Modeling the Irradiance and Temperature Dependence of PV Modules in PVsyst

Kenneth J. Sauer and Thomas Roessler

Yingli Solar

Clifford W. Hansen

Sandia National Laboratories

40th IEEE Photovoltaic Specialists Conference | June 8-13, 2014

Modeling the Irradiance and

Temperature Dependence of

PV Modules in PVsyst

2 2

Outline

• Basics

• Irradiance and Temperature Dependence in PVsyst

• Modeling Approaches

• Results

• Summary and Outlook

3 3

Basics: Main Components of PV Module Models

Electrical Parameters

at STC (I-V curve)

Irradiance Dependence

Temperature Dependence Incidence Angle Dependence

(see Poster 396-M20)

Module Model

…or physical modeling Provision via datasheet or

additional characterization

Additional characterization…

4 4

Basics: Irradiance and Temperature Dependence of

PV Modules

• Characterized via “relative efficiency deviation” ∆rel [%] as

function of irradiance G [W/m2] and cell temperature T [°C]:

Pmax(G,T) GSTC

Pmax,STC G

• Use Isc as a measure of G (self-reference method):

GSTC Isc(GSTC,T)

G Isc(G,T)

• Example behavior (c-Si):

∆rel (G,T) = x ‒ 1

=

5 5

Irradiance Dependence in PVsyst

Rsh,ref, Rs

Rsh,0, Rsh,exp

6 6

Temperature Dependence in PVsyst

“µPmpp,PVsyst”, µγ

γ(T) = γref + µγ·(T – TSTC)

7 7

Improved Model for Temperature Dependence:

“Then T" Approach

Measurement Data

Linear Fit to Measurement Data

PVsyst Standard Model (µPmpp,NP)(µPmpp,NP)

Pmax(Tmin)

Pmax,STC

Pmax(Tmax)

Tmin TSTC Tmax

Derivative

equal to

µPmpp,NP

8 8

Improved Model for Temperature Dependence:

“Then T" Approach

Measurement Data

Linear Fit to Measurement Data

PVsyst Standard Model (µPmpp,NP)

"Then T" Approach (Optimized µγ)

(µPmpp,NP)

µγ)

Pmax(Tmin)

Pmax,STC

Pmax(Tmax)

Tmin TSTC Tmax

Optimize uγ so that

µPmpp,slope ≈ µPmpp,NP

9 9

Overview of Modeling Approaches

10 10

-5%

-4%

-3%

-2%

-1%

0%

1%

200 300 400 500 600 700 800 900 1000

Δη

rel(G

, 2

5

C)

G [W/m2]

Measurement Data

PVsyst Standard Model

• (20) 310 W Yingli Solar mc-Si PV modules

• Yingli Americas PV Testing Lab (PVTL)

• Class AAA flash tester with temperature

control unit and insulated test chamber

• Pmax at STC

• 80 additional measurements at AM1.5G

– G = 200:100:1,000 W/m2 at each T

– T = 25:5:65 °C at each G

• Δηrel(200 W/m2, 25 °C)

– Nameplate / Measured Avg. = -3.3 %

– PVsyst Standard Model (V6) = -4.8 %

• µPmpp(1,000 W/m2, T)

– Nameplate / Measured Avg. = -0.429 %/°C

– PVsyst Standard Model (V6) = -0.442 %/°C

Example Data and PVsyst Standard Model

11 11

Results: Comparison of Modeling Approaches

Modeling

Approach

RMSD

[W]

MBE

[W]

Dev. from

µPmpp,NP [%]

Diff. from

Δηrel,LIC [%]

G-T Matrix 0.25 -0.18 0.78 -0.16

Only G Then T 0.40 -0.33 0.04 0.01

V6 Tool Then T 0.47 -0.39 0.52 0.12

Only G Input

µPmpp,NP 0.89 -0.73 2.68 0.01

V6 Tool Input

µPmpp,NP 0.97 -0.79 3.19 0.12

Standard

Model Then T 1.52 -1.39 0.27 -1.59

Standard

Model

Input

µPmpp,NP 1.92 -1.79 2.96 -1.59

12 12

-25%

-20%

-15%

-10%

-5%

0%

20

0

40

0

60

0

80

0

10

00

30

0

50

0

70

0

90

0

20

0

40

0

60

0

80

0

10

00

25 45 65

Δη

rel

G [W/m2] // T [ C]

Measurement Data PVsyst Standard Model

PVsyst "V6 Tool" Model "Only G - Then T" Model

"G-T Matrix" Model

Results: Visualization I

13 13

Results: Visualization II

14 14

Results: Impact on Energy Yield Forecasts

0.0%

0.2%

0.4%

0.6%

0.8%

1.0%

1.2%

1.4%

1.6%

1.8%

Antofagasta, Chile

Berlin, Germany

Phoenix, USA

Trivandrum, India

Gain

in

Fo

recaste

d A

nn

ual

En

erg

y Y

ield

over

PV

syst

Sta

nd

ard

Mo

del

PVsyst "V6 Tool" Model

"Only G - Then T" Model

"G-T Matrix" Model

15 15

Summary and Outlook

• Correct model for Pmax(G,T) more reliable energy forecast

• Introduced two new optimization approaches

1. “Only G - Then T” Model

– Exact fit of µPmpp,NP and Δηrel(LIC)

– Good fit of global Pmax(G,T)

2. “G-T Matrix” Model

– Improved global fit for Pmax(G,T)

– Slight deviation from µPmpp,NP and Δηrel(LIC)

• Both models clearly better than the PVsyst Standard Model

and still significantly better than the PVsyst “V6 Tool” Model

• Next up:

– Location-optimized models

– Optimized measurement data set for modeling c-Si

16

Thank You for Your Attention