Performance Evaluation of Unconcentrated Photovoltaic-Thermoelectric Generator Hybrid System under Tropical Climate

Indoor farming is among the approaches which can be adapted to improve agricultural sustainability and ensure the food security of countries. However, adopting indoor farming requires a high amount of energy to sustain the system. Incorporating a renewable energy system for supplying power to agricultural activities will allow the system to be more sustainable in the long run. An unconcentrated photovoltaic-thermoelectric generator (PV-TEG) hybrid system is designed based on the load requirement of an indoor farm. The hybrid system performance under Malaysia’s weather conditions was investigated and analyzed over three months. The designed system has shown its ability to provide sufficient power to the load, as well as supporting an additional load. Besides that, the TEGs power output was found to be dependent on the temperature as well as the types of connections used, where a higher temperature difference and series connection exhibited a better power output. Overall, the combination of the system showed that the addition of a TEG affects the efficiency and power output as compared to a standalone PV. Therefore, this implies that the hybrid system is able to exhibit a more positive outcome in certain weather conditions than a PV standalone system.


Introduction
Indoor farming has been considered as an alternative measure to dealing with low and inconsistent food supplies from conventional agriculture. Drastic weather changes caused by global warming have pushed farmers and the agriculture industry to seek other forms of farming which provide a Other factors, such as a TEG's figure of merit (zT) and weather conditions, play a crucial role in ensuring that the maximum power output is achievable for the hybrid system. Van Sark, 2011 [16], investigated the feasibility of PV-TEG hybrid modules where a model of the hybrid system was created and simulated. The author found that the hybrid system produces an increase in efficiency between 8% and 23% according to the type of PV module integration at a constant figure of merit value. The author emphasized on possible further improvement in the overall efficiency through increment on the figure of merit value of the TEG. This was supported by Chávez-Urbiola et al., 2012 [17], through theoretical analysis on the efficiency of a TEG, which was higher for zT = 4 than zT = 2.4. Other than that, the internal resistance of the combined system circuit must be minimized in order to avoid unnecessary power losses. Park et al., 2013 [18], reported an increase of 3.8% through the addition of a TEG to a sole PV circuit, given that coupling losses were made between the PV cell and the TEG. The author mentioned that the potential of upscaling the hybrid system would result in a better overall performance.
Moreover, the PV-TEG system has also been modelled to investigate the effect of actual weather conditions on the hybrid system. Rezania et al., 2016 [23], evaluated the potential of the hybrid system under the weather of several sample cities in Europe using a thermal equivalent model. The authors found that the summer weather condition produces a higher power output than in winter for the hybrid system. However, the opposite holds for the overall efficiency of the system where a slight drop in efficiency was observed in July as compared to in January. Yin et al., 2018 [24], used one day weather data in Nanjing, China to analyze the one day performance of the PV-TEG system. The findings showed an improved performance of the hybrid system, especially during 11:00 to 13:00, where the highest TEG efficiency was obtained.
In this paper, a functional prototype of an unconcentrated PV-TEG hybrid system is designed and sized based on the power output supply required by the load used for indoor farming. An experimental investigation to test the performance of the hybrid system under Malaysia's weather conditions was carried out. The PV-TEG hybrid system performance as a combined system and individually were obtained and are discussed. The possibility of night time power production from the TEGs are also examined. The effects of changing TEG circuit connections using a combination of series and parallel in fluctuating temperature differences are explored. The sufficiency of power output generated by the hybrid system with the load requirement is evaluated. Finally, the feasibility of the designed hybrid system is analyzed through comparison with a sole PV system in regard to the efficiency of systems.

Load Requirement
The hybrid system was designed based on the load applied for indoor farming, which was the lighting used in this research. High pressure sodium (HPS) is usually used as artificial lighting for indoor plant cultivation. However, due to its low efficiency and high power usage, the currently more preferred artificial lighting is light emitting diode (LED) [25]. There are variations in the amount of light required for optimal indoor plant growth. According to Al-Chalabi, most plants can thrive with the usage of artificial light for 18 h/day [26]. Figure 1 shows the designed LED arrangement in the indoor farming area.  [27]. Figure 1 illustrates the arrangement of LED lights used for the area, where the total length of LED used was 4.2 m. Based on the data sheet provided by the manufacturer, the LED lights rated power usage is assumed to be 14.4 W/m. Table 1 summarizes the power required from the LED lights. Based on the energy load requirement obtained in the previous section, system sizing can be calculated and determined. This study used a PV panel (model: ADL100-12V), a TEG (model: SP1848-27145), and a battery (model: PGEL 100-12). The specifications of PV panel, battery, and TEG are shown in Tables 2-4, respectively.   Figure 1 illustrates the arrangement of LED lights used for the area, where the total length of LED used was 4.2 m. Based on the data sheet provided by the manufacturer, the LED lights rated power usage is assumed to be 14.4 W/m. Table 1 summarizes the power required from the LED lights.   Malaysia has an estimated annual average daily solar irradiation (Irr) ranging from 4.21 kWh/m 2 to 5.56 kWh/m 2 [28]. For system sizing purposes, the value of solar irradiation (Irr) assumed was 4.21 kWh/m 2 . A solar PV panel has a standard testing condition with a solar irradiation (Irr STC ) value of 1000 W/m 2 . The following equations, Equations (1)-(5), used for system sizing refer to the equations given by Khatri [29]. The panel generation factor (ε gen ) can be determined using Equation (1), as follows: The Watt peak rating (W p ) for a PV module is the ratio of total energy required from load (E T ) to panel generation factor (ε gen ) as given by Equation (2), as follows: Thus, the number of PV modules required can be found using Equation (3), as follows: No. of modules = W p /ε gen .
Batteries are incorporated with the system to support the hybrid in cases where there is lack of sunlight, such as during nighttime and cloudy days. The required battery capacity (C) can be calculated with the known parameter of battery nominal voltage (V nom ), efficiency (η), and depth of discharge (DOD), as given in Table 3. Assuming that the days of autonomy (N) is one, the required battery capacity (C) is calculated using Equation (4), as follows: Thus, the number of batteries required is based on Equation (5), as follows: No. of battery = C/C nom .
A solar charge controller is required for managing the current received from PV modules. It acts as a surge protector and current management between the PV panels, battery, and load. A solar charge controller of 30 A was used based on the PV module's short circuit current (I sc ), with an added safety factor of 25%. The TEG module used in this study is a SP1848-27145, with a dimension of 40 mm × 40 mm × 3.4 mm. Table 5 summarizes the hybrid system sizing based on the equations given.

Experimental Setup
The overall experimental setup of the hybrid system is as shown in Figure 2.

Experimental Setup
The overall experimental setup of the hybrid system is as shown in Figure 2. The PV panels and TEGs were placed outside of the G3 building in Universiti Teknologi Malaysia (UTM), which is located in Kuala Lumpur, Malaysia (Longitude/Latitude: 101°43' E/ 3°10' N). Other equipment, such as the solar charge controller, LED light, batteries, and measuring equipment, were placed indoors to avoid any malfunctioning when subjected to weather conditions. The experiment was run for a duration of three months, which were in December, February, and March. The length of the experiment allowed for varying weather conditions. The average ambient temperature and typical weather conditions are summarized as shown in Table 6. The PV panels were positioned outside of the G3 building, with two panels facing south and another two panels facing north, as shown in Figures 3 and 4. The PV placement was chosen in order to fully maximized the solar energy harvested for the stationary PV installation. The panels and batteries were connected electrically in parallel and were directly connected to the solar charge controller. The batteries, LED load, and solar charge controller were placed indoors, as shown in Figure 5. The PV panels and TEGs were placed outside of the G3 building in Universiti Teknologi Malaysia (UTM), which is located in Kuala Lumpur, Malaysia (Longitude/Latitude: 101 • 43' E/ 3 • 10' N). Other equipment, such as the solar charge controller, LED light, batteries, and measuring equipment, were placed indoors to avoid any malfunctioning when subjected to weather conditions. The experiment was run for a duration of three months, which were in December, February, and March. The length of the experiment allowed for varying weather conditions. The average ambient temperature and typical weather conditions are summarized as shown in Table 6. The PV panels were positioned outside of the G3 building, with two panels facing south and another two panels facing north, as shown in Figures 3 and 4. The PV placement was chosen in order to fully maximized the solar energy harvested for the stationary PV installation. The panels and batteries were connected electrically in parallel and were directly connected to the solar charge controller. The batteries, LED load, and solar charge controller were placed indoors, as shown in Figure 5.

TEG Connection
The design of this hybrid system includes the TEGs to be arranged behind the solar PV panel, as shown in Figure 6. A layer of thermal paste was placed on the adhered surface of the TEG to the PV panel for minimizing the thermal contact resistance, thus allowing better thermal conductivity

TEG Connection
The design of this hybrid system includes the TEGs to be arranged behind the solar PV panel, as shown in Figure 6. A layer of thermal paste was placed on the adhered surface of the TEG to the PV panel for minimizing the thermal contact resistance, thus allowing better thermal conductivity

TEG Connection
The design of this hybrid system includes the TEGs to be arranged behind the solar PV panel, as shown in Figure 6. A layer of thermal paste was placed on the adhered surface of the TEG to the PV panel for minimizing the thermal contact resistance, thus allowing better thermal conductivity

TEG Connection
The design of this hybrid system includes the TEGs to be arranged behind the solar PV panel, as shown in Figure 6. A layer of thermal paste was placed on the adhered surface of the TEG to the PV panel for minimizing the thermal contact resistance, thus allowing better thermal conductivity between

Data Collection
The PV panels and TEG used different data acquisition systems due to the distinct output measured. For the PV, the power, voltage, and current were measured, whereas, for the TEG, the open circuit voltage (Voc) was measured. The PV panels power output, battery state of charge, load power output, and ambient temperature were logged using a maximum power point tracker (MPPT) solar charge controller. The MPPT solar charge controller has an efficiency of 99.5% and a converter efficiency of 98%. The TEGs Voc were logged using the NI-USB-6211 data acquisition device (DAQ). The DAQ has a maximum sampling rate of 250 kS/s and an analog input absolute accuracy of 0.088% ± 0.13 mV (for a nominal range of 0.2 V-10 V). The temperature of the PV panel was obtained using K-type thermocouple with specifications, as presented in Table 7, and the temperature obtained was logged using the DAQ.

Data Collection
The PV panels and TEG used different data acquisition systems due to the distinct output measured. For the PV, the power, voltage, and current were measured, whereas, for the TEG, the open circuit voltage (Voc) was measured. The PV panels power output, battery state of charge, load power output, and ambient temperature were logged using a maximum power point tracker (MPPT) solar charge controller. The MPPT solar charge controller has an efficiency of 99.5% and a converter efficiency of 98%. The TEGs Voc were logged using the NI-USB-6211 data acquisition device (DAQ). The DAQ has a maximum sampling rate of 250 kS/s and an analog input absolute accuracy of 0.088% ± 0.13 mV (for a nominal range of 0.2 V-10 V). The temperature of the PV panel was obtained using K-type thermocouple with specifications, as presented in Table 7, and the temperature obtained was logged using the DAQ.

Data Collection
The PV panels and TEG used different data acquisition systems due to the distinct output measured. For the PV, the power, voltage, and current were measured, whereas, for the TEG, the open circuit voltage (V oc ) was measured. The PV panels power output, battery state of charge, load power output, and ambient temperature were logged using a maximum power point tracker (MPPT) solar charge controller. The MPPT solar charge controller has an efficiency of 99.5% and a converter efficiency of 98%. The TEGs V oc were logged using the NI-USB-6211 data acquisition device (DAQ). The DAQ has a maximum sampling rate of 250 kS/s and an analog input absolute accuracy of 0.088% ± 0.13 mV (for a nominal range of 0.2 V-10 V). The temperature of the PV panel was obtained using K-type thermocouple with specifications, as presented in Table 7, and the temperature obtained was logged using the DAQ. TEG efficiency (η TEG ) relies on the temperature difference between the two junctions. It is the ratio of power output from the TEG (P TEG ) to the rate of heat transfer ( . Q), given by Equation (6), as follows: where ( . Q) is calculated based on temperature difference (∆T) and the thermal conductivity of bismuth telluride Bi 2 Te 3 , (K th ) in Equation (7), as follows: Whereas, PV module efficiency (η PV ) is dependent on the solar irradiance absorbed by the area of PV cells. It is calculated based on the ratio of power output from the PV (P PV ) to the area of PV (A p ) exposed to solar irradiance. Referring to the equation used by Rezania et al., it can be defined by the following Equation (8) [23]: The overall electrical efficiency of the combined PV-TEG system (η PVTEG ) can be denoted using Equation (9), as follows:

PV with Added Battery System Performance
This section elaborates on the PV with added battery system performance. The energy profile for load, PV, and battery were shown for all three months and the results were explained in detail. Furthermore, the effect of load addition to the system was discussed in the following subsection.

PV, Battery, and Load Energy Profile
A 5 m LED light was used as the load requirement for the indoor farm. On average, the LED light required about 20 W of electricity and the light was adjusted to be switched on and off at a specific time during the day. The power distribution between the batteries and the PV with load, as well as the battery's state of charge (SOC), are shown in Figures 8-10. In December, the usage of the LED light was only for 13.5 h. It was switched on from 08:00 to 16:00 and was switched off for 2.5 h before it was continually used during the night time. In February, LED light usage was increased to 18 h. The load had a more prolonged night time usage of 10 h compared to a mere 8 h usage during day time. Whereas, a continuous 24 h usage of the LED light was done in March.     Malaysia is blessed with an abundance of sunlight, which allowed the PV panel to produce consistent power for a duration of 12 h; thus, supporting the load during day time and charging the batteries during offload period. The amount of solar irradiance available, however, differed between the three months, with December having the lowest amount of irradiance followed by March, while February had the highest. Typically, the availability of solar irradiance begins from 07:30 h until 19:30 h. During night time, in the absence of power from PV panels, batteries were the sole power provider for running the load. As shown, the PV-battery system was correctly sized, thus enabling it to fully support the LED load under all conditions. However, in the real case for indoor farming, the LED load does not need to be in the 'ON' state for a duration of 24 h [26]. The breaks should preferably be during the availability of PV power in order to recharge the battery sufficiently for night time usage, as done in December and February.
The solar panel used has an open circuit voltage and a short circuit current of 22.45 V and 5.99 A. At standard testing conditions of G = 1000 W/m 2 and Top = 25 °C, the PV panel had a maximum power output of 100 W. The total active area of the PV module used was 2.5 m 2 under no shading conditions and 1.25 m 2 under partial shading conditions. These conditions changed due to the position of the PV panel, which was observed to be under partial shading conditions starting from 13:30 h. As a result, the shading condition may have contributed to an inconsistent power output of the panels. In December, higher power output was observed, ranging from 15 W to 20 W. Whereas, February and March had a slightly lower power output of around 15 W maximum. The performance of the solar panels was also dependent on many other factors such as dust accumulation, increased solar cell temperature, and losses from wiring and connections.

Effect of Load to the System
The PV-battery system was tested with an additional load from a submersible pump to emulate the usage of a pump for the irrigation of crops in the indoor farm, as shown in Figure 11. Firstly, the system was run with the usual LED load of about 20 W. A slight drop in the battery SOC by 5% was observed, which was similar to that as previously shown in Figures 8-10. Next, the system load was switched to the pump, with a power requirement of about 90 W. A further drop in the battery SOC, by 5%, was noted. The system managed to sustain the load without any further reduction in the battery SOC within the duration of 100 min. Finally, both the pump and LED load were used simultaneously for a short period of 60 min. Although this increased the power requirement from the loads to above 100 W, the battery maintained an SOC of 70% throughout the duration of usage. This Malaysia is blessed with an abundance of sunlight, which allowed the PV panel to produce consistent power for a duration of 12 h; thus, supporting the load during day time and charging the batteries during offload period. The amount of solar irradiance available, however, differed between the three months, with December having the lowest amount of irradiance followed by March, while February had the highest. Typically, the availability of solar irradiance begins from 07:30 h until 19:30 h. During night time, in the absence of power from PV panels, batteries were the sole power provider for running the load. As shown, the PV-battery system was correctly sized, thus enabling it to fully support the LED load under all conditions. However, in the real case for indoor farming, the LED load does not need to be in the 'ON' state for a duration of 24 h [26]. The breaks should preferably be during the availability of PV power in order to recharge the battery sufficiently for night time usage, as done in December and February.
The solar panel used has an open circuit voltage and a short circuit current of 22.45 V and 5.99 A. At standard testing conditions of G = 1000 W/m 2 and T op = 25 • C, the PV panel had a maximum power output of 100 W. The total active area of the PV module used was 2.5 m 2 under no shading conditions and 1.25 m 2 under partial shading conditions. These conditions changed due to the position of the PV panel, which was observed to be under partial shading conditions starting from 13:30 h. As a result, the shading condition may have contributed to an inconsistent power output of the panels. In December, higher power output was observed, ranging from 15 W to 20 W. Whereas, February and March had a slightly lower power output of around 15 W maximum. The performance of the solar panels was also dependent on many other factors such as dust accumulation, increased solar cell temperature, and losses from wiring and connections.

Effect of Load to the System
The PV-battery system was tested with an additional load from a submersible pump to emulate the usage of a pump for the irrigation of crops in the indoor farm, as shown in Figure 11. Firstly, the system was run with the usual LED load of about 20 W. A slight drop in the battery SOC by 5% was observed, which was similar to that as previously shown in Figures 8-10. Next, the system load was switched to the pump, with a power requirement of about 90 W. A further drop in the battery SOC, by 5%, was noted. The system managed to sustain the load without any further reduction in the battery SOC within the duration of 100 min. Finally, both the pump and LED load were used simultaneously for a short period of 60 min. Although this increased the power requirement from the loads to above 100 W, the battery maintained an SOC of 70% throughout the duration of usage. This proved that the system was able to perform even with the addition of a greater load, providing sufficient power to the load regardless of the low PV power output.
Sustainability 2019, 11, x FOR PEER REVIEW 12 of 21 proved that the system was able to perform even with the addition of a greater load, providing sufficient power to the load regardless of the low PV power output.

TEG Performance
This section discusses the TEGs performance for the three months. The temperature profile for both panel and ambient temperature obtained in the three months is shown and elaborated upon. Moreover, the power output obtained for the three different connections are depicted and compared. The effect of resistance to the three different connections are also illustrated and discussed in the final subsection. Temperature   Figures 12-14 show the average ambient temperature and panel temperature for the months of December, February, and March. The end of the year, between October and December, is the monsoon season in West Peninsular Malaysia, which means that a low ambient temperature is often observed during this duration of time. Thus, in December, a lower temperature difference was produced during the daytime compared to the other months, as shown in Figure 12. February had the highest average ambient temperature, followed by March, as shown in Figures 13 and 14. The panel temperature was highly affected by the ambient temperature and the amount of precipitation, which explained the low panel temperature in December, which reached only up to 35 °C, whereas in February and March the panel temperature reached 55 °C and 45 °C, respectively.

TEG Performance
This section discusses the TEGs performance for the three months. The temperature profile for both panel and ambient temperature obtained in the three months is shown and elaborated upon. Moreover, the power output obtained for the three different connections are depicted and compared. The effect of resistance to the three different connections are also illustrated and discussed in the final subsection. Temperature   Figures 12-14 show the average ambient temperature and panel temperature for the months of December, February, and March. The end of the year, between October and December, is the monsoon season in West Peninsular Malaysia, which means that a low ambient temperature is often observed during this duration of time. Thus, in December, a lower temperature difference was produced during the daytime compared to the other months, as shown in Figure 12. February had the highest average ambient temperature, followed by March, as shown in Figures 13 and 14. The panel temperature was highly affected by the ambient temperature and the amount of precipitation, which explained the low panel temperature in December, which reached only up to 35 • C, whereas in February and March the panel temperature reached 55 • C and 45 • C, respectively. observed during this duration of time. Thus, in December, a lower temperature difference was produced during the daytime compared to the other months, as shown in Figure 12. February had the highest average ambient temperature, followed by March, as shown in Figures 13 and 14. The panel temperature was highly affected by the ambient temperature and the amount of precipitation, which explained the low panel temperature in December, which reached only up to 35 °C, whereas in February and March the panel temperature reached 55 °C and 45 °C, respectively.    Figure 15 depicts the difference between the PV panel and ambient temperature for the three months. The temperature difference in February and March were very similar, where, during the night time, a temperature difference of 14 °C to 16 °C was reached. During day time, a maximum temperature difference of 20 °C and 18 °C were achieved in February and March, respectively. However, the temperature difference obtained for December was slightly less than other months, with a night time temperature range of 12 °C to 14 °C and a maximum day time temperature difference of 8 °C, as shown in Figure 15.   Figure 15 depicts the difference between the PV panel and ambient temperature for the three months. The temperature difference in February and March were very similar, where, during the night time, a temperature difference of 14 °C to 16 °C was reached. During day time, a maximum temperature difference of 20 °C and 18 °C were achieved in February and March, respectively. However, the temperature difference obtained for December was slightly less than other months, with a night time temperature range of 12 °C to 14 °C and a maximum day time temperature difference of 8 °C, as shown in Figure 15.  Figure 15 depicts the difference between the PV panel and ambient temperature for the three months. The temperature difference in February and March were very similar, where, during the night time, a temperature difference of 14 • C to 16 • C was reached. During day time, a maximum temperature difference of 20 • C and 18 • C were achieved in February and March, respectively. However, the temperature difference obtained for December was slightly less than other months, with a night time temperature range of 12 • C to 14 • C and a maximum day time temperature difference of 8 • C, as shown in Figure 15.

TEGs Power Output for December, February, and March
Different TEG connections were done for the three months, where it was firstly connected in series during December. Among the other months, it had the lowest power output, varying between 0 mW and 0.5 mW, as shown in Figure 16. Nonetheless, this connection shows the most promising in terms of power production during the night time. The TEGs were able to produce power starting from 19:30 to 07:30 the following day. Next, the TEG was connected in a combination of series and parallel in February, where four TEG were in series and the three groups were connected in parallel. The power output of the TEG in this month peaked at 1.6 mW at 13:00 h and maintained at a higher range than in December, as shown in Figure 17.

TEGs Power Output for December, February, and March
Different TEG connections were done for the three months, where it was firstly connected in series during December. Among the other months, it had the lowest power output, varying between 0 mW and 0.5 mW, as shown in Figure 16. Nonetheless, this connection shows the most promising in terms of power production during the night time. The TEGs were able to produce power starting from 19:30 to 07:30 the following day. Next, the TEG was connected in a combination of series and parallel in February, where four TEG were in series and the three groups were connected in parallel. The power output of the TEG in this month peaked at 1.6 mW at 13:00 h and maintained at a higher range than in December, as shown in Figure 17.

TEGs Power Output for December, February, and March
Different TEG connections were done for the three months, where it was firstly connected in series during December. Among the other months, it had the lowest power output, varying between 0 mW and 0.5 mW, as shown in Figure 16. Nonetheless, this connection shows the most promising in terms of power production during the night time. The TEGs were able to produce power starting from 19:30 to 07:30 the following day. Next, the TEG was connected in a combination of series and parallel in February, where four TEG were in series and the three groups were connected in parallel. The power output of the TEG in this month peaked at 1.6 mW at 13:00 h and maintained at a higher range than in December, as shown in Figure 17.  In March the TEGs were connected similar to in February, except that six TEG were in series and the two groups were connected in parallel. This month had the highest power output, which indicated its superiority compared to other connections, regardless of it having similar temperature differences during the day. The highest recorded power output reached 3 mW and maintained a power output of higher than 2 mW from 12:00 to 16:00, as shown in Figure 18. Generally, starting from 12:00 to 16:00, the system produced the most power from the TEG for the three months, regardless of the connections. It could be related to the high temperature difference between the PV panel and ambient temperature. Nonetheless, the connection of series and parallel combination is not recommended during night time due to the possibility of producing a short circuit from uneven temperature distribution amongst the TEGs.

Resistance Effect from Different TEG Configurations
TEG connections play an important role in determining total power output for TEG, besides the temperature difference between the hot and cold side. This is due to the variation in matched load resistance with changing TEG connections. Figures 19-21 show the changing resistance value of the TEG with differing TEG connections. Different connections seem to produce a different power output at the same temperature difference, wherein series connections produced the highest power output at 12 °C than the series and parallel combination connections. A similar pattern was also observed with the resistance for the different TEG connections, where an all series connection had the highest internal resistance, which approximately matched the optimal load at maximum power output [30]. The optimal load of about 55 Ω to 60 Ω was found for an all series connections, whereas a combination In March the TEGs were connected similar to in February, except that six TEG were in series and the two groups were connected in parallel. This month had the highest power output, which indicated its superiority compared to other connections, regardless of it having similar temperature differences during the day. The highest recorded power output reached 3 mW and maintained a power output of higher than 2 mW from 12:00 to 16:00, as shown in Figure 18. Generally, starting from 12:00 to 16:00, the system produced the most power from the TEG for the three months, regardless of the connections. It could be related to the high temperature difference between the PV panel and ambient temperature. Nonetheless, the connection of series and parallel combination is not recommended during night time due to the possibility of producing a short circuit from uneven temperature distribution amongst the TEGs. In March the TEGs were connected similar to in February, except that six TEG were in series and the two groups were connected in parallel. This month had the highest power output, which indicated its superiority compared to other connections, regardless of it having similar temperature differences during the day. The highest recorded power output reached 3 mW and maintained a power output of higher than 2 mW from 12:00 to 16:00, as shown in Figure 18. Generally, starting from 12:00 to 16:00, the system produced the most power from the TEG for the three months, regardless of the connections. It could be related to the high temperature difference between the PV panel and ambient temperature. Nonetheless, the connection of series and parallel combination is not recommended during night time due to the possibility of producing a short circuit from uneven temperature distribution amongst the TEGs.

Resistance Effect from Different TEG Configurations
TEG connections play an important role in determining total power output for TEG, besides the temperature difference between the hot and cold side. This is due to the variation in matched load resistance with changing TEG connections. Figures 19-21 show the changing resistance value of the TEG with differing TEG connections. Different connections seem to produce a different power output at the same temperature difference, wherein series connections produced the highest power output at 12 °C than the series and parallel combination connections. A similar pattern was also observed with the resistance for the different TEG connections, where an all series connection had the highest internal resistance, which approximately matched the optimal load at maximum power output [30]. The optimal load of about 55 Ω to 60 Ω was found for an all series connections, whereas a combination

Resistance Effect from Different TEG Configurations
TEG connections play an important role in determining total power output for TEG, besides the temperature difference between the hot and cold side. This is due to the variation in matched load resistance with changing TEG connections. Figures 19-21 show the changing resistance value of the TEG with differing TEG connections. Different connections seem to produce a different power output at the same temperature difference, wherein series connections produced the highest power output at 12 • C than the series and parallel combination connections. A similar pattern was also observed with the resistance for the different TEG connections, where an all series connection had the highest internal resistance, which approximately matched the optimal load at maximum power output [30]. The optimal load of about 55 Ω to 60 Ω was found for an all series connections, whereas a combination of series and parallel connections with six TEG in series had a smaller optimal load of about 15 Ω to 18 Ω. A smaller number of TEG used in series would result in a smaller optimal load, as shown in Figure 21, with four TEGs in series. of series and parallel connections with six TEG in series had a smaller optimal load of about 15 Ω to 18 Ω. A smaller number of TEG used in series would result in a smaller optimal load, as shown in Figure 21, with four TEGs in series.   It is worth noting that a circuit with all series connections may be ideal to be used for power production during night time. This connection has also shown its capability in producing a higher of series and parallel connections with six TEG in series had a smaller optimal load of about 15 Ω to 18 Ω. A smaller number of TEG used in series would result in a smaller optimal load, as shown in Figure 21, with four TEGs in series.   It is worth noting that a circuit with all series connections may be ideal to be used for power production during night time. This connection has also shown its capability in producing a higher  It is worth noting that a circuit with all series connections may be ideal to be used for power production during night time. This connection has also shown its capability in producing a higher power output than the combination of series and parallel, with respect to temperature difference. An all parallel circuit has been deemed as unsuitable in this case due to the heat dissipation produced from the flow of high current in the circuit, which may negatively impact the PV panel [31]. Besides that, a threshold for an optimum number of TEGs to be used is highly dependent on the type of TEG connections [31]. Only a specific number of TEGs can be connected in a circuit for maximum power output. Therefore, limiting the amount of TEG used in an upscale of this hybrid system may be more beneficial for the overall system performance.

PV-TEG Hybrid System Overall Performance
The PV-TEG hybrid system performance was determined based on the overall efficiency of the combined system. The efficiencies of the solar PV, TEG, and the PV-TEG hybrid systems are shown in Table 8. Power outputs from TEGs are much smaller compared to the power output of solar PV. Therefore, the efficiency of the overall hybrid system relies heavily on the efficiency of the PV panels. Solar irradiance can be seen to have a linear relationship with the panel temperature where higher levels of solar irradiance result in a higher PV panel temperature, as shown in the previous sections. Solar PV efficiency is directly related to the solar cell temperature during operation [32]. The mono-silicon PV panel has been known to have a lower temperature coefficient of the maximum power output ranging from 0.35 to 0.4%/ • C compared to other PV materials [33,34]. Nonetheless, a high PV panel temperature of 55 • C in February and 45 • C in March results in a significant drop of the PV efficiency. An average efficiency of 7.66% and 8.61% was obtained for February and March, respectively. December had the best PV panel efficiency of 12.49% as the panel temperature reached only up to a maximum of 35 • C.
On the other hand, TEG efficiency is also is inversely proportional to the efficiency of the PV panel. The efficiency of TEG relies on the temperature difference between the two junctions, where a higher temperature would result in improved efficiency for the TEG. The Bi 2 Te 3 TEG used in this study showed an increase in efficiency with increasing temperature difference, as can be seen in Figure 22. power output than the combination of series and parallel, with respect to temperature difference. An all parallel circuit has been deemed as unsuitable in this case due to the heat dissipation produced from the flow of high current in the circuit, which may negatively impact the PV panel [31]. Besides that, a threshold for an optimum number of TEGs to be used is highly dependent on the type of TEG connections [31]. Only a specific number of TEGs can be connected in a circuit for maximum power output. Therefore, limiting the amount of TEG used in an upscale of this hybrid system may be more beneficial for the overall system performance.

PV-TEG Hybrid System Overall Performance
The PV-TEG hybrid system performance was determined based on the overall efficiency of the combined system. The efficiencies of the solar PV, TEG, and the PV-TEG hybrid systems are shown in Table 8. Power outputs from TEGs are much smaller compared to the power output of solar PV. Therefore, the efficiency of the overall hybrid system relies heavily on the efficiency of the PV panels. Solar irradiance can be seen to have a linear relationship with the panel temperature where higher levels of solar irradiance result in a higher PV panel temperature, as shown in the previous sections. Solar PV efficiency is directly related to the solar cell temperature during operation [32]. The monosilicon PV panel has been known to have a lower temperature coefficient of the maximum power output ranging from 0.35 to 0.4%/°C compared to other PV materials [33,34]. Nonetheless, a high PV panel temperature of 55 °C in February and 45 °C in March results in a significant drop of the PV efficiency. An average efficiency of 7.66% and 8.61% was obtained for February and March, respectively. December had the best PV panel efficiency of 12.49% as the panel temperature reached only up to a maximum of 35 °C.
On the other hand, TEG efficiency is also is inversely proportional to the efficiency of the PV panel. The efficiency of TEG relies on the temperature difference between the two junctions, where a higher temperature would result in improved efficiency for the TEG. The Bi2Te3 TEG used in this study showed an increase in efficiency with increasing temperature difference, as can be seen in Figure 22.  According to Table 8, the average efficiency obtained by the TEGs was highest in March, with 4.92%, followed by in February, with 4.44%, and finally December, with 1.86%. The results are proportional with the average temperature difference values of 6.49 • C, 5.63 • C, and 2.06 • C attained in March, February, and December, respectively; thus, explaining the low efficiency obtained by the TEG in the month of December. Although the months of February and March had different maximum temperature differences, with February having a higher maximum temperature difference of 20 • C, on average March had a more significant temperature difference than in February; therefore, justifying the minor difference in TEG efficiency between the two months.
The overall efficiency of the hybrid system was similar to the PV panel efficiency, with zero to less than 1% efficiency increment. The hybrid system in December had the highest efficiency of 10.81%, whereas in February and March had an overall system electrical efficiency of 7.72% and 8.68%, respectively. February and March showed a slight improvement in efficiency while December showed a dropped in efficiency by 1.68%. This implies that, although the overall efficiency relied mostly on the PV panel efficiency, it was still affected by the efficiency of the TEGs. March and February had a higher TEG efficiency due to their larger temperature difference as opposed to the December temperature difference. Thus, March and February had a better overall system efficiency in contrast to a sole PV-battery system, while December had a poorer overall system efficiency. Nonetheless, the hybrid system fared better in the colder month of December than in March or February, as confirmed with Rezania et al. who found that the hybrid system performed better in January during the winter season than in July during summer [23].
The PV average daily energy production for the December, February, and March were 177.22 Wh, 148.1 Wh, and 132.57 Wh. Meanwhile, the average daily energy production of the TEGs for December, February, and March were 4.04 mWh, 3.02 mWh, and 6.16 mWh. The PV and TEG power output do not seem to have any correlation where the highest PV output observed was in December, whereas the highest TEG output was in March. This is because of the different input of the system, where PV relies on the availability of solar irradiance and is greatly affected by temperature, whereas TEG output is purely influenced by temperature difference. Previous studies have confirmed that TEG only contributed a small amount of power to an unconcentrated hybrid system and is incomparable to the amount produced from solar PV [23]. Nonetheless, the addition of TEG to a hybrid system of a lower efficiency PV panel, such as a-Si or polymer, may be able to contribute more to the overall power output of the system [19]. Due to the usage of mono silicon PV in this study, the addition of power produced from the TEGs was deemed insignificant as opposed to the high-power output generated from the PV panels. Proper modification for producing an upscale of the unconcentrated PV-TEG system, such as by incorporating a heat sink and using better semiconductor material with a higher figure of merit (zT), may result in further improvement to the system.

Conclusions
This research focused on the design of a PV-TEG hybrid system for potential indoor farming applications. In striving for agricultural sustainability, usage of renewable energy for agriculture activities was placed as a priority, especially with the continually improved methods used in agriculture, such as the indoor farming approach. Solar energy technologies, particularly solar PV, were the most suitable choice for countries with a hot and humid climate like Malaysia. In order to combat the unreliability of solar PV systems, a hybrid PV-TEG system was designed to further improve the efficiency of a sole PV system. The hybrid system was designed to provide sufficient load to the LED used in indoor farming. The system was experimentally tested under actual Malaysian weather conditions in the months of December, February, and March. The system performance was analyzed based on the changing ambient and panel temperature, solar irradiance, and the effect of TEG connections to the overall power output. From the experiment conducted, the significant findings are summarized as follows: The power outputs of PV for December, February, and March were slightly varied, ranging between 15 W and 20 W. The variation was caused by factors such as partial shading and solar module temperature while running the experiment. Whereas the TEGs power output was profoundly affected by the weather conditions, with the lowest power output obtained in December, followed by February and March. Different connections of the TEGs in the hybrid system were investigated in this study. The findings showed that the best connection for high and consistent power output from the TEGs throughout the day was a series connection. The matched optimal load for the series was considerably higher than a parallel and series combination, with a range between 50 Ω and 60 Ω, which affected the power output when used in the colder month. Although night time power production was possible with series connections, it also raised some uncertainties when applying for long term usage. Thus, for practical reasons, a combination of parallel and series connections with higher TEG numbers in series is preferred, such as that done for the month of March.
The hybrid system designed was able to successfully provide sufficient power output to the LED load for the months of December, February, and March. This study affirmed that the hybrid system main power output was from the PV panels, whereas additional power output produced from the TEGs was minor and incomparable to the PV output. Thus, the load was fully supported by the PV system instead of the hybrid system. The PV system was able to support not only the existing load, but also an additional load of a pump, despite it having a low power output of 15 W to 20 W in February and March. This suggests that a fully independent indoor farm relying on only renewable energy sources is possible with the hybrid system designed in Malaysia.
In the present result, the combination of PV-TEG has shown that the performance of the hybrid system was slightly improved when compared with a sole PV system. The hybrid system was found to have an increase in efficiency of 0.06%, as compared to stand alone PV. The hybrid system efficiency was also shown to be affected by TEG efficiency. A drop in efficiency by 1.68% was found when the hybrid system operated in cooler temperature, where the lowest TEG efficiency was obtained in the month of December. Regardless, this study showed the possibility of an improved hybrid system during favourable weather conditions, which were in hotter and less rainy months. Nonetheless, the hybrid system performance was better in a cooler month as compared to hotter months. Future work on improving the design of the PV-TEG system, such as by incorporating heat sinks, may provide a better outcome for the hybrid system.