Non-Fiction Archive

Investigating the Impact of Temperature on Solar Cells

Matthew Cooper

Researchers Say 'Anti-Solar Panels' Could Generate Power at Night ...

Investigating the impact of temperature on Solar Cells

This investigation analysed the relationship between temperature and solar panel efficiency in order to formulate an answer to the research question,  incorporating  two phases. Phase one was a test of real world data which involved collecting a large volume of power output data from an array of solar panels over 18 months, combining this data with temperatures measured by the Bureau of Meteorology to observe the correlation between temperature and efficiency. The second phase was implemented after the “real” data in phase one proved to be heavily affected by uncontrolled variables. Within phase two, power output data was recorded from a low voltage solar panel in different controlled temperatures ranging between 15OC and 50OC. After analysing all collected results, it was concluded with statistical significance that under controlled conditions, increasing temperature reduces the power efficiency of a solar panel with a linear trend. However, in uncontrolled conditions, the data showed that temperature has a positive impact on power output, which suggests that other variables correlated with temperature impact power efficiency positively.

Due to the steadily decreasing supply of fossil fuels and the resultant detrimental impacts on the environment due to the emission of greenhouse gases, the need for an efficient and dependable renewable energy source is becoming more prominent1. Solar power has been one of the most widely adopted forms of renewable energy since its commercial release2. However, too many fossil fuels are being used to sustain the world’s current power usage3. In order to use fewer finite resources, renewable energy sources such as solar panels must be optimised extensively, continuously searching for any means of increasing efficiency.

Function of Solar Panels
Solar panels make use of semiconductive materials in order to moderate current flow and make use of energy from photons. Semiconductors possess a unique property of an energy band gap in between insulators and conductors4. Due to this, free flowing electrons and electron “holes”, which are simply absences of electrons resulting in a net positive charge, within semiconductors can pass through this band gap providing they are provided the right conditions5. Semiconductors can be p-doped or n-doped, meaning an element such as silicon can be combined with another element to give an imbalance of free electrons or free electron holes, depending on the type of dopant5.

Solar panels typically contain  two layers of silicon, one p and one n-doped, forming a diode as visible in figure 1. A diode restricts the direction of current flow through the circuit by forming a PN junction in the joining region of the layers6. Upon joining the layers, several positively charged holes from the p-doped layer will move across the junction and be attracted to free moving electrons in the n-doped layer. Many electrons from within the n-doped layer will also be attracted to positively charged holes in the p-doped layer.This zone of attracted particles around the PN junction is referred to as the depletion region6. As positively charged holes and negatively charged electrons have moved across the PN junction, the depletion region gains an overall positive charge on the n-doped side and an overall negative charge on the p-doped side. Thus, an electric field is created as a result of the potential difference7.

1  (BACH, 1981)

2 (Ritchie and Roser, 2017)

3 (Jacobson and Delucchi, 2009)

4 (Van Zeghbroeck, 2011)

5 (Young et al., n.d.)


Figure 1 6

When photons from light hit the top of the cell, electrons within the silicon can be ejected free, leaving a hole, this free electron flows toward the n doped side due to the charge from the diode. Once it has reached the n doped side, the free electron must then flow further and be used in external electrical circuits to produce power before returning to a conductive sheet on the back of the photovoltaic cell5.

Issues with Solar Panels
One contemporary problem regarding solar panels is their power efficiency in comparison to other means of power generation. For example, coal-fired power stations generate power with 32-42% efficiency, whereas solar panels generate power with efficiency up to 24% depending on the type of panel7. The problem of solar efficiency is very relevant when considering feasible renewable energy sources8. As a result, any alteration that will increase the efficiency of solar panels will have a very large impact. 

As of today, more than 60% of the world’s solar panels are being manufactured and assembled in China9, with only a few major companies producing them in America and Canada. Thus, only a very small number of the world’s countries are producing their own solar panels. This means that different countries do not have the opportunity to optimise their solar panels according to their conditions such as climate and air quality.

If solar panels are not optimised for the needs of different locations, it is very likely that their efficiency is being affected10. There are a large variety of factors which could influence solar panel efficiency such as temperature, solar panel composition, humidity and cloud cover11. Temperature is a variable often overlooked in its impacts on the efficiency of solar panels due to its typical correlation with light intensity12. It is extremely important that the impacts of different variables like this on the efficiency of solar panels is tested so that as more panels are being installed over time, efficiency is optimised13.

6 (Fernandez, 2014)

7 (Vourvoulias, 2020)

8 (Barnard, 2017)

9 (Bahar, 2017)

10 (Hopkins and Li, 2016)

11 (Wiggins, 2016)

12 (Touati, Al-Hitmi and Bouchech, 2013)

13 (Mustafa, Gomaa, Al-Dhaifallah and Rezk, 2020)

Effect of Temperature on Solar Panels
Solar panels rely on photons from the sun ejecting electrons out of place to create a current in a circuit[1]. Thus, an increase in temperature will result in increased energy of electrons within solar cells prior to being ejected by photons[2]. As this thermal energy results in higher vibration of the lattice within the panel, the ability of electrons to flow through the semiconductive material is impeded. This is due to the fact that electrons become more likely to collide, increasing resistance of the circuit, inhibiting the electrons’ ability to follow the intended circuit and produce an electrical current. Due to this, the predicted trend is a negative linear trend proposing a constant for efficiency decrease per degree[3].

[1] (Girard, 2017)

[2] (Heim, 2011)

[3] (Honsberg and Bowden, 2020)


Phase 1:
How does temperature effect the power efficiency of solar panels in real environmental conditions?

Phase 2:
How does temperature effect the power efficiency of solar panels?

As temperature increases, the power efficiency of photovoltaic cells within solar panels will decrease following a linear trend.


Phase 1:
During the first phase of the investigation, data was collected historically from solar panels being used by a school in Menangle Park over a period of one year and seven months between July 2018 and March 2020[1], taking a large range of data in order to increase reliability by mapping a trend to more data points in order to evaluate consistency[2]. Throughout this time period, the total energy output of the solar array was recorded in intervals of five minutes from digital data loggers, improving accuracy by using precise equipment[3]. The first step in the process of formatting the data was cleansing. Any energy values less than 0kWh were excluded in along with any days where power efficiency was less than 5%. Daily maximum temperature data for the same time period as the power data was collected and used over average temperature data as it best reflected the temperature during hours of daylight[4]. This was provided from the Bureau of Meteorology using a weather station apparatus located the same suburb[5]. The daily maximum temperature values were placed in a corresponding fashion to the solar panel data. In addition to this, the amount of daylight hours in each day was calculated using a JavaScript program which retrieved the sunrise and sunset time of each day, taking the difference between these values as the amount of daylight hours. This was done to enforce validity, ensuring that the amount of sunlight was not an uncontrolled variable. After this was done, the previously found daily energy outputs were divided by the amount of sunlight hours to find a power value. Each power value was divided by the maximum power in the dataset to calculate efficiency. As the true power output of the solar panels could not be determined from the collected data, this estimation was used as per Field’s suggestion[6]. Efficiency was then graphed against the daily maximum temperature to observe a trend. From this trend, the R2 value was used to calculate a t-statistic according to Field’s method24. Using this t-statistic, a p-value was determined, requiring to be less than 0.05 in order to claim significance. From the determined trend and its significance, a response to the hypothesis could be formulated.

Phase 2:
In order to carry out the second phase of the investigation, an apparatus was constructed in a dark room. An ultraviolet (UV) lamp was then fixed in position such that the panel would be receiving maximum intensity of the UV light after taking repeated measurements in different positions. A multimeter was then attached in parallel to the circuit and set to measure voltage, providing digital readings to increase precision and hence improve accuracy21. After this setup was complete, the temperature of the room was altered using an electric heating device. The room was heated to a temperature of 45OC before the heating device was switched off. After the room reached the threshold temperature, recordings of the panel’s produced voltage were taken in intervals of 30 seconds until the room temperature remained constant. For each voltage measurement recorded, the temperature of the solar panel at that moment was found using an infrared temperature gun and recorded, again utilising digital measurement tools to increase precision21. This process was repeated three times in order to better gauge consistency of data.  The data was then input to a spreadsheet in order to be analysed. The collected voltage data was squared following the relationship: , assuming a constant resistance within the circuit. Each V2 value was divided by the maximum V2 value to determine efficiency, as done in phase 1 following Field’s method24. Plotting efficiency against temperature, a trend could be observed and analysed as done in phase 1 to find significance and formulate a response to the hypothesis.

[1] (, 2020)

[2] (Zamboni, 2018)

[3] (Reid, 2018)

[4] (How the Temperature Varies During the Day and Night | GLOBE Scientists’ Blog, 2020)

[5] (Historical weather observations and statistics, Bureau of Meteorology, 2020)

[6] (Field, 2018)

Summary Statistics (Phase One)

Figure 2

From phase one’s summary statistics within figure 2, it is visible that at higher temperatures, efficiency of solar panels increased, along with the overall power output of the solar panel.

Figure 3
Figure 4

Figure 3 and 4 show a normal distribution of data collected within phase one.

Figure 5

Figure 5 shows a linear trend in the data when efficiency is compared to maximum temperature. With an R2 value of 0.19, the trend occupies a correlation coefficient of 0.44. Using this correlation coefficient, the t-statistic can be determined as:

Providing a t-statistic of 12.14. Using this t-statistic to determine a p-value from socscistatistics, the determined p-value is below 0.05, indicating the significance of this trend.

Summary Statistics (Phase Two)

Figure 6

Figure 6 shows that as temperature decreases, mean voltage and efficiency of the solar panel increase.

Figure 7
Figure 8

Figures 7 and 8 show a right and left skew in the collected data respectively, likely caused by the solar panel reaching a maximum voltage for its supplied light intensity.

Figure 9 shows a negative linear trend in efficiency when compared with temperature. Using the R2  value of 0.79 from the graph to calculate a t-statistic as was done in phase one, a t-statistic of 14.82 is provided. Determining a p-value from this statistic yields a value less than 0.05, again indicating statistical significance of this trend25

Data Interpretation
Phase 1
From the first phase of the investigation, it can be found that as temperature increases, power output of photovoltaic cells also increases, going against what was expected in the hypothesis. This is observable in figure 5. As the trend within this data was shown to be significant after performing a t-test using the graph’s correlation coefficient, it can be concluded that in typical conditions, an increase in temperature corresponds to an increase in efficiency. Specifically, from a linear trend derived from the graph, this increase will be on average 1.4% per degree Celsius of maximum temperature for the school’s specific solar panel array. However, as this test was uncontrolled with many different variables being altered other than temperature, it is very unclear whether it is truly temperature causing this trend. This trend is likely because finding the daylight hours was not sufficient for controlling variables. Factors such as the brightness of the sun and cloud cover during seasons also contribute to efficiency and hence, during summer where the sun is brighter, temperatures are also hotter[1]. The impact of temperature is not truly visible due to the overarching impact on light intensity due to the sun’s increased brightness. Hence, in reality hotter days produce more power in solar cells, however this may not be due to alterations in temperature as several other factors also influence efficiency.

[1] (Bottom, 2017)

Phase 2
Within phase two of the experiment, all variables excluding temperature and voltage were attempted to be controlled in order to observe the true impact of temperature on the power output of photovoltaic cells. Within this phase, a very different result to that of the first phase was yielded. The second phase showed a clear linear relationship between voltage and temperature, however with a negative slope instead. This was the result expected within the investigation’s hypothesis and can be seen in figure 9.

The gradient of this graph shows that for each one degree increment of temperature, the efficiency of the particular solar panel used within the experiment will decrease by 0.5%. This result supports the hypothesis of the investigation and as the trend was proven significant after statistical analysis, the null hypothesis can be rejected.

The negative linear trend experienced within the investigation’s second phase very closely mirrors the trend deducted within Asif Javed’s research paper “The Effect of Temperatures on the Silicon Solar Cell”[1]. This trend is likely due to the energy gained by electrons within the panel’s lattice structure upon entering a higher energy state as temperature increases. Due to this increase in energy, electrons within the panel’s silicon begin colliding with one another more frequently, inhibiting their ability to flow through the established circuit, increasing resistance and thus resulting in a lowered power output.

[1] (Javed, 2014)

As there was no method of calculating true values in either phase of the investigation, accuracy must be estimated through calculation of uncertainty. As within both phases of the investigation, the limit of reading was quite precise, in most cases to two decimal places, this value is not the best estimate for uncertainty. The only known source of error within both phases of the investigation is calibration error due to the exclusive use of digital measuring tools such as the multimeter, infrared temperature reader and solar panel data loggers. Hence, the best estimate for uncertainty in both phases is the difference in reading from the mean for each value in each dataset[1]. The average uncertainty for each efficiency value within phase one was found to be ±16.2%. Thus, the accuracy of phase one was quite high. The average uncertainty for phase two of the investigation was ±3.6%, meaning this phase of the investigation was very accurate based on this estimation of uncertainty.

[1] (White, 2008)

The reliability of this investigation was different for the two phases. The results obtained during the first phase were found to be unreliable. With the linear trend derived from the graph of power divided by daylight hours against power output, an r value of 0.44 was determined, which is low[1]. Whilst the consistency of the results within this phase was quite poor, a trend was still observable from the data and due to the large amount of data sampled within the experiment. Taking power data from over one hundred thousand five-minute intervals allowed for many repetitions of measurements, however there were not many occasions where the conditions were identical and hence, environmental errors impacted the consistency of results, leading to poor reliability.

The reliability of the investigation’s second phase far greater than that of the first. Graphing power efficiency against temperature, an r value of 0.89 was determined, showing that the results of this test were much more consistent than the first phase. As many repetitions and measurements were conducted within this phase, a large sample size was provided. Due to this large sample size and the analysis of data using a graph, the consistency of data could be evaluated. As this phase of the investigation was conducted under identical conditions and produced results following a linear trend, the reliability of phase two was high.

[1] (Hayes, 2020)

In order for this investigation to be considered valid, adequately test the question being investigated [1]. In this case, that trend is the relationship between solar panel efficiency and solar panel temperature. In order to ensure that the test was valid and addressed the research question as claimed, a variety of measures were taken. The first of these was ensuring a fair test. In order for the investigation to test the relationship of the two variables claimed, all other variables would need to be controlled. With the temperature being the independent variable and efficiency being the dependent, a multitude of variables needed to be controlled. Within the first phase of the investigation, however, these variables were not controlled. Many uncontrolled variables such as cloud cover, light intensity, air current and precipitation likely impacted the investigation12. However, as the research question for phase one intended to evaluate the impact of temperature on power efficiency in real environmental conditions, this could not be avoided and thus the investigation still provided an answer to the research question, hence it was valid. Within the second phase of the investigation, all variables were controlled to ensure a fair test. This included environmental variables that were previously uncontrolled. As a result, this phase if the investigation did target the research question, showing the true relationship between power efficiency and solar panel temperature and thus was valid.

[1] (NESA, 2020)

Several limitations within this experiment impacted the results and restricted many possible research pathways which could have improved the findings. This investigation was likely impacted by bias, since a trend between temperature and power efficiency was expected from prior sources, hence a confirmation bias was likely present[1]. Another limitation within this investigation was the range of temperatures tested. As only a very small range of temperatures was tested throughout both phases, the impact of extremely high and low temperatures on power efficiency is unknown and could differ from the observed trend. In future research, this investigation could be conducted with a larger range of temperatures and with solar panels with known values of maximum power outputs in order to accurately calculate power efficiency.

[1] (confirmation bias | Definition, Background, History, & Facts, 2020)

This investigation addressed the research question, formulating a response to the hypothesis and doing so with statistical significance. As the primary intention of this investigation was to test the relationship between temperature and power efficiency in solar panels, a suitable two-phase experiment was devised and carried out to test the hypothesis that power will decrease linearly with temperature. As phase one of the investigation did not demonstrate the impact of temperature, the results of the second phase were used to formulate an answer to true impact of temperature in a controlled environment. As expected in the hypothesis, this experiment showed a decrease of 0.5% per degree Celsius, following a linear trend. Thus, the investigation’s hypothesis was supported and the null hypothesis rejected. After carrying out a statistical t-test on the data collected within the second phase, the trend was deemed significant and an answer to the research question could be justifiably drawn: an increasing temperature will result in a linearly decreasing power output following the formula:

Barnard, M., 2017. Is Solar Energy Less Efficient Than Non-Renewables?. [online] Forbes. Available at: <> [Accessed 15 August 2020].

Fernandez, A., 2014. How A Solar Cell Works. [image] Available at: <; [Accessed 25 August 2020].

Girard, L., 2017. How Do Solar Panels Produce Electricity?. [online] Sciencing. Available at: <; [Accessed 5 March 2020]. 2020. How The Temperature Varies During The Day And Night | GLOBE Scientists’ Blog. [online] Available at: <; [Accessed 17 August 2020].

Heim, D., 2011. The Simultaneous Effect of the Operating Temperature and Solar Radiation on the Efficiency of Photovoltaic Panels / Jednoczesny Wpływ Temperatury Pracy Oraz Promieniowania Słonecznego Na Wydajnosc Ogniw Fotowoltaicznych. Archives of Civil Engineering, 57(3), pp.261-274.

Honsberg, C. and Bowden, S., 2020. Effect Of Temperature | Pveducation. [online] Available at: <; [Accessed 3 March 2020].

Mustafa, R., Gomaa, M., Al-Dhaifallah, M. and Rezk, H., 2020. Environmental Impacts on the Performance of Solar Photovoltaic Systems. Sustainability, 12(2), p.608.

Reid, A., 2018. How To Improve Your Precision In The Lab. [online] Sciencing. Available at: <; [Accessed 25 August 2020].

Touati, F., Al-Hitmi, M. and Bouchech, H., 2013. Study of the Effects of Dust, Relative Humidity, and Temperature on Solar PV Performance in Doha: Comparison Between Monocrystalline and Amorphous PVS. International Journal of Green Energy, 10(7), pp.680-689.

Van Zeghbroeck, B., 2011. Principles Of Semiconductor Devices.

Vourvoulias, A., 2020. How Efficient Are Solar Panels? (2020) | Greenmatch. [online] Available at: <,is%20measured%20under%20laboratory%20conditions.&gt; [Accessed 18 August 2020].


Wilson, G., 2020. Not Too Hot, Not Too Cold. What’S ‘Just Right’ For Solar PV? – Solar Technologies. [online] Solar Technologies. Available at: <; [Accessed 12 June 2020].

Young, H., Freedman, R., Ford, L., Sears, F. and Young, H., n.d. University Physics With Modern Physics.

Zamboni, J., 2018. The Advantages Of A Large Sample Size. [online] Sciencing. Available at: <; [Accessed 25 August 2020] 2020. Historical Weather Observations And Statistics, Bureau Of Meteorology. [online] Available at: <; [Accessed 14 April 2020].

Bottom, S., 2017. Seasonal Changes in Light.

Encyclopedia Britannica. 2020. Confirmation Bias | Definition, Background, History, & Facts. [online] Available at: <; [Accessed 25 August 2020].

Field, A., 2018. Discovering Statistics Using IBM SPSS Statistics. 3rd ed. London [etc.]: Sage.

Hayes, A., 2020. R-Squared. [online] Investopedia. Available at: <; [Accessed 26 August 2020].

Javed, A., 2014. The Effect of Temperatures on the Silicon Solar Cell.

John Laird, K., 2020. [online] Solar Analytics WebPortal. Available at: <; [Accessed 12 May 2020]. 2020. Quick P Value From T Score Calculator. [online] Available at: <; [Accessed 11 July 2020].

White, G., 2008. Basics Of Estimating Measurement Uncertainty. [online] PubMed Central (PMC). Available at: <; [Accessed 23 August 2020].

%d bloggers like this: