instrumental variables approach

To solve the problem shown in the above Remark we

propose the analogous strategy as for the SISO dynamic

system identification (see e.g. [2] and [7]), i.e.,

generalization of (15) to the following form

( )

1. . . .

( , ) ,Ti v i v T T

i i iN iN iN iN a b Y W

=

(20)

where

],...,,[ )()2()1(N

iiiiN = (21)

is the additional matrix of instrumental variables, of the

same dimensions as WiN, i.e.

( ) ., )(2,

)(

1,

)( Tk

i

k

i

k

i = (22)

We impose the following two conditions oniN

:

(C1) The instrumental variables1,i

, and2,i

are

correlated with the inputi

u , such that

( )( )

1

1 1 TNT k TT kiN i iiN i i

k

W w E w

N N

=

= (23)

with probability 1 , as N , and the limit matrix

Eiw iT

is of full rank.

(C2)Simultaneously,1,i

, and2,i

are not correlated

with the aggregated output noisei

, i.e.,

,NiiN

EL= (24)

where

,0

0

=

i

i

iI

L (25)

and

.]0,...,0,1,0,...,0[=i

I (26)

Theorem: If the instrumental variables matrix

iN

fulfils (C1)and (C2)then. . . .

( , ) ( , )i v i v

i i i ia b a b

(27)

with probability 1 , as N .Proof:The estimation error has the form

( )

. . . .

1

1

( , ) ( , )

1 1

i v i v

i i i i

T TiNiN iN iN

T TiNiN iN iN

a b a b

W

WN N

= =

= =

=

(28)

where

iN i NW F E= (29)

and

.0

=

i

ii

iI

GHKHF (30)

Since, according to (C1) and (C2) it holds that

,1 ,1

,2 ,2

( , ) ( , )1,

( , ) ( , )

10,

i i i iTiN iN

i i i i

T

iN iN

x uW

N x u

N

(31)

from the Slutzky theorem we conclude (27).In real applications the procedure ofiN

-generation is

of fundamental meaning. Let us introduce the quality

index of instrumental variables

( ).)()()()(max iN

T

iNiNiNQ == (32)

The following theorem holds.

Theorem: The optimal instruments with respect to the

value of )(iN

Q has the form

,)|(where,),( KuHuxExuxwiii

T

iiii ==== (33)

i.e.,

.KHii = (34)Proof:

It is obvious that

( , ) ( , ) ( , ) ,i i ii i i i i i i i i i i iy a b w a b w a H G a b w z = + = + + = + whe

reiiii

GHaz += is a zero-mean disturbance,

uncorrelated with the elements of the 'expected' input

vector iw .

According to (28) we obtain that

1 1

1 1( ) ( )

1 1

1 1,

T T

iN iN iN iN

TT

iN iN iN iN

T

iN iN

N N

W WN N

N N

=

(35)

and making use of the property that

( ) ( ),)()()()(maxmax iNiN

T

iN

T

iN = (36)

for N large and ],...,,[)()2()1( N

iiiiNiNwwwW == we

simply get

.1

)(

1

max i

T

iNiNiNWW

NQ

= (37)

Under Lemma 6 in [4], for eachiN

it holds that

)()(iNiN

QQ (38)

with probability 1.Since the matrix K is unknown, the result (33) is not

constructive, but gives the general concept of using the

estimates of the noise-free interactionsi

x . In the

simulation we used the approximation of K obtained

by the least square method and implemented the

recursive version of the algorithm

. . . . . . . .

( ) ( 1)

. . . . ( )( ) ( )

( 1) ,

( , ) ( , )

( , ) ,T

i v i v i v i v

i i i ik k

i v i v k k k

i i ii k i i k

a b a b

y a b w P

= +

+

(39)

where

( ) ( )

, 1 , 1 , 1, ( )( )

, 1

.1

T

T

kk

ii k i k i i k i k k

kii i k

P P w PPP w

=

+ (40)

5. Simulation example

In this section we present the performance of the

algorithm on the example of the simple, two-element

cascade system with feedback (see Fig. 2). We set

)1,1(),(11 =ba , and )2,2(),(

22 =ba and the

interconnections are coded as follows

.01

10

=H (41)

The system is excited by two independent uniformly

random processes )1,0(,21

Uuu , and disturbed by

15


16/277

zero-mean noises )1.0,1.0(,21

U . For both

elements, the instrumental variables estimates are

computed and compared with the least squares results.

1u

2u

11,ba

22,ba 2y

1x

12 yx =

2

1

Fig. 2.The two-element cascade system with feedback

In Fig. 3 we present the Euclidean norm of the error for both algorithms.

0

0,01

0,02

0,03

0,04

0,05

0 100 200 300 400 500 600 700 800 900 1000

least squares

instrumental variables

Fig. 3. The estimation error versus number ofobservationsN

6. Summary

The idea of instrumental variables estimate was

successfully generalized for the complex,

interconnected systems. Since the production and

transportation systems usually work in steady state, welimited ourselves to the static blocks. However,

generalization for the FIR linear dynamic components

seems to be quite simple. For details, we refer the reader

to [6], where similar approach was applied for

identification of interconnected Hammerstein systems

(see Fig. 4).

ku ky

kz

kw( ) { }nii 0=

( ) { }p

jj 1=

kw'

kv

kv'

Fig. 4.The NARMAX system

In the contrary to the traditional least squares approach,

the proposed algorithm recovers true parameters of

subsystems. General conditions are imposed on the

instrumental variables for the estimate to be consistent,

and then the form of optimal values of instruments is

shown. We emphasize that the method works for any

structure of the system and for any distribution of the

random noises.

7. Appendix

In this section the subscript k is used for the time

instant. Consider the simple AR(1) linear dynamic

object with the inputk

u and the outputk

v , disturbed by

the random processk

, i.e.

.and,1 kkkkkk

vyavbuv +=+= (42)

Since the noise-free outputk

v is unknown, we must

base on the difference equation describing dependence

betweenk

u and the measured output

,),(1 kkkkkk zbazaybuy +=++= (43)

whereT

kkkuy ),(

1= and the resulting disturbance

1= kkk az is obviously correlated with 1ky ,

included in the generalized 'input'k

.

8. Acknowledgement

This research was supported by Polish National Science

Centre, Grant No. N N514 700640

9. References

1. W. Findeisen, F. N. Bailey, M. Brdy, K.

Malinowski, P. Tatjewski, A. Woniak, Control and

Coordination in Hierarchical Systems, Wiley, New

York, 1980.

2. Greblicki, W., Mzyk, G. Semiparametric approach

to Hammerstein system identification, In:

Proceedings of the 15th IFAC Symposium on

system Identification, Saint-Malo, France (2009)

1680-1685.

3. Z. Hasiewicz, "Applicability of least-squares to the

parameter estimation of large-scale no memory

linear composite systems", International Journal of

Systems Science, vol. 12, pp. 2427-2449, 1989.4. Z. Hasiewicz, G. Mzyk, "Hammerstein system

identification by non-parametric instrumental

variables",International Journal of Control, vol. 82,

No. 3, pp. 440-455, 2009.

5. M. D. Mesarowic, D. Macko, Y. Takahara, Theory

of Hierarchical, Multilevel Systems, Academic

Press, New York, 1970.

6. G. Mzyk, "Application of instrumental variable

method to the identification of Hammerstein-Wiener

systems", Proceedings of the 6th International

Conference on Methods and Models in Automation

and Robotics, vol. 2, pp. 951-956, Midzyzdroje,

Poland, 2000.7. G. Mzyk, "Nonlinearity recovering in Hammerstein

system from short measurement sequence", IEEE

Signal Processing Letters, vol. 16, No. 9, pp. 762-

765, September 2009.

8. P. Stoica, T. Sderstrm, "Instrumental variable

methods for system identification", Circuits Systems

Signal Processing, vol. 21, No. 1, pp. 1-9, 2002.

9. R. Ward, "Notes on the instrumental variable

method",IEEE Transactions on Automatic Control,

vol. 22, pp. 482-484, 1977.

10. K. Y. Wong, E. Polak, "Identification of linear

discrete time systems using the instrumental variable

method",IEEE Transactions on Automatic Control,

vol. AC-12, No. 6, pp. 707-717, 1967.

16


17/277

INVESTIGATION OF AUTOMATIC CONTROL SYSTEM

IN WOOD DRYING FACILITIES

Bronius KARALINAS*, Dalia LUKOIEN*, Liudmila ANDRIUIEN***Vilnius Gediminas Technical University, Lithuania

**Vilnius College of Technologies and Design, Lithuania

Abstract: The article analyses the automatic controlsystems in modern wood drying facilities as well astheir functions and possibilities.The analysis of the publications available [1, 2, 3]indicates that nowadays there are widely applied in ourcountry various types of wood drying facilities withdifferent level of automation. Although the possibilitiesof the fully computer-assisted control systems at thewood drying facilities are considered to be ratherextensive, but in the scientific publications availablethere has been noticed the shortage regarding theinformation on the structure of the systems, thealgorithm concerning the manufacturing processes,

programs, sensors and their interrelation andcorrelation.

Keywords: wood drying facility, automatic control,programmable logical controller, frequency converter,temperature, moisture, humidity, algorithm, heating.

1. Introduction

At present, in our country a huge amount of wood hasbeen processed by an industrial method, but the woodprocessing plants are not able to do without theupgraded wood drying facilities and equipment. Drying

of wood has been considered a complicated physicalprocess, which could be described as the interchange ofmass and heat in the wood and in the surroundingenvironment. The quality of drying depends on themajority of factors: the sort of wood, dimensions, thesize of the stocks, the technology applied in the dryingchambers and the level of the automation systemsinstalled [4].Wood drying could be carried out by means of thefollowing methods, namely, drying in chambers, dryingin the field of high frequency currents, drying by meansof infrared rays, drying in solutions and thermal drying[1, 2]. The most widely approved method is drying inspecial drying chambers under the natural andcompulsory circulation. Such type of chambers arecalled drying rooms. They are of two types [4]:

convection ventilation facilities; condensation facilities where there are preserved aclosed cycle of air circulation.Closed joint stock company ,,Lietmedis has proposeda new method of wood drying which consumes 10 timesless heat energy [5]. The convection condenser typedrying rooms were designed and introduced to serve thementioned above purpose, where the surplus humidityof the circulating air is condensed on the tubes of thecondenser. Then the heated water proceeds into theheater located in the manufacturing premises and heatsthe premises. There it is cooled down and then it isreturned into the tubes of the condenser located in the

drying room. The circulation pumps are installed insuch type of drying rooms. Thus, the whole heat energyaccumulated in the drying room is used efficiently.The purpose of the work is to analyse the automatedcontrol systems installed in modern wood dryingfacilities and present the description of thecharacteristics of the drying process.

2. The functional scheme of the control in the wood

drying room

The functional control schemes applied in theconvectional and condenser type drying rooms are verysimilar, but they differ by the installations andequipment assembled in them. The diagram in Fig.1displays the functional scheme of the system.The references indicated in this scheme Fig.1 have thefollowing meanings:VO the object of the control (pile of wood);VKD the equipment for the performance of operations(actuators);JUT sensors;

NKR the scheme for determining the deviations of theprogram concerning temperature, relative humidity,moisture of wood;SL the thresholds of decision making;SPR1 the scheme of decisions determining the logicaldeviations of the process;

17


18/277

Fig. 1 Functional control scheme used in the drying room

SPR2 the scheme of decisions forming the controlsignals of actuators;ROM the memory of the programs;VLD the scheme of the management of the control

programs;TMR is a stopwatch, performing the intervals of thecontrol signal, periods and duration.

The major equipment executing the processes in thecondenser type drying rooms are considered to be air

drying devices, which condense humidity and heat the

air, which is blown through them. Such devices at thebeginning of the drying process could eliminate morethan 500 litres of the condensate during the day andnight [4]. Besides that, in the automated drying roomsthere are required the controlled air vent valves in thechambers and the humidifying valves. The condensertype drying rooms have to be ventilated, when changingfrom the mode of wood steaming into the mode ofdrying. Such chambers are humidified when changingfrom drying mode into the steaming mode or when lessmoisture is disposed from the drying wood. All thedrying modes have to be controlled automatically.In modern drying rooms, there are used two circuit

control systems. One circuit is used to control theprogram, taking into consideration the time or realmoisture of the wood, the other circuit has to keep therequired accuracy when satisfying the pre-set values inthe program (temperature, relative air humidity etc.)

3. Determining the parameter deviations of the

drying process

During the process of manufacturing, there appear thedeviations of the microclimate parameters of the dryingroom from the pre set allowable limits of the control

program. The deviations could be eliminated by

applying at least three impacts directed towards eachother. In the condenser type drying rooms such impactsare considered to be the following, namely the heatingof the chamber, the condensation of the humidity,

watering, but in the convection type drying rooms airventilation is used as well.Various sensors and their versions keep watch andmonitor the parameters of the drying process. Theinformation received from the sensors is applied in thecompiling of the algorithm and the control program ofdrying as well as the determining of the allowabledeviations of drying parameters depends on theinformation obtained. The temperature of the dryingroom and the deviations of the relative air humidity aredetermined by the following methods [1]:

direct method, when the sensors of airtemperature and relative humidity of the dryingchamber are used:

k pT T T ; (1)

k p ; (2)

where kT and pT is air temperature in the chamberand in accordance with the program;

k and

p is

the relative air humidity in the chamber and accordingto the program.

indirect method, when the signals of the sensorsof air temperature and relative moisture are applied:

k pT T T ; (3)

pk ppW W , when 0T ; (4)

where pkW and ppW is the equilibrium moisture ofwood in the chamber and set in the program.

indirect method, when the signals of thethermometers indicating dry and humid air are used:

ks psT T T ; (5)

when T 0

( ),

ps pd ks kd

p ks kd

T T T T

T T T

(6)

where ksT and psT are the readings of the dry air

thermometers in the chamber and according to theprogram; kdT and pdT are the readings of the humid

air thermometers in the chamber and according to theprogram; pT the psychometric difference of

temperatures according to the program.When applying the first direct method, the determiningof the deviations of the temperature and relativehumidity is considered the most accurate and they donot depend on each other. According to the values of thedetermined deviations T and there are compiled

the areas of the allowable small and large deviations for

the other parameters of the drying process. Thesecalculations are required for compiling the algorithm ofthe control of the drying room and the program.

18


19/277

4. Selection of the sensors and performing devices

The technological process of wood drying consists ofseveral stages: warming up, initial steaming, one orseveral drying stages, final steaming and drying as wellas quenching. There are sometimes used the transitionalstages of steaming as well as the other drying stages.To measure the temperature in the drying chambersthere are applied expansion, manometric, electricresistance thermometers, thermoelectric and radiation

pyrometers. The temperature measuring devices andregistering equipment are presented in Fig.2.

Fig. 2. Temperature measuring devices

To measure the humidity in the chambers there are usedtwo main methods:

psychometric is meant to measure humidity athigher temperatures (up to 1000C);

hydroscopic is meant to measure humidity at

lower temperatures.At present, there are available humidity meters used formeasuring the moisture in wood, humidity in the

premises and the air of the environment. They arepresented in Fig.3.

Fig. 3. Wood moisture meters

The computer assisted system in the wood dryingfacility regulates the flow of air, by changing thevelocity of ventilators, when their motors are controlled

by the frequentative method.When selecting the electric motors for the ventilators ofthe convection ventilated type of drying rooms, it isimportant to evaluate the fact that motors have to workin hot and humid environment. In the ventilated drying

chambers, the temperatures of air and steam amount to(60 80)0 C. Because of that, the motors operating witha nominal load tend to be overheated and that is themain reason for the breakdowns appearing in the

ventilated drying rooms. The standardised electricmotors are applied in operation when the surroundingtemperature is +400 C. In case, the environmentaltemperature of the motors during the operation is higherthan +400C, then according to the requirements of thestandard [6] the superheat temperature of the insulationmaterials of the motors has to be reduced by theenvironment temperatures difference. Then in order tokeep the nominal operational efficiency of theventilators it is necessary to select higher capacityelectric motors.

5. The dynamic characteristics of wood

drying process

The computerized control systems nowadays applied inwood drying facilities are able automatically control andmonitor all the stages of drying. The total process ofwood drying may consists of 12 stages. For example,the new system EASY NEW is able to control at a time

even 32 drying chambers [7]. In the memory of thecomputer, dealing with the automatic control, there arestored many different programs of drying intended andspecified for every sort of wood. When selecting boththe mode of drying and the program there is determinedthe sort of wood, its thickness, and the size of theloading, the final humidity and other parameters.The wood drying computer program within the pre settime intervals via the consecutive communicationinterface RS485 collects all the data of the sensors fromthe drying chamber and processes it. The data of themeasurements are accumulated and stored in thememory and within the real time are displayed on the

screen of the monitor. It is possible to supervise theinstantaneous values of the drying parameters in theform of tables and curves. The program automaticallyregisters and displays all the dynamic characteristics ofthe drying process.The diagrams of the processes executed in the dryingroom, are presented in Fig.4. Fig.5 presents thedynamic characteristics when drying 30 mm thicknessoak boards.

Fig. 4.The diagrams of the processes held in the drying room:1 the set temperature; 2 the real temperature; 3 theaverage moisture of wood; 4 the average air humidity; 5 the equilibrium air humidity

19


20/277

Fig. 5. Drying characteristics for 30 mm thickness oakboards: 1 the set temperature; 2 real temperature in thechamber; 3, 4, 5 the curves of wood moisture; 6 theaverage air humidity; 7 the equilibrium air humidity

In Fig. 4 presented curves indicate that when drying the

moisture evaporates from the wood (curve 3), the air inthe chamber fills with water vapour and its temperatureis reduced (curve 2). Then wood stops drying. To

prolong the drying process it is necessary to increasethe temperature in the chamber as indicated in curve 2.In the control algorithms of the drying rooms there is

provided the possibility for the automatic system toreact to the deviations from in the program set

parameters of the drying processes. In case, the airtemperature in the chamber is lower, but the relativehumidity is higher than the standard, then within certainintervals the heating of the chamber is switched offautomatically. The mentioned above drying processes

are presented in Fig. 6.

Fig. 6. The diagrams exhibiting the change of heatingintensity of the chamber: 1 real temperature in the chamber;2 the set temperature; 3 the average wood moisture; 4 heating at certain periods of time; 5 the average air humidity

The decrease of wood moisture in the convectionventilated drying rooms is controlled through the valvesof fresh air. When air moisture in the chamberdecreases more than it is allowable, then the watering ofthe chamber is automatically switched on. Thecomputerassisted program controls these processes aswell.

6. Conclusions

1. Our country is extensively implementing varioustypes of wood drying facilities with different levelsof automation. However, there is a shortage ofinformation concerning the structure of the controlsystem for drying rooms, the algorithms ofoperation, programs, performing devices, sensorsand the other interrelated facilities.

2. From all the wellknown methods of wood drying,the best and the most efficient is considered to bethermal drying in special chambers, equipped withthe automatic control systems able to carry out allthe available stages of drying.

3. Closed joint stock company ,,Lietmedis hasintroduced the convection condenser type dryingrooms with the closed cycle of air and hot watercirculation. Because of that, the losses of heatenergy have been reduced and all the heat energyaccumulated in the drying room has been applied

efficiently.4. Modern computerized control systems available atwood drying facilities operate using the specifieddrying programs compiled individually for each sortof wood and they are able to control all the stages oftechnological process. The new system EASY

NEW manages to control at a time even 32 dryingchambers. The automated control systems applyvarious programmable logical controllers; however,the most popular are controllers from Germancompany ,,GANN HYDROMAT TK MP.

5. The results of the maintenance of the drying roomsindicate that the electric motors of the ventilation

devices tend to go breakdown most often becausethey operate under hot environmental conditions(60 80)0C. Following the requirements forstandard IEC 60034 -1, the ventilators operatingunder the mentioned above environmentalconditions have to be selected when motors bearhigher capacities by (20 30)%.

7. References

1. Kajalaviius A., Albrektas D. The ChamberConvection Drying of the Sawn Timber. Kaunas:Technologija, 2010. 212 p. (in Lithuanian).

2.

Hill C. A. S. Wood Modification: Chemical,Thermal and other Processes. Series: Wiley Seriesin Renewable Resources, Chichester, Sussex, UK,2006. 260 p.

3. Ciganas N., Raila A. Influence of Ambient Air onthe Drying Process of the Stored Wood Chips.Agricultural Engineering. Research Papers. 2010.Vol. 42, No. 4. P.71 85.

4. Kajalaviius A., Albrektas D. Hydrothermal WoodProcessing Theory and Equipment. Kaunas:Technologija, 2008. 120 p. (in Lithuanian).

5. http://www.lietmedis.lt/index_en.php.6. International standard IEC 60034-1. Rotating

Electrical Machines. Part 1. Rating andPerformance. 2004. 139 p.7. http://www.idna.lt/dziovyklos/easynew.html.

20


21/277

ARTIFICIAL INTELLIGENCE APPROACH TO SoC ESTIMATION FOR SMART

BMS

K.L. MAN1,2,3, C. CHEN5, T.O. TING1, T. KRILAVIIUS3,4, J. CHANG6and S.H. POON71Xi'an Jiaotong-Liverpool University, China; 2Myongji University, South Korea; 3Baltic Institute of Advanced

Technology, Lithuania; 4Vytautas Magnus University, Lithuania; 5Global Institute of Software Technology, China;6Texas Instruments, Inc, USA;

7National Tsing Hua University, Taiwan

Abstract: One of the most important and indispensableparameters of a Battery Management Systems (BMS) isaccurate estimates of the State of Charge (SoC) of thebattery. It can prevent battery from damage or

premature aging by avoiding over charge/discharge.Due to the limited capacity of a battery, advancedmethods must be used to estimate precisely the SoC inorder to keep battery safely being charged anddischarged at a suitable level and to prolong its lifecycle. In this paper, we review several effective

approaches: Coulomb counting, Open Circuit Voltage(OCV) and Kalman Filter method for performing the

SoC estimation; then we propose Artificial Intelligence(AI) approach that can be efficiently used to preciselydetermine the SoC estimation for the smart batterymanagement system as presented in [1]. By using our

proposed approach, a more accurate SoC measurementwill be obtained for the smart battery management

system.

Keywords:battery management systems (BMS), state ofcharge (SoC), artificial intelligence (AI).

1. Introduction

In the modern society, environment and transportationproblems are the main challenge for many countries.

Due to the increasing awareness of global warming, the

requirements for clean fuel are on the rise. Thus, there is

a continuous shift towards the Electric Vehicles (EVs)

and Hybrid Electric Vehicles (HEVs) [1]. Moreover,

battery-powered electronic devices have become

ubiquitous in modern society. Rapid expansion of the

use of portable devices (e.g. laptops, tablet computers

and cellular phones) creates a strong demand for a large

deployment of battery technologies at an unprecedented

rate. In addition, distinct requirements for batteries, such

as high energy storage density, no-memory effect, lowself-discharge and long cycling life, have drawn explicit

attention recently. Due to the above-mentioned facts,

Battery Management Systems (BMSs) become

indispensable for modern battery-powered applications

[11-13].

A BMS does not only monitor and protect the battery,but also provide the guidance on optimal usage of thebattery. One of the most important and indispensable

parameters of a BMS is to accurately estimate the Stateof Charge [5] of the battery. The SoC is defined as the

present capacity of the battery expressed as a percentageof some reference. Due to the limited capacity of abattery, advanced methods must be used to estimateprecisely the SoC of the battery in order to keep it safelybeing charged and discharged at a suitable level and to

prolong the life cycle of the battery. However, themeasurement of SoC is not a trivial task, because oneshould also consider the battery voltage, current,temperature, aging and so on. Accurate SoC estimationcan prevent the battery from damage or rapidly agingdue to unwanted overcharge and overdischarge on thebattery.

The conventional SoC estimation method such asCoulomb counting [3] suffers from an erroraccumulation glitch which leads to inaccurateestimation [2]. In addition, the finite battery efficiency[2] and the chemical reaction taking place during chargeand discharge cause temperature rise and badly

influence the SoC estimation. Therefore, efficientalgorithms are definitely needed for the accurate SoC

estimation. Furthermore, neither Coulomb counting norvoltage measurement alone is sufficient for highaccuracy SoC estimation, because the estimation of SoCis strongly influenced by many other factors such ascharge/discharge rates, hysteresis, temperature, cellaging, etc.

Asmart BMS for aged batteries and multi-cell batterieswas presented in [1] which aims to meet the followingrequirements:

Accurate estimation of SoC prevents battery

damage or premature aging by avoiding unsuitableover charge and over discharge.

SoC can be effectively used to deduce how wellthe battery system is functioning relative to its

21


22/277

nominal (rated) and end (failed) states.

The battery aging process needs to be reduced byconditioning the battery in a suitable manner (e.g.through controlling its charging and dischargingprofile), under various load conditions and harshenvironments.

Hardware implementation of the BMS is flexible

and adaptable in both Application-SpecificIntegrated Circuit (ASIC) and Field ProgrammableGate Arrays (FPGA) technology.

In this paper, we propose an Artificial Intelligence (AI)approach [15] that can be efficiently used to precisely

determine the SoC estimation for the smart batterymanagement system.This paper is organized as follows: Section 2 presentsand discusses several techniques that have been widelyapplied for the battery SoC estimation. Our intended AIapproach for estimating the SoC of the smart battery

management system is outlined in Section 3.Concluding remarks are given in Section 4 and

directions for future work are pointed out at the samesection.

2. Battery SoC Estimation

Many techniques have been proposed previously to

estimate the SoC of battery cells or battery packs, eachof them has merits and demerits.

2.1. Current Based SoC Estimation

Current based SoC estimation is also known as

Coulomb Counting [3-7], which takes integration of

current and time into account to estimate SoC. CoulombCounting requires an initial state namely SoC0, and ifthe initial state of the battery is known, from then the

SoC can be calculated though this method.For example, the initial state is SoC0, using I amperecurrent to charge the battery for thours, that will addI*tAh charge in the battery. Also, if the capacity of thebattery is C, then the final SoC can be calculated asfollows:

SoC SoC

(1)

Fig. 1. Estimating SoC by using Coulomb Counting

According to theory, if a battery was charged for 3

hours at 2A, the same energy can be released whendischarging. However, this is not the case in reality. Nomethodology is perfect. For instance, CoulombCounting suffers from a drift over time. As mentioned

in [3], battery aging causes a gradual small and constanterror in the variable. The small and constant errorcauses a tiny error for measurement of current, whichwill be magnified during each charging and dischargingcycle and will result in the SoC drift. Therefore, if thereis a way to re-calibrate the SoC on a regular basis, such

as reset the SoC to 100% when the battery is fullycharged, Coulomb Counting can be used to accurateestimate SoC and often enough to overcome drift.

2.2. Voltage Based SOC Estimation

There are already many applications that measure theSoC based on voltage, such as the charge balance shown

in cellular phones. The voltage is firstly measured andthen converted to SoC. When the battery is discharging,

the voltage drops more or less linearly [4]. In practice,there are two cases for the Lead Acid battery and theLi-ion battery [6]. For the Lead Acid battery, thevoltage diminishes significantly when it is discharged as

shown in Fig. 2 [4].However, the voltage is significantly affected by thecurrent, temperature, discharge rate and the age of cell.These factors need to be compensated, in order toachieve a higher accuracy of SoC.For the Li-ion battery, there are very small changes for

voltage between each charging and discharging cycle asshown in Fig.3. Due to the constant voltage of Li-ionbattery, it is difficult to estimate the SoC by voltage

based method. However, the voltage of the Li-ionbattery changes significantly at both the ends of SoCrange, which can be two important indicators ofimminent discharge. For instance, for many applicationsan early warning is required before the battery is

completely discharged or empty.

Fig. 2. The relationship between Voltage and SoC in Lead

Acid battery

22


23/277

Fig. 3. The relationship between Voltage and SoC in Li-ionbattery

2.3. Extended Kalman Filter

In 1960s, Kalman Filter theory was proposed [9] toaccurately estimate the state for the linear systems,especially for systems with multiple inputs, byremoving unwanted noise from a set of data. However,for the systems with specific requirements, such asnon-normality and non-linearity, the application of theKalman Filter method is not feasible.

Due to the time-variance, nonlinear model of the battery,noise assumption and the measurement error of theBMS, extended Kalman Filter (EKF) [14] is used toestimate the SoC for these nonlinear battery systems.With EKF method linearization process are used at eachtime step to approximate the nonlinear system with a

linear time-varying system [10]. EKF becomes anelegant and powerful solution to estimate the SoC.

3. Artificial Intelligence Approach to Estimate SoC

The battery pack charge and discharge processes are socomplex that it is essential to consider many factorssuch as cell voltage, current, internal impedance and

temperature gradients [16]-[18]. The battery packconnected in series presents a more complex problem.Careful monitoring and control is necessary to avoidany single cell within a Li-ion battery pack fromundergoing over-voltage or under-voltage. In a Li-ion

battery module-management system, the individualbattery SoC must be monitored because of overchargeand over-discharge issues. Therefore, it is essential tohave methods capable of estimating the battery SoC.As mentioned above, Kalman Filter (KF) is a powerfultool for the state estimation of systems. Some researchhave used this filter to estimate the open-circuit voltage

or other parameters of batteries that have a directrelationship with the SoC [19]. In [20, 21] the KF is

employed to estimate some physical quantities, whichhave direct effects on the SoC.To achieve the accurate estimation of SoC, ArtificialNeural Networks [22] (ANN) and Fuzzy Logic [23]

systems have been treated as the universalapproximators. Many techniques have been developed

to approximate the nonlinear functions for practical

applications [24, 25]. The BMF constructed in [24]possesses the property of local control and has beensuccessfully applied to Fuzzy Neural Control [26]. Also,the hybridization of fuzzy logic with ANN has been

used to improve the efficiency of function estimation.For instance, Fuzzy Neural Networks (FNN) have beenused in many applications, especially in identification of

unknown systems. The FNN can effectively model thenonlinear system by calculating the optimizedcoefficients of the learning mechanism [26-30].

The adaptive Neural Fuzzy method was proposed in[31] to estimate battery residual capacity. Although theestimation of battery residual is accurate, the algorithmutilizes the least-square method to identify the optimalvalues and hence, learning rate is computationallyexpensive; much time is wasted in training an ANN. Amore practical approach, called merged-FNN for SoCestimation is proposed in [32]. In merged-FNN, the

FNN strategy is combined with Reduced-form GeneticAlgorithm (RGA) [33] which performs effectively on

SoC estimation in a series-connected Li-ion batterystring. The merged-FNN achieved a faster learning rateand lower estimation error than the traditional ANNwith a back-propagation method.

Due to the above-mentioned facts, it is not hard to seethat AI is a promising approach for fast, precise andreliable SoC estimation.

4. Conclusions

An overview of current techniques for battery State ofCharge SoC (SoC) estimation has been given. Anintended Artificial Intelligence (AI) approach that can

be efficiently used to precisely determine the SoCestimation for a smart battery management system hasbeen presented. Our future will focus on theimplementation of a new AI technique to accuratelyestimate the SoC of various types of batteries.

5. Acknowledgment

The research work presented in this paper is partiallysponsored by SOLARI (HK) CO (www.solari-hk.com)and KATRI (www.katri.co.jp and www.katri.com.hk).

6. References

1.

Chen C., Man K.L., Ting T.O., Chi-Un Lei,Krilaviius T., Jeong T.T., Seon J.K., Sheng-UeiGuan and Prudence W.H. Wong, Design andRealization of Smart Battery Management System,Iin proceedings of the IAENG InternationalMultiConference of Engineers and Computer

Scientists - IMECS'12, Hong Kong, 2012, p.1173-1176.

2. Cheng K.W.E., Divakar B.P., Hongjie Wu, KaiDing, Ho Fai Ho. Battery-Management System(BMS) and SOC Development for ElectricalVehicles. Vehicular Technology, IEEE

Transactions on,vol.60, no.1, 2011, p. 76 883. Battery Management Systems (BMS).http://www.mpoweruk.com.

23


24/277

4. White Paper - Estimating the State Of Charge ofLi-Ion batteries.http://liionbms.com/php/wp_soc_estimate.php

5. State of Charge.

http://en.wikipedia.org/wiki/State_of_charge6. How to Measure State-of-charge.

http://batteryuniversity.com/learn/article/how_to_measure_state_of_charge

7.

Zhiwei He, MingyuGao, JieXu. EKF-Ah BasedState of Charge Online Estimation for Lithium-ion

Power Battery. Computational Intelligence andSecurity, 2009. CIS '09, 2009 p. 142 145

8. Christianson C.C., Bourke R.F. Battery state ofcharge gauge. US Patent 3,946,299, filed 11Feb.1975

9. Kalman R. A new approach to linear filtering andprediction problems, Transactions of theASMEJournal of Basic Engineering, vol. 82,

1960, p. 3545.10. Bucy R., Kalman R., Selin I. Comment on the

Kalman filter and nonlinear estimates ofmultivariate normal process. IEEE Transactions onAutomatic Control. Vol. 10, p. 118-119.

11. Duryea S., Islam S., Lawrance W. A battery

management system for stand-alone photovoltaicenergy systems. IEEE Ind. Appl. Mag., vol. 7, no.3, 2001, p. 6772.

12. Hauck B. BATTMANA Battery ManagementSystem, Milan, Italy, 1992.

13. Affanni A., Bellini A., Franceschini G., Guglielmi

P., Tassoni C. Battery choice and management fornew-generation electric vehicles. IEEE Trans. Ind.Electron., vol. 52, no. 5, 2005, p. 13431349.

14.

Extended Kalman Filters. Wikipedia.http://en.wikipedia.org/wiki/Extended_Kalman_filter

15. Cai C.H., Du D., Liu Z.Y. Battery state-of-charge(SOC) estimation using adaptive neuro-fuzzy

inference system (ANFIS). in Fuzzy Systems,2003. FUZZ '03. The 12th IEEE InternationalConference, vol.2, 2003, p. 1068-1073.

16. Caumont P., Moigne Le, Rombaut C., Muneret X.Energy gauge for lead-acid batteries in electricvehicles. IEEE Trans. Energy. Convers., vol. 15,

no. 3, 2000, p. 354-360.17. Garche, J., Jossen A. Battery management systems

(BMS) for increasing battery life time. In proc. ofthe 3rd Int. Conf. Telecommunications EnergySpecial, Dresden, Germany, 2000, p. 81-88.

18. Nagasubramanian G., Jungst R. G.. Energy and

power characteristics of tithium-ion cells. J. PowerSources, vol. 72, 1998, p. 189-193.

19. Barbarisi O., Vasca F., Glielmo L. State of chargeKalman filter estimator for automotive batteries.Control Eng. Pract., vol. 14, no. 3, 2006, p.267-275.

20. Lee J., Nam O., Cho B.H. Li-ion battery SoCestimation method based on the reduced orderextended Kalman filtering. J. Power Sources, vol.

174, no. 1, 2007, p. 9-15.21. Santhanagopalan S., White R.E. Online estimation

of the state of charge of a lithium ion cell. J. PowerSources, vol. 161, no. 2, 2006, p. 1346-1355.

22.

Hornik K., Stinchcombe M., White H. Multilayerfeedforward networks are universal approximators.

Neural Netw., vol. 2, 1989, p. 359-366.23. Wang L.X., Mendel J. M. Fuzzy basis functions,

universal approximation, and orthogonal least

squares learning. IEEE Trans. Neural Netw., vol. 3,no. 5, 1992, p. 807-814.

24. Wang C. H., Wang W.Y., Lee T.T., Tseng P.S.Fuzzy B-spline membership function (BMF) andits applications in fuzzy-neural control. IEEETrans. Syst., Man, Cybern. B, vol. 25, no. 5, 1995,

p. 841-851.25. Wang, L. X. Adaptive Fuzzy Systems and Control.

Englewood Cliffs, NJ: Prentice-Hall, 1994.26. Wang W.Y., Lee T.T., Liu C.L. Functionapproximation using fuzzy neural networks withrobust learning algorithm. IEEE Trans. Syst., Man,Cybern. B, vol. 27, no. 4, 1997, p. 740-747.

27. Karam M., Fadali M.S., White K. A Fourier /

Hopfield neural network for identification ofnonlinear periodic systems. In Proc. 35

thSoutheast.

Symp. Syst. Theory, Morgantown, NY, 2003, p.

53-57.28. Chang Q.Z., Sami Fadali, M. Nonlinear system

identification using a Gabor / Hopfield network,IEEE Trans. Syst., Man, Cybern. B, vol. 26, no. 1,

1996, p. 124-134.29. Park S.K.. Bearing estimation using Hopfield

neural network. in Proc. 22nd

Southeast. Symp.Syst. Theory, Cookville, Tn, Mar. 1990, p.

440-443.30. Kamiura N., Isokawa T., Matsui N. Learning based

on fault injection and weight restriction forfault-tolerant Hopfield neural networks. Proc.IEEE Int. Symp. Defect Fault Tolerance in VLSISyst., Cannes, France, Oct. 2004, p. 339-346.

31. Shen W.X., Chan C.C., Lo E.W.C., Chau K.T.Adaptive neuro-fuzzy modeling of battery residualfor electric vehicles. IEEE Trans. Ind. Electron.,

vol 49, 2002, p. 677-684.32. I-Hsum Li, Wei-Yen Wang, Shun-Feng Su,

Yuang-Shung Lee. A merged fuzzy neural network

and its applications in battery state-of-chargeestimation. IEEE Trans Energy Conv., vol. 22,2007, p. 689-708.

33. Wang W.Y., Li I.H. Evolutionary learning of BMFfuzzy-neural networks using a reduced-formgenetic algorithm. IEEE Trans. Syst., Man,

Cybern. B, Cybern., vol. 33. no. 6, 2003, p.966-976.

24


25/277

CLOSED-LOOP OPTIMIZATION ALGORITHMS IN

DIGITAL SELF-TUNING PID CONTROL OF PRESSURE PROCESS

Gediminas LIAUIUS, Vytautas KAMINSKASDepartment of Systems Analysis, Vytautas Magnus University, Lithuania

Abstract:At the design stage of digital self-tuning PID

control, the determination of closed-loop parametersand continuous-time sampling period is an issue. Thestated problem is proposed to solve by optimizing thegeneralized criterion of control quality in this paper.The methodology of sub-component search has beenused to solve optimization problem. The effectivenessof search algorithms has been investigatedexperimentally.

Keywords: pressure process, self-tuning PID control,closed-loop parameters, sampling period, optimization.

1. Introduction

The control quality of continuous-time processes bydigital controllers depends on the right choice of closed-loop characteristics and continuous-time samplingperiod [1]. Moreover, at the design stage of digital self-tuning PID control, the determination of thoseparameters is not straightforward. In the present paper,the stated problem is proposed to solve by optimizingthe generalized criterion of control quality via sub-component search methodology. The effectiveness ofsearch algorithms has been experimentally analysed onpressure process.

2. Digital self-tuning PID control

The scheme of pressure process is illustrated in Fig. 1.It consists of such components: the combined air inlet(no. 1) and air outlet (no. 4) tanks, two air chambers(no. 2) and two tubes (no. 3) with balls (no. 6) in them.The air from the inlet tank flows to air channels throughair chambers and leaves the equipment through theupper outlet tank. The distance to balls is measuredusing ultrasound distance sensors (no. 5). The fans(no. 7) are used to create pressure in the air channels inorder to lift the balls in tubes. The air chambers are

utilized for the purpose to stabilize oscillations of thepressure in each tube. The momentum of the fan, theinductance of the fan motor and an air turbulence in the

tube leads to complex physics governing ball's

behaviour. Slightly different weights of the balls and thelocation of air feeding vent additionally impacts thebehaviour of ball in the tubes.

Fig. 1.The scheme of pressure process

The input signals of the process are voltage values foreach fan from the range between 0 and 10 volts. Theintermediate values of voltage effects the power of fandisproportionately. The output

Documents

Proceedings ECT 2012