Condition Monitoring of a Large-scale PV Power Plant in Australia

Condition Monitoring of a Large‐scale PV Power Plant in Australia

Amit DhokeThe University of Queensland, Australia

[email protected], [email protected], [email protected]

1

Paper No: 16PESGM2224

Outline

• Background• System Review• System Modelling• Methodology• Simulation Results• Conclusions

2

Global Photovoltaic (PV) Penetration

3

• PV generation in Australia increased from 12 MWp to 5 GWp during last 9 years.

* Global demand for PV installation - 2004 to 2020

Solar star PV Plant, USA – 579 MWp

* www.greentechmedia.com

How PV system works?4

Connection topologies of the PV array. (a) Series array,(b) parallel array, (c) series-parallel array, (d) total-crosstied array, (e) bridge-link array and (f) honey-comb array.

(1) http://solar.smps.us/off-grid.html (2) http://www.yourhome.gov.au/energy/photovoltaic-systems (3) https://www.researchgate.net

(1)

(2)

(3)

Research Motivation5

Fire incident occurred 1 MWp Mount Holly, North Carolina, USA

• Conventional monitoring methodsare inefficient in large-scale PVfault detection.

• Most occurring incipient faultsand blind spots remain hidden inPV system.

• UQ Centre was affected due to2011 hailstorm and resulted inblind spots.

• UQ’s Prentice Building has stringfailure issue and it was there for10 months.

Fire incident occurred in 383 KWp Bakersfield, California, USA

UQ : The University of Queensland, Brisbane, Queensland, Australia

PV array faults6

PV system with various fault types in the array.

Blind spots: Faults whichremain hidden in protectiondevices- Overcurrent protection device

(OCPD)- Ground fault protection device

(GFPD)

Incipient faults: Low magnitudebut can cause large issue ifundetected

Zhao et al., 2014

PV system condition monitoring7

(1)

Block diagram of a PV array monitoring system Levels of PV system monitoring

Solar PV condition monitoring is the process of monitoring the condition of a PVsystem with the intent to analyse performance, detect operational issues and faults.

(1)

(1) http://www.renewableenergyworld.com/articles/print/rewna/solar-energy/solar-electric-facility-o-m-now-comes-the-hard-part.html

Model-based PV Output Prediction8

. ( )

. ( ) (1 ( )) / 2

[ ( ) ( ) ( ) ( ) ( )]-1

G H I D N I cos z D H I

PO AI D N I cos AO I D H I cos

AO I cos cos z cos sin z sin cos A

[ ( ) ( )] / ( )-1A s in co s s in w s in zWhere,

. .

. . .[1 ( 25)] / 1000

. .[1 ( 25)][ . ( / 1000)]

( . ) / ( ) ( . / )

oc p ocp m

m

out sc oc

sc s scp m

P I V FF

I N POAI I T

V N V T 1 ln POAI

FF a b POAI c ln POAI T d e POAI f POAI

Power calculations

Plane of Array Irradiance Calculations

GHI- Global Horizontal Irradiance DNI- Direct Normal Irradiance POAI- Plane of Array Irradiance AOI- Angle of IncidencePout- Power output Isc- Short circuit current Voc- Open circuit voltage FF- Fill factor

Block diagram of prediction model

This model has been utilised to detect faults using existing methods.

Field parameters

Proposed condition monitoring system

Condition Monitoring

Fault Detection

Irradiance

Temperature

String 1String 2

String n

GHI

Ta

String 1

String 2

String n

Health statusM

easu

red

para

met

ers

9

Condition monitoring based on measured and predicted power

10

5 8 11 14 17 200

2000

4000

6000

8000

10000

12000

Time (Hour)

Pow

er (W

)

PMeas

PSim

5 8 11 14 17 200

500

1000

1500

2000

2500

3000

Time (Hour)

Pow

er (W

)

PMeas

PSim

Power comparisons for clear and cloudy day

Results

• One minute resolution

• Measured and predicted output has good agreement for fault detection

• Change in irradiation and temperature affecting power output

Fault detection based on current and voltage indicators

11

Faulty String- Values of NRc are below the given threshold TNRcfs indicates fault in string

Faulty Module- Values of NRv are below the threshold TNRvbm indicates voltage reduction

Results Cont…

7 9 11 13 15 170

0.25

0.5

0.75

1

Time (Hour)

Cur

rent

Rat

io

NRcoTNRcfsNRc

7 9 11 13 15 170.6

0.65

0.7

0.75

0.8

0.85

0.9

Time (Hour)

Vol

tage

Rat

io

NRv NRvo TNRvbm

Current and voltage indicators for string and module fault

1 .02

/

. . (1 ( 25) / 1000

/

, 1 (1 / )

; co m o sc

m o p m pp m

cfs co p

c m p sc

TN R N R

N R I I

I N P O A I I T

N R I I

w here N

/

. .[1 ( 25)][1 . ( / 1000)]

1.02 . , 1 (1 / )

; /

vbm vo s

v mp oc vo mo oc

mo s mpp m

NR V V

V N V T ln POAI

TNR NR where N

NR V V

NRc-Current ratio NRco- Current at fault free operation TNRcfs- Threshold for current ratio in case of faulty stringNRv- voltage ratio NRvo- Voltage at fault free operation TNRvbm- Threshold for voltage ratio in case of faulty module

Results Cont…12

Current, Voltage and Power indicators for fault-free and string fault condition

/

//

o c m o m p p

o v m o m p p

o p s im m e a s

R I I

R P PR V V

7 9 11 13 15 170

1

2

3

4

5

Time (Hour)

Rat

io

RocRovRop

7 9 11 13 15 170.6

0.9

1.2

1.5

Time (Hour)

Rat

io

RocRovRop

Roc, Rov and Rop are ratio of predicted and measured current, voltage and power respectively

Imo, Vmo and Psim are predicted current, voltage and power respectively

Impp, Vmpp and Pmeas are measured current, voltage and power respectively

Fault Free Operation- All three current, voltage and power ratios are nearly same

Faulty String- Difference in Roc and Rov caused due to string fault

Fault detection based on parameter comparison

Results Cont…13

String Mean Value (µ) Std. Deviation (σ)

String 1 8.31 5.81

String 2 4.21 2.96

5 8 11 14 17 200

3

6

9

12

15

18

Time (Hour)

Cur

rent

(A)

String 1String 2

5 8 11 14 17 200

100

200

300

400

500

Time (Hour)

Vol

tage

(V)

String 1String 2

String comparison for string fault and module fault

Mean values and standard deviation for string current

5 8 11 14 17 200

5

10

15

Time (Hours)

Cur

rent

(A)

String 1String 2

String current comparison for partial shading

Fault detection based on string comparison

Conclusions

• A string level monitoring approach is adopted to detect faultsin an existing PV plant in Australia.

• The three methods analysed includes current and voltageratio comparison with thresholds, parameter ratio comparisonand string comparison.

• String failure, module failure and partial shading faults aredetected.

• String comparison method qualitatively outperforms theother two methods.

• String comparison method can be used for real‐time faultdetection with minimum sensors.

14