a photovoltaic module thermal model

Upload: bogdan-mondoc

Post on 03-Jun-2018

228 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/12/2019 A Photovoltaic Module Thermal Model

    1/7

    A Photovoltaic Module Thermal Model Using Observed Insolation andMeteorological Data to Support a Long Life, Highly Reliable Module-Integrated

    Inverter Design by Predicting Expected Operating Temperature

    Robert S. Balog1

    Senior Member, IEEETexas A&M University

    216P Zachary Engineering CenterCollege Station, TX 77843, USA

    [email protected]

    Yingying Kuai

    Student Member, IEEEUniversity of Illinois, Urbana-Champaign

    1406 W. Green StUrbana, IL61801, USA

    [email protected]

    Greg Uhrhan

    SolarBridge Technologies2111 S. Oak Street

    Champaign, IL 61821, [email protected]

    Abstract Accurate prediction of photovoltaic (PV) moduletemperature is needed to understand the expected electricalperformance, lifetime, and reliability of photovoltaic cells. Aphotovoltaic AC module (PVAC) integrated the inverter directlywith the PV module which exposes the power electroniccircuitry to the thermal environment of the PV module. This hasbeen reported to impose additional requirements on componentselection and circuit design. However, a worst-case stack upanalysis can lead to the conclusion that module-integratedinverters require industrial grade components or expensivethermal management.

    This paper presents a detailed thermal model for the PV

    module that uses real-world operating conditions, based onobserved data from the National Renewable Energy Laboratory(NREL) to calculate PV module temperature. Results from themodel confirm that the peak PV module temperature can reachover 80C, which was expected from other techniques, but thatthese peak temperatures occur on average for only 8 minutesper year in locations similar to Tucson, Az. Since the PV moduletemperature is found to be less than 70C for 99% of theoperation hours, thermal management is not onerous and thatthe use of lower cost, commercial grade components will providea mean time between failure (MTBF) to support an inverterwarranty equivalent to that of the PV module itself.

    Index Terms Photovoltaic power systems, photovoltaic cellthermal factors, solar energy, solar power generation, thermalmodeling.

    I. INTRODUCTION

    Accurate prediction of photovoltaic (PV) module

    temperature is needed to understand the expected electrical

    performance, lifetime, and reliability of photovoltaic cells.

    Recent interest in integrating the power electronic inverter

    directly with the PV module [1, 2] means that the power

    electronics circuitry and components will now be exposed to

    the thermal environment on or near the PV module which

    can impose additional requirements on component selectionand circuit design. Understanding the exact ambient thermal

    conditions expected both instantaneously as well as over the

    lifetime of the inverter is particularly important for circuit

    topologies that rely on aluminum electrolytic capacitors,

    which have well-known failure and degradation modes [3]

    that are accelerated under these elevated thermal conditions

    found near a PV module. While other researchers have sought

    to improve the reliability of the power inverter by eliminating

    these electrolytic capacitors altogether [4, 5], the expected

    operating temperature due to the proximity of the electronics

    to the PV module can have an effect on other component

    selection and reliability.

    While it is tempting to apply a worst-case stack upanalysis, this can lead to the conclusion that PV module-

    integrated inverters require industrial grade (105C)

    components [6]. This paper presents a thermal model for the

    PV module and simulation results that use real operating

    conditions, based on observed data from the National

    Renewable Energy Laboratory (NREL) [7] to calculate PV

    module temperature. Although the peak temperature was

    found to be 81C, the PV module temperature exceeded 80C

    for only 2 hours out of the 59,740 total operating hours in the

    15 year dataset an average of 8 minutes per year. These

    results suggest that since the PV module temperature is less

    than 70C for 99% of the operation hours, good thermal

    management will support the use of lower cost commercial

    grade (70C) components. The results can also be used in

    calculating the thermal stress for mean time between failure

    predictions.

    The thermal model presented in this paper uses a control-

    volume approach and takes into account incoming shortwave

    radiation (insolation), electrical conversion efficiency,

    longwave radiative exchange with the sky, earth, and roof, as

    well as free and forced convection from the top and bottom

    of the PV modules. Installation parameters include the roof

    pitch and the PV module tilt, which facilitates studying

    installation options other than parallel to the roof-line. The

    NREL database [7] provides actual meteorological data for

    1,454 locations in the United States and is used with themodel to generate a histogram of PV module operational

    temperatures.The model is designed for a roof-mounted PV

    module but is applicable to other mounting conditions,

    combined heat and power solar, and building-integrated

    solar applications.

    II. PREVIOUS WORK

    Recent work in predicting PV module operating

    The author was a Senior Engineer at SolarBridge Technologies, 2111 S.

    Oak Street, Champaign, IL, USA, 61821 when the work in this paper was

    originally performed.

    3343978-1-4244-2893-9/09/$25.00 2009 IEEE

    Authorized licensed use limited to: AALTO UNIVERSITY. Downloaded on February 8, 2010 at 09:06 from IEEE Xplore. Restrictions apply.

  • 8/12/2019 A Photovoltaic Module Thermal Model

    2/7

    temperature has applied a correlation approach where a

    known reference state, such as the normal operating cell

    temperature (NOCT) condition is extrapolated [8]. Since the

    NOCT is taken at a single operating point, this technique

    does not take into account heat loss that occurs under

    different sky, wind, and ambient conditions [9]. This paper

    extends the work in [10] which applies a first-principles

    based energy balance on the PV module control volume to

    solve for temperature. Accurate thermal models are important

    to understand and predict the performance of the silicon cells

    [11-13]. A recent trend has been to integrate the inverter

    electronics directly with the PV module [6]. Thermal stress

    limits the lifetime of critical components [3], overall mean

    time between failure (MTBF) and warranty programs [14].

    III. MODELING APPROACH

    The thermal model treats the PV module as a total energy

    balance on the control volume [10]:

    0mod =

    ACdt

    dT

    AqPAq

    s

    lossoutsw (1)

    where

    [ ]

    [ ]

    [ ]

    2

    2

    2

    2

    areapanel,

    capacityheat,

    etempertursurfacepanel,

    moduletheofoutfluxheattotal,

    outoutpowerelectrical,

    radiationincidentshortwave,

    mA

    C

    CT

    q

    WP

    q

    Km

    Jmod

    s

    m

    Wloss

    out

    m

    Wsw

    (2)

    The total heat flux out of the PV module is defined as:

    convforcedconvfreecondlwlossqqqqq +++= (3)

    where

    2

    2

    2

    2

    lossheatconvectionforced,

    lossheatconvectionfree,

    lossheatconductive,

    lossradiationlongwave,

    m

    Wconvforced

    m

    Wconvfree

    m

    Wcond

    m

    Wlw

    q

    q

    q

    q

    (4)

    A discussion of each term in (1-4) follows in section IV.

    In general, the PV module temperature, ( )tzyxTs ,,, , is afunction of spatial location on the PV module in addition to

    time-varying environmental conditions. A 3-D solution to (1)

    could be found using numerical finite-element analysis

    methods, which would need to be solved for each time value

    of insolation and weather data. However, since the thickness

    of the PV module is small compared to the area, it can be

    assumed that losses from the edges are small compared to the

    total surface area. Further, the heat fluxes in (3), are assumed

    to be constant across the x-yplane of the PV module. Thus

    the PV module temperature can be represented as a single,

    lumped value for each instant in time and the governing

    differential equation (1) is a function only of time.

    A. PV Module Heat Capacity

    In a standard transient thermal problem, stored energy, Q,is a function of the mass (m), specific heat (cp), and rate of

    temperature change (T) such that

    TcmQ p = . (5)

    In the case of the PV panel, which is a multi-layer laminate,

    the equation can be generalized:

    TcdAQ pm = , (6)

    where A is the PV module area, dm is the thickness of the

    laminate and and cp are composite values of density and

    specific heat, respectively. By considering the composition of

    each layer in the laminate, a composite heat capacity can be

    calculated:layers...1for,mod Nncdc

    nnpnnm == , (7)

    where

    [ ]

    [ ]

    KkgJth

    np

    m

    kgthn

    thnm

    nc

    n

    mnd

    layerofheatspecific,

    layerofdensity,

    layerofthickness,

    3 . (8)

    Table 1 provides an example calculation for a typical PV

    laminate [10] with material properties [15] representative of

    typical PV laminates. The PV module area, A, may be a freeparameter such that all calculations are on a per unit area

    basis, allowing the model to be scaled to arbitrary sized

    installations, which is useful when multiple identical PV

    modules are mounted side-by-side to form large arrays.

    B. NREL and NOAA Databases

    The input to the model is hourly insolation and

    meteorological data collected over the period of 1991- 2005

    [7]. A minimum level of insolation is needed to generate

    Table 1: PV material properties [10]

    PV moduleElement

    Densit

    [kg/m3]

    SpecificHeat

    cp[J/kg-K]

    Thicknessof Layer,

    dm[m]

    dmcp[J/m2-K]

    dmcp[WH/m2-K]

    Silicon PVcells

    2330 677 0.0003 473 0.13

    Polyester -Tedlar

    Trilaminate1200 1250 0.0005 750 0.21

    Glass Face 3000 500 0.003 4500 1.25

    Total 5723 1.59

    3344

    Authorized licensed use limited to: AALTO UNIVERSITY. Downloaded on February 8, 2010 at 09:06 from IEEE Xplore. Restrictions apply.

  • 8/12/2019 A Photovoltaic Module Thermal Model

    3/7

    enough electrical power to power-up the inverter controls. A

    realistic tare value is 50W/m2 (approximately 5% of full

    illumination) which yields a theoretical 10W from a typical

    200W, 1.5m2, 14% efficient PV module. Thus, the analysis

    of the 15 year dataset reveals a total of 59,739 inverter

    operational hours, or a duty cycle of about 45%, for a

    location with insolation and weather similar to Tucson, Az.

    IV. DISCUSSION OF SPECIFIC TERMS

    A. PV Module Electrical Output Power: Pout

    The model supports arbitrary electrical conversion

    efficiency of incoming short wave radiation. Typical values

    for electrical conversion efficiency are 10% to 14% for

    commercially available multi-crystalline cells. The results in

    this paper assume a constant efficiency. In general, the

    efficiency is a function of temperature which adds a non-

    linear term to the governing thermal equation (1).

    B. Long Wave Radiation: qlw

    Long wave radiation exchange is assumed to occurprimarily between the PV module and the sky, earth and roof.

    The heat flux for the top side of the PV module is

    ( )4444

    _

    earthsmeg

    skysmstgtoplw

    TTF

    TTFq

    +=

    , (9)

    where is the Stefan-Boltzmann constant for radiative heat

    transfer and gis the emissivity of the module top surface,typically 0.9 to 1.0. View factorsFmstis from the module tothe sky andFmeis from the module to the earth, both for thetop surface. Tearthis the ground surface temperature, which isassumed to be the ambient air temperature for locations such

    as residential subdivisions with significant vegetation. Thesolar temperature is

    ( ) 25.04ambientskysky TT = (10)where the emissivity of the sky, skyis determined by the dew

    point temperature Tdew[16]:

    +

    +=

    nighttimeduring,0.00620.741

    daytimeduring,0060.0727.0

    dew

    dewsky

    T

    T . (11)

    where Tdewis in C and is obtained from the NREL database

    [7], while Tskyand Tambientare in Kelvin, [K].

    Similarly, the heat flux for the bottom surface of the PV

    module is

    ( )4444

    _

    roofsmrb

    skysmsbbbotlw

    TTF

    TTFq

    +=

    . (12)

    where, bis the emissivity of the module bottom surface and

    is assumed to be equal to g. The roof temperature Troofismore complicated to compute since it is related to theambient temperature, roofing material and construction, andsolar heating of the roofing material near the PV module [15].

    The roof temperature can easily be 20C or greater thanambient temperature on a clear sunny day.

    C. Conductive Loss: qcond

    A thermal conduction path exists from the edge of the PV

    module, through the frame. Given the large surface area of

    the PV module and the small contact area of the frame,

    conductive heat flux can be assumed to be negligible.

    D. Convective losses: qconv

    Convective heat transfer occurs due to free (natural

    buoyancy of hot air) convection qconvfree and forced

    convection qconvforcedcaused by wind. In standard heat transfer

    calculation, if one convective mode dominates, the other is

    usually ignored. In this model, both modes are currently

    taken into account at all times. The average temperature of

    the surface and ambient, Tavg, is used to calculate film

    properties:

    2

    )( ambientsavg

    TTT

    += . (13)

    with a characteristic length of

    )(*2 WH

    AL

    += . (14)

    The kinematic viscosity, , conductivity, k, and Prandtl

    number, Pr, are all determined over the range from 0C to

    90C using curve-fitting software such as EES. Other

    coefficients needed to compute the Grasshof number are

    avgT

    1= (15)

    and

    ambientsdiff TTT = . (16)

    The Grasshof number can then be computed using the

    standard definition

    2

    3)cos(81.9

    LTGr

    diffL

    = (17)

    where is the angle of the PV panel to the vertical. Then, the

    Rayleigh number is calculated by

    PrGrRa LL = . (18)

    and a Function is tabulated as

    9

    16

    16

    9

    ])492.0(1[

    +=Pr

    . (19)

    The Nusselt number is then calculated as

    25.0)(67.068.0 += LL RauN (20)

    which yielding a free convection heat transfer coefficient of

    L

    kuNh Lfree

    = . (21)

    3345

    Authorized licensed use limited to: AALTO UNIVERSITY. Downloaded on February 8, 2010 at 09:06 from IEEE Xplore. Restrictions apply.

  • 8/12/2019 A Photovoltaic Module Thermal Model

    4/7

    In this model the heat transfer is assumed to be the same for

    both top and bottom surfaces, which is a fair assumption at

    near vertical angles but introduce error for PV modules

    mounted flat.

    A method of calculating heat transfer under forced

    convection, regardless of PV module orientation (inclination

    or yaw) can be found in [17], which is useful since the wind

    speed is a scalar value on a horizontal plate [7]. The specific

    heat and density product is found from a curve fit:

    20116391.56864.437.1300avgavg

    TTcprho += (22)

    From this an average heat transfer coefficient can be

    determined:

    3

    2

    5.0

    ,

    Pr

    931.0

    =L

    windcprho

    h tforced

    (23)

    Invoking Newtons Law of Cooling results in

    difftforcedconvforced Thq = ,2 . (24)

    which assumes that both the top and bottom have the same

    heat transfer rates and are both exposed to the wind. In case

    of the panel being flush mounted with no air gap to the roof,

    the 2 factor would be eliminated.

    Free and forced convection are combined into an effective

    convective heat transfer coefficient:

    333freeforcedeff

    NuNuNu = (25)

    such that

    L

    kNuh

    effeff

    = (26)

    and

    diffeffconv Thq = . (27)

    V. IMPLEMENTATION

    Rearranging (1) into a form suitable for simulation yields

    Ac

    AqqqPAq

    dt

    dT

    mod

    convcondlwoutsws

    ++= (28)

    which is a nonlinear, time-varying first-order ordinary

    differential equation (ODE) in terms of PV module

    temperature Ts since the terms qsw, qlw, and qconv are allfunctions of Ts and time t. MATLAB was used as the

    simulation environment and numerical solver. The Backward

    Euler method provided accurate results, was relatively fast to

    converge, and was stable whereas higher order methods, such

    as Runge-Kutta and Adams methods were more robust but

    took longer to simulate and did not provide significantly

    better results.

    The simulation contains three functional blocks as shown

    in Figure 1: data loading, ODE solving, and post-processing.

    The data loading block allows the user to select system

    parameters such as location and PV panel type. The ODE

    solving block solves the above defined model and outputs PV

    module temperature as an hourly data array in C. Once the

    simulation has run for the entire data set the post-processing

    block performs statistical analysis on the data which includes

    generating histograms, cumulative distribution functions, and

    exceedence plots.

    Parameter inputs to the system comprise of insolation and

    meteorological data, PV module characteristic data, and

    installation parameters. Meteorological data included solar

    insolation, S, wind speed, WS, ambient temperature, Tambient,

    and dewpoint temperature, Tdewwhich are available from the

    NREL database [7] and represent hourly measured/estimated

    data for 15 years from 1991 to 2005. PV module data areobtained from manufacturer datasheets. Installation

    User Input:

    Location

    Selection

    User Input:

    PV module

    Selection

    Run thermal model

    Load Insolation and

    Meteorological Data

    Load PV Module

    parameters

    Start

    User Input:

    Installation

    parameters

    Calculate view factors

    and mount angles

    Increment hour

    Solution

    Converged

    NO

    Last data point

    NO

    End

    Post-processing

    statisitcal analysis and

    plotting

    Data Loading

    ODE Solver

    Post Processing

    Figure 1: Simulation flowchart.

    3346

    Authorized licensed use limited to: AALTO UNIVERSITY. Downloaded on February 8, 2010 at 09:06 from IEEE Xplore. Restrictions apply.

  • 8/12/2019 A Photovoltaic Module Thermal Model

    5/7

    parameters are set by the user and can accommodatecorrection for geographic location, time-of-day peaking, as

    well as roof pitch and orientation.

    VI. RESULTS

    Simulations have been performed for a wide variety of

    sites throughout the United States. The following set of

    figures illustrates the type of input and output data for the

    model. The results are based on a typical commercially

    available 195W PV module installed on a roof a 14.0 pitch

    and a PV module tilt of 14.3. The model allows for these

    installation parameters to be arbitrarily varied.

    Figure 2 shows the time traces for the simulation input

    data and the resulting PV module temperature. Figure 3shows the histogram of the PV module temperature during

    the time the inverter is delivering power. In practice, a

    minimum level of insolation is needed to generate sufficient

    electrical power for the inverter controls, which results an

    inverter duty cycle of about 45% for a location with

    insolation and weather similar to Tucson, Az

    The overall shape of the curve reveals that the PV module

    temperature is less that 60C for more than 95% of the time

    and the cumulative distribution function (CDF), in Figure 4,shows that the PV module temperature is less than 70C for

    99% of operation hours. Although the peak PV module

    temperature was found to be 81C, the PV module

    temperature exceeded 80C for only 2 hours out of the 59,740

    total operating hours as shown in Figure 5. The CDF and

    exceedence plots reveal that the module-integrated inverter

    need not be designed for continuous duty at or above 80C.

    Using the temperature histogram the design engineer can

    compute a composite MTBF, which may be a weighted

    average of the MTBF at a number of different temperatures,

    which will reflect actual expected operation.

    The PV module temperature is affected by different

    installation methods such as if mounted close to the roofwhich blocks airflow on the bottom side or on a rack system

    with unimpeded air flow. Further, different roofing materials

    also affect the module temperature. All of these installation-

    specific variations may be considered simultaneously,

    individually, or by using a statistical technique such as

    MonteCarlo analysis to determine the design margin

    sufficient for an intended application.

    When multiple geographic regions are compared, the data

    reveals that the hottest ambient air temperature does not

    2 4 6 8 10 12

    x 104

    0

    500

    1000

    Solar Insolation for 15 year(s) (NREL dataset: tucson )

    [W/m2]

    2 4 6 8 10 12

    x 104

    0

    20

    40Ambient Temperature for 15 year(s) (NREL dataset: tucson )

    [oC]

    2 4 6 8 10 12

    x 104

    0

    5

    10Sky cover for 15 year(s) (NREL dataset: tucson )

    [numberoftenth]

    2 4 6 8 10 12

    x 104

    0

    10

    Wind Speed for 15 year(s) (NREL dataset: tucson )

    [m/s]

    2 4 6 8 10 12

    x 104

    020

    4060

    PV module temperature for 15 year(s) with Heat Capacity = 1.59 (NREL dataset: tucson )Range: -12 ~ 79oC

    [oC]

    Time (hour)

    Figure 2: Time trace for temperature and weather data for Tucson, Az.

    3347

    Authorized licensed use limited to: AALTO UNIVERSITY. Downloaded on February 8, 2010 at 09:06 from IEEE Xplore. Restrictions apply.

  • 8/12/2019 A Photovoltaic Module Thermal Model

    6/7

    always result in the hottest PV module temperature. This is

    significant since many empirical system-design rules predict

    worst-case PV module temperature by adding a constant

    offset to the worst-case ambient air temperature. However,

    sky conditions and wind speed are not considered in these

    approaches yet the model shows that they have a significant

    impact on module temperature.

    Results from the thermal model and simulation also can

    provide daily temperature cycle data, as shown in Figure 6.

    Diurnal thermal cycling can have a significant impact oninverter reliability due to mis-matched coefficient of thermal

    expansion as well as stress and fatigue on mounting

    hardware, connectors, and solder connections. Comparing

    diurnal PV module temperature cycles in different geographic

    regions revealed that the region with the worst-case peak

    module temperature did not necessarily have the greater daily

    thermal cycle.

    VII. CONCLUSION

    Accurate prediction of PV module temperature is

    important to understand performance, reliability and lifetime

    of a PV module-integrated inverter. Whereas other techniques

    correlate and extend a single measured operating point, this

    paper presents a first-principles thermal model and simulation

    methodology that uses an energy-balance approach along

    with observed insolation and meteorological data to compute

    the temperature of the PV module using historical, measured

    data. The result is an hour-by-hour PV module temperature

    which is post-processed to produce a histogram of

    temperature distribution, cumulative-distribution function,

    and exceedence plot. Additional post-processing on the PV

    module temperature data can produce distributions of

    expected diurnal temperature cycles. The complete set of

    governing thermal equations needed to directly implement

    the model has been included.

    The data provided from this thermal model can be used by

    design engineers to optimize the component selection, cost

    and thermal management of the power electronic inverter in

    -20 -10 0 10 20 30 40 50 60 70 80 900

    0.5

    1

    1.5

    2

    2.5

    3His togram of PV panel tem perature for 15 years during inverter operation - 59740 operating hours

    Temperature [oC]

    Percentageoveroperatinghours[%]

    Figure 3: Example of a PV module temperature histogram for inverter operational hours.

    -20 0 20 40 60 800

    0.2

    0.4

    0.6

    0.8

    1

    Cumulative Dis tribution Function

    Temperature [

    o

    C]

    Percentageoftotalhours

    PVpaneltemperature

    temperature

    Figure 4: Example of a cumulative distribution function for PV moduletemperature.

    74 76 78 80 820

    20

    40

    60

    80

    100High Temperature

    Exceedence Temperature [o

    C]

    Numberofhours

    Figure 5: Example of an exceedence temperate plot.

    3348

    Authorized licensed use limited to: AALTO UNIVERSITY. Downloaded on February 8, 2010 at 09:06 from IEEE Xplore. Restrictions apply.

  • 8/12/2019 A Photovoltaic Module Thermal Model

    7/7

    order to meet cost and reliability objective and achieve the

    long lifetime and high reliability needed from a PV module-

    integrated inverter.

    REFERENCES

    [1] Q. Li and P. Wolfs, "A Review of the Single Phase PhotovoltaicModule Integrated Converter Topologies With Three Different DCLink Configurations," IEEE Transactions on Power Electronics, vol.23, no. 3, pp. 1320-1333, May 2008.

    [2] S. B. Kjaer, J. K. Pedersen, and F. Blaabjerg, "A review of single-phasegrid-connected inverters for photovoltaic modules,"IEEE Transactionson Industry Applications, vol. 41, no. 5, pp. 1292-1306, Sept.-Oct.2005.

    [3] J. L. Stevens, J. S. Shaffer, and J. T. Vandenham, "The service life oflarge aluminum electrolytic capacitors: effects of construction andapplication," IEEE Transactions on Industry Applications, vol. 38, no.5, pp. 1441-1446, Sept.-Oct. 2002.

    [4] P. T. Krein and R. S. Balog, "Cost-Effective Hundred-Year Life forSingle-Phase Inverters and Rectifiers in Solar and LED LightingApplications Based on Minimum Capacitance Requirements and aRipple Power Port," in Applied Power Electronics Conference and

    Exposition, 2009. APEC 2009. Twenty-Fourth Annual IEEE, pp. 620-625.

    [5] A. C. Kyritsis, N. P. Papanikolaou, and E. C. Tatakis, "A novel ParallelActive Filter for Current Pulsation Smoothing on single stage grid-connected AC-PV modules," in Proceedings, IEEE EuropeanConference on Power Electronics and Applications, 2007, pp. 1-10.

    [6] B. Sahan, N. Henze, A. Engler, et al., "System design of compact low-power inverters for the application in photovoltaic AC-modules," in 5th

    International Conference on Integrated Power Electronic Systems,CIPS 2008, VDE Verlag GmbH, Berlin, Germany, March 2008, pp.187-92.

    [7] NREL, "National Solar Radiation Data Base (NSRDB) 1991- 2005Update," National Renewable Energy Laboratory (NREL). Available:http://rredc.nrel.gov/solar/old_data/nsrdb/1991-2005/.

    [8] E. Skoplaki and J. A. Palyvos, "Operating temperature of photovoltaicmodules: a survey of pertinent correlations," Renewable Energy, vol.34, no. 1, pp. 23-29, 2009.

    [9] E. Skoplaki, A. G. Boudouvis, and J. A. Palyvos, "A simple correlationfor the operating temperature of photovoltaic modules of arbitrary

    mounting," Solar Energy Materials and Solar Cells, vol. 92, no. 11, pp.1393-1402, Nov 2008.

    [10] A. D. Jones and C. P. Underwood, "A thermal model for photovoltaicsystems," Solar Energy, vol. 70, no. 4, pp. 349-359, Mar 8, 2001.

    [11] K. Emery, J. Burdick, Y. Caiyem, et al., "Temperature dependence ofphotovoltaic cells, modules, and systems," in Record, IEEEPhotovoltaic Specialists Conference, 1996, pp. 1275-1278.

    [12] A. H. Fanney, M. W. Davis, B. P. Dougherty, et al., "Comparison ofphotovoltaic module performance measurements," Transactions of theASME. Journal of Solar Energy Engineering, vol. 128, no. 2, pp. 152-9, May 2006.

    [13] G. Notton, M. Mattei, C. Cristofari, et al., "Calculation of thepolycrystalline PV module temperature using a simple method ofenergy balance," Renewable Energy, vol. 31, no. 4, pp. 553-67, Apr.2006.

    [14] S. Vittal and R. Phillips, "Modeling and Optimization of ExtendedWarranties Using Probabilistic Design," in Proceedings, IEEE

    Reliability and Maintainability Symposium, 2007, pp. 41-47.[15] "ASHRAE Fundamentals Handbook." Atlanta, Ga.: American Society

    of Heating Refrigerating and Air-Conditioning Engineers, 2001.[16] A. F. Mills,Heat and mass transfer.Burr Ridge, Ill.: Irwin Inc., 1995.[17] E. M. Sparrow and K. K. Tien, "Forced convection heat transfer at an

    inclined and yawed square plate-application to solar collectors,"Transactions of the ASME, Journal of Heat Transfer, vol. 99, no. 4, pp.507-12, Nov 1977.

    -5 0 5 10 15 20 25 30 35 40 45 50 55 60 650

    5

    10

    15

    20

    25PV Panel Daily Temperature Cycles for 15 years (NREL dataset: tucson)

    Temperature Difference |T| [oC]

    Percentageofcyclesfor15years

    Figure 6: Example of PV module daily temperature cycles.

    3349