Since the introduction of the photovoltaic policy in 2018, the domestic solar industry has transitioned from initial government subsidies to a market-driven phase of grid parity, fostering competition-driven qualitative improvements in photovoltaic technology. Bifacial solar panels, which enhance power generation and efficiency by absorbing additional light radiation from the rear side, have emerged as a key technology supporting grid parity for photovoltaic modules. According to statistics from the China Photovoltaic Association and the International Technology Roadmap for Photovoltaic (ITRPV), the demand for bifacial photovoltaic modules in downstream applications is steadily increasing. Their market share in the domestic photovoltaic sector was 14% in 2019 and is projected to exceed 50% by 2025 and 70% by 2030. Research indicates that reflectors made of specific materials placed beneath bifacial solar panels can increase their power output. Furthermore, power generation can be optimized through solar tracking to maximize the effective light-receiving area, with tilted single-axis tracking being a common method for bifacial photovoltaic panels.
To further enhance the photoelectric conversion efficiency of bifacial solar panels, we designed a novel photosensitive sensor device that integrates reflectors with front and rear sensor cartridges. Utilizing low-frequency wake-up and wireless communication technologies, we achieved dual-precision dual-axis sun tracking for bifacial photovoltaic panels. By studying the circuit model of bifacial solar panels and improving the particle swarm optimization algorithm, we accomplished rapid maximum power point tracking under sudden shadow conditions. This article details the structural design, circuit implementation, software algorithms, and experimental validation of the system.
The tracking device for solar panels consists of two layers: an upper layer for mounting two bifacial photovoltaic panels and a lower layer connected to tracking motors. To ensure effective reflection, the four sides of the支架 are unobstructed and equipped with reflectors. The bottom edge of each reflector is connected to the支架 base, extending upward and outward at a predetermined tilt angle. Front and rear sensor cartridges are positioned between the two photovoltaic panels, both on the front and back sides. The reflectors comprise four symmetrically arranged mirrors on the front, back, left, and right sides. Their functions include: (1) reflecting sunlight from the front to the rear of the photovoltaic panel to increase irradiance, and (2) creating an angular difference between the reflected beams from the left-right and front-rear sides, which are captured by the rear sensor cartridge. To ensure that reflected light fully irradiates the rear photovoltaic panel under vertical incidence, the angles and dimensions of the reflectors must satisfy specific geometric relationships.
The key parameters for the reflector design are defined as follows: let \( h \) be the vertical height from the reflector base to the bifacial photovoltaic panel, \( l \) the length of the photovoltaic支架, \( w \) the width of the photovoltaic支架, \( L_1 \) the length of the left and right reflectors (equal to the支架 width), \( L_2 \) the length of the front and rear reflectors, and \( K_1 \) and \( K_2 \) the widths of the left-right and front-rear reflector groups, respectively. The angles \( \theta_1 \) and \( \theta_2 \) are critical for reflection efficiency and are derived from the following equations:
$$ \tan \theta_1 = \frac{w}{2h} $$
$$ \tan \theta_2 = \frac{l}{2h} $$
These equations ensure optimal light reflection onto the rear surface of the solar panels, enhancing the overall performance of the photovoltaic system. The dimensions \( L_1 \) and \( L_2 \) are calculated based on these angles to cover the required reflective area.
Light signal acquisition is implemented through coarse adjustment via the rear sensor cartridge and fine adjustment via the front sensor cartridge. The front cartridge is a square box with a square skylight, containing five photoresistors. When light is perpendicular, only the central photoresistor is exposed, used for overcast judgment. During tilted illumination, the front or rear, left or right photoresistors are exposed, generating differential signals that are amplified and input to the microcontroller to control the elevation and azimuth tracking motors for fine adjustment. The rear cartridge is a hollow rectangular structure with strip-shaped skylights on all four sides, each side equipped with five photoresistors spaced along the skylight direction. The hollow section houses the photoresistor wiring connected to the control circuit.
For the left-side photoresistors in the rear cartridge, the coarse adjustment angle \( \beta \) is calculated based on the critical points where reflected light barely reaches each photoresistor. For instance, when vertical incident light is reflected to the topmost photoresistor, the distance from the top \( H = 0 \), and both left and right first photoresistors are illuminated, with no rotation of the azimuth motor. As the incident angle increases, the corresponding height \( H \) and angle \( \beta \) vary, as summarized in the table below.
| Photoresistor Number (n) | Height (H) | Angle (β) |
|---|---|---|
| 1 | 0 | 0° |
| 2 | \( h – \frac{w}{4} \) | \( 90^\circ – 4\theta_1 \) |
| 3 | \( h – \frac{w \tan \theta_1}{2} \) | \( 90^\circ – 3\theta_1 \) |
| 4 | \( h \) | \( 90^\circ – 2\theta_1 \) |
| 5 | \( h + \frac{w \tan \theta_1}{2} \) | \( 90^\circ – \theta_1 \) |
This table illustrates the relationship between the photoresistor position and the required adjustment angles, enabling precise control of the solar tracking mechanism. The use of these sensors ensures that the photovoltaic panels maintain optimal orientation towards the sun, maximizing energy capture throughout the day.
The circuit system is divided into two parts: a reader and photovoltaic tracking device nodes (tags), with the STM32F103VBT6 microcontroller serving as the main controller. The reader functions for both low-frequency wake-up and data feedback. When inactive, the tracking device remains in sleep mode. Upon receiving a low-frequency wake-up signal from the reader, the device activates and controls the motors for dual-axis tracking based on photosensitive information. Simultaneously, the rear end of the photovoltaic panel connects to an adaptive charging control circuit to charge the battery. During tracking, the orientation angles and output power of the bifacial solar panels are transmitted wirelessly to the reader for display, allowing operators to compare tracking accuracy and power generation.
In the low-frequency transceiver circuit, when the reader activates the tracking button, the STM32 generates a PWM wave for On-Off Keying (OOK) modulation of the power monitor MAX9930. This signal is then level-shifted and power-amplified by the MOS driver MIC4424 and voltage converter IRF7839, producing a 125 kHz low-frequency signal sent to the low-frequency receiver part of the main circuit. The low-frequency receiver circuit AS3933 wakes the main circuit’s STM32 via the SPI interface, transitioning the solar tracking device from sleep to active state for dual-axis tracking.
During operation, the tracking device collects the elevation and azimuth angles of the photovoltaic panels at different times using angle sensors, along with output current and voltage from the current and voltage acquisition modules. These data are stored in the STM32 registers and can be read by operators via the wireless communication chip CC1101. The CC1101, integrated with external crystals, filters, and matching circuits, connects to the STM32’s SPI interface. The reader similarly receives this information through a CC1101 and peripheral circuits, displaying it on an LCD screen.
To mitigate overvoltage or overcurrent effects on battery lifespan during charging of bifacial solar panels and reduce self-discharge losses, an adaptive charging control chip UC3906 and its peripheral circuits are employed for intelligent battery charging. The chip’s internal circuit manages three charging states: float, bulk, and overcharge. A resistor divider network detects the battery voltage, comparing it with a reference voltage to determine the charging state. With temperature-adaptive adjustment, the chip flexibly adjusts the reference voltage based on real-time temperature, maximizing battery capacity and extending service life.
The software design encompasses low-frequency wake-up procedures, motor tracking control, and maximum power point tracking (MPPT) algorithms. The wireless communication part uses SmartRFStudio software to configure the CC1101, with data read via register windows. The low-frequency wake-up program involves four steps: initialization, sleep setting, low-frequency wake-up, and wireless transmission. After microcontroller initialization, it enters sleep mode with interrupts enabled. Upon receiving wake-up data encoded in Manchester code by the AS3933, the STM32 is awakened, parses the data, and controls the motors for dual-axis tracking.
Motor control for coarse and fine adjustments is based on photosensitive signal data from the rear and front cartridges input to the STM32’s I/O ports, generating corresponding level signals. Motors B1 and B2 use worm gear reducers with encoders. The STM32’s timer, capable of reading encoder pulse counts, provides feedback on rotation angles. Using dual-edge counting, with two encoder channels and one pulse counting as four, the motor reduction ratio is 322. The count value \( CT \) for a given rotation angle \( \beta \) is calculated as:
$$ CT = \frac{\beta \times 4 \times 9 \times 322}{360} = 32.2\beta $$
For fine adjustment, the step size is set to 1°, corresponding to a \( CT \) of 32.2. For coarse adjustment, taking the left-side photosensitive signals of the rear cartridge as an example, the input control signals from PA0 to PA4 are written to the register Left_Num. With \( \theta_1 = 12^\circ \), the control logic is summarized in the following table.
| PA4~PA0 Value | Left_Num Range | B1 Rotation Angle (β) | Count Value (CT) |
|---|---|---|---|
| 00001 | 1 | No rotation | 0 |
| 00010 | >1 && ≤3 | Left rotation \( 90^\circ – 4\theta_1 \) | 1352 |
| 00011 | >3 && ≤7 | Left rotation \( 90^\circ – 3\theta_1 \) | 1739 |
| 00100 | >7 && ≤15 | Left rotation \( 90^\circ – 2\theta_1 \) | 2125 |
| 00101 | >16 && ≤31 | Left rotation \( 90^\circ – \theta_1 \) | 2512 |
The overall motor control program includes judgments for overcast conditions and logic for coarse and fine adjustments of elevation and azimuth angles, ensuring the solar panels accurately follow the sun’s path.
For MPPT, the circuit model of bifacial photovoltaic modules is derived from the “five-parameter model” of single-sided modules. Using the TUN NORD test method, rear irradiance is compensated to the front, resulting in effective irradiance \( G_{\text{eff}} \):
$$ G_{\text{eff}} = G_{\text{front}} + B \times G_{\text{rear}} \times R $$
where the bifacial coefficient \( B = \frac{I_{\text{sc,rear}} – I_{\text{sc,front}}}{I_{\text{sc,front}}} \), with \( I_{\text{sc,rear}} \) and \( I_{\text{sc,front}} \) being the short-circuit currents of the rear and front sides, respectively. \( R \) is the reflection coefficient, which can reach up to 1.8 under vertical illumination due to the reflectors and overlapping reflection areas.
The particle swarm optimization algorithm is employed for MPPT under multi-peak conditions. Assume a randomly generated population \( X = (X_1, X_2, \ldots, X_n) \) in a d-dimensional search space, where \( n \) is the total number of particles. For the i-th particle, its position is:
$$ X_i = (x_{i1}, x_{i2}, \ldots, x_{id})^T $$
and its velocity is:
$$ V_i = (v_{i1}, v_{i2}, \ldots, v_{id})^T $$
The velocity and position update equations are:
$$ v_{id}^{k+1} = \omega v_{id}^k + c_1 r_1 (P_{\text{best},i} – x_{id}^k) + c_2 r_2 (G_{\text{best}} – x_{id}^k) $$
$$ x_{id}^{k+1} = x_{id}^k + v_{id}^{k+1} $$
where \( P_{\text{best},i} \) is the personal best solution of the i-th particle, \( G_{\text{best}} \) is the global best solution, \( k \) is the current iteration number, \( \omega \) is the inertia weight, \( c_1 \) and \( c_2 \) are learning factors, and \( r_1 \) and \( r_2 \) are independent random numbers in (0,1).
To avoid converging to local optima in multi-peak scenarios, \( \omega \), \( c_1 \), and \( c_2 \) are improved as follows:
$$ \omega = \omega_{\text{max}} – \frac{(\omega_{\text{max}} – \omega_{\text{min}}) \times k}{M} $$
$$ c_1 = c_{1,\text{start}} + (c_{1,\text{end}} – c_{1,\text{start}}) \times \left[ 1 – \cos\left(\frac{\pi k}{M}\right) \right] / 2 $$
$$ c_2 = c_{2,\text{start}} + (c_{2,\text{end}} – c_{2,\text{start}}) \times \left[ 1 – \cos\left(\frac{\pi k}{M}\right) \right] / 2 $$
where \( M \) is the maximum number of iterations, with \( \omega_{\text{max}} = 0.9 \), \( \omega_{\text{min}} = 0.4 \), \( c_{1,\text{start}} = 1.2 \), \( c_{1,\text{end}} = 2.7 \), \( c_{2,\text{start}} = 2.2 \), and \( c_{2,\text{end}} = 0.5 \). These modifications enhance the algorithm’s ability to escape local optima and converge to the global maximum power point efficiently.
Experimental results were obtained to evaluate the tracking accuracy and power generation efficiency of the reflective dual-axis tracking device. Based on the horizontal coordinate system and Cooper’s algorithm, theoretical values of the sun’s elevation and azimuth angles were calculated. Data were collected at 38 time points from 6:20 to 17:40 over 50 days in a specific location (latitude 31.43°N, longitude 119.54°E) during March, May, July, September, and November 2023. The elevation and azimuth angles displayed by the reader were compared with theoretical values for error analysis. The average tracking error for elevation angle was within 1.47°, and for azimuth angle within 1.59° over the 50 days.
To compare the power generation of the reflective dual-axis tracking device with traditional tilted single-axis tracking, two bifacial photovoltaic panels with a front maximum output power of 300 W and rear maximum output power of 226 W were used under the same location and time conditions for 50 days. Due to the enhanced rear irradiance from reflectors and improved tracking accuracy, the new device increased average power generation by 67.6% compared to the traditional system, meeting design expectations.
Additionally, a \( 2 \times 4 \) bifacial photovoltaic array model was built on the Matlab/Simulink platform based on a single bifacial photovoltaic panel circuit to simulate MPPT tracking under sudden shadow conditions using the improved PSO algorithm. The output power of the photovoltaic array is shown in the following analysis: when shadow conditions changed at 0.2 s, the algorithm escaped local optima and tracked the maximum power point at 0.75 s, with a relative power tracking error of less than 1%. Simulation experiments demonstrated that the improved PSO algorithm offers strong anti-interference capability and tracking efficiency, thereby enhancing the power generation efficiency of bifacial photovoltaic arrays.
In summary, we designed a novel reflective sensor device for bifacial solar panels, incorporating low-frequency wake-up, wireless communication, and intelligent charging technologies. This dual-precision dual-axis tracking system improves photoelectric conversion efficiency by maximizing light absorption and power output. Experimental comparisons confirmed tracking errors within 1.59° and increased power generation compared to traditional tilted single-axis tracking. The improved PSO algorithm, tested on a Matlab-simulated \( 2 \times 4 \) photovoltaic array circuit, effectively addressed local optima issues under multi-peak or sudden shadow conditions, achieving fast and accurate MPPT tracking. From the design of the new photovoltaic tracking sensor group to the development of supporting circuits and software systems, this work provides a solution for enhancing the efficiency of bifacial photovoltaic panels, contributing to advancements in solar energy technology.