An Intelligent Fault Detection Method of a Photovoltaic Module Array Using Wireless Sensor Networks

Research ArticleAn Intelligent Fault Detection Method of a Photovoltaic ModuleArray Using Wireless Sensor Networks

Kuei-Hsiang Chao, Pi-Yun Chen, Meng-Hui Wang, and Chao-Ting Chen

Department of Electrical Engineering, National Chin-Yi University of Technology, No. 57, Section 2, Zhongshan Road,Taiping District, Taichung 41170, Taiwan

Correspondence should be addressed to Kuei-Hsiang Chao; [email protected]

Received 3 November 2013; Accepted 30 March 2014; Published 15 May 2014

Academic Editor: Wan-Young Chung

Copyright 2014 Kuei-Hsiang Chao et al. This is an open access article distributed under the Creative Commons AttributionLicense, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properlycited.

This study developed a fault diagnosis meter based on a ZigBee wireless sensor network (WSN) for photovoltaic power generationsystems. First, the Solar Pro software was used to simulate the 9-series, 2-parallel photovoltaic module array formed with theSharp NT-R5E3E photovoltaic module as well as record the power generation data of the photovoltaic module array at differentlevels of solar radiation, module temperature, and fault conditions. The derived data were used to establish the weights of theextension neural network (ENN).The fault diagnosis in the photovoltaic power generation system required extracting the systemspower generation data and real-time solar radiation and module temperatures; this study thus developed an acquisition circuitfor measuring these characteristic values. This study implemented extension neural network theory using a PIC single chipmicrocontroller and incorporated the ZigBee wireless sensor network module to construct a portable fault diagnosis meter toassess the photovoltaic power generation system.The experimental results showed that the proposed portable fault diagnosis meterbased on the extension neural network for the photovoltaic power generation system possessed a high level of accuracy in faultidentification.

1. Introduction

The photovoltaic power generation system is a minimallypolluting renewable energy source. To improve power-generation efficiency, the system must be installed in aspacious and unshaded area. However, the systems outputpower may decrease with prolonged operation. Furthermore,as the system is outdoors for long time periods, naturaldisasters could also cause malfunction in the module. A mal-function in any module of the photovoltaic power generationsystem reduces output power drastically. Therefore, devel-oping fault diagnosis technology for PV power generationsystems can not only increase testing and repair efficiencyand improve system power generation reliability but alsoreduce operating costs. As for current fault diagnosis systemsof PV power generation systems, relevant PVSAT projectsare proceeding in Germany, Netherlands, and Switzerland[1, 2]. These projects establish a self-fault detection systemfor grid-connected PV power generation systems for the use

in detecting system errors and analyzing causes of errors.However, the development of this system involves the useof meteorological satellites to transmit atmospheric data forcross-checking with data measurements from ground-basedmeteorological stations in order to provide atmosphericparameters for simulation analysis to raise the overall recog-nition rate. Using this system for fault detection in stand-alone PV power generation systems is prohibited by the highcost, restricting the range of applications for this diagnosissystem and reducing its practicality. In addition, experts havealso suggested the use of high frequency reaction measure-ment [3] and time domain reflectometry (TDR) [4] in faultdiagnosis of PV power generation systems. These diagnosismethods involve the use of identified and reflected feedbacksignals to determine the occurrence of faults and the areasin which faults occur. However, this diagnosis technologyrequires the use of network analyzers and other additionalmeasurement equipment. The technologies applied in bothmethods require detailed calculations of the internal wire

Hindawi Publishing CorporationInternational Journal of Distributed Sensor NetworksVolume 2014, Article ID 540147, 12 pageshttp://dx.doi.org/10.1155/2014/540147

2 International Journal of Distributed Sensor Networks

and cable lengths of PVmodules. Consequently,maintenanceengineersmust possess sufficient PV power generator-relatedexpert knowledge. Furthermore, if this fault diagnosis tech-nology is applied to large-scale PV power generation systems,cable lengths must be calculated accurately or misjudgmentof module fault regions will easily occur. Therefore, theapplication of this diagnosis technology does not facilitatethe convenience of system repair. While the author recentlyproposed an intelligent PV power generation system faultdiagnosis method [5] which achieved a fairly good fault typeidentification rate, the method remains in the simulationstage and is limited to the types of faults it can identify.

A photovoltaic module malfunction is difficult to detectwith naked eyes, and, therefore, eachmodule requiresmanualtesting, taking long hours. Additionally, system operationsmust be terminated while conducting manual testing toensure the safety of maintenance personnel. Since the pho-tovoltaic module is expensive, the longer the operations arehalted, the more extensive the losses become. Furthermore,photovoltaic power generation systems are installed in lessaccessible regions, thus compounding the difficulty ofmanualrepairs. The development of a fault diagnosis meter in aphotovoltaic power generation system, operated via awirelesssensor network transmission, is therefore a key issue fordiscussion.

To promote the fault diagnosis accuracy rate, an intelli-gent fault diagnosis method based on extension theory withneural networks was proposed in this paper. The relatedworks of the proposed extension neural network have beenapplied in [610]. The proposed method has the advantagesof less learning time, higher accuracy, and less memoryconsumption. In addition, a PIC microcontroller and aZigBee wireless sensor network are combined to reduce thehardware circuit size and conduct remote fault diagnosis.

2. Capture of PV Power GenerationSystem Fault Data

In this paper, a 3.15 kW PV power generation system con-sisting of NT-R5E3E PV modules connected into 9 2series-parallel is constructed by Solar Pro software packagefor testing [11]. The Solar Pro software can set the moduletype, quantity, and installation angle for PV power generationsystems. This software is also capable of simulating the I-Vand P-V curves of PV power generation systems in shadowor with faults and even power generation over one day, onemonth, or one year.

Figure 1(a) shows the I-V and P-V curves provided inthe datasheet for Sharp PVmodule NT-R5E3E for irradiationfrom 200W/m2 to 1,000W/m2 and surface temperature of25C [12]; Figure 1(b) shows the I-V and P-V curves assimulated by Solar Pro PV system analysis software forSharp solar powermodule NT-R5E3E under irradiation from200W/m2 to 1,000W/m2 and surface temperature of 25C.Comparison of Figures 1(a) and 1(b) demonstrates that theresults simulated by Solar Pro are completely consistentwith the figures of the datasheet. It is confirmed that thecharacteristic curves simulated by Solar Pro are consistent

with the actual working characteristic curves and powergeneration data.

Figure 2 shows the simulated I-V and P-V curves whentwo modules in a 3.15 kW PV module array are completelyshadowed under irradiation of 900W/m2 and surface tem-perature of 51.6C. It can be seen from Figure 2 that when aPVmodule is shadowed, the I-V curve is no longer a smoothcurve and that multiple peaks are produced in the P-V curve.Figure 3 shows the actual measured results under irradiationof 945W/m2 and surfacemodule temperature of 51.6Cwhentwo modules in a series array of PV modules are shadowed.It can be seen from Figure 2 that the I-V curve changesand multiple peaks are produced in the P-V curve. It isevident from Figures 2 and 3 that simulation results are fairlyconsistent with actual testing results, demonstrating that I-V and P-V curves experience changes when any modulein a PV power generation system is shadowed or faulted;the voltage, current, and power of the maximum powerpoint also change. Consequently, voltage of maximum powerpoint

m, current of maximum power point m, power ofmaximum power point m, and open circuit voltageoc wereselected as the fault diagnosis characteristics by which faultsand fault types of PV module array are diagnosed in thispaper.

3. Fault Diagnosis Configuration ofa Photovoltaic Power Generation System

3.1. Division of Operation Regions. As irradiation and tem-perature change over time, the module temperature andirradiation ranges that potentially appear through a dayare categorized into 21 categories. Between 300W/m2 and1,000W/m2, every 100W/m2 is designated as one interval.Each interval is then divided into three subintervals every10C between 31C and 60C. Categories are shown inTable 1.

The fault diagnosis method for PV power generationsystems discussed in this paper first requires obtainingthe current irradiation and module temperature of the PVmodule array and then using that information together withthe characteristics captured for PV power generation systemfault diagnosis, including the maximum output power m,voltage of maximum power point

m, current of maximumpower point

m, and open circuit voltage oc of the photo-voltaic module array during operation. Finally, the proposedextension neural network fault diagnosis method is used toidentifywhether the PVpower generation system is operatingnormally or a fault has occurred.

3.2. Fault Types of PVPowerGeneration Systems. Under iden-tical irradiation and module temperature conditions, thispaper divides fault categories of PV power generation sys-tems into 10 different types, as shown in Table 2. In addi-tion, the upper and lower limits of the classical domainbetween irradiation of 301W/m2 and 1,000W/m2 and mod-ule temperature between 31C and 40C for 21 regionsunder 10 different types can be obtained from simulationresults.

International Journal of Distributed Sensor Networks 3

Voltage (V)

Curr

ent (

A)

Pow

er (W

)

Current versus voltagePower versus voltage

0 10 20 30 40 500

30

60

90

120

150

180

0

1

2

3

4

5

6

Characteristic curves: current/power versus voltage(cell temperature: 25 C)

800W/m2

600W/m2

400W/m2

1,000W/m2

200W/m2

(a)

0 10 20 30 40 500

1

2

3

4

5

6

0 10 20 30 40 500

20

40

60

80

100

120

140

160

180

200

Irradiation 1000W/m2

Irradiation 800W/m2

Irradiation 600W/m2

Irradiation 400W/m2

Irradiation 200W/m2Irradiation 1000W/m2

Irradiation 800W/m2

Irradiation 600W/m2

Irradiation 400W/m2

Irradiation 200W/m2

Curr

ent (

A)

Voltage (V) Voltage (V)

Pow

er (W

)

(b)

Figure 1: Characteristic curves for surface temperature of 25C and irradiation from 200W/m2 to 1000W/m2 (a) provided in Sharp NT-R5E3E datasheet and (b) simulated by Solar Pro software.

3.3. PIC Single Chip Microcontroller. The single chip micro-controller used in this study was the PIC18F8720 microcon-troller manufactured by Microchip [13]. This study used auniversal asynchronous receiver/transmitter (USART) in thePIC18F8720 single chip microcontroller to transmit data andcontrol operation of a ZigBee module. The ZigBee wirelesssensor network module uses a RS-232 communication pro-tocol to transmit data, to contribute to the communicationbetween PIC and ZigBee module. A RS-232 transmissionprogram was coded in the PIC to facilitate communicationwith the ZigBeemodule, to control the ZigBeemodules oper-ations, and to transmit the photovoltaic power generationsystem data from the ZigBee wireless sensor network routerto the PIC microcontroller for processing.

3.4. ZigBee Wireless Sensor Network. ZigBee is a wirelessnetwork protocol based on IEEE 802.15.4 and formulated bythe ZigBee Alliance [14, 15]. This wireless sensor networktechnology is short in distance, simple in structure, low inpower consumption, and slow in transmission speed [1619]. This study used a durable FT-6260 ZigBee Mesh rapiddevelopment kit [20] with a transmission distance three tofive times greater than the regular ZigBee module. Since eachZigBee module is equipped with a RS-232 transmission port,users can connect the ZigBee module to a computer, facilitat-ing the setting of ZigBee module parameters. Additionally,each ZigBee development module is equipped with fourdigital input/output (DIO) ports and four analog to digitalconverters (ADCs). On the DIO side, the ZigBee coordinator


Curr

ent (

A)

0

3

6

9

3000

2000

1000

0

Pow

er (W

)

Voltage (V)

Corp.: Sharp Module: NT-R5E3E String (Series parallel): 92capacity: 3.15 (kW)

Official Parameters: Pmax = 175.00 (W) Vpmax = 35.40 (V) Ipmax = 4.95 (A) Voc = 44.40 (V) Isc = 9.84 (A)

Simulation Results: Pmax = 2085.17 (W) Vpmax = 233.70 (V) Ipmax = 8.92 (A) Voc = 382.03 (V) Isc = 9.84 (A)

(irradiance 0.90 (kW/m2), Shade Density 100(%))

Figure 2: I-V and P-V curves obtained in simulation under 900W/m2 irradiation and surface temperature of 51.6C where two modules ina PV module array are shadowed.

10

9

8

7

6

5

4

3

2

1

0

2000

1800

1600

1400

1200

1000

800

600

400

200

0

Power curveMPP marking

Curr

ent (

A)

Pow

er (W

)

Voltage (V)0 40 80 120 160 200 240 280 320 360

9.09A

10.43A

MPP:1962.5W

I-V-curve

Figure 3: Actual I-V and P-V curves obtained under 945W/m2irradiation and surface temperature of 51.6C where two modulesin a PV module array are shadowed.

terminal can control the DIO pins at any router end or end-device end on the same network and can also change thevoltage level at DIO pins from high voltage to low voltage, orvice versa. On theADC side, the selectedADC resolutionwas12 bits, which allowed for conversion of the 02.4V voltagerange into 0004095(000FFF) digital data.

3.5.The Sensor Circuit of Characteristic Extraction. The char-acteristics used to diagnose faults in the photovoltaic powergeneration system are the maximum output power m,voltage of maximum power point

m, current of maximumpower point

m, and open circuit voltage oc of the photo-voltaic module array. Figure 4 shows the characteristic data

Table 1: 21 regions divided by temperature and irradiation intervals.

Irradiation Module temperature Region

301400W/m23140C A14150C A25160C A3

401500W/m23140C B14150C B25160C B3

501600W/m23140C C14150C C25160C C3

601700W/m23140C D14150C D25160C D3

701800W/m23140C E14150C E25160C E3

801900W/m23140C F14150C F25160C F3

9011000W/m23140C G14150C G25160C G3

extraction circuits. This study uses the ZigBee coordinatormodule of the fault diagnosismeter in the photovoltaic powergeneration system to control the remote router DIO at lowvoltage level, allowing the solid state relay (SSR) to dissociatethe photovoltaic module array and the power conditioner, tomeasure the open circuit voltage of the photovoltaic modulearray. However, the 400V maximum open circuit voltage ofthe photovoltaic module array exceeds the maximum inputvalue of 2.4 V for the ZigBee module. To limit the maximumopen circuit voltage of the photovoltaic module array towithin themaximumADC input value of 2.4V for the ZigBee


Table 2: Fault types of PV power generation systems.

Fault type Fault conditionPF1 Normal operationPF2 Of two series branches, a fault has occurred in a module in either of the two branchesPF3 Of two series branches, a fault has occurred in two modules in one branchPF4 Of two series branches, a fault has occurred in three modules in one branchPF5 Of two series branches, a fault has occurred in one module in each branchPF6 Of two series branches, a fault has occurred in two modules in each branchPF7 Of two series branches, a fault has occurred in three modules in each branchPF8 Of two series branches, a fault has occurred in one module in one branch and in two modules in the other branchPF9 Of two series branches, a fault has occurred in one module in one branch and in three modules in the other branchPF10 Of two series branches, a fault has occurred in two modules in one branch and in three modules in the other branch

module, the photovoltaic module arrays output voltage isprocessed through the series connection of 2M and 12K,compressing the maximum open circuit voltage of 400Vinto 2.4V. Following the above method, through ZigBeewireless sensor network transmission, we are able to inputthe photovoltaic module arrays open circuit voltage into thefault diagnosis meter in the photovoltaic power generationsystem to conduct fault diagnosis. Additionally, 2M and12K resistors prevent loading effects and improve thevoltage division accuracy. Using the same method to set theZigBee wireless sensor network router DIO at high voltagelevel and then to facilitate SSR conduction and connect thephotovoltaic module array and power conditioner in parallel.Owing to the ZigBee modules minimal power output, theSSR cannot be turned on directly, constructing an opera-tion amplifier to boost power and connect the photovoltaicmodule array to the power conditioner. Then, the powerconditioner is equippedwithmaximumpower point tracking(MPPT) and can obtain the voltage ofmaximumpower point,the current of maximum power point, and the maximumpower (obtained by multiplying the previous two entities) ofthe photovoltaicmodule array by using the voltage sensor andthe hall current sensor.

4. Fault Diagnosis in a Photovoltaic PowerGeneration System

4.1. The Overall Scheme of the Proposed Portable FaultDiagnosis Meter. Figure 5 displays the overall scheme of theproposed portable fault diagnosis meter in the photovoltaicpower generation system. The ZigBee module in the faultdiagnosis meter controls the conduction of the SSR on thecharacteristic value acquisition circuit to extract the voltageof maximum power point, current of maximum power point,open circuit voltage, solar radiation, and module tempera-tures of the photovoltaic module array. The ZigBee wirelessInternet transmits the voltage, current, solar radiation, andmodule temperature values to the fault diagnosis meter, andthe diagnosis results and extracted statistics are shown on theLCD.

4.2. The Proposed Extension Neural Network Fault Diag-nosis Method. Most of the previously presented results in

intelligent neural network design used traditional learningmethods for tuning the network parameters [2125]. How-ever, in this paper, an intelligent method with more efficientlearning process was proposed. The characteristics used todiagnose faults in the photovoltaic power generation systemare the maximum output power m, voltage of maximumpower point

m, current of maximum power point m, andopen circuit voltage oc of the photovoltaic module array.Because multiple characteristics were used, the proposedmethod can avoid the inaccuracy in diagnosis caused by thechanges in internal electric quantity with temperature andage of a photovoltaic module. The selected characteristicsof photovoltaic module arrays are utilized to construct anintelligent diagnosis method according to the extensiontheory [26] and a modified neural network [27].

Thematter-element model and extension distance are themain principles of the proposed extension neural network,which can indicate the alterable relations between quality andquantity by matter-element transformation. The proposedfault detection method will first create a set of fault matter-element model of a photovoltaic module array by extensionneural network learning process. And, a regular extensiondistance will then identify the fault type of a photovoltaicmodule array by calculating the distance between test dataand each matter-element model. According to these results,the proposed fault detection method can detect the fault typeof a photovoltaic module array correctly and promptly. Therelated works of the proposed extension neural network havebeen successfully applied in [610].

Figure 6 presents a structural diagram of the proposedextension neural network.The data are classified and enteredinto the neural cells of input layers.The number of input layerneural cells is decided by the number of the characteristics ofthe matter-elements to be identified. Conversely, the outputlayers are used to store calculated extension distances, andthe lines connecting the two layers display the weight, whichcomprises the upper weight, the center weight, and the lowerweight. Finally, the minimum extension distance is identifiedvia the various extension distances stored in the output layersfor classification [610].

4.3. Learning Process of Extension Neural Network. Beforeconducting the learning processes, the samples were defined


Inverter

AD202J

1

2 338

20 22 18

19

SSR75A

+

+SSR

4

75A

+

+

PV array

10K

40K

15V

15V

A 741

1

2

3

4

8

7

6

5

ZigBee DIOinput

200

0.1

50K100K

OP07

1

2

3

4

8

7

6

5

0.1

Zero adjustmentHall current sensor2M

12K

2K

32V DC 432V DC

5240V DC 5240V DC

Vpmax and Voc

Ipmax

+15V

+15V

15V

15V

Figure 4: Acquisition circuit of maximum power, maximum power voltage, maximum power current, and open circuit voltage.

Temperature

Volta

gese

nsor

Irradiation

meter

Photovoltaic arrayThermalcouple

Irradiation

ZigBee

router

ZigBee

Coordinator

PIC18F8720

Wireless network

transmission

SSR

Command

Display irradiation, temperature,

Irradiation, temperature,

Hall current sensor

Inve

rter

Load

Extension neural

network fault

diagonsis method

SSR control signal

RS-232

Vpm and Voc

Ipm

Vpm, Ipm, Voc

Vpm, Ipm, Pm, Voc and fault type

Figure 5: Overall hardware construct of the fault diagnosis meter of portable photovoltaic power generation system.


1 2 3

1 2 3

n

n

c

Input layer(characteristics)

Output layer(extension distance, ED)

Weighting factor wkj

Data input

Find minimum EDED2

ED1

Decision output category

1n

wL1 3

xpi3

xpi2x

pi1

Xi

z13wU13

wL1 1

z11wU11

wL1 2

z12wU12

wUkj,weighting center zkj and lowerlimit of weighting wLkj)

EDn

wU

wL1n

xp1n

z1n

(upper limit of weighting

Figure 6: Structure of the proposed extension neural network.

as = {1, 2, 3, . . . ,

}. refers to the total number

of samples, each comprised of data characteristics and clas-sifications. Consider

= {

1,

2,

3, . . . ,

}, where the

learning samples are denoted by = 1, 2, 3, . . . , , n refers

to the total number of characteristics in the matter-elementmodel, k refers to the sample data classification, and the errorratio

is defined as

=

, (1)

where refers to the total error number of classification.

The supervised learning algorithmprocesses of the exten-sion neural network are detailed as follows.

Step 1. Input the learning data to establish matter-elementmodels that correspond to the different classifications andestablishweights between the input and output layers.The kthmatter-element is presented as follows:

=

[[[[

[

, 1, V1

2, V2

......

, V

]]]]

]

, = 1, 2, 3, . . . , ; = 1, 2, 3, . . . , ,

(2)

where is the total number of data classifications, is the

jth characteristic of the Nkth classification, and V is theclassical region

,

of the

characteristic of the Nkth

classification.The range of the classical region can be decidedby the learning data and

and

are defined as

= max

{

} ,

= min

{

} , (3)

where

refers to the learning data inputted into theextension neural network.

Step 2. Calculate the center weight of each classificationcharacteristic. Consider

= {1, 2, 3, . . . ,

} ,

=

(

+

)

2, (4)

where = 1, 2, . . . , and = 1, 2, . . . , .

Step 3. Extract the th learning sample of the kth classificationas shown in

= {

1,

2,

3, . . . ,

} . (5)

Step 4. Use (6) to calculate the extension distance betweenthe learning sample

and the various classifications. Con-

sider

ED= [

[

=1

((

) /2)

(

) /2

+ 1]

]

,

= 1, 2, . . . , ,

(6)


where refers to the jth characteristic of the th learning

sample of the kth classification, refers to the center weight

between the jth input and the kth output, and and

refer to the upper and lower weight limits between the jthinput and the kth output.

Step 5. Find theminimum extension distance between all theclassifications in the output layer. If ED

= min{ED},

then classification and data classification are the same( = ), denoting the identificationwas correctly conducted,and Step 7 should be conducted directly. If the extensiondistance ED

=min{ED}, denoting the classification

differs from data classification k ( = ), the learningprocesses in Step 6 should be fulfilled.

Step 6 (adjust the weights). (1) Modify the upper and lowerweights.

Modify the weight of the classification to obtain

new =

old + (

old) ,

new =

old + (

old) .

(7)

Modify the weight of the classification to obtain

new =

old (

old) ,

new =

old (

old) .

(8)

(2) Modify the center weight

new =

(

new +

new)

2,

new =

(

new +

new)

2,

(9)

where is the learning rate of the extension neural net-work,

new,

new are the upper and lower weights ofthe th characteristic in the th classification after learn-ing,

new,

new are the upper and lower weights of

the th characteristic in the th classification after learn-ing,

old,

old are the upper and lower weights of theth characteristic in the th classification before learning,

old,

old are the upper and lower weights of the

th characteristic in the th classification before learning, old, old are the center weight of the th characteris-tic in the th and th classification before learning, and new, new are the center weight of the th characteristicin the th and th classification after learning.

Step 7. The learning processes are terminated upon com-pleted classification of all the learning samples. Otherwise,repeat Steps 3 to 6.

Step 8. When the total error ratio reaches the expected

value, the learning classification procedures are terminated.Otherwise, repeat Step 3 to resume the learning and trainingprocedures.

Figure 7 shows the adjustment process of the weightingfactors, as described in Step 6. The learning sample

, in

Figure 7(a), belongs to cluster B. However, the result ofthe extension distance equation is ED

< ED

; thus, it

is classified as cluster A. As shown in Figure 7(b), afterthe adjustment of the weighting factor, the new extensiondistance is ED

> ED

; hence, the learning sample

can

be classified as cluster B.

4.4. The Recognition Algorithm Processes of the ENN. Uponcompletion of the extension neural network, the learningprocesses, recognition, diagnosis, and classification can beimplemented. The procedures are as follows [610].

Step 1. Extract the weight matrix of the trained extensionneural network.

Step 2. Extract the samples awaiting recognition.

Step 3. Use (6) to calculate the extension distance betweenthe samples awaiting identification and each classification.

Step 4. Find the minimum extension distance to identify theclassification of each data awaiting identification.

Step 5. Check that all samples awaiting identification aretested before the algorithm is terminated. Otherwise, repeatStep 2 to process the next sample awaiting identification.

The characteristics used to diagnose faults in the pho-tovoltaic power generation system are the maximum outputpower m, voltage of maximum power point m, currentof maximum power point

m, and open circuit voltageoc of the photovoltaic module array. Therefore, there arefour parameters fed to the neural network. Under identicalirradiation and module temperature conditions, this paperdivides fault categories of PV power generation systems into10 different types, as shown in Table 2.There are 2,196 entriesof data collected from the photovoltaic module array for 10different types that were divided into 1,098 entries of trainingdata and 1,098 test data. The 1,098 entries of training datawere used for the extension neural network training processintroduced in Section 4.3. Therefore, the vector lengths ofdata from the cells unit are 1,098. In this study, (6) is usedto calculate the extension distance ED

between the learning

sample and the various classifications. After that, the

minimum extension distance between all the classificationsin the output layer is found. If ED

= min

{ED},

then classification and data classification are the same( = ), denoting the identificationwas correctly conducted,and the learning processes are terminated upon completedclassification of all the learning samples. However, if theextension distance ED

=min{ED}, denoting the clas-

sification differs from data classification k ( = ), theweights adjusting process should be fulfilled. Therefore, theextension distances ED

between the learning sample

and the various classifications are the feedback controlledelements.


EDA

EDB

ED

zBzA xij

Cluster A Cluster B

x0

(a)

ED

EDA

EDB

zBzA

Cluster A Cluster B

x0

xij

(b)

Figure 7: Adjustment process of cluster weighting: (a) before learning and (b) after learning.

(a) (b)

Figure 8:Hardware photos of the PV fault diagnosismeter: (a) the characteristic value acquisition circuit andZigBeewireless internetmodulecombination and (b) the combined ZigBee coordinator and PIC microcontroller circuits.

5. Experimental Results

Figure 8(a) demonstrates the hardware photo of the com-bined characteristic value acquisition circuit and ZigBeewireless internet module. In addition to requiring a faultdiagnosis program on the PIC single chip microcontrollerof the photovoltaic power generation system, a transmittedcontrol signal is also essential for controlling the ZigBeemodule. ZigBee uses the 12V input and output voltage RS-232 communication protocol. Without an input and outputvoltage lower than 12V, the PIC single chip microcontrollerrequires a Max232IC on the fault diagnosis meter to com-municate with the ZigBee module and control its operations.The generator data transmitted by the ZigBee module tothe photovoltaic power generation system is inputted intothe PIC microcontroller chip for concurrent processing.Figure 8(b) shows the hardware photo of the combined

ZigBee coordinator and PIC microcontroller circuits in thephotovoltaic power generation system.

To determine the proposed extension neural networksfeasibility for photovoltaic power generation system faultdiagnosis, this study employed test data with a solar radiationbetween 901W/m2 and 1,000W/m2 and module tempera-tures between 31C and 40C. Table 3 displays the simulateddata from Solar Pro [11] and their transmitted data fromthe ZigBee module ADC, making the transmitted statisticsthe data for actual test. The table demonstrates that theproposed extension neural network accurately diagnosedthe photovoltaic power generation system faults. Despitediscrepancies between the data generated via the Solar Prosoftware and the transmitted data from the ZigBee moduleADC, Table 3 indicates only one datum (item number 3) wasincorrectly assessed, thereby proving a high-identificationrate.


Table 3: Test data with a solar radiation between 901W/m2 and 1,000W/m2 and module temperatures between 31C and 40C, datatransmitted to ZigBee, known fault types, and the results of extension neural network fault diagnosis.

Item number Simulated data of Solar Pro Data transmitted of ZigBee Known fault type Diagnosed fault typem m m oc m m m oc

1 3116.6 314.66 9.9 396.08 3134 316 9.8 397 PF1 PF12 2872.7 291.21 9.86 396.02 2890 293 9.8 397 PF2 PF23 2539.8 255.57 9.94 396.03 2559 257 9.9 397 PF3 PF24 1944.71 218.99 8.8 390.39 1961 220 8.8 392 PF4 PF45 2463.83 278.09 8.86 347.05 2480 279 8.8 348 PF5 PF56 2156.42 242.16 8.9 330.68 2172 244 8.8 305 PF6 PF67 1847.94 206.17 8.96 260.29 1864 208 8.9 262 PF7 PF78 2241.85 250.56 8.95 347.01 2260 252 8.9 348 PF8 PF89 1940.26 218.82 8.87 347.01 1958 220 8.8 348 PF9 PF910 1925.53 218.11 8.83 303.63 1942 220 8.8 305 PF10 PF10

Upon conducting a local feasibility analysis of the exten-sion neural network for fault diagnosis in a photovoltaicpower generation system, the methods accuracy under dif-ferent solar radiation and module temperature conditionsmust be measured. Table 4 shows the fault diagnosis resultsof randomly chosen test data and indicates that the diagnosisaccuracy remained high under various solar radiation andmodule temperature conditions. The proposed use of theextension neural network for fault diagnosis in a photovoltaicpower generation system involves minimal training andsimple learning processes and is, therefore, suitable for faultdiagnosis in a photovoltaic power generation system.

The fault classification characteristic values of testdata with a solar radiation value between 901W/m2 and1,000W/m2 and module temperatures between 31C and40C and those with a solar radiation value between601W/m2 and 700W/m2 and module temperatures between31C and 40C are too similar. This resulted in diagnosiserrors and reduced accuracy rates to only 96% and 92%, lowerthan those in other regions. However, only these two regionsobserved lower accuracy rates; the rates in other regions were100%.

To display the superiority of the proposed extensionneural network, fuzzy neural network [25] and multilayerperceptrons and back propagations (MLP) [21] with threenetwork structures, namely, 4-7-10, 4-8-10, and 4-9-10 (inputlayer-hidden layer-output layer), were constructed to diag-nose photovoltaic module array faults and to compare diag-nosis results of the extension neural network, as seen inTable 5. The fault diagnosis of photovoltaic module arraysdivided solar radiation and module temperatures into 21regions, conducting a diagnostic simulation of each region.The fault diagnosis displayed inTable 5 demonstrates a higheridentification performance by the proposed extension neuralnetwork compared with that of multilayer perceptrons andback propagation diagnosis and only requires 22 iterations,indicating that each region only requires approximately 1iteration to reach a 100% identification rate. Conversely,fuzzy neural network and multilayer perceptrons and backpropagations require at least 5,584 and 8,507 iterations,respectively, indicating that each region requires an average of

250 and 400 iterations to reach a neural network convergence.The comparison results indicate that an extension neuralnetwork conducting fault diagnosis is highly accurate andrequires minimal iterations and is, therefore, more efficientthan traditionalmultilayer perceptrons and back propagationdiagnosis.

Table 5 shows that the proposed method suits betterthe practical situation due to the effects of photovoltaicmodule aging being considered. The proposed extensionneural network only needs the basic computing rules suchas addition, subtraction, multiplication, and division. Com-pared with the neural network method presented in [21]and fuzzy neural network proposed in [25], the learning anddiagnosis response of the proposed extension neural networkis also faster under the same conditions. It is because morecomplex computing instruction such as exponential functionis needed using the traditional neural network and fuzzyneural network methods. However, the proposed extensionneural network diagnosis method does not need complexlearning procedure, such that it can easily be implementedon a single chip microcontroller. However, when the capacityof photovoltaic modules increases or reduces, a fractionalamount of the classical region data should be modifiedto promote the diagnosis accuracy. Because the diagnosisscheme is simple, a PIC microcontroller can be utilized toimplement the hardware for real-time fault diagnosis of aphotovoltaic module array. There are a lot of analog anddigital modules in the PIC microcontroller, so it makes thehardware circuit size very small.

The characteristics used to diagnose faults in the pho-tovoltaic power generation system are the maximum outputpower m, voltage of maximum power point m, current ofmaximum power point

m, and open circuit voltage oc ofthe photovoltaic module array. Because multiple characteris-tics were used, the proposedmethod can avoid the inaccuracyin diagnosis caused by the changes in internal electricquantity with temperature and age of a photovoltaic module.However, the proposed fault diagnosis system cannot handlesparks caused by lighting. Fortunately, the lighting sparkscan be absorbed by lighting arrester installed on photovoltaicpower generation system.


Table 4: Fault diagnosis accuracy rates of the extension neural network in photovoltaic power generation system in different regions.

Irradiation Module surface temperature Diagnosis accuracy 1(error number/total number)

9011000W/m231C40C 1 2/50 (=96%)41C50C 1 0/50 (=100%)51C60C 1 0/58 (=100%)

801900W/m231C40C 1 0/50 (=100%)41C50C 1 0/50 (=100%)51C60C 1 0/50 (=100%)

701800W/m231C40C 1 0/50 (=100%)41C50C 1 0/50 (=100%)51C60C 1 0/50 (=100%)

601700W/m231C40C 1 4/50 (=92%)41C50C 1 0/50 (=100%)51C60C 1 0/50 (=100%)

501600W/m231C40C 1 0/55 (=100%)41C50C 1 0/50 (=100%)51C60C 1 0/52 (=100%)

401500W/m231C40C 1 0/55 (=100%)41C50C 1 0/55 (=100%)51C60C 1 0/55 (=100%)

301400W/m231C40C 1 0/55 (=100%)41C50C 1 0/56 (=100%)51C60C 1 0/57 (=100%)

Table 5: Comparison of fault diagnosis accuracy rates between test datawith a solar radiation between 301W/m2 and 1,000W/m2 andmoduletemperatures between 31C and 40C, when applied in extension neural networks and traditional neural networks.

Method Comparison itemIteration number Learning accuracy rate Diagnosis accuracy rate

Proposed extension neural network (ENN) 22 100% 100%MLP (4-7-10) [21] 8,507 90.85% 93.33%MLP (4-8-10) [21] 11,089 85.65% 90%MLP (4-9-10) [21] 8,597 96.64% 93.33%Fuzzy neural network [25] 5,584 98.84% 97.70%

6. Conclusions

This study conducted fault diagnosis on photovoltaic powergeneration systems. First, Solar Pro software was used todevelop a 3.15 kW photovoltaic power generation systemsimulating regular operation and operation under variousfault classifications. The characteristic values of the photo-voltaic power generation system under both conditions wererecorded. Half the simulation results were used as trainingdata to test the weight of the extension neural network, andthe PIC single chip microcontroller was used to developthe photovoltaic power generation systems fault diagnosismeter based on an extension neural network. The resultsindicate a high extension neural network recognition rate,and that minimal learning data is required to completeweight training. Additionally, the combination of a PIC singlechip microcontroller and a ZigBee wireless sensor networktransmitter greatly reduced the risks involved in manualrepairs and reduced the time required for system diagnosis.

Furthermore, a ZigBee wireless sensor network transmissioncan conduct a remote fault diagnosis. Finally, the proposedfault diagnosis meter for photovoltaic power generationsystems and the fault characteristic value acquisition circuitare both small in size and light in weight and are thereforeportable.

Conflict of Interests

The authors of the paper declare that there is no conflict ofinterests with any of the commercial identities mentioned inthe paper.

Acknowledgment

This work was supported by National Science Council,Taiwan, under Grants NSC 102-2221-E-167-009 and NSC 101-ET-E-167-001-ET.


References

[1] J. Betcke,V.V.Dijk, C. Reise et al., PVSAT: remote performancecheck for grid connected PV systems using satellite data,evaluation of one year field-testing, in Proceedings of the 17thEuropean Photovoltaic Solar Energy Conference, 2001.

[2] E. Lorenz, J. Betcke, A. Drews et al., PVSAT-2: intelligentperformance check of PV system operation based on satellitedata, in Proceedings of the 19th European Photovoltaic SolarEnergy Conference, 2004.

[3] L. Schirone, F. P. Califano, U. Moschella, and U. Rocca, Faultfinding in a 1MW photovoltaic plant by reflectometry, inProceedings of the 1st IEEE Photovoltaic Specialists Conference,pp. 846849, December 1994.

[4] T. Takashima, K. Otani, K. Sakuta et al., Electrical detectionand specification of failed modules in PV array, in Proceddingsof the 3rd World Conference on Photovoltaic Energy Conversion,pp. 22762279, jpn, May 2003.

[5] K. H. Chao, S. H. Ho, and M. H. Wang, Modeling and faultdiagnosis of a photovoltaic system, Electric Power SystemsResearch, vol. 78, no. 1, pp. 97105, 2008.

[6] M. H.Wang, Extension neural network for power transformerincipient fault diagnosis, IET Generation, Transmission &Distribution, vol. 150, pp. 679685, 2003.

[7] M. H. Wang and C. P. Hung, Extension neural network,in Proceedings of the International Joint Conference on NeuralNetworks, pp. 399403, July 2003.

[8] K. H. Chao, C. L. Chiu, C. J. Li, and Y. C. Chang, A novelneural network with simple learning algorithm for islandingphenomenondetection of photovoltaic systems,Expert Systemswith Applications, vol. 38, no. 10, pp. 1210712115, 2011.

[9] K. H. Chao and J. W. Chen, State-of-health estimator based-on extension theory with a learning mechanism for lead-acidbatteries, Expert Systems with Applications, vol. 38, no. 12, pp.1518315193, 2011.

[10] K. H. Chao and J. W. Chen, Design and implementation of acapacity estimator using simple neural networks for lead-acidbatteries, International Review of Electrical Engineering, vol. 6,no. 7, pp. 29652973, 2011.

[11] Solar Pro Brochure and Laplace System Co., 2007, http://www.lapsys.co.jp/english/e products/e pamphlet/sp e.pdf.

[12] Sharp Corporation of Australia, Sharp NT-R5E3E PV mod-ule user menu, 2009, http://www.sharp.net.au/catalogue/bro-chures/NTR5E3E 1127 brochure.pdf.

[13] PIC18Fxx20 Data sheet, Microchip Technology Inc., 2003,http://ww1.microchip.com/downloads/cn/DeviceDoc/cn0118-97.pdf.

[14] J. T. Adams, An introduction to IEEE STD 802.15.4, inProceedings of the IEEE Aerospace Conference, March 2006.

[15] ZigBee Alliance, ZigBee specification version 1.0, 2004, http://www.zigbee.org.

[16] X. Q. Zhu and J. M. Wang, The research and implementationof ZigBee protocol network, Electron Technologies, vol. 1, pp.129132, 2006.

[17] D. Geer, Users make a beeline for ZigBee sensor technology,Computer, vol. 38, no. 12, pp. 1619, 2005.

[18] L. Y. Yuan and Y. L. Zhu, Traffic monitoring network sim-ulation of wireless sensor based on NS2, Journal of SystemSimulation, vol. 19, no. 3, pp. 660664, 2007.

[19] W. Andrew, Commercial applications of wireless sensor net-works using ZigBee, IEEE Communications Magazine, vol. 45,no. 4, pp. 7077, 2007.

[20] FT-6260 High power ZigBee mesh rapid development kit, inUser Manual, Surewin Technology Ltd., 2009.

[21] Y. Wu, Q. Lan, and Y. Sun, Application of BP neural networkfault diagnosis in solar photovoltaic system, in Proceedings ofthe International Conference on Mechatronics and Automation(ICMA 09), pp. 25812585, 2009.

[22] C. H. Wang, W. Y. Wang, T. T. Lee, and P. S. Tseng, FuzzyB-spline membership function (BMF) and its applications infuzzy-neural control, IEEE Transactions on Systems, Man andCybernetics, vol. 25, no. 5, pp. 841851, 1995.

[23] J. S. R. Jang, Input selection for ANFIS learning, inProceedingsof the 5th IEEE International Conference on Fuzzy System, 1996.

[24] C. T. Lin and C. S. G. Lee, Neural Fuzzy Systems: A Neuro-Fuzzy Synergism to Intelligent Systems, Prentice-Hall, Engle-wood Cliffs, NJ, USA, 1999.

[25] S. K. Oh, W. Pedrycz, and H. S. Park, A new approachto the development of genetically optimized multilayer fuzzypolynomial neural networks, IEEE Transactions on IndustrialElectronics, vol. 53, no. 4, pp. 13091321, 2006.

[26] W. Cai, The extension set and incompatibility problem,Journal of Scientific Exploration, vol. 1, pp. 8193, 1983.

[27] K. H. Chao, C. J. Li, and M. H. Wang, A maximum powerpoint tracking method based on extension neural network for PVsystem, vol. 5551 of Lecture Notes in Computer Science, Springer,2009.

Submit your manuscripts athttp://www.hindawi.com

VLSI Design

Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014

International Journal of

RotatingMachinery


Hindawi Publishing Corporation http://www.hindawi.com

Journal of

EngineeringVolume 2014


Shock and Vibration


Mechanical Engineering

Advances in


Civil EngineeringAdvances in

Acoustics and VibrationAdvances in



Electrical and Computer Engineering

Journal of


Distributed Sensor Networks


The Scientific World JournalHindawi Publishing Corporation http://www.hindawi.com Volume 2014

SensorsJournal of


Modelling & Simulation in EngineeringHindawi Publishing Corporation http://www.hindawi.com Volume 2014


Active and Passive Electronic Components

Advances inOptoElectronics

Hindawi Publishing Corporation http://www.hindawi.com

Volume 2014

RoboticsJournal of


Chemical EngineeringInternational Journal of


Control Scienceand Engineering

Journal of


Antennas andPropagation




Navigation and Observation


Documents

An Intelligent Fault Detection Method of a Photovoltaic Module Array Using Wireless Sensor Networks