characterizing shading losses on partially shaded pv ... · pdf file•shade simulation...

NREL is a national laboratory of the U.S. Department of Energy, Office of Energy Efficiency and Renewable Energy, operated by the Alliance for Sustainable Energy, LLC.

CHARACTERIZING SHADING LOSSES ON PARTIALLY SHADED PV SYSTEMS

Chris Deline

PV Performance Modeling Workshop

September 23, 2010

Albuquerque, NMNREL/PR-520-49504

Innovation for Our Energy Future

Overview

2

Introduction: Shading on PV systemsTheory: Shaded PV power lossPractical issues with modeling shaded PV

• Shade Estimation• IV curve analysis

Methods of implementing partially shaded PV modelingSome experimental resultsCurrent and future work


Introduction – Shading on PV systems

3

Shading and mismatch occur on all types of PV installations.

• Nearby shade obstructions like trees and telephone poles

• Horizon shading from faraway structures

• Self-shading from adjacent rows

• Imp mismatch from orientation, manufacturing tolerance, differential aging or soiling

Some types of shading are easier to quantify and model than others.

1

2

31: Lakewood, CO. 2: Maumee, OH. 3: Arlington, VA

Credit: NREL

Credit: NREL PIX 15617

Credit: NREL PIX 08558


Introduction – Impact of Shade

4

Shade impact depends on e.g. module type (fill factor, bypass diode placement), severity of shade, and string configuration.Power loss occurs from shade, also current mismatch within a PV string and voltage mismatch between parallel strings.Power lost is greater than proportional to the amount of shade on the system

‘Shade Impact Factor’ (ratio of power lost to area of shade) for a single module in a single string PV system [1]

[1] C. Deline, IEEE PVSC, 2009


Bypass diode operation in most modules

5

+-+-

-

+

+-

I+V-

NN-12

1

+

D1

Shade

Bypass diodes typically protectsubstrings of 15-20 cells.Shade on one of these cells cancause the diode to turn on,removing those cells electricallyfrom the string.Current is continuous in the PVstring; a small amount of shadecan greatly reduce outputpower.On typical Si modules, reducing1 cell’s irradiance by 25% canlead to bypass diode turn-on.

1 substring


I

V

Imp

Shaded cell

Vrev

Unshaded cell

Theory – Partially shaded substring of cells

6

A shaded cell has reduced Isc. Inorder to pass the string current Impthe cell will operate in reversebias. The total substring voltageis a sum of the various operatingvoltages including the reversebiased cell.

If the total substring voltage < 0, the bypass diode turns on and the shaded cell will operate near Vrev.

Variability exists in the reverse-bias characteristics of differentcells – the same shading couldresult in different outcomes.

Full I-V curve of a partially shaded cell. Current continuity requires the shaded cell to operate in reverse bias to pass the Imp current of the rest of the substring.


0

10

20

VoltsC

urre

nt0 50 100 150

0

500

1000

Pow

er

Unshaded IV + Shaded

IV

Theory – system level IV curve

7

System IV curve is built from individual substring IV curves in series and parallel.

+ =


0

10

20

VoltsC

urre

nt0 50 100 150

0

500

1000

Pow

er

Global max [A]Local max [B]

Theory – system level IV curve

8

System IV curve is built from individual substring IV curves in series and parallel.Partial shading can lead to Local [B] and Global [A] maxima.Bypass diode turn-on depends on the peak power point chosen. For instance, operating at point [A] requires shaded bypass diode turn-on while point [B] does not.


Practical matters – shade estimation

9

Shading trees

Proposed array

Rooftop survey

Shading site survey typically relies on aerial imagery andfisheye shade analysis e.g. SunEyeTM, or Solar PathfinderTM.Some issues include: foliage changes throughout the year,spatial resolution requires multiple pictures, shading objectsare considered 100% opaque, nearby objects have moreposition uncertainty, 3D CAD modeling is time intensive.

Credit: Chris Deline / NREL



Methods of modeling substring IV curves

10

Full 5-parameter IV curve• High accuracy, but slow (when

calculated 1000’s of times)

Simplified IV curve (3-parameter)• Computationally less intense,

reduced accuracy for V < Vmp.

• I = C1 – C2 exp(C3 * V)

Empirical ‘Shade impact factor’• System-specific lookup table, based

on shade % and diffuse / global ratio.

0 5 10 15 20 25 30 350

2

4

6

8

Volts [V]

Cur

rent

[A]

5-parameter3-parameter

Comparison of full 5-parameter IV curve with a simplified 3-parameter IV curve for an Evergreen ES-200 PV module. Accuracy is better for V > Vmp .


Real-world application of shade modeling

11

Site survey conducted on a ‘typical residential installation’[3]

• Most shade from 6-10am, 2-6pm• ~21% annual irradiance loss• 2 strings of 7 mSi modules @ 3kWSite survey picture taken at each PV module substring• 3 images per module = 42 total• Provides # of shaded substrings

for a given hour and date

‘Typical residential installation’. 2x7 mSi panels

Site survey:~20% irradiance loss due to shade[3] R. Levinson, Solar Energy 83, 2009

Credit: Brent Nelson / NREL



Numerical shade simulation

12

Simulation uses PVWatts with additional shade derating[4]

• Derating based on empirical relationships between shade extent and power loss, as determined in a scale experiment at NREL.• TMY3 weather data and the default PVWatts AC to DC factor (0.77)

Two shade conditions are simulated: 1) both strings are shaded as per the survey, and 2) one string is entirely unshaded• Two-string shading is more realistic, but some installations may have

more limited shading.

[4] C. Deline, IEEE PVSC, 2010


Σ

Simulation method - overview

13

Site survey: one image for each substring TMY3 database

Experimental Shade Impact Factor

Beam/Global Irradiance

PVWatts modelX

DC Derating (hourly)

Annual shaded power production

# Shaded Substrings


Simulation results

14

Annual power produced Power lost to shade

Unshaded baseline 4.4 MWh 0Site survey estimate -21%2 strings shaded 3.5 MWh -22%1 string shaded 3.7 MWh -17%PVWatts simulation results using site survey data for a two-string PV system.

5 10 15 200

500

1000

1500

2000

2500

Time

DC

pow

er (W

)

March modeledMarch actualDec. modeledDec. actual

Modeled results (dots) and measured data (lines) for two representative sunny dates

Modeled results compare favorably with measured data on representative sunny days.Annual results show close agreement with site survey’s ‘solar resource fraction’ (but this is not always the case)


Large commercial installation

15

0 5 10 15 20 250

0.2

0.4

0.6

0.8

1

Percent of string shaded

Stri

ng p

ower

FF = 0.65FF = 0.73Observed

6 8 10 12 14 16 18-0.5

0

0.5

1

1.5

2

2.5

3

Time

Stri

ng p

ower

(kW

)

AM shading

30% loss from shade

Morning power loss is monitored with DC current transducers. 30% loss is coincident with 12.5% string shading

Modeled shade impact for large parallel systems. Note that higher FF is more sensitive to shade.

This 1MW installation has 16 PV modules per string. Periodic shading occurs from nearby light poles.



Current / Future work at NREL

16

• Shade simulation feature going into Solar Advisor Model, specifically for inter-row shading of large utility-scale systems.• Further work on developing and validating shaded PV models• Test & Evaluation of DC-DC converter devices and micro-inverters to determine the performance improvement


Questions / Comments?

17

AcknowledgmentsThis work was supported by the U.S. Department of Energy under Contract No. DOE-AC36-08GO28308 with the National Renewable Energy Laboratory.

Chris [email protected]: (303) 384-6359

mailto:[email protected]�

Innovation for Our Energy Future18

Backup slides


0 10 20 30 40

0

0.2

0.4

0.6

0.8

1

Cell shade %

Rel

ativ

e su

bstri

ng p

ower

No DC-DC

Bypass diode turn-on

0 10 20 30 40

0

0.2

0.4

0.6

0.8

1

Cell shade %

Rel

ativ

e su

bstri

ng p

ower

Bypass diode turn-on

With DC-DC

No DC-DC

1 2 30

50

100

150

200

250

300

# of M10's substrings shaded

Sys

tem

pow

er lo

ss (W

)

2126

60

4122

117

78

18

175

M1-M5M6-M9M10

Mismatch

(No DC-DC)(With DC-DC, 50% shade)

1 2 30

50

100

150

200

250

300

# of M10's substrings shaded

Sys

tem

pow

er lo

ss (W

)

M1-M5M6-M9M10

Results – single module shading

19

Without DC-DC devices:• Single-cell shading of 25%

causes bypass diode turn-on• Mismatch loss accounts for

~40% of the total shade lossWith DC-DC devices:• Bypass diode turn-on can be

delayed• Mismatch losses reduced• Shaded module output

proportional to shade opacity


0 10 20 30 40 50 60 7050

60

70

80

90

100

Single PV string shade(%)

Sys

tem

pow

er(%

)

>60% shadeSIF = 1.63

Results – Shade Impact Factor

20

Shade Impact Factor without DC-DC= 1.63With DC-DC, SIF = shade opacity

With DC-DC

Shade %

Power loss

m = SIF

0 10 20 30 40 50 60 7050

60

70

80

90

100

Single PV string shade(%)

Sys

tem

pow

er(%

)

75% shadeSIF = 0.73

100% shadeSIF = 0.97

50% shadeSIF = 0.48

No DC-DC

SIF = 1.63

characterizing shading losses on partially shaded pv ... · pdf file•shade simulation...

Documents