Electrical Energy Producing Greenhouse Shading System with a Semi-Transparent Photovoltaic Blind Based on Micro-Spherical Solar Cells

: An increasing population and limited arable land area endanger sufﬁcient and variegated food supplies worldwide. Greenhouse cultivation enables highly intensive plant production and thereby enables the production of abundant fresh vegetables and fruits. The salient beneﬁts of greenhouse cultivation are supported by ingenious management of crop environments, assisted by fossil fuel and grid electricity supplies. To reduce dependence on traditional energy resources, various studies have investigated exploitation of renewable energies for greenhouse environment management. Among them, solar photovoltaic (PV) technologies are anticipated to feed electrical energy to greenhouse appliances for microclimate control. This study proposes a venetian-blind-type shading system consisting of semi-transparent PV modules as blind blades based on micro-spherical solar cell technology to achieve greenhouse shading and electricity production concurrently. In response to the solar irradiance level, the PV blind inclination was altered automatically using a direct current (DC) motor driven by electrical energy generated by the PV blind itself. The PV blind was operated continuously during a ﬁve-month test period without outage. Moreover, the PV blind generated surplus electrical energy of 2125 kJ for blind system operations during the test period. The annual surplus energy calculated under the present experimental condition was 7.8 kWh m − 2 year − 1 , suggesting that application of the PV blind to a greenhouse roof enables sunlight level control and electrical appliance operations in the greenhouse with a diminished fuel and grid electricity supply, particularly in high-insolation regions.


Introduction
Greenhouse cultivation allows intensive plant production supported by ingenious management of crop environments, assisted with fossil fuel and grid electricity supplies. Demands for fuel and electricity have increased as growers strive to improve crop yield and quality and to extend cultivation seasons and geography, partly because of expectations to feed the increasing populations [1]. As the dependence on fuel and grid electricity increases, the risk of losing stability of growers' profits increases because of the fluctuating prices of energy resources. Furthermore, the use of fossil fuels Among them, dynamic regulation of greenhouse-roof PV shading is expected to provide a better balance between crop cultivation and electricity production. In fact, the performance can be improved further by exploiting semi-transparent PV technologies. For this reason, we developed a greenhouse venetian-blind-type shading system in this study using semi-transparent PV modules as blind blades. The PV blind characteristics and the energetic performances of the PV blind operations are reported herein.
Active blind-type shading systems have been investigated widely for building applications [54]. However, such systems have been investigated only rarely for greenhouse applications, although shading control is extremely important in greenhouse cultivations [12]. In the present manuscript, a novel PV blind system specific for greenhouse applications is presented. The system is designed with an automatic blind-angle control in response to the solar irradiance level at the greenhouse site. The PV blind can rotate parallel or perpendicular to the greenhouse roof, according to desired and pre-chosen irradiation levels. This function enhances the efficacy of the reported PV blind systems [51,52], providing more hospitable cultivation conditions in greenhouses under fluctuating sky conditions. The semi-transparency of the PV blind provides the benefits of sunlight availability to crops. Moreover, the energy generated by the PV blind can compensate the electricity demands for greenhouse environment management in addition to the PV blind operations.

Bifacial Semi-Transparent PV Modules Used as Blind Blades
The prototype PV blind had three semi-transparent PV modules as blind blades ( Figure 1). For this study, only three identical PV modules were assembled because of the difficulty of manufacturing the special semi-transparent PV modules, which thereby increased manufacturing costs. The specifications of the PV modules are presented in Table 1. In the bifacial PV module, numerous spherical micro-PV cells (Sphelar ® ; Sphelar Power Corp., Kyoto, Japan) were embedded in the transparent resin layer with conductor wires (0.38 mm-wide and 0.1 mm-diameter) connecting each cell [55]. The PV cells were aligned between the conductors (Figure 1a), which draw generated electric power to the external circuit. Each cell had a p-type semiconductor inner core coated with an n-type semiconductor outer shell [26,27,[55][56][57]. The see-through semi-transparency is a particular merit of using the dispersed numerous micro-PV cells. Between a pair of the linear conductors, 62 PV cells were aligned (Figure 1a). This arrangement was repeated 74 times in the 154 mm × 158 mm rectangular area. Each PV module had three rectangular semi-transparent areas. The odd and even lines of the cell series were defined as a i and b i (i = 1-37), respectively. The 124 cells in each a i and b i pair were connected electrically in parallel. The 37 pairs of a i and b i were connected in series. The cell arrangements in the three rectangular areas were connected in parallel to constitute the 1.2 W rated power per single PV module. Finally, the electricity output terminals of the three PV modules were connected in parallel.
The PV module rim was enclosed within an aluminum frame. Three PV modules were aligned on the common rotation axis (Figure 1b). A shaft joined the PV modules and a geared direct current (DC) motor (SS23F-LH-860-DC12V; Sawamura Denki Ind. Co., Ltd., Kanagawa, Japan) with rated specifications of 12 V voltage, 0.3 A current, 2.0 N m torque, and 4 rpm rotation speed. The PV modules were supported with a 1872 mm × 825 mm frame. Two pyranometers, P PVT and P PVB (ML-01; Eko Instruments Co. Ltd., Tokyo, Japan), were installed proximally at the long side of the central PV module (Figure 1b). The accuracy of the pyranometer was ±1.70% for irradiance measurements. The P PVT aligned as its normal coincided with the PV module normal to measure global irradiance on the PV top surface I PVT . The P PVB directed 180 • , thereby aligned as opposite, measured global irradiance on the PV back surface I PVB . A pyranometer P Cell was installed at 0.2 m below the PV module with its normal direction aligned to the PV module normal. P Cell measured the irradiance in the semi-transparent PV module shadow I Cell to determine the sunlight transmittance of the PV blind. PPVT and PPVB are pyranometers facing opposite directions to measure the total incident irradiance on the bifacial PV module. S1 and S2 are mechanical switches to stop the module rotation at the perpendicular or parallel position to the greenhouse roof when the module contacts with S1 or S2. CW-clockwise; CCW-counter-clockwise.  Figure 1. Semi-transparent photovoltaic (PV) module as a venetian-blind-blade for the greenhouse shading application: overview of the PV module with close-up photograph and cross-sectional structure of the spherical Si-PV cell; (a): block diagram of the prototype PV blind system (b): P PVT and P PVB are pyranometers facing opposite directions to measure the total incident irradiance on the bifacial PV module. S 1 and S 2 are mechanical switches to stop the module rotation at the perpendicular or parallel position to the greenhouse roof when the module contacts with S 1 or S 2 . CW-clockwise; CCW-counter-clockwise.

Sunlight to Electricity Conversion Characteristics of the PV Blind
The sunlight to electricity conversion characteristics of the PV blind were measured at a field plot on the Shimane University campus (35 • 29 N, 133 • 04 E) on 10 October 2017; a sunny day. This experiment was designed to evaluate basic characteristics of the PV blind irradiated with natural sunlight without obstructions such as greenhouse glazing. The PV modules were supported 2 m above the ground (Figure 2). The PV module inclination was fixed at 26.5 • , corresponding to the common slope of a conventional greenhouse glazing roof of this region. The azimuth of the PV-module normal was directed to true south. I PVT and I PVB were measured using pyranometers P PVT and P PVB (Figure 1b). Another pyranometer (ML-01) was positioned horizontally 2 m above the ground on the frame to measure the horizontal global irradiance I H . The current i PV -voltage V PV characteristics of the PV blind were measured using a voltage and current source/meter (6241A; ADC Corp., Tokyo, Japan) and a data acquisition unit (34970A; Agilent Technologies Inc., Santa Clara, CA, USA) at 1 min intervals. The accuracies of the voltage and current source/meter were ±0.02% for voltage and ±0.05% for current. The accuracy of the data acquisition unit was ±0.005% for DC voltage measurements. The module efficiency η M was determined by the percentage of the PV modules' power output P PV to the impinging irradiance I PVT + I PVB on the area S PV of 500 mm × 200 mm × 3 modules as The angle γ between direct beam sunlight irradiating on the PV module and the PV-module normal (Figure 1b) was calculated at 1 min intervals for the experimental date and site using the following geometric formula [58].
where h, ϕ, α, and β represent the solar altitude, the PV module inclination, the azimuth of the sun, and the azimuth of the PV module normal, respectively.

Sunlight to Electricity Conversion Characteristics of the PV Blind
The sunlight to electricity conversion characteristics of the PV blind were measured at a field plot on the Shimane University campus (35°29′ N, 133°04′ E) on 10 October 2017; a sunny day. This experiment was designed to evaluate basic characteristics of the PV blind irradiated with natural sunlight without obstructions such as greenhouse glazing. The PV modules were supported 2 m above the ground (Figure 2). The PV module inclination was fixed at 26.5°, corresponding to the common slope of a conventional greenhouse glazing roof of this region. The azimuth of the PV-module normal was directed to true south. IPVT and IPVB were measured using pyranometers PPVT and PPVB (Figure 1b). Another pyranometer (ML-01) was positioned horizontally 2 m above the ground on the frame to measure the horizontal global irradiance IH. The current iPV-voltage VPV characteristics of the PV blind were measured using a voltage and current source/meter (6241A; ADC Corp., Tokyo, Japan) and a data acquisition unit (34970A; Agilent Technologies Inc., Santa Clara, CA, USA) at 1 min intervals. The accuracies of the voltage and current source/meter were ±0.02% for voltage and ±0.05% for current. The accuracy of the data acquisition unit was ±0.005% for DC voltage measurements. The module efficiency ηM was determined by the percentage of the PV modules' power output PPV to the impinging irradiance IPVT + IPVB on the area SPV of 500 mm × 200 mm × 3 modules as The angle γ between direct beam sunlight irradiating on the PV module and the PV-module normal (Figure 1b) was calculated at 1 min intervals for the experimental date and site using the following geometric formula [58]. cos = sin ℎ cos + cos ℎ sin cos | − | , where h, , α, and β represent the solar altitude, the PV module inclination, the azimuth of the sun, and the azimuth of the PV module normal, respectively.

PV-Blind Control Circuit
A motor drive circuit ( Figure 3) was developed to turn the PV blind according to irradiance level. A pyranometer PSIGNAL (ML-01) transformed IH into voltage as the input signal of the control circuit. An operational amplifier (LM358; Texas Instruments Inc., Dallas, TX, USA) linearly amplified the PSIGNAL output voltage. The amplification factor was regulated by an R0 value of a variable resistor (Figure 3). The output voltage of the operational amplifier drove a transistor to control voltage VCW and VCCW at the IN1 and IN2 terminals of a DC motor full bridge driver (TB6643KQ; Toshiba Corp., Tokyo, Japan). The motor rotation direction was reversed according to the balance of VCW and VCCW [55]. The PV blind rotation from the perpendicular (θ = 90°, Figure 4b) to the parallel (θ = 0°, Figure 4a) angle relative to a greenhouse roof surface was defined as clockwise (CW). The reverse rotation was denoted as counter-clockwise (CCW). The R0 value set a threshold IH value for the blind rotation. The relation between the control voltage and the blind angle is

PV-Blind Control Circuit
A motor drive circuit ( Figure 3) was developed to turn the PV blind according to irradiance level. A pyranometer P SIGNAL (ML-01) transformed I H into voltage as the input signal of the control circuit. An operational amplifier (LM358; Texas Instruments Inc., Dallas, TX, USA) linearly amplified the P SIGNAL output voltage. The amplification factor was regulated by an R 0 value of a variable resistor ( Figure 3). The output voltage of the operational amplifier drove a transistor to control voltage V CW and V CCW at the IN 1 and IN 2 terminals of a DC motor full bridge driver (TB6643KQ; Toshiba Corp., Tokyo, Japan). The motor rotation direction was reversed according to the balance of V CW and V CCW [55]. The PV blind rotation from the perpendicular (θ = 90 • , Figure 4b) to the parallel (θ = 0 • , Figure 4a) angle relative to a greenhouse roof surface was defined as clockwise (CW). The reverse rotation was denoted as counter-clockwise (CCW). The R 0 value set a threshold I H value for the blind rotation. The relation between the control voltage and the blind angle is The motor control circuit, a battery with 50 Ah rated capacity (JC50-12; Denryo Co. Ltd., Tokyo, Japan), and the PV module output terminals were connected to a charge-discharge controller (SA-MN05-8; Denryo Co. Ltd., Tokyo, Japan) to integrate the stand-alone power system (Figures 1b and 3).
The motor control circuit, a battery with 50 Ah rated capacity (JC50-12; Denryo Co. Ltd., Tokyo, Japan), and the PV module output terminals were connected to a charge-discharge controller (SA-MN05-8; Denryo Co. Ltd., Tokyo, Japan) to integrate the stand-alone power system (Figures 1b  and 3).

Verification of the PV Blind Characteristics at the Test Greenhouse
The PV blind was installed underneath the 26.5°-sloped east-sky-facing roof of a north-south oriented greenhouse (5.65 m × 4.25 m) at the Shimane University campus ( Figure 5). The PV blind was installed underneath the roof because the prototype system was not water resistant. In the greenhouse, no crops were cultivated on the concrete floor. Pyranometer PH (ML-01) was used for measuring exterior horizontal global irradiance IH. Pyranometer PT (ML-01) measured global irradiance on the inclined greenhouse roof surface IT (facing 6° to the north from the true east with 26.5° inclination). Pyranometers PH, PT, and PSIGNAL were positioned on the greenhouse ridge at 3.5 m height.
Electrical power generation characteristics of the PV blind inclined at θ = 0° ( Figure 4a) were measured at 1 min intervals from 10:00 to 12:30 on 1 November 2017 using a data acquisition unit (34972A; Keysight Technologies Inc., Santa Rosa, CA, USA) and the voltage and current  Figure 3. PV blind (PV 1 -PV 3 ) control circuit: P SIGNAL transforms the horizontal global irradiance into voltage. A variable resistor R 0 sets a threshold irradiance level for blind rotations. An additional PV module (PV N ) supplied current to a MOS FET relay to turn on the gate at dawn and to turn off the gate at dusk, in order to eliminate standby energy consumption at the motor drive circuit during night-time [55].
The motor control circuit, a battery with 50 Ah rated capacity (JC50-12; Denryo Co. Ltd., Tokyo, Japan), and the PV module output terminals were connected to a charge-discharge controller (SA-MN05-8; Denryo Co. Ltd., Tokyo, Japan) to integrate the stand-alone power system (Figures 1b  and 3).

Verification of the PV Blind Characteristics at the Test Greenhouse
The PV blind was installed underneath the 26.5°-sloped east-sky-facing roof of a north-south oriented greenhouse (5.65 m × 4.25 m) at the Shimane University campus ( Figure 5). The PV blind was installed underneath the roof because the prototype system was not water resistant. In the greenhouse, no crops were cultivated on the concrete floor. Pyranometer PH (ML-01) was used for measuring exterior horizontal global irradiance IH. Pyranometer PT (ML-01) measured global irradiance on the inclined greenhouse roof surface IT (facing 6° to the north from the true east with 26.5° inclination). Pyranometers PH, PT, and PSIGNAL were positioned on the greenhouse ridge at 3.5 m height.
Electrical power generation characteristics of the PV blind inclined at θ = 0° ( Figure 4a) were measured at 1 min intervals from 10:00 to 12:30 on 1 November 2017 using a data acquisition unit (34972A; Keysight Technologies Inc., Santa Rosa, CA, USA) and the voltage and current

Verification of the PV Blind Characteristics at the Test Greenhouse
The PV blind was installed underneath the 26.5 • -sloped east-sky-facing roof of a north-south oriented greenhouse (5.65 m × 4.25 m) at the Shimane University campus ( Figure 5). The PV blind was installed underneath the roof because the prototype system was not water resistant. In the greenhouse, no crops were cultivated on the concrete floor. Pyranometer P H (ML-01) was used for measuring exterior horizontal global irradiance I H . Pyranometer P T (ML-01) measured global irradiance on the inclined greenhouse roof surface I T (facing 6 • to the north from the true east with 26.5 • inclination). Pyranometers P H , P T , and P SIGNAL were positioned on the greenhouse ridge at 3.5 m height.
Electrical power generation characteristics of the PV blind inclined at θ = 0 • (Figure 4a) were measured at 1 min intervals from 10:00 to 12:30 on 1 November 2017 using a data acquisition unit (34972A; Keysight Technologies Inc., Santa Rosa, CA, USA) and the voltage and current source/meter (6241A). The accuracy of the data acquisition unit was ±0.005% for DC voltage measurements.
Shading characteristics of the PV blind inclined at θ = 0 • were measured at 1 min intervals on 6 November 2017 using pyranometers P T , P PVT , and P Cell , the output terminals of which were connected with the data acquisition unit (34972A). The sunlight transmittance of the greenhouse glazing was determined as I PVT divided by I T . The sunlight transmittance of the PV module was determined as I Cell divided by I PVT .
Then, the PV blind was connected to the control circuit ( Figure 3). The values of I H , motor current i M , and motor voltage V M were measured on 27 November 2017 at 10 ms intervals using a data-logger (HIOKI8430; Hioki E. E. Corp., Nagano, Japan) to determine the instantaneous electrical energy consumption during motor rotations. The accuracy of the data-logger was ±0.1% for DC voltage and ±0.01% for time. The target threshold I H values for triggering the PV module rotations were set at 370, 400, and 440 W m −2 , respectively, achieved by changing the R 0 values to 661, 597, and 534 Ω, respectively.
Energies 2018, 11, x FOR PEER REVIEW 7 of 23 source/meter (6241A). The accuracy of the data acquisition unit was ±0.005% for DC voltage measurements. Shading characteristics of the PV blind inclined at θ = 0° were measured at 1 min intervals on 6 November 2017 using pyranometers PT, PPVT, and PCell, the output terminals of which were connected with the data acquisition unit (34972A). The sunlight transmittance of the greenhouse glazing was determined as IPVT divided by IT. The sunlight transmittance of the PV module was determined as ICell divided by IPVT.
Then, the PV blind was connected to the control circuit ( Figure 3). The values of IH, motor current iM, and motor voltage VM were measured on 27 November 2017 at 10 ms intervals using a data-logger (HIOKI8430; Hioki E. E. Corp., Nagano, Japan) to determine the instantaneous electrical energy consumption during motor rotations. The accuracy of the data-logger was ±0.1% for DC voltage and ±0.01% for time. The target threshold IH values for triggering the PV module rotations were set at 370, 400, and 440 W m −2 , respectively, achieved by changing the R0 values to 661, 597, and 534 Ω, respectively.

Energy Balance of PV Blind System Operations
The energy balance of the PV blind system installed underneath the greenhouse roof ( Figure 5) was measured from 8 December 2017 through 25 April 2018. The threshold IH value for PV blind rotations was set at 500 W m −2 to turn the PV blind actively during winter-spring. The PV-generated current iPV, the charging current into the battery iC, the control current to the motor drive circuit iD, IH, IT, IPVT, IPVB, and voltages at the PV module, the battery, and the circuit terminals were measured at 10 s intervals using the data acquisition unit (34972A). By integrating the product of the voltage and current with the duration, the cumulative energy generated by the PV module EPV, charged into the battery EC, and supplied to the motor drive circuit ED, was calculated. Subsequently, total cumulative energy consumption Ei of the charge/discharge controller, conducting wires, shunt resistances, and the internal resistance of the battery (i.e., internal system energy-dissipation) were calculated using the following formula: As described in the Results section, the energy generated by the PV blind exceeded the demand of the PV-blind system operations. Consequently, the battery voltage approached the upper limit (13 V) of the charging cutoff in mid-February 2018. Accordingly, discharge of energy stored in the battery to some loads was demanded. For this reason, on 14 February 2018, light-emitting diodes (LEDs, HLMP-1540; Broadcom Ltd., San Jose, CA, USA) were connected additionally in parallel to the load terminal of the charge-discharge controller (SA-MN05-8) as a model greenhouse load. The electrical power consumption of the additional LED load with current regulating diodes was 0.4 W. It was operated 24 h per day. The cumulative energy consumed at the additional LED load EL had been incorporated into Equation (4) since the LED load was added. Therefore, the energy balance equation became

Energy Balance of PV Blind System Operations
The energy balance of the PV blind system installed underneath the greenhouse roof ( Figure 5) was measured from 8 December 2017 through 25 April 2018. The threshold I H value for PV blind rotations was set at 500 W m −2 to turn the PV blind actively during winter-spring. The PV-generated current i PV , the charging current into the battery i C , the control current to the motor drive circuit i D , I H , I T , I PVT , I PVB , and voltages at the PV module, the battery, and the circuit terminals were measured at 10 s intervals using the data acquisition unit (34972A). By integrating the product of the voltage and current with the duration, the cumulative energy generated by the PV module E PV , charged into the battery E C , and supplied to the motor drive circuit E D , was calculated. Subsequently, total cumulative energy consumption E i of the charge/discharge controller, conducting wires, shunt resistances, and the internal resistance of the battery (i.e., internal system energy-dissipation) were calculated using the following formula: As described in the Results section, the energy generated by the PV blind exceeded the demand of the PV-blind system operations. Consequently, the battery voltage approached the upper limit (13 V) of the charging cutoff in mid-February 2018. Accordingly, discharge of energy stored in the battery to some loads was demanded. For this reason, on 14 February 2018, light-emitting diodes (LEDs, HLMP-1540; Broadcom Ltd., San Jose, CA, USA) were connected additionally in parallel to the load terminal of the charge-discharge controller (SA-MN05-8) as a model greenhouse load. The electrical power consumption of the additional LED load with current regulating diodes was 0.4 W. It was operated 24 h per day. The cumulative energy consumed at the additional LED load E L had been incorporated into Equation (4) since the LED load was added. Therefore, the energy balance equation became The last terms (E C + E L ) represent the surplus of electrical energy generated by the PV blind. On 14 February, all measurements were suspended to add the LED load.

Sunlight to Electricity Conversion Characteristics of the PV Blind
The sunlight to electricity conversion characteristics of the PV blind under natural sunlight were measured in a field plot on the Shimane University campus on 10 October 2017; a sunny day. I H , I PVT , and I PVB are depicted in Figure 6a. The PV blind received approximately 85% of the total irradiance from the sky and 15% from ground-reflected radiation. The power P PV -voltage V PV curves of the PV blind are depicted in Figure 6b. The peak values of each P PV curve were defined as P max . The maximum P max value of the day was 3.75 W (V PV = 14.9 V, i PV = 251 mA) at 11:59 when the total incident irradiance I PVB + I PVB = 1113 W m −2 (Figure 6c). The P max value of 3.75 W corresponds to 12.5 W m −2 . The P max values generally followed the course of bifacial incident irradiance I PVT + I PVB (Figure 6c). The PV module efficiency η M was determined by Formula (1) (Figure 6c). The average η M during 9:02-16:00 was 1.2% over a wide incident angle γ of direct sunlight (Figure 7a,b).
On 1 November 2017, the PV blind was installed underneath the eastern roof glazing of the north-south oriented greenhouse. The PV modules faced the eastern sky with 26.5 • inclination parallel to the greenhouse roof. Irradiance and P PV -V PV characteristics of the PV blind are depicted in Figure 6d,e. The PV modules were irradiated with direct sunlight from 10:00 to 12:30 through the greenhouse glazing. For other hours, the PV modules were frequently shaded by the greenhouse frame structures. The maximum P max value of the day was 2.04 W (V PV = 14.6 V, i PV = 139 mA) at 10:00 when incident irradiance I PVB + I PVB = 583 W m −2 (Figure 6f). The P max value of 2.04 W corresponds to 6.8 W m −2 . The average η M during 10:00-12:30 was 1.0% (Figure 6f). The loss of η M under the greenhouse roof was attributable to the partial shadow of the greenhouse narrow frames on part of the PV module area.
The last terms (EC + EL) represent the surplus of electrical energy generated by the PV blind. On 14 February, all measurements were suspended to add the LED load.

Sunlight to Electricity Conversion Characteristics of the PV Blind
The sunlight to electricity conversion characteristics of the PV blind under natural sunlight were measured in a field plot on the Shimane University campus on 10 October 2017; a sunny day. IH, IPVT, and IPVB are depicted in Figure 6a. The PV blind received approximately 85% of the total irradiance from the sky and 15% from ground-reflected radiation. The power PPV-voltage VPV curves of the PV blind are depicted in Figure 6b

Sunlight Transmittance of the PV Blind
The sunlight transmittance of the PV blind and the greenhouse glazing was measured on 6 November 2017. The PV blind was positioned underneath the eastern roof glazing of the greenhouse ( Figure 5). The PV blind was irradiated with direct beam sunlight through the greenhouse glazing from 10:00 to 12:30. Graphs of IT, IPVT, and ICell are portrayed in Figure 8a. IT peaked at 10:14. IPVT generally followed the course of IT with constantly attenuated values according to the partial sunlight scattering at the greenhouse glazing. However, around 10:25, IPVT decreased sharply because only pyranometer PPVT was shaded by a narrow greenhouse frame. PCell was in the shadow of the semi-transparent area (154 mm × 158 mm square, as depicted in Figure 1a) of the PV module from 10:01 to 11:17. Direct beam sunlight reached PCell from outside of the PV module during other hours. The sunlight transmittance of the greenhouse roof glass was ascertained as IPVT over IT when PPVT was not shaded by any opaque obstruction during 10:34-11:17. The sunlight transmittance of the PV module was determined by ICell over IPVT during 10:34-11:17. During the period, γ was 52-55° ( Figure 8b). The average sunlight transmittances of the greenhouse glass roof and the PV module were 85% and 60%, respectively (Figure 8c). The average value of the total transmittance of the overlapping greenhouse glazing and the PV module was 51% (Figure 8c).

Sunlight Transmittance of the PV Blind
The sunlight transmittance of the PV blind and the greenhouse glazing was measured on 6 November 2017. The PV blind was positioned underneath the eastern roof glazing of the greenhouse ( Figure 5). The PV blind was irradiated with direct beam sunlight through the greenhouse glazing from 10:00 to 12:30. Graphs of IT, IPVT, and ICell are portrayed in Figure 8a. IT peaked at 10:14. IPVT generally followed the course of IT with constantly attenuated values according to the partial sunlight scattering at the greenhouse glazing. However, around 10:25, IPVT decreased sharply because only pyranometer PPVT was shaded by a narrow greenhouse frame. PCell was in the shadow of the semi-transparent area (154 mm × 158 mm square, as depicted in Figure 1a) of the PV module from 10:01 to 11:17. Direct beam sunlight reached PCell from outside of the PV module during other hours. The sunlight transmittance of the greenhouse roof glass was ascertained as IPVT over IT when PPVT was not shaded by any opaque obstruction during 10:34-11:17. The sunlight transmittance of the PV module was determined by ICell over IPVT during 10:34-11:17. During the period, γ was 52-55° (Figure 8b). The average sunlight transmittances of the greenhouse glass roof and the PV module were 85% and 60%, respectively (Figure 8c). The average value of the total transmittance of the overlapping greenhouse glazing and the PV module was 51% (Figure 8c).

Sunlight Transmittance of the PV Blind
The sunlight transmittance of the PV blind and the greenhouse glazing was measured on 6 November 2017. The PV blind was positioned underneath the eastern roof glazing of the greenhouse ( Figure 5). The PV blind was irradiated with direct beam sunlight through the greenhouse glazing from 10:00 to 12:30. Graphs of I T , I PVT , and I Cell are portrayed in Figure 8a. I T peaked at 10:14. I PVT generally followed the course of I T with constantly attenuated values according to the partial sunlight scattering at the greenhouse glazing. However, around 10:25, I PVT decreased sharply because only pyranometer P PVT was shaded by a narrow greenhouse frame. P Cell was in the shadow of the semi-transparent area (154 mm × 158 mm square, as depicted in Figure 1a) of the PV module from 10:01 to 11:17. Direct beam sunlight reached P Cell from outside of the PV module during other hours. The sunlight transmittance of the greenhouse roof glass was ascertained as I PVT over I T when P PVT was not shaded by any opaque obstruction during 10:34-11:17. The sunlight transmittance of the PV module was determined by I Cell over I PVT during 10:34-11:17. During the period, γ was 52-55 • (Figure 8b). The average sunlight transmittances of the greenhouse glass roof and the PV module were 85% and 60%, respectively (Figure 8c). The average value of the total transmittance of the overlapping greenhouse glazing and the PV module was 51% (Figure 8c).

Observations of the PV Blind Rotations in Response to Irradiance
The PV blind operations were tested on 27 November 2017. The PV blind, connected with the control circuit (Figure 3), was positioned underneath the eastern glazing roof of the test greenhouse ( Figure 5). The PV blind turned automatically in response to IH. The desired threshold IH value for the CW or CCW rotation of the PV blind was achieved by implementing the proper R0 value. Table 2 presents the threshold IHs of the CW or CCW rotations at three R0 values. The motor drive-circuit operations are described herein using data obtained at R0 = 597 Ω. When

Observations of the PV Blind Rotations in Response to Irradiance
The PV blind operations were tested on 27 November 2017. The PV blind, connected with the control circuit (Figure 3), was positioned underneath the eastern glazing roof of the test greenhouse ( Figure 5). The PV blind turned automatically in response to I H . The desired threshold I H value for the CW or CCW rotation of the PV blind was achieved by implementing the proper R 0 value. Table 2 presents the threshold I H s of the CW or CCW rotations at three R 0 values. The motor drive-circuit operations are described herein using data obtained at R 0 = 597 Ω.   Table 2. Threshold irradiance IH, rotating duration t of the direct current (DC) motor, and energy consumption e during the clockwise (CW) and counter-clockwise (CCW) rotations at each R0 value of the PV blind control circuit.

Energy Balance of the PV Blind System
The PV blind was operated without outage for five months from 8 December 2017 through 25 April 2018 (Figure 10), although the PV blind and measurement systems were suspended during LED load installation on 14 February. The IH values in winter were generally low at the greenhouse location because of actual cloudy sky conditions and sunlight shading by university buildings during the low solar altitude season. For example, on 15 December, the IH curve was suppressed in the morning and afternoon by the building shade, although the sky was fair throughout the day. After mid-March, the IH curve became smooth on sunny days as the sun path became higher than the buildings. The PV blind was rotated 2255 times in the CW direction and 2255 times in the CCW direction during five months. On heavy cloudy days, the blind did not rotate. It retained θ = 90°, prioritizing sunlight intake into the greenhouse. On sunny days, PPV reached 2 W in December, 3 W in January and February, and 4 W in March and April; PPV was less than 0.5 W on heavy cloudy

Energy Balance of the PV Blind System
The PV blind was operated without outage for five months from 8 December 2017 through 25 April 2018 (Figure 10), although the PV blind and measurement systems were suspended during LED load installation on 14 February. The I H values in winter were generally low at the greenhouse location because of actual cloudy sky conditions and sunlight shading by university buildings during the low solar altitude season. For example, on 15 December, the I H curve was suppressed in the morning and afternoon by the building shade, although the sky was fair throughout the day. After mid-March, the I H curve became smooth on sunny days as the sun path became higher than the buildings. The PV blind was rotated 2255 times in the CW direction and 2255 times in the CCW direction during five months. On heavy cloudy days, the blind did not rotate. It retained θ = 90 • , prioritizing sunlight intake into the greenhouse. On sunny days, P PV reached 2 W in December, 3 W in January and February, and 4 W in March and April; P PV was less than 0.5 W on heavy cloudy days. The daily total system efficiency η 0 was calculated using Formula (6) according to the surplus energy of PV blind operations against the daily sunlight energy received by the horizontal occupation of the PV blind.
where P C and P L denote power input to the battery and power supplied to the load LEDs, respectively. P L had been zero until the LED load was installed to the system on 14 February.   (Figure 10a). It was a partially cloudy day. I H frequently exceeded 500 W m −2 during the day (Figure 11a). Thereby, the blind often turned to θ = 0 • . The PV blind generated electricity up to 3 W. Sunny day data obtained on 19 April 2018 are presented in Figure 11b. They show that P PV increased in the early morning because the PV blind faced the eastern sky. At 8:16, the PV blind turned to θ = 0 • when I H reached at 494 W m −2 . At that moment, P PV jumped up to 2.7 W. The P PV value was maintained as higher than 2 W until P PV dropped at 13:47, when direct sunlight failed to shine on the PV blind because of the reflection at the western greenhouse roof. The PV blind turned to θ = 90 • at 16:00 when I H decreased to 460 W m −2 . η 0 = 0.93% on 19 April. The lowest value of η 0 was recorded on 10 January when the greenhouse was covered with snow. I H was less than 100 W m −2 . Thereby, P PV was nearly 0 W. The blind system consumed stand-by operation energy that led to the least η 0 of −1.00%. The negative value means that the system maintenance energy was supplied by the battery. The mean and median values of η 0 were 0.22% and 0.25%, respectively, during winter (8 December to 13 February). From mid-February to mid-March, irradiance sometimes approached 1 kW m −2 instantaneously, but the sky was covered with discrete clouds. Accordingly, the number of blind turns increased as the irradiance frequently changed, as it did on 1 and 6 March. As the solar elevation increased, the sun path was not blocked by university buildings, which led to more available insolation and greater η 0 . From mid-March, the number of clear sky days increased. On cloudless days, blinds usually turned once in the morning to θ = 0 • and once in the afternoon to θ = 90 • . Cloudless days η 0 were constantly 1%. Mean and median values of η 0 were 0.58% and 0.68%, respectively, during spring (15 February through 25 April).
The PV blind was parallel to the roof for only 2% duration in the winter (Figure 12a). During the 2% duration, the PV blind generated 32% of the total electrical energy generated during the season (Figure 12b), indicating that the PV orientation parallel to the greenhouse roof is the best mode for electricity generation. The PV blind at the perpendicular angle to the roof produced 68% of electrical energy by virtue of the bifacial validity of the PV modules and isotropic susceptibility of the spherical PV cells. The duration at θ = 0 • increased to 14% in spring (Figure 12c). Accordingly, 72% of electrical energy was produced when the blind was parallel to the roof (Figure 12d).
At the end of the operation test, E PV , E C + E L , E D , and E i were 2659 kJ, 2125 kJ, 399 kJ, and 135 kJ, respectively ( Figure 13 and Table 3). Table 2 indicated that a single rotation of the PV blind consumes up to 15 J. The PV blind rotated 4510 times during the period. In all, 15 × 4510 = 67.7 kJ was approximately consumed for the blind rotations. This value corresponded to 17% of E D . E i was 1.0 kJ day −1 almost constantly (Table 3). Although the sky condition was generally overcast in winter, E C was greater than E D . Along with the daily insolation increased in spring, E PV increased gradually. Consequently, surplus energy E C + E L also increased ( Figure 13). Surplus electrical energy of 2125 kJ had been obtained by the end of the experiments.     Figure 13. Energy balance of the PV blind system during the five-month operation test: EPV, EC + EL, and ED represent the cumulative PV generated energy, surplus energy, and drive-circuit supplied energy, respectively.

Discussion
The PV system concurrently functioned as a blind to regulate the shading level in the greenhouse and as an electricity power generator. The PV blind system prioritized the sunlight intake in the greenhouse by virtue of its movability and semi-transparency, whereas conventional large-scale PV greenhouses with full [59][60][61][62] or half [63][64][65][66][67][68] roof area coverage prioritized electrical power generation. The prototype PV blind system operated successfully according to the threshold   Figure 13. Energy balance of the PV blind system during the five-month operation test: EPV, EC + EL, and ED represent the cumulative PV generated energy, surplus energy, and drive-circuit supplied energy, respectively.

Discussion
The PV system concurrently functioned as a blind to regulate the shading level in the greenhouse and as an electricity power generator. The PV blind system prioritized the sunlight intake in the greenhouse by virtue of its movability and semi-transparency, whereas conventional large-scale PV greenhouses with full [59][60][61][62] or half [63][64][65][66][67][68] roof area coverage prioritized electrical power generation. The prototype PV blind system operated successfully according to the threshold  Figure 13. Energy balance of the PV blind system during the five-month operation test: E PV , E C + E L , and E D represent the cumulative PV generated energy, surplus energy, and drive-circuit supplied energy, respectively.

Discussion
The PV system concurrently functioned as a blind to regulate the shading level in the greenhouse and as an electricity power generator. The PV blind system prioritized the sunlight intake in the greenhouse by virtue of its movability and semi-transparency, whereas conventional large-scale PV greenhouses with full [59][60][61][62] or half [63][64][65][66][67][68] roof area coverage prioritized electrical power generation. The prototype PV blind system operated successfully according to the threshold level of solar irradiance during winter-spring. The threshold level was regulated by changing the value of the variable resistor in the control circuit ( Table 2). The blind operation was automatic and energetically stand-alone with 2125 kJ of surplus electrical energy at the end of the experiments.
The annual energy balance of the PV-blind system operation was estimated using experimental and local meteorological statistical data. Annual I H was estimated based on local insolation data obtained during the prior nine years (2009-2017) [69]. Three levels of η 0 were assumed: 0.2%, 0.6%, and 1.0%. Assuming that replications of the present PV blind system cover the entire greenhouse roof underneath the glazing, the surplus electrical energy per unit greenhouse area was estimated annually (Figure 14). For η 0 = 0.2%, 2.6 kWh m −2 year −1 of greenhouse loads can be supplied. Yano et al. [70,71] demonstrated that a single-span plastic greenhouse equipped with a side ventilation controller demanded electrical energy of 0.1 kWh m −2 year −1 for automatic window operations in response to the greenhouse interior temperature ( Table 4). The present PV blind system would be sufficient to compensate such small-scale greenhouse demand for microclimate control. The actual surplus electrical energy obtained from the present experiments is also depicted in Figure 14. That is approximately on the line of η 0 = 0.6% estimation. Another study estimated that 7 kWh m −2 year −1 of electrical energy is demanded in a Spanish pepper greenhouse equipped with fans, irrigation and fertilization systems, fuel burners, window-operation motors, screen motors, climate control automatism, compressors, and fuel reservoirs [72]. In a Spanish greenhouse with window and pump operations, 3 kWh m −2 per eight months of electrical energy was demanded [29]. The surplus of electrical energy producible by the PV blind system underneath the roof glazing with η 0 = 0.6% was estimated as compensating the electricity demand in such Mediterranean greenhouses. The total compensation of electricity demands is particularly valuable in rural areas where greenhouses are often far from power lines [73]. The E C value of 13 kWh m −2 year −1 can be a surplus assuming that η 0 = 1%. Some additional greenhouse loads can be operated using this electrical energy. Campiotti et al. [74] reported that electrical power requirements range from 9 kWh m −2 year −1 for Mediterranean greenhouses with advanced climate control (heating, cooling or ventilation) to 2 kWh m −2 year −1 for low-technology greenhouses (Table 4). Some other reports have described 20-30 kWh m −2 year −1 of electricity demand ( Table 4). The present PV blind would partially compensate such greater demands.
As estimated in Figure 14, this system would produce a surplus of electricity in high-insolation greenhouses, which demand several kilowatt hours per square meter per year of electricity. In addition, Figure 6b suggests that external installation would greatly improve electricity production. For these reasons, externally mounted PV blinds with less PV cell density might compensate electricity demands of conventional greenhouses in high-insolation regions. The decreased cell density is expected to increase agronomic sustainability for greenhouse crop cultivation. Waterproofing and mechanical reliability against severe weather conditions should be included with these systems to be installed externally. However, in very high-irradiation regions, a higher PV cell density might be acceptable. This might result in higher electricity production for greater load operations with moderate shading inside the greenhouse.
To apply the present PV blind systems to actual scale greenhouses, decreasing the PV module production costs is absolutely necessary by manufacturing automation. The PV module production cost, because of its current hand-made manufacturing process, is extremely expensive, such that it cannot be compensated by the value of the PV producible electricity. The other components of the blind systems, including the control circuit parts, are readily available from the market with reasonable prices.    Greece low-level energy consumption greenhouse 25 [33] China cooling and ventilation 30 [75] Saudi Arabia fans, cooling pump, PC 56 [6] * Eight-month data.
The shading percentage of the PV blind was 40% at θ = 0 • inclination. Shading affects the crop growth, the yield, and the choice of the greenhouse species that can be cultivated underneath the PV panels. Some reports of earlier studies have described crop growth under certain levels of PV shading. Wild rocket crops were cultivated properly in an Italian greenhouse under a PV panel installation that provided 32% roof coverage [35]. Under 50% coverage of a greenhouse roof by PV cells arranged in a checkerboard formation, lettuce was cultivated during summer and autumn in Japan [31]. Specific translucent film was placed underneath the solar cells in their test greenhouse. The film assisted scattering of light impinged across PV cell intervals to wider area of crops below [31]. More recently, the qualities of berries cultivated under the PV modules, which covered 32% of the greenhouse roof, were investigated in Italy [76]. Reportedly, the concentrations of total anthocyanins and total phenols in raspberry and blackberry fruits grown under the PV shading were greater than those of the control berries grown in the conventional greenhouse. These results of earlier studies suggest that controlled shading by the semi-transparent PV blind might be beneficial for the cultivation of some crop species.
The concept of shading percentage of the see-through PV modules used in the present study differed from that of opaque planar PV cells installed in other PV greenhouses. The installation of conventional planar PV cells on a greenhouse roof casts a distinct shadow area on the ground in the greenhouse. The shadow area is dark without direct sunlight, the remaining area is bright. As the earth rotates, the shadow moves in the greenhouse. Accordingly, plants at a fixed position in the greenhouse are irradiated with fluctuating intensities of sunlight [20]. By contrast, see-through semi-transparency of 40% shading does not cast explicit shadow patterns in the greenhouse. Instead, part of the apparent sun area seen from plants is eclipsed by the micro PV cells [26,27,55]. The remaining part of direct sunlight beam can irradiate plants. In addition, the dynamic shading control of the present PV blind enables moderate shading only during high-insolation hours. Yield and quality of some crop species might be improved if the excessive sunlight were partly shaded by the semi-transparent PV blinds, or if the blind rotates when irradiance is less than the threshold level, letting more sunlight irradiate the plants. Figure 15 depicts the calculated solar energy inside the PV blind greenhouse with the north-south ridge orientation assuming that the entire roof was covered with the PV blinds and that the sunlight transmittance of the PV module is 0% at θ = 90 • and 60% at θ = 0 • . Eight-percent of the sky would be covered with the PV modules of 0% transmittance at θ = 90 • according to the geometry calculation presented in Figure 15a. Cloudless sky with atmospheric transmittance of 0.65 was assumed. Solar energy was shown as parameters of threshold I H levels for the PV blind turns. For that calculation, the total sunlight transmittance of overlapping greenhouse glazing and the PV blind was assumed to be 51% (Figure 8c) when θ = 0 • . It was assumed to be 78% (=0.85 of glazing transmittance ×0.92 of unshaded zone as depicted in Figure 15a ×100%) when θ = 90 • . As the threshold I H value for the blind operation increases, solar energy in the greenhouse increases. As presented in Table 5, the annual solar energy in the greenhouse would be 59% of that outside if the threshold I H was set at 500 W m −2 . It would increase to 73% if the threshold I H was set at 900 W m -2 . By contrast, electricity generation would decrease as the threshold I H increases. Compared with the annual solar energy received in the greenhouse without the PV blind installation, 70%, 77%, and 86% of solar energy would be received in the PV blind greenhouse with the set threshold I H s of 500, 700, and 900 W m −2 , respectively. Figure 14 shows an estimation of 7.8 kWh m −2 year −1 of electrical energy production for η 0 = 0.6%, although the greenhouse crops receive 1170 kWh m −2 year −1 if threshold I H is set at 500 W m −2 (Figure 15b). Incorporating the actual sky conditions (2009-2017, Japan Meteorological Agency) into calculations, 1170 kWh m −2 year −1 is declined to 765 kWh m −2 year −1 . Assuming that 1% of the solar energy can be converted into plant biomass [77,78], the solar-energy use ratio for plant biomass production to electricity production would be approximately 1:1 under η 0 = 0.6% estimation. To validate the estimation, plant growth under the PV blinds should be investigated in future experiments. of the present PV blind enables moderate shading only during high-insolation hours. Yield and quality of some crop species might be improved if the excessive sunlight were partly shaded by the semi-transparent PV blinds, or if the blind rotates when irradiance is less than the threshold level, letting more sunlight irradiate the plants. Figure 15 depicts the calculated solar energy inside the PV blind greenhouse with the northsouth ridge orientation assuming that the entire roof was covered with the PV blinds and that the sunlight transmittance of the PV module is 0% at θ = 90° and 60% at θ = 0°. Eight-percent of the sky would be covered with the PV modules of 0% transmittance at θ = 90° according to the geometry calculation presented in Figure 15a. Cloudless sky with atmospheric transmittance of 0.65 was assumed. Solar energy was shown as parameters of threshold IH levels for the PV blind turns. For that calculation, the total sunlight transmittance of overlapping greenhouse glazing and the PV blind was assumed to be 51% (Figure 8c) when θ = 0°. It was assumed to be 78% (=0.85 of glazing transmittance ×0.92 of unshaded zone as depicted in Figure 15a ×100%) when θ = 90°. As the threshold IH value for the blind operation increases, solar energy in the greenhouse increases. As presented in Table 5, the annual solar energy in the greenhouse would be 59% of that outside if the threshold IH was set at 500 W m −2 . It would increase to 73% if the threshold IH was set at 900 W m -2 . By contrast, electricity generation would decrease as the threshold IH increases. Compared with the annual solar energy received in the greenhouse without the PV blind installation, 70%, 77%, and 86% of solar energy would be received in the PV blind greenhouse with the set threshold IHs of 500, 700, and 900 W m −2 , respectively. Figure 14 shows an estimation of 7.8 kWh m −2 year −1 of electrical energy production for η0 = 0.6%, although the greenhouse crops receive 1170 kWh m −2 year −1 if threshold IH is set at 500 W m −2 (Figure 15b). Incorporating the actual sky conditions (2009-2017, Japan Meteorological Agency) into calculations, 1170 kWh m −2 year −1 is declined to 765 kWh m −2 year −1 . Assuming that 1% of the solar energy can be converted into plant biomass [77,78], the solar-energy use ratio for plant biomass production to electricity production would be approximately 1:1 under η0 = 0.6% estimation. To validate the estimation, plant growth under the PV blinds should be investigated in future experiments.
(a) (b) Figure 15. Estimated cumulative solar energy in the north-south oriented greenhouse equipped with a PV blind entirely covering the roof: (a) the shading zones by the PV blinds were painted in gray assuming that the sunlight transmittance of the PV module at θ = 90° is 0; (b) the thresholds IHs were chosen at the three levels of 500, 700, and 900 W m −2 . The transmittance of the greenhouse glazing and the blind at parallel and perpendicular positions were assumed as 85%, 60%, and 92% (8% is PV shadow as depicted in (a)), respectively. Cloudless sky with atmospheric transmittance of 0.65 was assumed. For comparison, solar energy outside the greenhouse and inside the greenhouse without PV blinds (greenhouse frames were omitted) are also presented. Figure 15. Estimated cumulative solar energy in the north-south oriented greenhouse equipped with a PV blind entirely covering the roof: (a) the shading zones by the PV blinds were painted in gray assuming that the sunlight transmittance of the PV module at θ = 90 • is 0; (b) the thresholds I H s were chosen at the three levels of 500, 700, and 900 W m −2 . The transmittance of the greenhouse glazing and the blind at parallel and perpendicular positions were assumed as 85%, 60%, and 92% (8% is PV shadow as depicted in (a)), respectively. Cloudless sky with atmospheric transmittance of 0.65 was assumed. For comparison, solar energy outside the greenhouse and inside the greenhouse without PV blinds (greenhouse frames were omitted) are also presented.

Conclusions
A semi-transparent PV blind system was prototyped based on micro-spherical solar cell technology for application to greenhouse shading control. The PV blind inclination had been altered autonomously according to the set points of the irradiance threshold values that were adjustable by a grower by modulating the variable resistor value in the control circuit. The cumulative energy balance of the PV blind system became positive after the five-month operation test during winter to spring on the west coast of Japan. This result suggests that the surplus electrical energy is useful for greenhouse environment management to achieve better crop yield and quality with the mitigation of fuel and grid electricity consumption, particularly in high-insolation regions. Assuming that 1% of the solar energy can be converted into plant biomass, the solar-energy use ratio for plant biomass production to electricity production would be approximately 1:1 using the present PV blind system at the greenhouse under the actual sky condition. That ratio is adjustable by the set-point of the threshold irradiance for the blind rotations. This performance might provide a new strategic cultivation opportunity to growers. Although this system is still a prototype, the obtained experimental data clearly indicate that the concept of the semi-transparent PV blind system for application to greenhouses is sufficiently valuable to be tested further for its feasibility, including effects on crop yield and quality.