Multi-Climate Simulation of Temperature-Driven Efficiency Losses in Crystalline Silicon PV Modules with Cost–Benefit Thresholds for Evaluating Cooling Strategies

Abstract

We explored the impact of high operating temperatures for monocrystalline silicon photovoltaic (PV) modules which dominate the market. Using nine years of hourly climate data with the System Advisor Model (SAM), we examined temperature impacts and cooling potential benefits across three climate zones in the United States. Assuming that cooling approaches can achieve a constant temperature decrease of ΔT independent of irradiance and environmental conditions, our simulations show that a ΔT = 10 °C temperature reduction could improve energy yield by almost 3% annually. Cooling technologies have the strongest impact during the hottest months, with even a 5 °C reduction raising efficiency by nearly 10%. When the minimum temperature of the cooled module is constrained to the ambient temperature, ΔT = 20 °C boosts the hottest month energy yield by over 25%. For economically viable cooling systems, the cooling cost should be much less than the break-even cost. We estimate break-even costs of USD 25–40/m² for 10 °C and USD 40–60/m² for 20 °C cooling for the locations simulated. For ΔT > 20 °C, the added energy yield shows diminishing returns with minimum increase in break-even costs.

1. Introduction

The National Renewable Energy Laboratory’s (NREL) champion module efficiency chart [1] shows that over the last decade, the highest efficiencies for monocrystalline silicon (c-Si) PV modules have increased from 22.8% to 25.4%, representing an improvement of 2.6%. These efficiencies are measured at Standard Test Conditions (STC) with 1000 W/m² irradiance and a cell temperature of 25 °C. The temperature coefficient, which quantifies the relative change in the output power versus temperature changes, has improved from around −0.5%/°C [2] to a range of −0.25%/°C to −0.4%/°C [3]. In practice, Nominal Operating Cell Temperatures (NOCTs) are about 45 °C with an ambient temperature of 20 °C, making the temperature-dependent contribution to the efficiency a dominant factor in practice. As further improvements in optical and electrical performance become increasingly expensive, mitigating thermal losses is gaining attention as a cost-effective path to improving system output [4]. Various cooling strategies have been proposed; however, practical implementations have been slow. We propose a cost–benefit strategy to help evaluate cooling technologies, using the extra electricity from cooling and local electricity prices to estimate the acceptable cost.

We chose c-Si PV modules as the focus of this study because, despite the availability of multiple PV technologies, monocrystalline silicon dominates both the U.S. and global solar markets. According to the NREL’s Fall 2024 Solar Industry Update [5], monocrystalline silicon modules continue to account for more than 90% of global PV capacity, showing their market dominance and technological maturity.

Various cooling strategies, including spray cooling, liquid cooling, phase-change materials (PCMs), and radiative coatings, have shown significant temperature reductions in lab tests [6,7,8]. Passive cooling methods like air and water spray cooling are simple and can lower cell temperatures by about 10 °C. In parallel, optical strategies such as spectral-selective coatings and nanophotonic structures have been developed to suppress heat generation at the source [9], with radiative sky cooling demonstrating sub-ambient daytime cooling without energy input [10]. Modeling studies further suggest that integrating radiative cooling with sub-bandgap reflection may enhance efficiency without compromising PV surface compatibility [11]. To achieve more substantial temperature reductions up to 30 °C or more, water cooling, PCMs, or hybrid designs are typically required. Under test conditions, these advanced strategies can boost relative PV efficiency well over 10%, and even up to 50% [12]. In contrast, practical implementations often deliver more modest improvements: fins and free convection can enhance efficiency by 2–2.5%, and adding fans raises this to about 3%, albeit with added operational costs [13]. However, deployment feasibility also depends heavily on system complexity and cost. Passive solutions like aluminum fins or thermosiphon cooling remain relatively affordable (USD 25/m² to USD 58/m²), while advanced systems like PCM or forced convection fans can cost from USD 68/m² to over USD 1000/m², depending on their design [14]. This big cost difference highlights the need to weigh performance gains against real-world economic viability.

In evaluating the technical and economic feasibility of cooling strategies, accurate temperature- and irradiance-dependent performance models are essential that are more comprehensive than STC, covering the full range of operating environments. The linear temperature coefficient does not include environmental and mounting condition effects on cell temperature and can introduce significant prediction errors, impacting long-term energy yield and levelized cost of energy (LCOE) calculations. Furthermore, regional differences in climate require location-specific performance assessment [15].

Recognizing these limitations, this paper focuses on establishing a performance evaluation framework for PV systems operating in high-temperature environments. Using the Faiman cell temperature model, which considers wind speed and provides a more realistic temperature estimation than STC-based models, we analyzed high-resolution, multi-climate data to evaluate how cooling can reduce temperature-induced efficiency losses. Our emphasis was on quantifying the economic and technical feasibility of cooling strategies to support decision-making for PV system deployment in hot climates. To provide representative yet comparable scenarios, three cities with similar latitudes but distinct climate types were selected. These locations not only reduce variation in sun position, but also represent relatively warm regions, making them suitable for evaluating heat-related efficiency losses.

2. Climate-Dependent Cell Temperature and Energy Yield Simulations

The simulation workflow in this study is illustrated in Figure 1. For a given location, weather and irradiance information are input along with module and mounting characteristics.

Temperature models were used to estimate the cell temperature using ambient temperature and wind speed information. Cooling strategies were simulated by a constant decrease in cell operating temperature $\left(\Delta T\right)$, independent of irradiance and environmental conditions. The simulated outputs were the annual energy yield and the LCOE. For annualized calculations of the LCOE, energy yield was defined over a one-year period. In addition, the average module efficiency can be calculated for any time period. The red box highlights the critical need for location-specific cell temperature (T_cell) model selection. We obtained weather data from the National Solar Radiation Database (NSRDB). The modules and mounting configurations were defined in SAM.

2.1. Climate Sites and Simulation Setup

We selected three American cities with similar latitudes (around 30° N) but different climate types: El Paso, TX (hot and dry), College Station, TX (subtropical humid), and Jacksonville, FL (coastal humid). We used system-level simulations in SAM to assess temperature effects on silicon PV efficiency in these climates. We gathered weather data from the NSRDB and used the NSRDB-GOES typical meteorological year files (v4-0-0) from 2015 to 2023, which include average ambient temperature, wind speed, and global horizontal irradiation (GHI). We summarized key weather and irradiance information in

Table 1. Average weather and insolation values for the three U.S. cities.

Location	Yearly Tamb (°C)	Daily GHI (kWh/m2/day)	Annual Insolation (kWh/m2)	Yearly Wind Speed (m/s)	July Tamb (°C)	Latitude
College Station, TX	20.53	4.89	1784.9	2.67	33	30.61° N
El Paso, TX	18.77	5.92	2160.8	3.44	33	31.77° N
Jacksonville, FL	21.68	4.85	1770.2	1.34	30	30.33° N

, averaged for the 9-year data period.

In the PV system parameter settings of SAM, we used the IEC 61853 [16]-based model and default module test data as the basic PV parameters, namely mono-Si 96 cells with an area of 1.68 m². The results are shown in Figure 2, and the module and mounting-related parameters used in the SAM simulation are summarized in Table 2. The NOCT is 45 °C and the temperature coefficient at STC is −0.275%. The NOCT-based cell temperature is expressed as follows:

$${\mathrm{T}}_{\mathrm{cell}}={ \mathrm{T}}_{\mathrm{amb}}+{\displaystyle \frac{\mathrm{G}\times \left(\mathrm{NOCT}-20\right)}{800}}$$

(1)

where

• G is irradiance in W/m²;
• T_cell is cell temperature in °C;
• T_amb is ambient temperature in °C.

The NOCT, measured at open circuit, estimates T_cell. A modified version is used in SAM that includes wind velocity and mounting-related dependencies.

Figure 2. Fitted efficiency data from module test data (IEC 61853) across different irradiance and cell temperatures.

Figure 2. Fitted efficiency data from module test data (IEC 61853) across different irradiance and cell temperatures.

Table 2. Module and mounting-related parameters used in SAM.

Parameter	Value	Unit	Note
Module type	Mono-Si 96-cell	–	SAM IEC 61853 model
Power at STC (Pmp)	322.3	W
Open circuit voltage (Voc)	70.21	V
Voltage at Pmp (Vmp)	58.54	V
Short-circuit current (Isc)	5.903	A
Current at Pmp (Imp)	5.506	A
Efficiency at STC	19.19	%
Temperature coefficient at STC	−0.275	%
Module area	1.68	m2
Installation	Fixed tilt	–	Tilt angle = latitude
Orientation	South-facing	–	Azimuth = 180°

Table 2. Module and mounting-related parameters used in SAM.

Parameter	Value	Unit	Note
Module type	Mono-Si 96-cell	–	SAM IEC 61853 model
Power at STC (Pmp)	322.3	W
Open circuit voltage (Voc)	70.21	V
Voltage at Pmp (Vmp)	58.54	V
Short-circuit current (Isc)	5.903	A
Current at Pmp (Imp)	5.506	A
Efficiency at STC	19.19	%
Temperature coefficient at STC	−0.275	%
Module area	1.68	m2
Installation	Fixed tilt	–	Tilt angle = latitude
Orientation	South-facing	–	Azimuth = 180°

Since there was only one PV module, there was no shading. For the module deployment, we selected ground-mounted, south-facing with fixed-tilt angle as same as its latitude. The key parameters T_cell and I_POA were calculated directly through SAM’s built-in model. To evaluate the cell temperature independently using the weather and irradiance data, we used the Faiman model included in the IEC 61853-2 standard [16]. The Faiman model considers ambient temperature, wind speed, and irradiance as follows:

$${\mathrm{T}}_{\mathrm{cell}}-{ \mathrm{T}}_{\mathrm{amb}}={\displaystyle \frac{{\mathrm{I}}_{\mathrm{POA}} }{\left({\mathrm{U}}_0+{ \mathrm{U}}_1·\mathrm{v}\right)}}$$

(2)

where

• $I_{POA}$ is irradiance in W/m²;
• v is wind speed in m/s;
• U₀ and U₁ are heat loss coefficients in W/(m²·°C) and W/(m²·°C·(m/s)).

By calculating the Faiman model separately in MATLAB, we can perform regression analysis and extract empirical thermal loss coefficients (U₀ and U₁) based on SAM output data [17], and compare them to measured results reported in the literature. The Faiman model is widely used in PV research due to its simplicity and dependence on readily available environmental parameters. According to the IEC 61853-2 standard [16], the accuracy of the Faiman model depends on proper coefficient extraction using multiple days of data with significant wind speed variation (at least 4 m/s) to ensure reliability.

After completing the simulation in SAM, the output data were imported into MATLAB R2021b (Version 9.11.0.1873467, Update 3), for further data filtering, and then the filtered data were used to calculate monthly and annual averages to observe the relationship between efficiency and temperature, as well as the variation in power generation with seasons.

2.2. Efficiency and Energy Yield

We calculated the instantaneous efficiency using P_out, I_POA and the area of the solar panel as follows:

$$\mathsf{\eta }={\displaystyle \frac{{\mathrm{P}}_{\mathrm{out}}}{{\mathrm{I}}_{\mathrm{POA}}\times \mathrm{A}}}$$

(3)

where P_out (kW) is the electrical power generated by the PV system, and A (m²) is the area of the solar panel(s).

We preprocessed the data by removing nighttime values (GHI < 20 W/m²) and outliers. Then, we calculated the energy yield (EY) for different time periods at each location. The EY was calculated by summing the hourly power outputs from the SAM simulations after filtering out nighttime and outlier data points.

$$\mathrm{EY}\left(\mathrm{kWh}/\mathrm{yr}\right)={\displaystyle \sum }_{\mathrm{yr}}\mathsf{\eta }\left({\mathrm{I}}_{\mathrm{POA}},{\mathrm{T}}_{\mathrm{c}}\right){\mathrm{I}}_{\mathrm{POA}}\left({\displaystyle \frac{\mathrm{kW}}{{\mathrm{m}}^2}}\right)\mathrm{A} \Delta \mathrm{t}$$

(4)

where $\Delta t$ = 1 h. It was helpful to further normalize EY to the total system power at STC to obtain the energy yield in kWh per kW of system capacity.

The annual insolation provides the total yearly input solar energy to the system as follows:

$${\mathrm{E}}_{\mathrm{in}}\left(\mathrm{kWh}/\mathrm{yr}\right)={\displaystyle \sum }_{\mathrm{yr}}{\mathrm{I}}_{\mathrm{POA}}\left({\displaystyle \frac{\mathrm{kW}}{{\mathrm{m}}^2}}\right)\mathrm{A} \Delta \mathrm{t}$$

(5)

so that an average efficiency for a year can easily be defined as ${\mathsf{\eta }}_{\mathrm{avg}}=\mathrm{EY}/{\mathrm{E}}_{\mathrm{in}}$. Similarly, other time periods may be used such as monthly averages.

2.3. LCOE Calculation and Cooling Comparison

We considered the LCOE as a key economic indicator to evaluate PV systems. The 2024 NREL report [5] gives a benchmark LCOE of USD 0.046/kWh for utility-scale PV systems, with a capital cost of USD 1.12/W_dc, while residential-scale systems have more than twice the capital cost at USD 2.74/W_dc. This benchmark allows us to assess the cost-effectiveness of temperature control strategies. Higher module temperatures reduce energy conversion efficiency and thus increase the LCOE. Cooling technologies that lower cell temperatures (ΔT) can increase energy output and reduce the LCOE. Although SAM offers more detailed LCOE calculations, we simplified the calculation in this study as follows:

$$\mathrm{LCOE}={\displaystyle \frac{\mathrm{FCR}\times \mathrm{CC}+\mathrm{FOC}}{\mathrm{EY}}}$$

(6)

where

• FCR: fixed charge rate (default = 0.09);
• CC: system capital cost (USD/kW_dc);
• FOC: annual fixed operation and maintenance cost;
• EY: annual energy yield (kWh) from SAM simulation.

FOC is estimated at USD 30/kW_dc/yr for residential and USD 19/kW_dc/yr for utility scale PV systems, respectively. To evaluate the impact of temperature on the energy efficiency and economics of PV systems, we adopted the following approach. We simulated the annual energy yield of the system without cooling and averaged the annual values over nine years to determine the 9-year average EY for each location. We then considered an idealized scenario where the PV cell temperature was reduced by ΔT (e.g., 10 °C) under all environmental and irradiance conditions. We performed MATLAB-based simulations using the SAM data to obtain the corresponding annual energy yield and evaluated the annual efficiency and LCOE of the system accordingly. We calculated the difference in LCOE (ΔLCOE) for each city before and after cooling and quantified the relative reduction in LCOE per °C temperature decrease. We estimated the break-even cost of the cooling device using the following formula (in USD/m²):

Break-even cost = ΔLCOE × Lifetime Yield

(7)

where

• ΔLCOE: the difference in LCOE ($/kWh) before and after cooling;
• Lifetime yield: the total energy produced per unit area (kWh/m²) over the system’s lifetime.

To further assess the applicability of cooling strategies at different scales, we used typical unit costs for residential and utility-scale systems while keeping other configurations unchanged. We calculated the break-even cooling costs for the same cooling gain to analyze how system scale affects the return on cooling investment. This analysis helps determine the economic feasibility of cooling strategies for both small and large PV deployments.

3. Simulation Results

This section presents the simulation and thermal model fitting results for the three selected climates.

3.1. Temperature-Driven Efficiency Losses

Figure 3 shows that the trend of the multi-year monthly average efficiency values of the three cities is consistent. This figure corresponds to the seasonal thermal impacts modeled using SAM weather files from 2015 to 2023. Regardless of the location, the efficiency is the highest in winter (December to February), the lowest in summer (June to August), and rebounds when the temperature starts to drop at the end of summer, presenting a U-shaped curve for each city. The decline in summer is due to the significant increase in the cell temperature, which leads to a drop in voltage that is not offset by the increase in photocurrent. Among the three cities, El Paso had the lowest efficiency in summer but outperforms the others during non-summer months, likely due to its dry and sunny climate. Jacksonville showed the lowest overall efficiency.

The error bars in Figure 3 represent the standard deviation of monthly average efficiency across the 9-year dataset (2015–2023), indicating year-to-year variation for each location. The U-shaped seasonal trend reflects reduced efficiency during hotter summer months due to higher cell temperatures.

The figure also includes a constant STC-based efficiency baseline (19.19%), representing the rated efficiency under fixed conditions. Unlike the simulated results reflecting dynamic weather conditions, the STC line remains flat throughout the year. Even during winter, a performance gap of nearly 2% remains, and in the hottest months, this deviation nearly doubles, underscoring the importance of temperature effects. Figure 3 visually verifies the temperature sensitivity predicted by the model, indicating that thermal impact must be considered in evaluating efficiency or LCOE.

3.2. Regression of U₀/U₁

In Section 2, we introduced the Faiman model. Using the SAM data, we reproduced the linear regression for all three cities, extracting U₀ and U₁ coefficients from the SAM output after applying the same nighttime and filtering as before. We performed the regression for each year (2015–2023) and for the combined 9-year dataset. Figure 4 shows the multi-year average regression results.

For comparison,

Table 3. Regression coefficient comparison.

Location/Study	U0 [W/(°C·m2)]	U1 [W·s/(°C·m3)]	Module Type	Data Period
College Station, TX (This study)	26.88	8.88	mono-Si	2015–2023
El Paso, TX (This study)	26.23	8.90	mono-Si	2015–2023
Jacksonville, FL (This study)	26.17	8.93	mono-Si	2015–2023
Cologne, DE ([15])	35.7	8.22	poly-Si	6 months
Tempe, US ([15])	32.1	6.08	poly-Si	6 months
Ancona, IT ([15])	41.9	3.95	poly-Si	6 months
Chennai, IN ([15])	30.1	4.75	poly-Si	6 months
Thuwal, SA ([15])	39.7	3.06	poly-Si	6 months

shows our Faiman model coefficients and those reported by Barykina and Hammer, who measured polycrystalline silicon modules at five international test sites under different climatic conditions. Their research found that the range of U₀ is 28.6 to 41.9 W/(°C·m²), and the range of U₁ is 3.06 to 8.22 W·s/(°C·m³), depending on different time periods, the local wind profile, irradiation level and specific on-site installation conditions in their Table 3 [15]. In contrast, our U₀ values in College Station, El Paso, and Jacksonville ranged from 26.17 to 26.88 W/(°C·m²), and the U₁ values were all close to 8.9 W·s/(°C·m³), showing smaller variations.

We expected smaller variation because we used a relatively long-time span. Compared with the data in Barykina’s article, we used the average daily data for 9 years and conducted screening and filtering. This effectively reduced the interference of short-term weather anomalies and measurement noise. In addition, monocrystalline silicon and polycrystalline silicon solar panels themselves also have certain performance differences. Compared with the reported measurement data, our simulation data show consistent values for the Faiman parameters, which are a bit smaller than reported for the polysilicon measurements.

The residuals and confidence intervals across all three sites confirmed the stability of the Faiman model fitting. For our data, Jacksonville had the lowest mean absolute error MAE ≤ 0.48, where MAE refers to the average absolute difference between the observed and fitted values of G/(T_cell − T_amb), with units of W/(°C·m²). In comparison, College Station had slightly larger variations, with U₀ and U₁ confidence intervals exceeding 2.4 W/(°C·m²) in some years and MAE reaching 0.74.

3.3. LCOE and Cost–Benefit Simulation Cooling Cost Break-Even Estimation

In Section 2.3, we introduced a cost effectiveness metric for evaluating cooling strategies using the LCOE framework. To quantitatively assess the economic impact of temperature on the performance of photovoltaic systems, in this section, we analyze the specific impact of temperature on LCOE by comparing the baseline and the results of cooling enhancement.

Before presenting the simulation results, we briefly related the cooling performance using fixed temperature reductions of $\Delta T$ to changes in the effective heat transfer coefficient. In the previous section, we used the Faiman model to fit cell temperature data. Starting from the Faiman equation (Equation (2)), the cell-to-ambient temperature difference T_cell − T_amb was determined by both I_POA and the wind speed. The denominator of the expression, which includes the effect of wind speed, represents the system’s ability to dissipate heat and can be interpreted as an effective heat transfer coefficient h that encompasses the combined influence of convection, radiation, electrical power generation, and environmental conditions. We can relate the cooling impact $\Delta T$ to a change in the heat transfer coefficient, h + $\delta h,$ as follows:

$${\mathrm{T}}_{\mathrm{Cell}}-{\mathrm{T}}_{\mathrm{amb}}-\Delta \mathrm{T}={\displaystyle \frac{{\mathrm{I}}_{\mathrm{POA}}}{\left(\mathrm{h}+\mathsf{\delta }\mathrm{h}\right)}}$$

(8)

To further quantify the effect of additional cooling, we rearranged Equation (8) to isolate the incremental heat transfer coefficient, δh, under a given fixed temperature reduction ΔT. By substituting the Faiman model expression

$$\mathrm{h}={\displaystyle \frac{{\mathrm{I}}_{\mathrm{POA}}}{{\mathrm{T}}_{\mathrm{cell}}-{\mathrm{T}}_{\mathrm{amb}}}}$$

(9)

Rearranging Equation (8) and substituting Equation (9), we find an expression for $\delta h$ as follows:

$$\mathsf{\delta }\mathrm{h}={\displaystyle \frac{\mathrm{h}\times \Delta \mathrm{T}}{\left({\mathrm{T}}_{\mathrm{Cell}}-{\mathrm{T}}_{\mathrm{amb}}-\Delta \mathrm{T}\right)}}=\mathrm{h}\left({\displaystyle \frac{\mathrm{x}}{1-\mathrm{x}}}\right)$$

(10)

where x = ΔT/(T_cell − T_amb). Equation (10) works well for cell temperatures above ambient, and other formulations will be needed for sub-ambient cooling. For practicality, when simulating the impact of fixed ΔT, we ensure that T_cell > T_amb. According to Equation (10), when x = 0.5, the required additional heat transfer coefficient δh = h. Cooling becomes harder to achieve as the module temperature nears ambient as expected. In practical cases, passive cooling methods operate in the range where ΔT is smaller than T_cell − T_amb. When the temperature difference is small, even a ΔT of less than 5 °C may push the system into a nonlinear regime. Conversely, if T_cell − T_amb is large enough, a temperature reduction of up to 30 °C may still fall within the quasi-linear region. As ΔT approaches T_cell − T_amb, sub-ambient cooling techniques may be needed such as radiative cooling or active approaches that introduce external energy.

Based on this understanding, we simulated temperature reductions of 5 °C, 10 °C, and 20 °C which should be reasonably achieved through passive or low-cost cooling methods. In contrast, a 30 °C reduction likely requires the support of more complex and costly systems. This assumption serves as a practical reference for the following simulation scenarios.

Accordingly, we simulated the energy yield (EY) under normal operating conditions and the resulting LCOE simultaneously and compared them with a module temperature reduction of 5 °C, 10 °C, 20 °C and 30 °C. The changes in LCOE for each city were quantified, and the percentage decrease in LCOE with temperature was evaluated.

We compared the obtained LCOE value with the local electricity price to evaluate the competitiveness of the photovoltaic system under thermal impact. Based on the reduction in LCOE achieved through cooling, we further estimated the break-even cost of implementing the cooling strategy, expressed in US dollars per square meter. Module and other systems cost are often quoted in USD/m², so it is a handy metric for comparison. Using this universal unit as a benchmark can facilitate the derivation of quantitative indicators for photovoltaic systems of various scales and models. Investment cost for cooling generally does not vary with power, so such a unit selection is more reasonable.

Based on the simulations for EY, we then calculated the residential-scale LCOE values for each city, as shown in

Table 4. LCOE without cooling compared with electricity prices and with cooling. Relative energy yield change is calculated as the ratio of ΔEY to the reference EY without cooling.

City	LCOE No Cooling (USD/kWh)	Local Electricity Price (USD/kWh)	$\mathbf{LCOE} \mathbf{for} \Delta \mathit{T}$ = 10 °C (USD/kWh)	$\mathbf{Relative} \mathbf{EY} \mathbf{Change} (\%) \mathbf{for} \Delta \mathit{T}$ = 10 °C
College Station	0.165	0.119	0.161	2.68
Jacksonville	0.167	0.13	0.162	2.83
El Paso	0.131	0.16	0.128	2.59

. The table also presents how LCOE changes when the system temperature is reduced by 10 °C.

Table 4 compares the LCOE without cooling to the local electricity prices in each city, using data retrieved from public sources. Although the average LCOE reduction resulting from cooling PV modules by 10 °C is small, 0.005 USD/kWh, the annual impact is an increase in energy yield of 2.6% to 2.8%.

Figure 5 shows the energy yield gained from cooling. Each bar represents the simulated additional energy yield (in kWh/kW/year) under different cooling levels, based on 9-year hourly data. Baseline energy yields without cooling (Base EY) are labeled above each city. The results show that even modest cooling yields considerable energy gains, especially in hotter climates such as El Paso.

A comparison of the relative LCOE reductions in Table 4 and the relative energy yield changes in Figure 5 shows that the percentage decrease in LCOE is the same as the percentage increase in energy yield and average efficiency. This consistency results from the simplified cost–benefit model and highlights its ability to effectively capture the impact of cooling on energy yield, cost, and efficiency.

Figure 6 illustrates the simulated energy gain in the least efficient month (July) under varying cooling levels up to 30 °C. While the overall trend shows that increasing the cooling amplitude leads to larger gains across all three cities, these results are based on an idealized assumption in which a fixed temperature reduction (ΔT) is applied regardless of environmental conditions. When a lower bound of T_cell − ΔT >T_amb is enforced, a 20 °C cooling still yields over 25% energy gain in July across all three cities. According to the annual trends summarized in Table 4, a 5 °C reduction in module temperature corresponds to an approximate efficiency improvement of 1.3–1.4% suggesting a potential recovery of around 6.5% with 30 °C cooling. Comparing these gains to the STC benchmark in Figure 3 suggests that reducing the module temperature by at least 15 °C is required to restore real-world performance to STC levels, underscoring the impact of temperature-related losses in field conditions.

On the other hand, as shown in Figure 5, the ΔEnergy of photovoltaic systems under different cooling magnitudes varied across the three cities. For example, in College Station, applying a 10 °C cooling strategy results in an annual energy gain of approximately 45 kWh per kilowatt of system capacity. When combined with the local electricity price (assumed to be USD 0.12/kWh), this corresponds to an annual cost saving of about USD5.4/kW. Over a 25-year system lifetime, the cumulative savings would reach USD 135/kW.

To achieve economic viability, the total cost of the cooling system must not exceed this threshold. In other words, this value represents the break-even cost of the cooling strategy under the given scenario. Alternatively, the benefit can be expressed as the cost saving per degree Celsius of cooling per year, with the following formula:

$$\mathrm{Savings} \mathrm{Rate}={\displaystyle \frac{\Delta \mathrm{E}\times \mathrm{P}}{\Delta \mathrm{T}}} \left[{\displaystyle \frac{\mathrm{USD}}{°\mathrm{C}\times \mathrm{kW}\times \mathrm{year}}}\right]$$

(11)

where

• ΔE represents the annual power generation gain (kWh/kW/year);
• P represents the local electricity price (USD/kWh);
• ΔT represents the cooling amplitude (°C).

Although LCOE-based metrics (e.g., ∆LCOE/°C) offer a normalized cost indicator, Equation (11) offers an energy-centric perspective that directly links technical gains to monetary savings.

The simulation results shown in Figure 5 can also be used to estimate the break-even cost threshold for implementing cooling strategies. Since the economic justification of any cooling strategy relies fundamentally on the additional energy it can generate, the break-even analysis for a specific location can be formulated as follows:

$$\mathrm{Break} \mathrm{even} \cos \mathrm{t} \left({\displaystyle \frac{\mathrm{USD}}{{\mathrm{m}}^2}}\right)={\displaystyle \frac{\Delta \mathrm{E}\times \mathrm{N}\times \mathrm{P}}{\mathrm{module} \mathrm{capacity}}}$$

(12)

where

• ΔE: the annual power generation gain (kWh/kW/year);
• P: the local electricity price (USD/kWh);
• N: system lifetime (year);
• Module capacity: module capacity per unit area (kW/m²).

This formula converts the energy gain enabled by cooling into the allowable investment per square meter, providing a practical reference for evaluating the economic feasibility of surface cooling methods in different climate zones and system configurations.

Figure 7 illustrates the upper limit of the cooling cost per unit area of the three cities under different cooling values (ΔT), calculated based on Equation (12) and the simulation data of this study. This cost threshold was determined based on the local electricity price listed in Table 4. The overall trend indicates that when the electricity price is high and the energy gain resulting from cooling is substantial, the acceptable cooling input cost (USD/m²) of the system also increases accordingly. It is worth noting that this break-even cost is an “upper limit” on the cost of cooling, i.e., when the return on the cooling cost investment is zero.

Among the three locations, College Station consistently showed the lowest break-even cooling cost across all tested ΔT values. Table 1 shows that College Station has middle-range values in both annual temperature and irradiance compared to the other two cities, and its lower electricity price results in the lowest break-even cooling cost. However, in regions like Texas where grid stability concerns may drive PV adoption for energy reliability rather than direct financial return, thermal management may still offer performance advantages by lowering LCOE during prolonged high-temperature periods. While cooling technologies can help improve PV system performance, whether they are worth using depends on local conditions such as electricity prices, temperature, and system goals.

While Figure 7 presents the idealized break-even cooling costs using all data points with T_cell > T_amb, this assumption may not reflect real-world limitations. Extreme cooling levels are achievable with advanced systems, but the implementation cost may exceed the energy benefit, reducing their practical value for PV applications. To address this, we introduced a constraint on the minimum cooled temperature in Figure 8 by limiting the actual cooling effect to the available temperature difference T_cell − T_amb at each time point. As shown in Figure 8, this constraint reduces the effective cooling gain at higher ΔT values, leading to lower estimated break-even costs. For moderate cooling levels at 5 °C or 10 °C, the impact is minimal, indicating these are more practical cooling cases. Even for ΔT = 20 °C, the results begin to diverge, but not significantly, revealing the increasing difficulty of maintaining large temperature drops under realistic conditions. At 30 °C, the break-even cost still shows a slight improvement over the 20 °C case, suggesting that such extreme cooling could still offer some efficiency gains. However, due to the limited number of hours where T_cell − T_amb exceeds 30 °C in the selected locations, this improvement is marginal and subject to diminishing returns. The historical average value for T_cell − T_amb is between 10 °C and 15 °C for these locations, supporting Figure 8 in identifying the 10 °C case as the most practical break-even point. However, the results also suggest that if the cooling cost remains acceptable relative to the electricity price, then aiming for a threshold of 15–20 °C could still be justified in practice.

4. Discussion

4.1. STC-Based Efficiency Benchmark and Real-World Deviation

The summer deviation in Figure 3 highlights the strong temperature sensitivity of crystalline silicon (c-Si) modules, which directly affects annual energy yield and LCOE. While technological advances in c-Si modules have steadily increased STC efficiency, NREL data suggest a 2.6% improvement over the last decade [1], real-world high-temperature conditions can easily reduce efficiency by a similar amount, erasing the gains of a decade of innovation. In this context, cooling can effectively “recover” these laboratory-level improvements.

4.2. Temperature Models

The Faiman model, a well-established thermal model, is helpful to simulate cell temperature deviation from ambient temperature based on irradiance and wind speed. The model’s thermal coefficients (U₀ and U₁) can be easily extracted from provided or measured data where the underlying cell and ambient temperature dependence is not known. In Jacksonville, the limited amount of strong wind data slightly reduced the fitting accuracy, likely because of the city’s relatively low average annual wind speed. The fit parameters were almost identical for the other two locations.

Our results for the 10 °C scenario showed a 2.6–2.8% annual temperature correction across the locations, aligning closely with the selected module temperature coefficient at STC (−0.275%/°C). This agreement, coupled with knowledge of the average annual location temperatures, confirms the average behavior. The temperature dependence of the efficiency for the module data fits well to a quadratic curve, showing that the efficiency losses increase more significantly as the cell temperature increases above STC. Hence, module efficiency data over a range of irradiances are critical for accurate high temperature simulations.

4.3. Interpreting LCOE Improvements and the Applicability of Cooling Strategies

The average LCOE reduction from 10 °C cooling was approximately USD 0.005/kWh, corresponding to a 2.7% decrease across the cities. When 2.75% is considered in terms of the corresponding energy and average efficiency gains, the impact appears more substantial. Figure 6 analyzes the additional energy gain achieved through different cooling amplitudes during the least efficient month (July) across all three cities. Although we focused on July, these energy gains can extend to all high-temperature months, indicating that cooling would offer substantial performance recovery throughout the hottest quarter of the year. Figure 6 and Figure 7 illustrate that cooling in El Paso increased the annual energy yield by more than 50 kWh/kW, representing a relative gain of over 50% compared to the baseline. Although Figure 6 focuses on the worst-performing month (July), similar trends are expected in other peak-temperature months, indicating that cooling would offer substantial performance recovery during the worst three months of the year rather than just a single month.

Figure 7 further shows the cooling threshold across different cooling amplitudes, visually reflecting how the economic viability of cooling strategies varies with temperature reduction. Extrapolating this gain over a typical 25-year lifetime would result in over 1200 kWh/kW of additional energy production, potentially translating into savings of more than USD 140 at standard residential electricity rates. These figures highlight the significant cooling potential in hot and dry climates and underscore the economic value of targeted temperature management.

However, real-world rooftop systems face restricted ventilation and higher module temperatures compared to open-rack test configurations. Previous studies (e.g., Yadav et al. [18]) noted that rooftop-mounted PV systems can have longer payback periods due to increased temperatures. New cooling strategies [19] have also found that even improved ventilation does not fully eliminate rooftop temperature penalties. These findings underscore that our simulated results—based on ideal open-rack conditions—represent an upper-limit scenario. Actual performance may fall short without site-specific adaptation.

Importantly, cooling should not be viewed as a universal add-on for new PV installations. Instead, these findings support retrofitted or passive cooling as a cost-effective strategy for existing systems in hot climates. Even modest energy gains can offer long-term benefits without requiring structural or material upgrades.

Cooling benefits also depend heavily on climate context. El Paso showed clear economic potential for cooling, while Jacksonville’s humid and mild climate yields minimal LCOE improvements, making active cooling economically unjustifiable there. These differences highlight the need to tailor cooling approaches to regional conditions, rather than applying uniform designs.

4.4. Limitations of Simulation-Based Evaluation and the Need for Experimental Validation

Model simulations can quickly point out promising directions for optimization, but there are clear restrictions. They do not fully see the details of real-world PV systems, which can create differences between our model results and field test data. As a next step, we plan to do field tests with our system setup to better understand cooling strategies and their impact on crystalline silicon modules.

5. Conclusions

This study quantitatively analyzed temperature-driven efficiency losses in crystalline silicon photovoltaic (PV) systems across three representative climate zones in the United States: College Station (humid subtropical), El Paso (hot and dry), and Jacksonville (coastal humid). Using multi-year SAM simulation data and applying a thermal modeling method based on the Faiman model, the results show that an increase in ambient temperature has a significant inhibitory effect on photovoltaic performance, with efficiency during peak summer periods decreasing dramatically compared with module performance under STC.

To evaluate the economic implications of cooling, we simulated 5 °C to 30 °C temperature reduction scenarios. Under idealized conditions with no temperature constraint, allowing sub ambient cooling in some cases, the energy yield improvement scales nearly linearly with cooling amplitude. When the cooled module temperature is constrained to be above ambient temperature, the curve flattens slightly, but a 20 °C reduction still delivers over 25% gain in July across all locations. By expressing energy revenue as an investment threshold per unit area, we established a cost–benefit framework for evaluating cooling strategies. The findings highlight that the economic viability of active or passive cooling strongly depends on climate and must be assessed regionally rather than universally.

Given the simulation-based nature of this study, experimental verification is planned. This work will involve controlled field tests under IEC-compliant conditions to validate the correlation between temperature reductions and energy gains, as well as to assess the performance of cooling systems under various temperatures, irradiance, wind and humidity conditions. Such validation is expected to enhance the engineering relevance of the simulation framework, refine the parameters developed in this study, and lay the groundwork for integrating thermal cooling in future PV systems. Dramatic improvements can be achieved in the hottest months with even a few degrees of cooling. While cooling costs typically increase with ΔT, cooling in the range of 10–20 °C may provide the most economical impact in terms of the break-even cost. The simulation framework highlights the need for critical testing of cooling technologies in a manner similar to the multi-irradiance and multi-temperature efficiency testing for PV modules.