Research Article Optimization and Coordination of HAFDV ...downloads.hindawi.com/journals/jcse/2013/872624.pdfMotor Motor Motor Motor Motor Motor Motor Motor ... Particle swarm optimization,

Hindawi Publishing CorporationJournal of Control Science and EngineeringVolume 2013 Article ID 872624 7 pageshttpdxdoiorg1011552013872624

Research ArticleOptimization and Coordination of HAFDV PINN Control byImproved PSO

Bin Huang Nengling Tai and Wentao Huang

School of Electronic Information and Electrical Engineering Shanghai Jiao Tong University Shanghai 200240 China

Correspondence should be addressed to Bin Huang hbsjtugmailcom

Received 8 June 2012 Accepted 25 January 2013

Academic Editor Pierluigi Siano

Copyright copy 2013 Bin Huang et al This is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

The new hybrid active filter (HAF) is composed of the larger-capacity passive filter banks and the smaller-capacity active filter It isdifficult to tune the parameters of a PI controller using the DC capacitor voltage control In this paper the improved particle swarmoptimization (improved PSO) algorithm is proposed to solve the coordinated design problem and the neural network weights asthe particle swarm optimization are adopted to optimize the system parameters Comparing with the conventional PI controllerthe results of PINN controller prove the effectiveness of designed method on both the transient and steady-state performance ofthe hybrid active filter DC bus voltage (HAFDV) controllers

1 Introduction

Microgrid an entity of various distributed energy resources(DERs) with different characteristics and mutual comple-mentation can operate in phase withmacrogrid and improveenergy utilization efficiency and electricity supply reliabilityDue to the applications of nonlinear elements the harmonicinterference on the micronetwork is growing In order toensure the safety of the macrogrid we use the hybrid activefilter to remove the harmonics of the micronetwork

Hybrid active filter which has dynamic better compen-sation characteristics and a greater compensation capacity isused to suppress harmonics and reactive power compensa-tion of the macrogrid Reference [1] studied the capacitanceand voltage of the active filter it is not suitable for alarger-capacity macrogrid References [2ndash4] designed the PIcontroller of the DC bus voltage without optimizing the con-trollers Therefore improved PSO optimizes the parametersof HPF controller using PI neural network instead of theconventional PI The time-domain results demonstrate theeffectiveness of the proposed design method of the HAF DCbus voltage controllers

2 Problem Formulations

In this section the mathematical model of HPF is presentedin detail and then the HAFDV controllers will be consideredas an example to verify the effects of this method

21 Modeling of HAF The structure of the proposed hybridactive power filter (HAPF) is shown in Figure 1 It is com-posed of a smaller-scale three-phase voltage source inverter(VSI) employing pulse-width modulation (PWM) with alarge capacitor at the DC side and an output filter to eliminatethe high-frequency ripples at the output side A couplingtransformer connects VSI in series with a group of passivepower filters And all of them are connected in shunt withMicrogrid

22 Formulations of the HAF DC Bus Voltage ControllersThe main objective of the control scheme is to force all theharmonics of load current to flow into passive filter [5ndash8] Ahysteresis current controller is used to switch the bridge armof the active filter [9ndash11]

2 Journal of Control Science and Engineering

AC

Active filterOutput filter

Coupling transformer

PF

119885119904

+

minus

Microgrid119868119904

119880dc

Figure 1 Configuration of the hybrid active power filter

The control scheme is indicated in Figure 2 It builds thereference currents for the HAF based on the instantaneousactive and reactive power theory The synchronous referenceframe (SRF) theory is used to extract the fundamentalcomponent of the source current by transformation of thesupply currents (119894

119906119894119907119894119908) into the reference frameThe system

under study is a three-wire system where the zero sequencemay be neglectedThe following are basic equations for thesetransformations

[119894119901

119894119902

] = radic2

3[sin120596119905 minus cos120596119905

minus cos120596119905 minus sin120596119905]

times[[[

[

1 minus1

2minus1

2

0radic3

2minusradic3

2

]]]

]

[

[

119894119906

119894119907

119894119908

]

]

(1)

where the two-phase sinusoidal signals sin120596119905 and cos120596119905 areresulted from the phase-locked loop (PLL) circuit

The lowpass filter is used to extract the DC componentsrealized by moving average to 80Hz The extracted DCcomponents are transformed back into a-b-c coordinates toobtain the fundamental components (119894

119906119891119894119907119891119894119908119891) of source

currents as shown in

[119894119901119895

119894119902119895

] = [sin120596119905 minus cos120596119905

minus cos120596119905 minus sin120596119905] [

119894119901

119894119902

]

[

[

119894119906119891

119894119907119891

119894119908119891

]

]

= radic2

3

[[[[[[[

[

1 0

minus1

2

radic3

2

minus1

2minusradic3

2

]]]]]]]

]

[119894119901119895

119894119902119895

]

(2)

So each current harmonic component (119894119906ℎ119894119907ℎ119894119908ℎ) is given

by

[

[

119894119906ℎ

119894119907ℎ

119894119908ℎ

]

]

= [

[

119894119906

119894119907

119894119908

]

]

minus [

[

119894119906119891

119894119907119891

119894119908119891

]

]

(3)

The source current is controlled to follow this referencecurrent by switching the voltage source inverter with ahysteresis current controller

23 The Objective Function Conventional PI controller isused to maintain the DC bus voltage of the HAF by compen-sating the voltage losses [12ndash14] Conventional PI controllerparameters are often obtained by several experiments It isdifficult to adjust all the parameters This paper designs PIneural network controller combined with improved PSOThe interlayer and output layer linking weights of PI neuralnetwork (PINN) are used as the optical parameters of PSO

According to the formulae (1)ndash(3) A-phase fundamentalcurrent increment values are obtained bys

Δ119894119886= radic

2

3sin120596119905 sdot Δ119894pf (4)

where Δ119894pf is the increment of the grid active currentcomponent

So the exchange active power with HAF and supplysystem is calculated as

119875119891

= radic3119864119886sdot Δ119894119886

int

Δ119905

0

119875119891119889119905 =

119862dc [(119880dc + Δ119880119862dc

)2

minus 119880dc2

]

2

+ int

Δ119905

0

119870119894ca119889119905

(5)

where119864119886is A-phase line voltageRMS119862dc is a capacitor value

119880dc is the ideal steady-state value of the HAF DC voltageΔ119880119862dc

is the increment of capacitor voltage during Δ119905 119870 isa constant with switching characteristics of capacitor and the119894ca is RMS in A-phase of 119894

119888

According to the formulae (4)-(5) with neglecting theitem (Δ119880

119862dc)2 and linearization near the operating point after

Laplace transform the PI regulator 119866pi(119904) controlled object119866119907(119904) and voltage detection 119866

119891(119904) are given by

119866pi (119904) =

119870119901(119879119894+ 1)

119879119894119904

119866119907(119904) =

119870119894119864119886

119862dc119880dc119904

119866119891(119904) =

119870119891

119879119891119904 + 1

(6)

where 119870119875is a proportional coefficient 119879

119894is an integral

constantFigure 3 shows the system transfer function it is

expressed as

119866 (119904) =

119866pi (119904) 119866119907 (119904)

1 + 119866pi (119904) 119866119907 (119904) 119866119891 (119904) (7)

The disadvantage of conventional PI controller lies inthe difficulty to resolve the trade-off among smoothness

Journal of Control Science and Engineering 3

LPF

Three

TwoThree

Two

transform inversetransform

PI

PLL

119894119906

119894119907

119894119908

119894119886

119894119887

D-q D-q

119894119889

119894119902

119906119907

119894119901119891

119894119902119891

+

+

minus

minusΔ

119889

119902119889

119902

119894119901119895

119894119902119895

119894119906119891

119894119907119891

119894119908119891

minus

minus

minus +

+

+119894119906ℎ

119894119907ℎ

119894119908ℎ

sin 120596119905 and cos 120596119905generators

minus1

lowast

119880dc

119880dc119880dc

Figure 2 The control scheme of HAF

PINN

Δ119880

minus119866119901119894(119904) 119866119907(119904)

119866119891(119904)

119880lowastdc119880dc

Figure 3 System transfer function

Controlled object

119879PINN

119870PINN

119903

119910

119910

1199061 1199091

1199062 1199092

11987111

11987112

11987121

11987122

119907119906998400998400 119909998400998400

11990619984001199091

998400

11990629984001199092

998400

1198711998400

1198712998400

Figure 4 The structure of PINN

fastness and accuracyThemajor parameters of the HAF DCbus voltage controller require experiments and high skills insystem debugging process and will be fixed eventually

A PINN controller shown in the dotted line in Figure 3is adopted as the DC bus voltage controller instead of theconventional PI controller so the system transfer function ismodified into

119866 (119904) =

PINN (119878)119870119894119864119886119879119891119904 + PINN (119878)119870

119894119864119886

119862dc119880dc1198791198911199042 + 119862dc119880dc + PINN (119878)119870

119894119864119886119870119891

(8)

Initial improved PSO and PINN parameters

Generate position and velocity and calculate adaptive value

Start

Evaluate the fitness and calculate the new adaptive value

Update position and velocity

Yes

Get the PINN structure

No

PINN-BP learning and predicting

Find the local optimum position 119875119894119889 andglobe optimum position 119892119894119889

Achieving condition

Figure 5 The flow chart of improved PSO

The improved PSO algorithm is applied to the optimiza-tion of PINN parameters An objective optimization problemcan be defined as follows

119883PINN = NN(119870PINN 119879PINN)119883PINN are neural networkvectors representing the decision variables 119891

119899(119883PINN) is


G G

M M

G

GM M

G G

M M

G G

M M

M M M M M M M M

Gen Gen GenG

Gen Gen Gen Gen Gen

Motor Motor Motor Motor Motor Motor MotorMotor

MotorMotorMotorMotorMotorMotorMotorMotor

Transformer Transformer

11 kV 11 kV 11 kV 11 kVMV1 MV2 MV3 MV4

400 V LV1 400 V LV2 400 V LV3 400 V LV4

400 V ESB1

Emergency gen

Figure 6 Microgrid system single line figure

15

1

05

0

minus05

minus1

minus150 01 02 03 04 05 06 07 08 09 1

Curr

ent (

)

Time (s)

(a) The current of one phase at the 11 kV level

6

5

4

3

2

1

0

Curr

ent s

pect

rum

()

0 5 10 15 20 25 30 35 40 45 50Harmonic order

(b) The spectrum of one phase at the 11 kV level

Figure 7 The current and spectrum of one phase at the 11 kV level

the objective function 119892119895(119870PINN) and ℎ

119896(119879PINN) represent

the constraints

3 Improved PSO

Particle swarm optimization inspired by the social behaviorof a bird flock was originally designed by Kennedy andEberhart in 1995 The original idea is to simulate the birdsrsquofood-hunting for the global optimal point One particlepresents a bird and it can calculate the adaptive value of itscurrent position and every particle records the optimal valuesearched by itself [15 16]

The particles have their memory and each particle keepstracking of its local best position (119901id) and its correspondingfitness The particle with the greatest fitness is called theglobal best position (119892id) of the swarm The basic PSO

algorithm has many defects such as falling into local opti-mum and being impacted by the fact that it is insensitiveto environment variables precocious and so on The paperadopted improved PSO which has multigroup particles tosearch different parts of the solution space When 119901id and119892id are obtained a particle updates its velocity and positionbased on (9) In the end the algorithm will check the resultsuntil the best solution is found or termination conditions aresatisfied

119907119896+1

id = 120582 times 119907119896

id + 1205731times random () times (119901

119896

id minus 119909119896

id)

+ 1205732times random () times (119892

119896

id minus 119909119896

id)

119909119896+1

id = 119909119896

id + 119907119896+1

id

(9)

where 1205731 1205732

ge 0 120582 is the inertia weight factor 1205731and 120573

2

are acceleration constants random() is a random number


1

08

06

04

02

020 40 60 80 100 120 140 160 180 200

Simple (119899)

119910(119905)

PINN

Figure 8 Improved PSO optimizes PINN

700600500400300200100

0

119890(119905)

0 5 10 15 20 25 30 35 40 45 50Step (119899)

Figure 9 Convergence curve of adaptive value

between 0 and 1 119907id and 119909id are the velocity and the currentposition of particle at iteration id

The improved PSO algorithmoptimizes PINNcontrollerswhich is defined as proportional integral function of neu-rons PINN is a dynamic multilayer feed forward neural net-work The parameters of PINN (119870PINN119879PINN) are optimizedby improved PSO

Neural network of PINN applies 2 times 2 times 1 structureshown in Figure 4 The input layer of the PINN has twoneurons of proportion The two neurons (119870PINN119879PINN) inthemiddle layer are proportional and integral elements of theinput signal and the weights from input layer to the middlelayer are remained of constant value (minus1 +1) In order tominimize the objective function value the mean square error(MSE) function has been employed in (10) The improvedPSO adjusts the network weights by the back-propagation(BP) algorithm [17] The flow chart of the improved PSO isshown in Figure 5 Consider the following

119869 =1

119898

119899

sum

ℎ=1

[119903 (119896) minus 119910 (119896)]2

(10)

The improved PSO shows many advantages Firstly itinitializes the improved PSO and PINN parameters whichcontain the initial particles their positions and velocities Foreach particle it evaluates the fitness function [18] Secondly itgenerates position and velocity and calculates adaptive valueconsidering the steady error and setting time It calculatesthe value of every particle in all subgroups Besides it cantrain PINN structure with current positions of the particlesby BP learning and predicting In addition improved PSO

Table 1 Harmonic of microgrid system

Harmonic order A percentage of content ()5 1224717 08434311 58320313 45484223 21118625 21000735 13269937 13059947 11776349 114342All 849

can find the local optimum position 119901id and globe optimumposition 119892id through selecting the particles with minimal 119869Furthermore it evaluates and calculates the fitness and thenew adaptive value by using (9) What is more is that it canupdate the velocity and the position of each particle Thenew position is kept if the current position is dominated bypositioning the 119901id otherwise the current position replacesit in space neither of them is randomly selected Finally itcan judge the termination criterion the procedure goes to thethird step until it is satisfied

4 Results

The proposed island Microgrid system is shown in Figure 6and the primary parameters of system are as follows Thesystem comprises four 11 kV medium-voltage buses with two4000 kW generator engines and two 3000 kWmotors drivenby 12-pulse inverters linking each other through the circuitbreakers and four 400V low-voltage buses with some motorsand loads connected by the circuit breaker In addition a1500 kW of emergency engine is installed on the low-voltageside of power supply in case of emergency

The harmonic parameters of the proposed island Micro-grid system are summarized in Table 1 while the total THD

119894

is 849 it exceeds the national standard of China Figure 7shows the current and spectrum of one phase at the 11 kVlevel For the security of the grid HAF is used between thebus and inverter


12

1

08

06

04

02

0

minus020 005 01 015 02 025 03 035

Time (s)

119880dc

(pu)

(a) DC bus voltage control with PINN

25

2

15

1

05

00 005 01 015 02 025

Time (s)

119880dc

(pu)

(b) DC bus voltage control with conventional PI

Figure 10 Compare PINN with conventional PI

150

100

50

0

minus50

minus100

minus1500 01 02 03 04 05 06 07 08 09 1

Curr

ent (

)

Time (s)

(a) The current of one phase at the 11 kV level with HAF

10

9876543210

Curr

ent s

pect

rum

()

0 5 10 15 20 25 30 35 40 45 50Harmonic order

(b) The spectrum of one phase at the 11 kV level with HAF

Figure 11 The current and spectrum of one phase at the 11 kV level with HAF

Table 2 Parameters of PF

The PF of HAF 119871ohm 119862120583F 119876

11th 19185 1371 4013th 11301 1667 40

Theperformance of theHAFdepends on theDCbus volt-age control optimization shown in Figure 1 The simulationparameters are total power 119875all = 32MW a single motorpower factor cos120593 = 909 119862

119880cd= 3000 120583F out filter 119871out =

05mH 119862out = 500 120583F and 119880dc = 6000V The parameters ofimproved PSO are set to be max119881id = 20 min119881id = minus20 thepopulation size is 100 the iteration times are 50 the inertiaweight factor is 08 the acceleration constants 120573

1and 120573

2both

are 3 random() is a random number between 0 and 1 and thesample and saturation of BP are 200 and 2 respectively So theoptimum weights of PINN which are trained by improved

PSO are 830376 and 000283987 The parameters of PF aresummarized in Table 2

The step response of 119880lowastdc(PU) jumps from 07 to 10 andthe 119880dc tracks the given value with PINN by improved PSOoptimization as shown in Figure 8 Training PINN to 26 stepsthe error 119890(119905) of algorithm is 14169119890

minus005 shown in Figure 9The optimization result for the DC bus voltage control ofPINN is effective

The HAF with DC bus voltage control by PINN isinstalled in the island microgrid system It can be seen fromFigure 10 that PINN has advantages over the conventionalPI and that the filtering functions are well performed bythe DC bus voltage controller under the different operatingconditions The maximum overshoot of the traditional PI isclose to 19 compared with 19 the maximum overshoot of119880dc with PINN is lowered to about 095 when 119880

lowast

dc is 07 TheDC bus voltage with PINN is stable at 013 s which is thelower than conventional PI and the response is quicker andsmoother than the conventional PI


Table 3 Harmonic of microgrid system with HAF

Harmonic order A percentage of content ()5 03007797 011124111 0049815213 0019656523 00076863325 00046405535 027077737 0279686All 096

By using the HAF with DC bus voltage PINN control thetotal THD

119894of the island Microgrid system is reduced from

849 to 096 Figure 11 presents the results of the smoothcurrent and reduced harmonic All the harmonics are thatlower than the national standards are summarized in Table 3

5 Conclusions

Due to the harmonic interference of nonlinear loads theHAF is used to protect the security of the island Microgridsystem An optimizationmethod based on improved PSO fordesign ofHAFDVcontrol is developed by usingPINN insteadof conventional PI The coordinated design problem of DCbus voltage control is formulated as a nonlinear constrainedobjective optimization problem where improved PSO isemployed to search for the optimal solutions Comparingcontrollers of PINN and conventional PI simulation resultsshow that HAFDV control with PINN has more effectivecontrol results better stability and lower overshoot of DCvoltage and more rapid response than conventional PI

References

[1] Z Ke L An X Xiang-yang and Z Wei ldquoDC side capacitorrsquosdesign and voltage control in high-capacity active power filterrdquoPower Electronics vol 4 no 41 2007

[2] GMing zhen R Zhen T Zhuo yao and L Q zhan ldquoOptimiza-tion design for capacity on passive active hybrid power filtersrdquoJournal of the China Railway Society vol 5 no 21 pp 43ndash461999

[3] L Shiguo ldquoOptimal design of DC voltage close loop control foran active power filterrdquo inProceedings of International ConferenceonPower Electronics andDrive Systems vol 2 pp 565ndash570 1995

[4] Z Dong L Zheng-yu and C Guo-zhu ldquoCapacitor voltagecontrol of shunt active power filterrdquo Power Electronics vol 10no 41 2007

[5] H Akagi and T Hatada ldquoVoltage balancing control for a three-level diode-clamped converter in a medium-voltage trans-formerless hybrid active filterrdquo IEEE Transactions on PowerElectronics vol 24 no 3 pp 571ndash579 2009

[6] V F Corasaniti M B Barbieri P L Arnera and M I VallaldquoHybrid active filter for reactive and harmonics compensationin a distribution networkrdquo IEEE Transactions on IndustrialElectronics vol 56 no 3 pp 670ndash677 2009

[7] H Akagi and R Kondo ldquoA transformerless hybrid active filterusing a three-level Pulsewidth Modulation (PWM) converterfor amedium-voltagemotor driverdquo IEEE Transactions on PowerElectronics vol 25 no 6 pp 1365ndash1374 2010

[8] A Luo W Zhu R Fan and K Zhou ldquoStudy on a novelhybrid active power filter applied to a high-voltage gridrdquo IEEETransactions on Power Delivery vol 24 no 4 pp 2344ndash23522009

[9] P T Cheng S Bhattacharya and D M Divan ldquoControlof square-wave inverters in high-power hybrid active filtersystemsrdquo IEEE Transactions on Industry Applications vol 34no 3 pp 458ndash472 1998

[10] R Inzunza and H Akagi ldquoA 66-kV transformerless shunthybrid active filter for installation on a power distributionsystemrdquo in Proceedings of the 35th Annual Power ElectronicsSpecialists Conference (PESCrsquo04) pp 4630ndash4636 June 2004

[11] A Luo Z Shuai W Zhu Z J Shen and C Tu ldquoDesign andapplication of a hybrid active power filter with injection circuitrdquoIET Power Electronics vol 3 no 1 pp 54ndash64 2010

[12] V F Corasaniti M B Barbieri P L Arnera and M I VallaldquoHybrid power filter to enhance power quality in a medium-voltage distribution networkrdquo IEEE Transactions on IndustrialElectronics vol 56 no 8 pp 2885ndash2893 2009

[13] J Dixon DC Link ldquoFuzzy control for an active power filtersensing the line current onlyrdquo in Proceedings of the AnnualPower Electronics Specialists Conference (PESC rsquo97) vol 2 pp1109ndash1114 1997

[14] N Hatti K Hasegawa andH Akagi ldquoA 66-kV transformerlessmotor drive using a five-level diode-clamped PWM inverter forenergy savings of pumps and blowersrdquo IEEE Transactions onPower Electronics vol 24 no 3 pp 796ndash803 2009

[15] I M Supratid ldquoA multi-subpopulation particle swarm opti-mization a hybrid intelligent computing for function opti-mizationrdquo in Proceedings of the 3rd International Conference onNatural Computation (ICNC rsquo07) vol 5 pp 679ndash684 August2007

[16] C Fan and Y Wan ldquoAn adaptive simple particle swarmoptimization algorithmrdquo in Proceedinhs of the Chinese Controland Decision Conference (CCDC rsquo08) pp 3067ndash3072 July 2008

[17] N Dongxiao G Zhihong and X Mian ldquoResearch on neuralnetworks based on culture particle swarm optimization and itsapplication in power load forecastingrdquo in Proceedings of the 3rdInternational Conference on Natural Computation (ICNC rsquo07)pp 270ndash274 August 2007

[18] L Rui G Yirong X Yujuan and L Ming ldquoA novel multi-swarm particle swarm optimization algorithm applied in activecontour modelrdquo in Proceedings of the WRI Global Congress onIntelligent Systems (GCIS rsquo09) pp 139ndash143 May 2009

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014


Active and Passive Electronic Components

Control Scienceand Engineering

Journal of



RotatingMachinery


Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design



Shock and Vibration


Civil EngineeringAdvances in

Acoustics and VibrationAdvances in



Electrical and Computer Engineering

Journal of

Advances inOptoElectronics


Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of


Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014


Chemical EngineeringInternational Journal of Antennas and

Propagation




Navigation and Observation



DistributedSensor Networks



AC

Active filterOutput filter

Coupling transformer

PF

119885119904

+

minus

Microgrid119868119904

119880dc

Figure 1 Configuration of the hybrid active power filter

The control scheme is indicated in Figure 2 It builds thereference currents for the HAF based on the instantaneousactive and reactive power theory The synchronous referenceframe (SRF) theory is used to extract the fundamentalcomponent of the source current by transformation of thesupply currents (119894

119906119894119907119894119908) into the reference frameThe system

under study is a three-wire system where the zero sequencemay be neglectedThe following are basic equations for thesetransformations

[119894119901

119894119902

] = radic2

3[sin120596119905 minus cos120596119905

minus cos120596119905 minus sin120596119905]

times[[[

[

1 minus1

2minus1

2

0radic3

2minusradic3

2

]]]

]

[

[

119894119906

119894119907

119894119908

]

]

(1)

where the two-phase sinusoidal signals sin120596119905 and cos120596119905 areresulted from the phase-locked loop (PLL) circuit

The lowpass filter is used to extract the DC componentsrealized by moving average to 80Hz The extracted DCcomponents are transformed back into a-b-c coordinates toobtain the fundamental components (119894

119906119891119894119907119891119894119908119891) of source

currents as shown in

[119894119901119895

119894119902119895

] = [sin120596119905 minus cos120596119905

minus cos120596119905 minus sin120596119905] [

119894119901

119894119902

]

[

[

119894119906119891

119894119907119891

119894119908119891

]

]

= radic2

3

[[[[[[[

[

1 0

minus1

2

radic3

2

minus1

2minusradic3

2

]]]]]]]

]

[119894119901119895

119894119902119895

]

(2)

So each current harmonic component (119894119906ℎ119894119907ℎ119894119908ℎ) is given

by

[

[

119894119906ℎ

119894119907ℎ

119894119908ℎ

]

]

= [

[

119894119906

119894119907

119894119908

]

]

minus [

[

119894119906119891

119894119907119891

119894119908119891

]

]

(3)

The source current is controlled to follow this referencecurrent by switching the voltage source inverter with ahysteresis current controller

23 The Objective Function Conventional PI controller isused to maintain the DC bus voltage of the HAF by compen-sating the voltage losses [12ndash14] Conventional PI controllerparameters are often obtained by several experiments It isdifficult to adjust all the parameters This paper designs PIneural network controller combined with improved PSOThe interlayer and output layer linking weights of PI neuralnetwork (PINN) are used as the optical parameters of PSO

According to the formulae (1)ndash(3) A-phase fundamentalcurrent increment values are obtained bys

Δ119894119886= radic

2

3sin120596119905 sdot Δ119894pf (4)

where Δ119894pf is the increment of the grid active currentcomponent

So the exchange active power with HAF and supplysystem is calculated as

119875119891

= radic3119864119886sdot Δ119894119886

int

Δ119905

0

119875119891119889119905 =

119862dc [(119880dc + Δ119880119862dc

)2

minus 119880dc2

]

2

+ int

Δ119905

0

119870119894ca119889119905

(5)

where119864119886is A-phase line voltageRMS119862dc is a capacitor value

119880dc is the ideal steady-state value of the HAF DC voltageΔ119880119862dc

is the increment of capacitor voltage during Δ119905 119870 isa constant with switching characteristics of capacitor and the119894ca is RMS in A-phase of 119894

119888

According to the formulae (4)-(5) with neglecting theitem (Δ119880

119862dc)2 and linearization near the operating point after

Laplace transform the PI regulator 119866pi(119904) controlled object119866119907(119904) and voltage detection 119866

119891(119904) are given by

119866pi (119904) =

119870119901(119879119894+ 1)

119879119894119904

119866119907(119904) =

119870119894119864119886

119862dc119880dc119904

119866119891(119904) =

119870119891

119879119891119904 + 1

(6)

where 119870119875is a proportional coefficient 119879

119894is an integral

constantFigure 3 shows the system transfer function it is

expressed as

119866 (119904) =

119866pi (119904) 119866119907 (119904)

1 + 119866pi (119904) 119866119907 (119904) 119866119891 (119904) (7)

The disadvantage of conventional PI controller lies inthe difficulty to resolve the trade-off among smoothness


LPF

Three

TwoThree

Two


PI

PLL

119894119906

119894119907

119894119908

119894119886

119894119887

D-q D-q

119894119889

119894119902

119906119907

119894119901119891

119894119902119891

+

+

minus

minusΔ

119889

119902119889

119902

119894119901119895

119894119902119895

119894119906119891

119894119907119891

119894119908119891

minus

minus

minus +

+

+119894119906ℎ

119894119907ℎ

119894119908ℎ


minus1

lowast

119880dc

119880dc119880dc


PINN

Δ119880

minus119866119901119894(119904) 119866119907(119904)

119866119891(119904)



Controlled object

119879PINN

119870PINN

119903

119910

119910

1199061 1199091

1199062 1199092

11987111

11987112

11987121

11987122

119907119906998400998400 119909998400998400

11990619984001199091

998400

11990629984001199092

998400

1198711998400

1198712998400




119866 (119904) =

PINN (119878)119870119894119864119886119879119891119904 + PINN (119878)119870

119894119864119886

119862dc119880dc1198791198911199042 + 119862dc119880dc + PINN (119878)119870

119894119864119886119870119891

(8)



Start



Yes


No



Achieving condition




119899(119883PINN) is


G G

M M

G

GM M

G G

M M

G G

M M

M M M M M M M M

Gen Gen GenG

Gen Gen Gen Gen Gen





400 V LV1 400 V LV2 400 V LV3 400 V LV4

400 V ESB1

Emergency gen


15

1

05

0

minus05

minus1

minus150 01 02 03 04 05 06 07 08 09 1

Curr

ent (

)

Time (s)


6

5

4

3

2

1

0

Curr

ent s

pect

rum

()

0 5 10 15 20 25 30 35 40 45 50Harmonic order





the constraints

3 Improved PSO




119907119896+1

id = 120582 times 119907119896


119896

id minus 119909119896

id)


119896

id minus 119909119896

id)

119909119896+1

id = 119909119896

id + 119907119896+1

id

(9)

where 1205731 1205732


2



1

08

06

04

02

020 40 60 80 100 120 140 160 180 200

Simple (119899)

119910(119905)

PINN


700600500400300200100

0

119890(119905)

0 5 10 15 20 25 30 35 40 45 50Step (119899)





119869 =1

119898

119899

sum

ℎ=1

[119903 (119896) minus 119910 (119896)]2

(10)





4 Results



119894



12

1

08

06

04

02

0

minus020 005 01 015 02 025 03 035

Time (s)

119880dc

(pu)


25

2

15

1

05

00 005 01 015 02 025

Time (s)

119880dc

(pu)



150

100

50

0

minus50

minus100

minus1500 01 02 03 04 05 06 07 08 09 1

Curr

ent (

)

Time (s)


10

9876543210

Curr

ent s

pect

rum

()

0 5 10 15 20 25 30 35 40 45 50Harmonic order




The PF of HAF 119871ohm 119862120583F 119876

11th 19185 1371 4013th 11301 1667 40




1and 120573

2both






lowast








5 Conclusions


References





















RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










LPF

Three

TwoThree

Two


PI

PLL

119894119906

119894119907

119894119908

119894119886

119894119887

D-q D-q

119894119889

119894119902

119906119907

119894119901119891

119894119902119891

+

+

minus

minusΔ

119889

119902119889

119902

119894119901119895

119894119902119895

119894119906119891

119894119907119891

119894119908119891

minus

minus

minus +

+

+119894119906ℎ

119894119907ℎ

119894119908ℎ


minus1

lowast

119880dc

119880dc119880dc


PINN

Δ119880

minus119866119901119894(119904) 119866119907(119904)

119866119891(119904)



Controlled object

119879PINN

119870PINN

119903

119910

119910

1199061 1199091

1199062 1199092

11987111

11987112

11987121

11987122

119907119906998400998400 119909998400998400

11990619984001199091

998400

11990629984001199092

998400

1198711998400

1198712998400




119866 (119904) =

PINN (119878)119870119894119864119886119879119891119904 + PINN (119878)119870

119894119864119886

119862dc119880dc1198791198911199042 + 119862dc119880dc + PINN (119878)119870

119894119864119886119870119891

(8)



Start



Yes


No



Achieving condition




119899(119883PINN) is


G G

M M

G

GM M

G G

M M

G G

M M

M M M M M M M M

Gen Gen GenG

Gen Gen Gen Gen Gen





400 V LV1 400 V LV2 400 V LV3 400 V LV4

400 V ESB1

Emergency gen


15

1

05

0

minus05

minus1

minus150 01 02 03 04 05 06 07 08 09 1

Curr

ent (

)

Time (s)


6

5

4

3

2

1

0

Curr

ent s

pect

rum

()

0 5 10 15 20 25 30 35 40 45 50Harmonic order





the constraints

3 Improved PSO




119907119896+1

id = 120582 times 119907119896


119896

id minus 119909119896

id)


119896

id minus 119909119896

id)

119909119896+1

id = 119909119896

id + 119907119896+1

id

(9)

where 1205731 1205732


2



1

08

06

04

02

020 40 60 80 100 120 140 160 180 200

Simple (119899)

119910(119905)

PINN


700600500400300200100

0

119890(119905)

0 5 10 15 20 25 30 35 40 45 50Step (119899)





119869 =1

119898

119899

sum

ℎ=1

[119903 (119896) minus 119910 (119896)]2

(10)





4 Results



119894



12

1

08

06

04

02

0

minus020 005 01 015 02 025 03 035

Time (s)

119880dc

(pu)


25

2

15

1

05

00 005 01 015 02 025

Time (s)

119880dc

(pu)



150

100

50

0

minus50

minus100

minus1500 01 02 03 04 05 06 07 08 09 1

Curr

ent (

)

Time (s)


10

9876543210

Curr

ent s

pect

rum

()

0 5 10 15 20 25 30 35 40 45 50Harmonic order




The PF of HAF 119871ohm 119862120583F 119876

11th 19185 1371 4013th 11301 1667 40




1and 120573

2both






lowast








5 Conclusions


References





















RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










G G

M M

G

GM M

G G

M M

G G

M M

M M M M M M M M

Gen Gen GenG

Gen Gen Gen Gen Gen





400 V LV1 400 V LV2 400 V LV3 400 V LV4

400 V ESB1

Emergency gen


15

1

05

0

minus05

minus1

minus150 01 02 03 04 05 06 07 08 09 1

Curr

ent (

)

Time (s)


6

5

4

3

2

1

0

Curr

ent s

pect

rum

()

0 5 10 15 20 25 30 35 40 45 50Harmonic order





the constraints

3 Improved PSO




119907119896+1

id = 120582 times 119907119896


119896

id minus 119909119896

id)


119896

id minus 119909119896

id)

119909119896+1

id = 119909119896

id + 119907119896+1

id

(9)

where 1205731 1205732


2



1

08

06

04

02

020 40 60 80 100 120 140 160 180 200

Simple (119899)

119910(119905)

PINN


700600500400300200100

0

119890(119905)

0 5 10 15 20 25 30 35 40 45 50Step (119899)





119869 =1

119898

119899

sum

ℎ=1

[119903 (119896) minus 119910 (119896)]2

(10)





4 Results



119894



12

1

08

06

04

02

0

minus020 005 01 015 02 025 03 035

Time (s)

119880dc

(pu)


25

2

15

1

05

00 005 01 015 02 025

Time (s)

119880dc

(pu)



150

100

50

0

minus50

minus100

minus1500 01 02 03 04 05 06 07 08 09 1

Curr

ent (

)

Time (s)


10

9876543210

Curr

ent s

pect

rum

()

0 5 10 15 20 25 30 35 40 45 50Harmonic order




The PF of HAF 119871ohm 119862120583F 119876

11th 19185 1371 4013th 11301 1667 40




1and 120573

2both






lowast








5 Conclusions


References





















RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










1

08

06

04

02

020 40 60 80 100 120 140 160 180 200

Simple (119899)

119910(119905)

PINN


700600500400300200100

0

119890(119905)

0 5 10 15 20 25 30 35 40 45 50Step (119899)





119869 =1

119898

119899

sum

ℎ=1

[119903 (119896) minus 119910 (119896)]2

(10)





4 Results



119894



12

1

08

06

04

02

0

minus020 005 01 015 02 025 03 035

Time (s)

119880dc

(pu)


25

2

15

1

05

00 005 01 015 02 025

Time (s)

119880dc

(pu)



150

100

50

0

minus50

minus100

minus1500 01 02 03 04 05 06 07 08 09 1

Curr

ent (

)

Time (s)


10

9876543210

Curr

ent s

pect

rum

()

0 5 10 15 20 25 30 35 40 45 50Harmonic order




The PF of HAF 119871ohm 119862120583F 119876

11th 19185 1371 4013th 11301 1667 40




1and 120573

2both






lowast








5 Conclusions


References





















RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










12

1

08

06

04

02

0

minus020 005 01 015 02 025 03 035

Time (s)

119880dc

(pu)


25

2

15

1

05

00 005 01 015 02 025

Time (s)

119880dc

(pu)



150

100

50

0

minus50

minus100

minus1500 01 02 03 04 05 06 07 08 09 1

Curr

ent (

)

Time (s)


10

9876543210

Curr

ent s

pect

rum

()

0 5 10 15 20 25 30 35 40 45 50Harmonic order




The PF of HAF 119871ohm 119862120583F 119876

11th 19185 1371 4013th 11301 1667 40




1and 120573

2both






lowast








5 Conclusions


References





















RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation















5 Conclusions


References





















RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation









Research Article Optimization and Coordination of HAFDV ...downloads.hindawi.com/journals/jcse/2013/872624.pdfMotor Motor Motor Motor Motor Motor Motor Motor ... Particle swarm optimization,

Documents