Effect of Temperature and Humidity on the Degradation Rate of Multicrystalline Silicon Photovoltaic Module

Abstract

In a PV module, the relative humidity (rh) of a front encapsulant is different from that of a backside encapsulant (rh_back). In this study, the effective humidity (rh_eff) in a PV module was investigated to study the effects of moisture variation on the degradation rate (𝑅_𝐷). rh_eff represents uniform humidity in a PV module when it is exposed to certain damp heat conditions. Five types of accelerated tests were conducted to derive the relation between rh_eff and rh_back. rh_eff showed a linear relationship with rh_back at constant temperature. Two types of models, namely, Eyring and Peck models, were used for predicting the 𝑅_𝐷 of PV modules, and their results were compared. The 𝑅_𝐷 of PV modules was thermally activated at 0.49 eV. Furthermore, the temperature and rh_eff history of PV modules over one year were determined at two locations: Miami (FL, USA) and Phoenix (AZ, USA). The accumulated 𝑅_𝐷 values based on the temperature and rh_eff of the modules were calculated by summing the hourly degradation amounts over the time history.

1. Introduction

Moisture can diffuse into photovoltaic (PV) modules through their breathable back sheets or their ethylene vinyl acetate (EVA) sheets [1]. When in service in hot and humid climates, PV modules experience changes in the moisture content, the overall history of which is correlated with the degradation of the module performance [1]. If moisture begins to penetrate the polymer and reaches the solar cell, it can weaken the interfacial adhesive bonds, resulting in delamination [2] and increased numbers of ingress paths, loss of passivation [3], and corrosion of solder joints [4, 5]. Of these possibilities, the occurrence of corrosion has one of the highest frequencies in outdoor-exposed PV modules [6]. Significant losses in PV module performance are caused by the corrosion of the cell, that is, the SiNx antireflection coating, or the corrosion of metallic materials, that is, solder bonds and Ag fingers [7, 8]. Corrosion is defined as the destructive chemical or electrochemical reaction of a metal with its environment. The moisture from the environment may lead to electrochemical reactions that can result in corrosion. For the electrochemical reaction of metals with their environment, an aqueous, ion-conduction enabling environment is necessary; moreover, at high temperatures, gas-metal reactions are possible [9]. The International Electrotechnical Commission (IEC) 61215 test defines a damp heat (DH) test in 10.13. The DH test is conducted for determining the effect of long-term penetration of humidity on materials. Therefore, many researchers have studied the reliability of PV modules on the basis of IEC 61215. Laronde et al. [10] have employed DH testing to study the degradation of PV modules subjected to corrosion. Peike et al. [11] have reported that grid corrosion or reduced conductivity between the emitter and grid is the most likely cause of DH-induced degradation. They have also shown that high temperatures accelerate water vapor permeation into the module and the subsequent degradation reactions. Furthermore, it has been reported that loss of adhesion strength is exacerbated by exposure to high humidity environments [2] and facilitates the delamination of EVA from the cell [3], which results in grid corrosion [12]. Therefore, in order to understand the effect of grid corrosion on the degradation of PV modules, one needs to comprehend how the surrounding environment affects the module temperature (𝑇_𝑚) and moisture content of the module. Kempe [1] has suggested that analytical equations can be used to determine timescales for moisture ingress with breathable back sheets. Koehl et al. [13] have showed that the maximum moisture concentration in front of a cell was not reached after 1,000 h under DH conditions at 85°C and 85% rh because of the long path to the back sheet. They have also documented that the humidity in front of the cell is not directly influenced by ambient fluctuations in actual weather conditions.

As mentioned above, PV modules are degraded by ambient temperature and humidity; moreover, these factors can accelerate the degradation. This degradation is mainly caused by corrosion [4, 5, 7, 8, 10, 11]. It can be assumed that the temperature of a PV module is uniform; however, moisture concentration in a PV module is not uniform. Therefore, it is difficult to predict moisture-induced degradation.

Therefore, in this study, effective humidity (rh_eff) in a PV module was investigated to study the effect of moisture variation in a PV module. rh_eff represents uniform humidity in a PV module when it is exposed to certain DH conditions. This paper documents the relation of rh_eff with ambient temperature and humidity. Moreover, the resultant module degradation rate (𝑅_𝐷) for various rh_eff values is also reported in this paper. These data allow the prediction of real-world thermal- and moisture-induced 𝑅_𝐷 values of PV modules and enable the computation of acceleration factor (AF) for the DH test.

2. Experiments

Six-inch multicrystalline Si (m-Si) solar cells were used in this study. The typical characteristics of the cells at a light intensity of 1 sun were approximately as follows: open-circuit voltage (𝑉_oc), 0.60 V; short-circuit current density (𝐽_sc), 33.9 mA/cm²; fill factor (FF), 0.72; and conversion efficiency, 16.8%.

A copper ribbon wire, which was plated with 62Sn36Pb2Ag solder, was used for cell interconnection. The dimensions of the ribbon wire were 0.15 mm × 1.5 mm. The samples were divided into two groups on the basis of lamination conditions. Type 1 was arranged with EVA, cell, and EVA. Type 2 was laminated with low-iron glass of area 180 mm × 180 mm and thickness 3.2 mm, an EVA of thickness 0.35 mm, a cell, an EVA, and a TPT back sheet of thickness 0.35 mm, as shown in Figures 1(a), 1(b), and 1(c). The solar cells were laminated with EVA by heating them up to 150°C for 12 min. After encapsulation, the samples were divided into five groups and exposed to accelerated stresses, as shown in Table 1. Five kinds of test conditions were selected: 85°C, 85% rh (8585), 65°C, 85% rh (6585), and 45°C, 85% rh (4585) for temperature acceleration and 65°C, 75% rh (6575), and 65°C, 65% rh (6565) for humidity acceleration. A total of five cases, each with ten samples, were tested. DH conditions were obtained in an environmental chamber (Hygros340C, ACS Co., Massa Martaba, Italy).

Table 1 

Test conditions for accelerated tests.

Figure 1 

Test samples for DH tests: (a) PV module with EVA/cell/EVA structure (Type 1), (b) PV module with glass/EVA/cell/EVA/back sheet structure (Type 2).

The electrical performance of each sample was measured every 200 h using a solar simulation system (K202 Lab200, Mac Science, Seoul, South Korea). The standard test conditions were (a) irradiance: 1000 W/m², (b) cell temperature: 25°C, and (c) spectral distribution of irradiance: AM 1.5 G (IEC 60904-3).

Accelerated tests (ATs) were conducted for 4,500 h for 8585, 6585, and 4585 and for 3,200 h for 6575 and 6565. The maximum power of all samples decreased by over 5% after the ATs. Using the results, the average 𝑅_𝐷 for each condition was calculated.

3. 𝑅_𝐷 Prediction Models and Effective Humidity

𝑅_𝐷 data can be postulated with an empirical kinetic model by assuming that the rate of degradation is proportional to the concentration of water in PV modules, and that the rate constant has Arrhenius temperature dependence.

Escobar and Meeker [14] has proposed two degradation models: variations of Eyring and Peck models, which are used for operating conditions when temperature and humidity are the accelerated stresses in a test. The degradation rate based on the Eyring model (𝑅_𝐷,Eyring) is given by 𝑅_𝐷,Eyring=𝐴exp−𝐸_𝑎−𝑏𝑘𝑇rh, (1) where 𝐸_𝑎 is the thermal activation energy of the degradation process (eV), 𝑘 is the Boltzmann constant (8.62×10⁻⁵ eV/K), 𝑇 is the temperature (°K), and rh is the relative humidity (%). 𝐴 and 𝑏 are two constants dependent on the failure mode. 𝑅_𝐷 (%/h) is the inverse of the mean time to failure (MTTF) at a given condition. In order to obtain 𝑅_𝐷, we determined the time to 5% reduction in the initial maximum power (𝑃_max) at all samples. The life distributions were determined with ALTA 7 software. Using the results, the MTTF was calculated.

The other degradation model based on the Peck model (𝑅_𝐷,Peck) is expressed as follows: 𝑅_𝐷,Peck=𝐵exp−𝐸_𝑎⋅()𝑘𝑇rh^𝑛, (2) where 𝐵 and 𝑛 are two constants dependent on the failure mode.

In order to obtain 𝐴, 𝐸_𝑎 in (1) and 𝐵, 𝑛 in (2), the equations can be represented on a logarithmic scale by a straight line, using the following equations: ln𝑅_𝐷,Eyring(𝐴)−𝐸=ln^𝑎−𝑏𝑘𝑇,𝑅rhln_𝐷,Peck(𝐵)−𝐸=ln^𝑎().𝑘𝑇+𝑛lnrh (3)

A plot of the left-hand side of (3) versus 1/𝑇 (°K) gives an Arrhenius plot with a slope 𝐸_𝑎/𝑘 and an intercept ln(𝑘₀).

ln(𝑅_𝐷) versus 1/𝑇 (°K) gives an Arrhenius plot with a slope 𝐸_𝑎/𝑘 and intercepts ln(𝐴) and ln(𝐵).

The rh in PV modules depends on their ambient climate, material (encapsulant, back sheet) properties, and the operation conditions. If all information is available, moisture concentration in the backside encapsulant can be calculated using the diffusion model [1]. However, the moisture concentration in the front encapsulant is different from that at the back of the cell because of the long path from the back sheet [13]. In a 85°C, 85% rh test, the maximum moisture content in the backside encapsulant was reached quickly; however, the maximum moisture content of the front encapsulant was not reached even after 1,000 h of exposure time [13]. Therefore, rh_eff of a PV module was considered to determine the uniform moisture content in the module.

In order to derive the effect of uniform humidity in a PV module on its degradation, PV modules with a EVA/cell/EVA structure (Type 1) were used as shown in Figure 1(a). This is because the cell and EVA layer structure allow water vapor to permeate the surface of the cell within minutes during the DH test [15].

In case of Type 1, it can be assumed that the rh in the PV module is uniform. If 𝐸_𝑎 and constants (𝐴, 𝐵, etc.) are determined, the degradation rate of Type 1 (𝑅_𝐷,Type1) can be predicted with 𝑅_{𝐷,Type1,Eyring}=𝐴₁exp−𝐸_𝑎,Type1𝑘𝑇₁−𝑏₁rh₁,𝑅_{𝐷,Type1,Peck}=𝐵₁exp−𝐸_𝑎,Type1𝑘𝑇₁⋅rh₁^𝑛1. (4) However, in case of glass/EVA/cell/EVA/back sheet structure (Type 2), rh is not uniform in the PV module. Therefore, it is difficult to fix the rh in the PV module. If 𝑅_𝐷,Type2 is known, 𝑅_𝐷,Type2 is expressed with the right-hand side of (4), and rh_eff is as follows: 𝑅_𝐷,Type2=𝐴₁exp−𝐸_𝑎,Type1𝑘𝑇₁−𝑏₁rh_eff,Eyring,𝑅_𝐷,Type2=𝐵₁exp−𝐸_𝑎,Type1𝑘𝑇₁⋅rh_eff,Peck^𝑛1, (5) where 𝐸_𝑎,Type1 and other parameters (𝐴₁, 𝑏₁, 𝐵₁, and 𝑛₁) can be determined by ATs as shown in Table 1. Using (5), rh_eff is rearranged as follows: rh_eff,Eyring=−𝑏₁ln𝑅_𝐷,Type2−ln𝐴₁+𝐸_𝑎,Type1/𝑘𝑇₁,rh_eff,Peck=𝑅_𝐷,Type2𝐵₁exp(−𝐸_𝑎,Type1/𝑘𝑇₁)^1/𝑛1. (6)(7)

4. Results

4.1. Results of Accelerated Tests

The changes in average normalized 𝑃_max of the modules in Type 1 as a function of time are plotted in Figure 2. The figure shows that 𝑃_max decreases linearly over time. The results correspond well with those found in earlier studies [6, 16]. Several authors [17, 18] argue that the limited experimental evidence available is not enough to take the linear 𝑅_𝐷 for granted and suggest that an exponential degradation rate could be a more suitable trend as is the case in some optoelectronic devices. However, it should be noted that both trends exhibit very similar evolution during the first 10–15 years; if similar initial annual degradation rates are assumed, then the linear degradation rate is a more pessimistic estimate [19].

Figure 2 

Average normalized 𝑃max of Type 1 as a function of time.

In order to keep the discussions simple, we will not consider the exponential degradation rate here.

Using AT results of Type 1 samples, the 𝑅_𝐷 was determined for each set of conditions.

𝐸_𝑎 is obtained by fitting the 𝑅_𝐷 data in (1) and (2) for three values of temperature. 𝐸_𝑎,Type1 and other constants were calculated as summarized in Table 2.

Table 2 

Activation energy and constants of Type 1 samples for Eyring and Peck models.

𝑅_𝐷,Type2 of the modules was also determined for each set of conditions. The 𝑅_𝐷 values (%/h) of 8585, 6585, 4585, 6575, and 6565 were 0.00611, 0.00208, 0.00087, 0.00132, and 0.00079, respectively.

In order to derive the relationship between rh_eff and rh in the backside encapsulant, 𝑅_𝐷,Type2 in (4) was replaced with 𝑅_𝐷 values. Figure 3 shows a plot of rh_eff versus rh in the backside encapsulant at a constant temperature and constant humidity. In both models, rh_eff versus rh has a linear relation at a constant temperature, and rh_eff decreases linearly with temperature at constant humidity (inset in Figure 3). Therefore, rh_eff is expressed as follows: rh_eff,Eyring=⋅[(73.521−0.0275⋅𝑇)]0.8555rh_back+0.4114,71.2rh_eff,Peck=⋅[(74.417−0.0300⋅𝑇)]0.8552rh_back+0.1584.71.9 (8)(9)

(a)

(b)

(c)

(d)

Figure 3 

rheff versus rh for the backside EVA and rheff versus module temperature are plotted for two 𝑅𝐷 prediction models: (a) rheff versus rh for the backside EVA at constant temperature for Eyring model, (b) rheff versus module temperature at constant humidity for Eyring model, (c) rheff versus rh for the backside EVA at constant temperature for Peck model, (d) rheff versus module temperature at constant humidity for Peck model.

Using rh_eff values at AT conditions, predicted 𝑅_𝐷s_,Type2 of (5) are plotted as a function of time in Figure 4. It shows that the predicted 𝑅_𝐷,Type2 is almost identical to the measured data.

Figure 4 

Average normalized 𝑃max of Type 2 as a function of time. □ shows the measurement data, dotted line (black) shows the predicted degradation rate obtained using (1), and straight line (red) shows the predicted degradation rate obtained using (2).

4.2. Acceleration Factor

AF is defined as the ratio between the 𝑅_𝐷 at a given temperature, rh_eff and 𝑅_𝐷 at a reference temperature, and rh_eff (in our case 25°C, 50% rh) [20]. AF is expressed as AF_Eyring−𝐸=exp^𝑎,Type1𝑘1𝑇₀−1𝑇+𝑏₁1rh₀−1rh_eff,AF_Peck−𝐸=exp^𝑎,Type1𝑘1𝑇₀−1𝑇rh₀rh_eff^𝑛1, (10) where 𝑇₀ and rh₀ are the reference temperature and reference humidity, respectively.

Figure 5 shows the AF contours on plots of rh_eff versus module-temperature data. Estimated AFs for the 8585 condition versus 25°C, 50% rh range from 190× to 250× for the Peck and Eyring models, respectively. However, in real-world operation, a constant temperature and rh are not realistic. To project field 𝑅_𝐷 at a specific location, it was characterized by the annual module temperature and rh of the backside encapsulant.

(a)

(b)

Figure 5 

AF Contours of module temperature and rheff for two 𝑅𝐷 prediction models: (a) Eyring model, (b) Peck model.

5. 𝑅_𝐷 in Two Benchmark Climates

Two benchmark climates (BMCs) were selected to quantify the stress: Miami (FL, USA) and Phoenix (AZ, USA). The module-temperature history exposed to the two BMCs was derived from meteorological data [21]. Meteorological data for Miami, FL, for 2005 and Phoenix, AZ, for 2002 were obtained from the National Climatic Data Center.

The rh of the backside encapsulant was calculated using the model of moisture ingress [1]. We determined the activation energy for encapsulant solubility in a previous research [22]. 𝐸_𝑎 for maximum water vapor transmission rate (WVTR_max) of back sheets was evaluated. The transient WVTR was measured using a WVTR instrument (Permatran-W^3/33, Mocon, Minneapolis, MN, USA) at 25°C, 37.8°C, and 50°C. Assuming an Arrhenius equation between WVTR_max and temperature, WVTR_max can be described by WVTR_max=𝐴exp−𝐸_𝑤.𝑘𝑇 (11) Parameters 𝐴 and 𝐸_𝑤 were obtained from a curve fit of a plot of of ln(WVTR_max) versus 1/𝑘𝑇. The 𝐴 and 𝐸_𝑤 values are 2.703×10⁶ g/m²/day and 0.367 eV, respectively.

The accumulated 𝑅_𝐷 based on the module temperature and rh_eff can be calculated by summing the hourly degradation amounts over the time history, as given by 𝑅_𝐷,Type2⋅𝑡=𝐴₁exp−𝐸_𝑎,Type1𝑘𝑇₁−𝑏₁rh_eff,Eyring⋅𝑡,𝑅_𝐷,Type2⋅𝑡=𝐵₁exp−𝐸_𝑎,Type1𝑘𝑇₁⋅rh_eff,Peck^𝑛1⋅𝑡. (12)

Thermal and rh_eff,Eyring history of a PV module over one year were calculated at two BMCs as shown in Figure 6. Compared with Miami, the module temperature at Phoenix is shifted more than 10°C higher and the rh_eff has a relatively low distribution.

Figure 6 

Module temperature and rheff over one year in 2 benchmark climates. rheff is determined with (8) for Eyring model.

Accumulated 𝑅_𝐷 in the two BMCs is calculated using (12), which are based on the module temperature and rh_eff of Figure 6.

The normalized 𝑅_𝐷,Eyring and 𝑅_𝐷,Peck of the PV module exposed to two BMCs are plotted over one year in Figure 7(a). The accumulated 𝑅_𝐷,Peck was 3.0% greater than 𝑅_𝐷,Eyring in Miami. The conditions of 8585 are normally used for DH tests as IEC 61215. Therefore, the accumulated 𝑅_𝐷 exposed to 8585 was evaluated and compared to the accumulated 𝑅_𝐷 in the Miami. The accumulated 𝑅_𝐷,Eyring and 𝑅_𝐷,Peck for exposure to 8585 were about 23.7 and 23.1 times greater than those in Miami, as shown in Figure 7(b). Koehl et al. [13] have reported that AF of 8585 is about 23 for a tropical site (Serpong, Indonesia), 40 for an arid site (Sede Boqer, Israel), and 106 for an alpine site (Zugspitze, Germany) for a degradation process with an activation energy of 0.416 eV.

(a)

(b)

Figure 7 

Normalized 𝑅𝐷 exposure to actual weather conditions for 1 year and AF of accelerated conditions: (a) normalized 𝑅𝐷,Eyring and 𝑅𝐷,Peck for an initially dry module after exposure to two BMCs, (b) AF exposure to 85°C, 85% rh and 65°C, 85% rh is compared to that at Miami (FL, 2005).

In case of 6585, the 𝑅_𝐷 values according to Eyring and Peck models are 9.5 and 9.3 times greater than those in Miami. Therefore, it can be supposed that the Peck model is a slightly conservative estimation for 𝑅_𝐷 prediction.

6. Conclusions

The objectives of this study were to investigate the relation of ambient temperature and humidity with the rh_eff of a PV module and to use the rh_eff values to predict the 𝑅_𝐷 values under actual weather conditions.

The degradation of PV modules is accelerated by temperature and humidity [1, 4, 5, 10, 11]. It can be assumed that the temperature in a PV module is uniform. However, the moisture concentration is not uniform [13]. Therefore, we used a Type 1 module with a EVA/cell/EVA structure for uniform humidity in the module.

Two types of models, namely, Eyring and Peck models, were used for 𝑅_𝐷 prediction, and their results were compared with each other. Five types of ATs were conducted to determine 𝐸_𝑎 and humidity dependence. The 𝑅_𝐷s of PV modules were thermally activated at 0.49 eV.

The moisture content in a PV module is dependent on material properties such as back sheets and EVA. Therefore, we determined the relation between the rh of the backside encapsulant and rh_eff. rh of the backside encapsulant was calculated using the moisture ingress model. Thermal and rh_eff history of a PV module over one year were calculated in two BMCs. For exposure to 8585, the accumulated 𝑅_𝐷,Eyring and 𝑅_𝐷,Peck values at Phoenix were, respectively, about 23.7 and 23.1 times greater than those at Miami.

Abbreviation:

AF:	Acceleration factor
AT:	Accelerated tests
BMC:	Benchmark climate
DH:	Damp heat
EVA:	Ethylene vinyl acetate
FF:	Fill factor
IEC:	International Electrotechnical Commission
𝐽sc:	Short-circuit current density
MTTF:	Mean time to failure
PV:	Photovoltaic
𝑅𝐷:	Degradation rate
𝑅𝐷,Eyring:	Degradation rate based on the Eyring model
𝑅𝐷,Peck:	Degradation rate based on the Peck model
rh:	Relative humidity
rhback:	Relative humidity of a backside encapsulant
rheff:	Effective humidity
𝑇𝑚:	Module temperature
𝑉oc:	Open-circuit voltage.

Acknowledgments

This work was supported by the New and Renewable Energy of the Korea Institute of Energy Technology Evaluation and Planning (KETEP) Grant funded by the Ministry of Trade, Industry and Energy (MOTIE) (no. 2012T100100605). Furthermore, this work was supported by the Human Resources Development of the Korea Institute of Energy Technology Evaluation and Planning (KETEP) Grant funded by the Korea Governments Ministry of Knowledge Economy (no. 20104010100640).