Solar photovoltaic automatic tracking method, tracking system and storage medium

Abstract

The application provides a solar photovoltaic automatic tracking method, a tracking system and a storage medium, and the tracking method comprises the following steps: predicting the position of the sun at a future time T; collecting the light intensity, photovoltaic output power, photovoltaic panel temperature and wind speed in real time; calculating the light uniformity according to the real-time collected light intensity; generating a light matrix by using a bilinear interpolation according to the real-time collected light; performing convolution calculation on the light matrix to obtain the light gravity center; calculating the prediction weight and the feedback weight according to the light uniformity; and obtaining the final angle signal of the horizontal shaft and the final angle signal of the pitch shaft according to the horizontal shaft prediction angle, the pitch shaft prediction angle, the horizontal shaft feedback correction amount, the pitch shaft feedback correction amount, the prediction weight and the feedback weight. Through the double-shaft dynamic prediction-feedback fusion algorithm, the improved sun trajectory prediction model and the multi-parameter feedback correction result are fused, and the high-precision sensor and the actuator are combined, so that the tracking precision is better than 0.1 DEG, which is significantly higher than that of the traditional system.

Description

Technical Field

[0001] This invention relates to the field of solar photovoltaic power generation technology, specifically a solar photovoltaic automatic tracking system. Background Technology

[0002] Traditional photovoltaic (PV) tracking systems primarily rely on a single sensor or purely astronomical trajectory algorithms to track the sun's position. This simplistic design leads to insufficient adaptability and stability in complex environments. As patent CN121143475A points out, when traditional systems use a single photoelectric sensor for sun position detection, limitations imposed by sensor performance and environmental conditions mean that even in open areas, uneven illumination of the PV panels can occur due to cloud cover, dust, and mutual shading, making it difficult to accurately determine the sun's position and compromising tracking accuracy and stability. Experimental data shows that the average deviation of the solar tracking angle in traditional systems under cloudy weather conditions can reach ±5°, significantly exceeding the requirements for high-precision applications. This directly results in a power generation fluctuation range of ±20W, severely impacting the stability of power generation efficiency.

[0003] Meanwhile, traditional photovoltaic tracking systems generally lack flexible environmental adaptability mechanisms. When faced with sudden environmental changes such as cloudy skies and sudden wind speed changes, they exhibit significant response lag and accumulated angle adjustment errors. In partially shaded scenarios, traditional systems cannot effectively predict the shadow's movement path, often resorting to passive adjustments, leading to a significant drop in photovoltaic panel power generation. In experiments, power generation during shaded periods can drop to around 80W, a significant decrease compared to normal operating conditions. Under strong wind conditions, traditional systems often employ a fixed angle locking strategy, lacking targeted wind-resistant attitude optimization. The average drag coefficient is as high as 1.2, and the maximum stress value of the support structure can reach 150MPa, which not only affects system stability but also shortens the equipment's lifespan.

[0004] Furthermore, traditional system control strategies often focus solely on maximizing power generation, neglecting the balancing control of mechanical losses. Over long-term operation, this can lead to accelerated equipment wear and tear, further increasing maintenance costs. These issues collectively limit the effectiveness of photovoltaic tracking systems in complex climatic environments. Therefore, developing a photovoltaic tracking technology that combines high precision, strong adaptability, and long-term reliability has become a crucial direction. Summary of the Invention

[0005] The technical problem to be solved by this invention is to propose an automatic tracking method and system for solar photovoltaic panels, so as to achieve high-precision and highly adaptable automatic tracking of photovoltaic panels.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: This invention first provides a method for automatic solar photovoltaic tracking, comprising the following steps: Step S1: Predict the position of the sun at a future time T; Step S2: Real-time acquisition of light intensity, photovoltaic output power, photovoltaic panel temperature, and wind speed; calculation of light uniformity based on the real-time acquired light intensity. Step S3: Based on the real-time collected illumination, generate an illumination matrix using bilinear interpolation; perform convolution calculation on the illumination matrix to obtain the illumination centroid; Step S4: Calculate the prediction weight and feedback weight based on the illumination uniformity; obtain the final horizontal axis angle signal and the final pitch axis angle signal based on the horizontal axis prediction angle, pitch axis prediction angle, horizontal axis feedback correction amount, pitch axis feedback correction amount, prediction weight, and feedback weight.

[0007] The present invention also provides a solar photovoltaic automatic tracking system for performing the above-described photovoltaic automatic tracking method, the photovoltaic automatic tracking system comprising: Data acquisition module: Used to collect real-time data through gridded light sensors, wind speed, temperature, voltage and current sensors, transmit the data via CAN bus and complete environmental calibration to build a high-precision input dataset; Trajectory prediction module: It adopts an improved declination angle and hour angle model, combined with seasonal correction factors, regional offset and cloud layer influence correction, to achieve dynamic prediction of the sun's position; Feedback correction module: Generates local occlusion adaptive feedback correction amount through power gradient calculation, bilinear interpolation illumination matrix, and illumination centroid offset compensation; Fusion decision module: dynamically adjusts the prediction-feedback weights based on illumination uniformity and environmental parameters, and outputs the optimal dual-axis control command using fuzzy logic; The drive module employs a dual-axis brushless DC motor and a high-precision encoder, combined with PID closed-loop control to achieve precise angle execution and error compensation. Power supply module: Composed of photovoltaic panels, DC-DC converter and lithium iron phosphate battery, to achieve system energy self-sufficiency and uninterrupted power supply; Anomaly protection module: It has functions such as sensor fault diagnosis and mode switching, automatic wind resistance and retraction in high wind speed, and attitude error self-calibration, which improves system reliability.

[0008] Furthermore, the data acquisition module adopts a grid arrangement of 4 light sensors, supports temperature-wind speed coupled data calibration, has a sampling frequency of 1Hz, and transmits data synchronously at a high speed of 1Mbps via a CAN bus.

[0009] Furthermore, the trajectory prediction module adopts a cubic curve dynamic model, which is trained based on historical solar position and cloud data, and the model coefficients are updated continuously to predict the solar position in the next 10 minutes.

[0010] Furthermore, the feedback correction module divides the photovoltaic panel surface into a 10×10 grid, generates an illumination matrix through bilinear interpolation, and calculates the illumination centroid to achieve adaptive compensation for shading.

[0011] Furthermore, the fusion decision module dynamically allocates weights based on the uniformity of illumination U. When U>0.8, prediction is the primary method, and when U<0.3, feedback is the primary method, thus achieving all-weather adaptive tracking.

[0012] Furthermore, the execution drive module has a horizontal axis rotation range of 0-360° and a pitch axis range of 0-90°, and is equipped with a 16-bit absolute encoder, with a positioning accuracy better than 0.05°.

[0013] Furthermore, the anomaly protection module activates protection when the wind speed exceeds the threshold, placing the photovoltaic panel in a safe posture to reduce wind load; and automatically switches to pure astronomical tracking mode when the sensor malfunctions.

[0014] Furthermore, the power supply module adopts a synchronous rectification DC-DC converter with a conversion efficiency of more than 95%, and is equipped with a lithium iron phosphate battery to achieve stable power supply on cloudy days and at night.

[0015] Beneficial effects Compared with existing technologies, this technology has the following advantages: (1) High tracking accuracy: By using a dual-axis dynamic prediction-feedback fusion algorithm, the improved solar trajectory prediction model and multi-parameter feedback correction results are fused together. Combined with high-precision sensors and actuators, the tracking accuracy is better than 0.1°, which is significantly higher than that of traditional systems.

[0016] (2) Strong adaptability: It can cope with complex weather conditions and partial shading scenarios, and dynamically adjust the weights through illumination uniformity to quickly respond to environmental changes and avoid getting trapped in local optima.

[0017] (3) High reliability: It adopts a modular design and has a complete abnormal handling mechanism, such as sensor fault switching and wind speed protection.

[0018] (4) Energy self-sufficiency: Equipped with a lithium battery energy storage unit and combined with a high-efficiency DC-DC converter, the system achieves energy self-sufficiency without the need for additional power supply and has a wide range of applications. Attached Figure Description

[0019] Figure 1 This is a structural block diagram of the solar photovoltaic automatic tracking system of the present invention; Figure 2 Here is the algorithm flowchart for an automatic solar photovoltaic tracking system; In the diagram: 1 is a light sensor, 2 is a photovoltaic panel, 3 is a DC-DC converter, 4 is a lithium iron phosphate battery, 5 is a temperature sensor, 6 is a voltage / current sensor, 7 is a wind speed sensor, 8 is a CAN bus, 9 is a processor, 10 is a pitch axis motor, 11 is a pitch axis motor encoder, 12 is a horizontal axis motor, 13 is a horizontal axis motor encoder, and 14 is a dual-axis mechanism. Detailed Implementation

[0020] Example 1 This embodiment provides a solar photovoltaic automatic tracking system based on a dual-axis dynamic prediction-feedback fusion algorithm, employing a modular design. For system composition and connection methods, please refer to [link / reference]. Figure 1 It includes a data acquisition module, an algorithm processing module, an execution driver module, and a power supply module. These modules work together to achieve real-time tracking and attitude adjustment of the sun's position.

[0021] Data acquisition module A four-channel high-precision light intensity sensor 1 is used, positioned at the four corners of the photovoltaic panel 2, to collect real-time light intensity data at each location, with a sampling frequency of 1Hz. The light intensity sensor uses photodiodes as the sensing element, with a response time of less than 10 ms and a measurement range of 0-2000 W / m². 2 The measurement accuracy is ±5%.

[0022] It integrates one wind speed sensor (7), one voltage / current sensor (6), and one temperature sensor (5) to comprehensively monitor the environment and system status. The wind speed sensor uses ultrasonic principles, with a measurement range of 0-60 m / s and an accuracy of ±0.1 m / s; the voltage / current sensor uses the Hall effect, with measurement ranges of 0-1000V and 0-100A respectively, and an accuracy of ±0.5%; the temperature sensor uses a thermocouple, with a measurement range of -40-125℃ and an accuracy of ±0.5℃.

[0023] Data from each sensor is transmitted to the processor 9 via the CAN bus 8 at a rate of 1 Mbps, ensuring data synchronization and real-time performance. The CAN bus uses differential signal transmission, providing strong anti-interference capabilities and supporting multi-node communication, with a maximum of 110 nodes.

[0024] Sensor data calibration formula:

[0025] in, For the calibrated light intensity, Original light intensity Ambient temperature; This refers to wind speed.

[0026] This formula takes into account the effects of temperature and wind speed on the measurement accuracy of the light sensor, thus improving data reliability.

[0027]

[0028] in, , , , The light intensity is collected by the light sensors located at the four corners of the photovoltaic panel.

[0029] The system collects a set of multiple parameters in real time:

[0030] in, The output voltage of the photovoltaic panel. To output current to the photovoltaic panel, For ambient temperature, For ambient wind speed.

[0031] Algorithm processing module Based on a high-performance microprocessor, it runs a dual-axis dynamic prediction-feedback fusion algorithm with a computation cycle of ≤100 ms, and has strong floating-point computing power and real-time response capability.

[0032] Processor performance evaluation: Power consumption formula:

[0033] in, P For processor power consumption, f Main frequency, C For switched capacitors, V This refers to the power supply voltage. Formula for floating-point arithmetic capability:

[0034] in, FLOPS The number of floating-point operations per second. N Number of floating-point operations per cycle.

[0035] This formula is used to evaluate processor performance and power consumption, providing a reference for system design.

[0036] The algorithm processing module includes five sub-units: a solar trajectory dynamic prediction unit, a real-time data preprocessing unit, a power gradient and illumination centroid calculation unit, a dual-axis angle decision unit, and an anomaly and protection processing unit. These units work together to achieve high-precision, adaptive solar tracking control.

[0037] The solar trajectory dynamic prediction unit is used to dynamically predict the sun's position; the real-time data preprocessing unit is responsible for calibrating, filtering, and validating the raw data collected by the sensors to construct a high-precision input dataset; the power gradient and illumination centroid calculation unit is responsible for calculating the power gradient and illumination centroid offset, generating dual-axis angle feedback correction values ​​to solve the local optimum problem in partially occluded scenarios; the dual-axis angle decision unit dynamically adjusts the prediction weight and feedback weight based on illumination uniformity and environmental parameters, merges the prediction and feedback results, and outputs the final dual-axis control command; the anomaly and protection handling unit is responsible for real-time monitoring of the system status and triggering corresponding protection and fault handling mechanisms.

[0038] The solar trajectory dynamic prediction unit performs the following processing: By dynamically predicting the sun's position using a solar trajectory prediction unit, the sun's azimuth can be predicted in advance for the next 10 minutes. This invention proposes an improved dynamic declination and hour angle calculation model, introducing a seasonal correction factor. and regional offset Used to calculate the solar declination angle and hour angle: The seasonal correction factor is:

[0039] The regional offset is:

[0040] in, n For years, t It is true solar time.

[0041] Based on the above parameters, the solar declination angle can be calculated. Sum of time angles Then, combining the local latitude and longitude, the solar altitude angle and azimuth angle are calculated. This invention further proposes a cubic curve dynamic prediction model, trained based on five years of historical solar position and cloud movement data, dynamically updating coefficients to achieve dynamic prediction of the solar declination and hour angle for the next 10 minutes. The model uses a rolling window method, updating model parameters every 5 minutes to improve prediction accuracy.

[0042]

[0043]

[0044] in, To predict the declination angle, To predict the hour angle, These are time-varying coefficients; The time-varying coefficients are updated in real time using the least squares method for fitting:

[0045] in, These are historical declination angle sample values; These are sample values ​​of the hour angle. For the corresponding time period; N is the number of sample points involved in the fitting. Solar altitude angle and solar azimuth Calculate using the following formulas respectively:

[0046] in, The local geographical latitude; The effects of cloud cover are corrected for by the rate of change in illumination and cloud thickness at adjacent time points:

[0047] in, This is the correction amount for the angle prediction caused by cloud cover. k 1 The cloud layer influence coefficient. The rate of change of light intensity The speed of cloud movement. This refers to the thickness of the cloud layer.

[0048] The cloud movement speed and cloud thickness are both calculated from image sensors or meteorological data.

[0049] The cloud layer correction is superimposed on the predicted angle to obtain the corrected horizontal axis predicted angle. and pitch axis prediction angle :

[0050]

[0051] Real-time data preprocessing unit: Real-time acquisition of data such as light intensity, photovoltaic panel output power, photovoltaic panel temperature, and wind speed. The light intensity is collected in real time by four sensors located at the four corners of the photovoltaic panel; the output power of the photovoltaic panel is measured by the voltage / current sensor outputting voltage U. pv With current I pv Calculate P=U pv ·I pv The temperature of the photovoltaic panel is collected by a temperature sensor attached to the back of the photovoltaic panel.

[0052] The power gradient and illumination centroid calculation unit performs the following processing: Power gradient calculation Based on the real-time output power of the photovoltaic panel, the partial derivatives of the power with respect to the angles of the horizontal and pitch axes, i.e. the power gradient, are calculated using an improved forward difference method, providing a directional basis for feedback correction.

[0053]

[0054]

[0055] Where P is the output power of the photovoltaic panel. This represents the real-time angle of the horizontal axis. This represents the real-time angle of the pitch axis. For the angle, it is a small variable; For fine-tuning the horizontal axis The actual output power after that, For pitch axis fine-tuning The actual output power after that.

[0056] The sign of the gradient directly indicates the trend of power change with angle: if the gradient is positive, it means that the power increases with the increase of angle, and the adjustment should be made in that direction; if the gradient is negative, the adjustment should be made in the opposite direction, thus providing a clear directional basis for subsequent corrections.

[0057] Lighting matrix blending: The surface of the photovoltaic panel is divided into The grid is constructed using real-time light intensities I1, I2, I3, and I4 measured by light sensors placed at the four corners of the photovoltaic panel. A bilinear interpolation method is then used to generate a light matrix for each grid point.

[0058] in, This is the lighting matrix; The weights are bilinear interpolation weights, which are dynamically adjusted based on the distance between the light sensor position and the grid points. Let be the light intensity of the k-th light sensor.

[0059] Bilinear interpolation weight calculation:

[0060]

[0061]

[0062]

[0063] in, For grid point coordinates, , , , The coordinates of the four illumination sensors are given. This formula is used to calculate the bilinear interpolation weights, improving the accuracy of the illumination matrix. Finally, the complete... Lighting Matrix Each element represents the light intensity at the corresponding location on the photovoltaic panel.

[0064] Solving for the centroid of illumination and generating feedback corrections The generated illumination matrix is ​​convolved and moment-operated to solve for the coordinates of the illumination centroid. Then, combined with the power gradient direction, the angle feedback corrections for the horizontal and pitch axes are generated.

[0065] Calculate the centroid of illumination using convolution:

[0066]

[0067] in, The x-coordinate of the centroid of illumination; The vertical coordinate of the centroid is the center of gravity of the solar panel; under uniform illumination, the centroid is located at the geometric center of the solar panel. G x0 , G y0 When partial occlusion occurs, the center of gravity will deviate from the geometric center.

[0068] When there is partial occlusion, the center of gravity offset is converted into an angle correction amount:

[0069]

[0070] in, For the horizontal axis illumination centroid correction angle; This is the horizontal axis centroidal correction factor; The reference coordinates for the centroid of uniform illumination are the coordinates of the photovoltaic panel's geometric center. For the pitch axis illumination centroid correction angle; Horizontal axis centroidal correction factor and pitch axis center of gravity correction factor This is the proportional coefficient used to convert the centroid offset of illumination into an angle correction, expressed in degrees per grid. It can be determined through theoretical calculations or system calibration: theoretically, it can be calculated based on the photovoltaic panel size, the number of grid cells in the illumination matrix, and the maximum tracking field of view; in practical engineering, it can be determined through multiple sets of local shading calibration tests, taking the coefficient value that maximizes the system's average power generation, typically ranging from 0.5° to 10° per grid. To improve system robustness, adaptive corrections to the coefficient can also be introduced based on illumination intensity and wind speed to avoid system oscillations in low-light or high-wind scenarios. In this example, we take... = =2.0° / grid.

[0071] The dual-axis angle decision unit performs the following processing: Based on the uniformity of illumination U, the prediction and feedback weights are dynamically adjusted:

[0072]

[0073]

[0074] in, The standard deviation of the intensity of the four-way illumination; The average light intensity of the four light sensors; To predict weights; For feedback weights; Based on the calculated illumination uniformity U. If U ≥ 0.8, it is considered uniform illumination, and centroid correction is not required; if U < 0.8, it is determined that there is local occlusion, and the centroid correction angle calculated in step S3-4 is used. , As a feedback correction, it is used for angle adjustment.

[0075] To improve robustness under occlusion conditions, this invention proposes an improved formula for illumination uniformity:

[0076] in, Let the light intensity be the light intensity of the k-th light sensor. This represents the average light intensity obtained from the four light sensors.

[0077] This formula uses the mean absolute deviation instead of the standard deviation, and is not sensitive to single-channel sensor errors and local occlusion. It can more accurately describe the uniformity of light distribution and provide a more reliable basis for weight adjustment.

[0078] This unit dynamically integrates prediction and feedback results based on real-time environmental parameters:

[0079]

[0080] in, This is the final angle signal of the horizontal axis. This is the final angle signal of the pitch axis. Predict the angle for the horizontal axis. To predict the pitch axis angle, This is the horizontal axis feedback correction amount. This is the pitch axis feedback correction amount. To predict the weight, For feedback weights.

[0081] Execution driver module The dual-axis mechanism 14 is driven by a brushless DC motor, with a horizontal axis (azimuth angle) rotation range of 0–360° and a pitch axis (elevation angle) rotation range of 0–90°. The motor is powered by 24V DC, with a rated speed of 3000rpm, a rated torque of 0.5N·m, and a stall torque of 1.5 N·m. It is equipped with a 16-bit absolute encoder with a resolution of 0.0055°, enabling closed-loop angle control and a positioning accuracy better than 0.05°.

[0082] Kinematic model of the actuator: The equation of motion for the horizontal axis is:

[0083] in, This represents the real-time angle of the horizontal axis. The initial angle of the horizontal axis. The horizontal axis angular velocity, For time.

[0084] The equation of motion for the pitch axis is:

[0085] in, This represents the real-time angle of the pitch axis. The initial angle of the pitch axis. The pitch axis angular velocity, For time.

[0086] This model is used to describe the motion laws of the actuator, providing a foundation for the design of control algorithms.

[0087]

[0088] Power supply module The system outputs voltage from the photovoltaic panels, which is then regulated to 12V by a DC-DC converter. It is also equipped with a lithium battery energy storage unit to ensure normal operation on cloudy days or at night, achieving energy self-sufficiency. The DC-DC converter uses synchronous rectification technology, achieving a conversion efficiency greater than 95% and an output ripple of less than 100 mV. The lithium battery is a lithium iron phosphate battery with a capacity of 100 Ah and a cycle life greater than 2000 cycles.

[0089] The energy balance equation for the power supply system is:

[0090] The power generation of the photovoltaic panels is:

[0091] Lithium-ion battery energy storage is:

[0092] in, The total energy of the system. For the power generation of photovoltaic panels, For energy storage of lithium batteries, Consumption due to system load. Real-time power of the photovoltaic panel. For lithium battery capacity, This refers to the lithium battery voltage.

[0093] This formula is used to evaluate the energy balance of a power supply system and provides a reference for system design and operation.

[0094]

[0095] in, This is the system power supply voltage status assessment value. For DC-DC converter efficiency parameters, This is the output voltage signal for the photovoltaic panel.

[0096] Example 2 This embodiment provides a solar photovoltaic tracking method, including the following steps: S1. Dynamic Prediction of Solar Trajectory: The dynamic prediction unit of solar trajectory dynamically predicts the position of the sun, enabling advance prediction of the sun's position for the next 10 minutes. This invention proposes an improved dynamic declination and hour angle calculation model, introducing a seasonal correction factor. and regional offset Used to calculate the solar declination angle and hour angle: The seasonal correction factor is: (13) The regional offset is:

[0097] in, n For years, t It is true solar time.

[0098] Based on the above parameters, the solar declination angle can be calculated. Sum of time angles Then, combining the local latitude and longitude, the solar altitude angle and azimuth angle are calculated. This invention further proposes a cubic curve dynamic prediction model, trained based on five years of historical solar position and cloud movement data, dynamically updating coefficients to achieve dynamic prediction of the solar declination and hour angle for the next 10 minutes. The model uses a rolling window method, updating model parameters every 5 minutes to improve prediction accuracy.

[0099]

[0100]

[0101] in, To predict the declination angle, To predict the hour angle, These are time-varying coefficients; The time-varying coefficients are updated in real time using the least squares method for fitting:

[0102] in, i These are historical declination angle sample values; These are sample values ​​of the hour angle. t i For the corresponding time period; N is the number of sample points involved in the fitting. Solar altitude angle and solar azimuth Calculate using the following formulas respectively:

[0103] in, The local geographical latitude; The effects of cloud cover are corrected for by the rate of change in illumination and cloud thickness at adjacent time points:

[0104] in, This is the correction amount for the angle prediction caused by cloud cover. k 1 The cloud layer influence coefficient. The rate of change of light intensity The speed of cloud movement. This refers to the thickness of the cloud layer.

[0105] The cloud movement speed and cloud thickness are both calculated from image sensors or meteorological data.

[0106] The cloud layer correction is superimposed on the predicted angle to obtain the corrected horizontal axis predicted angle. and pitch axis prediction angle :

[0107]

[0108] S2. Real-time data acquisition: Real-time acquisition of data such as light intensity, photovoltaic panel output power, photovoltaic panel temperature, and wind speed. The light intensity is collected in real time by four sensors located at the four corners of the photovoltaic panel; the output power of the photovoltaic panel is measured by the voltage / current sensor outputting voltage U. pv With current I pv Calculate P=U pv ·I pv The temperature of the photovoltaic panel is collected by a temperature sensor attached to the back of the photovoltaic panel.

[0109] S3. This step takes the photovoltaic output power and four-channel illumination sensor data collected in step S2 as input, and generates the angle feedback correction of the horizontal axis and pitch axis through power gradient calculation, global illumination matrix reconstruction and illumination centroid solution. This solves the problem that traditional power extremum search algorithms are prone to getting trapped in local optima and improves the tracking robustness in partially occluded scenarios.

[0110] S3.1 Power Gradient Calculation Based on the real-time output power of the photovoltaic panel collected in step S2, the partial derivatives of the power with respect to the angles of the horizontal and pitch axes, i.e. the power gradient, are calculated using the improved forward difference method, providing a directional basis for feedback correction.

[0111]

[0112]

[0113] Where P is the output power of the photovoltaic panel. This represents the real-time angle of the horizontal axis. This represents the real-time angle of the pitch axis. For the angle, it is a small variable; For fine-tuning the horizontal axis The actual output power after that, For pitch axis fine-tuning The actual output power after that.

[0114] The sign of the gradient directly indicates the trend of power change with angle: if the gradient is positive, it means that the power increases with the increase of angle, and the adjustment should be made in that direction; if the gradient is negative, the adjustment should be made in the opposite direction, thus providing a clear directional basis for subsequent corrections.

[0115] S3.2 Lighting Matrix Blending: The surface of the photovoltaic panel is divided into The grid is constructed using real-time light intensities I1, I2, I3, and I4 measured by light sensors placed at the four corners of the photovoltaic panel. A bilinear interpolation method is then used to generate a light matrix for each grid point.

[0116] in, This is the lighting matrix; The weights are bilinear interpolation weights, which are dynamically adjusted based on the distance between the light sensor position and the grid points. Let be the light intensity of the k-th light sensor.

[0117] Bilinear interpolation weight calculation:

[0118]

[0119]

[0120]

[0121] in, For grid point coordinates, , , , The coordinates of the four illumination sensors are given. This formula is used to calculate the bilinear interpolation weights, improving the accuracy of the illumination matrix. Finally, the complete... Lighting Matrix Each element represents the light intensity at the corresponding location on the photovoltaic panel.

[0122] S3.3 Solving for the Center of Gravity of Illumination and Generating Feedback Correction Amounts Perform convolution and moment operations on the illumination matrix generated in step S3.2 to solve for the coordinates of the illumination centroid. Then, combine the power gradient direction to generate the angle feedback correction values ​​for the horizontal axis and pitch axis.

[0123] Calculate the centroid of illumination using convolution:

[0124]

[0125] in, The x-coordinate of the centroid of illumination; The vertical coordinate of the centroid is the center of gravity of the solar panel; under uniform illumination, the centroid is located at the geometric center of the solar panel. G x0 , G y0 When partial occlusion occurs, the center of gravity will deviate from the geometric center.

[0126] When partial occlusion occurs, the center of gravity offset is converted into an angle correction amount and calculated according to formulas (30) and (31).

[0127]

[0128]

[0129] in, For the horizontal axis illumination centroid correction angle; This is the horizontal axis centroidal correction factor; The reference coordinates for the centroid of uniform illumination are the coordinates of the photovoltaic panel's geometric center. For the pitch axis illumination centroid correction angle; Horizontal axis centroidal correction factor and pitch axis center of gravity correction factor This is the proportional coefficient used to convert the centroid offset of illumination into an angle correction, expressed in degrees per grid. It can be determined through theoretical calculations or system calibration: theoretically, it can be calculated based on the photovoltaic panel size, the number of grid cells in the illumination matrix, and the maximum tracking field of view; in practical engineering, it can be determined through multiple sets of local shading calibration tests, taking the coefficient value that maximizes the system's average power generation, typically ranging from 0.5° to 10° per grid. To improve system robustness, adaptive corrections to the coefficient can also be introduced based on illumination intensity and wind speed to avoid system oscillations in low-light or high-wind scenarios. In this example, we take... = =2.0° / grid.

[0130] S3.4 Correction fusion output Based on the power gradient direction obtained in step 3.1, a weighted correction is applied to the center of gravity offset angle to finally generate the angle feedback correction amounts for the horizontal and pitch axes:

[0131]

[0132] in, This is the horizontal axis feedback correction amount. This is the pitch axis feedback correction amount; This is the sign function, used to ensure that the correction direction is consistent with the power gradient direction.

[0133] S4. Calculate the prediction / feedback weights based on the uniformity of illumination (higher illumination uniformity results in higher prediction weights, and vice versa), and combine the predicted angle of the solar trajectory with the feedback correction amount to generate the final angle command for the dual axes.

[0134] Dynamic weight adjustment mechanism: Based on the uniformity of illumination U, the prediction and feedback weights are dynamically adjusted:

[0135]

[0136]

[0137] in, The standard deviation of the intensity of the four-way illumination; The average light intensity of the four light sensors; To predict weights; For feedback weights; Based on the calculated illumination uniformity U. If U ≥ 0.8, it is considered uniform illumination, and centroid correction is not required; if U < 0.8, it is determined that there is local occlusion, and the centroid correction angle calculated in step S3-4 is used. , As a feedback correction, it is used for angle adjustment.

[0138] To improve robustness under occlusion conditions, this invention proposes an improved formula for illumination uniformity:

[0139] in, Let the light intensity be the light intensity of the k-th light sensor. This represents the average light intensity obtained from the four light sensors.

[0140] This formula uses the mean absolute deviation instead of the standard deviation, and is not sensitive to single-channel sensor errors and local occlusion. It can more accurately describe the uniformity of light distribution and provide a more reliable basis for weight adjustment.

[0141] Dual-axis angle decision unit This unit dynamically integrates prediction and feedback results based on real-time environmental parameters:

[0142]

[0143] in, This is the final angle signal of the horizontal axis. This is the final angle signal of the pitch axis. Predict the angle for the horizontal axis. To predict the pitch axis angle, This is the horizontal axis feedback correction amount. This is the pitch axis feedback correction amount. To predict the weight, For feedback weights.

[0144] S5: The dual-axis mechanism is driven by a brushless DC motor to execute angle commands. It is equipped with a 16-bit absolute encoder for PID closed-loop error compensation. At the same time, it detects system abnormalities in real time and triggers mechanisms such as sensor fault switching and wind speed too high closing protection.

[0145] For system workflow, please refer to [link / reference]. Figure 2 .

[0146] (1) System Initialization: After power-on, the power supply module starts, the sensor performs self-calibration, and collects initial environmental parameters. The sensor self-calibration adopts a three-point calibration method, with calibration points at 0W / m² and 0W / m² respectively. 2 1000W / m 2 and 2000W / m 2 Perform calibration to ensure measurement accuracy.

[0147] (2) Position prediction: Based on the local latitude and longitude and real-time time, the dynamic prediction method of the solar trajectory described in step S1 is used to calculate the position of the sun in the next 10 minutes, including the calculation of declination angle and hour angle, dynamic fitting of cubic curve, and cloud movement correction. The location prediction uses a rolling window method, updating model parameters every 5 minutes to improve prediction accuracy.

[0148] (3) Pre-adjustment: The actuator drives the photovoltaic panel to the predicted position. The pre-adjustment adopts a feedforward control strategy, which directly calculates the motor output based on the predicted angle to improve the adjustment speed.

[0149] (4) Real-time data acquisition: Continuously collect parameters such as illumination, power, and temperature. Data acquisition adopts multi-threading technology to ensure data synchronization and real-time performance.

[0150] (5) Feedback calculation: Calculate the power gradient and illumination centroid to generate feedback correction. The feedback calculation employs parallel computing technology to improve computational efficiency.

[0151] (6) Decision Fusion: Calculate weights based on environmental parameters to generate the final angle command. Decision fusion uses a fuzzy logic algorithm to comprehensively consider the prediction results and feedback corrections to generate the optimal angle command.

[0152] (7) Execution and Compensation: The drive motor adjusts the angle, and the encoder performs closed-loop error compensation. The execution compensation adopts a PID control algorithm, which adjusts the motor output in real time according to the encoder feedback to improve positioning accuracy.

[0153] PID control algorithm formula:

[0154] in, For error signals, This is the proportionality coefficient. The integral coefficient; These are the differential coefficients. To control the output signal.

[0155] This formula is used in PID control algorithm design to improve system stability and response speed.

[0156] (8) Loop execution: Repeat steps (2)–(7) to achieve continuous, adaptive tracking. Loop execution uses timer interrupt technology to ensure real-time system response.

[0157] The system also has an anomaly handling mechanism, such as switching to pure astronomical mode when the sensor malfunctions, and entering a retraction protection state when the wind speed is too high:

[0158] Exception handling process: (9) Sensor fault detection: Sensor faults are detected in a timely manner through sensor data consistency detection and fault diagnosis algorithms; (10) Fault handling: When the sensor fails, switch to pure astronomical mode and use astronomical formulas to calculate the position of the sun; (11) Wind speed protection: When the wind speed exceeds the safety threshold, the system enters the retraction protection state and the photovoltaic panel is adjusted to the 0° position to avoid damage.

[0159] System performance evaluation To evaluate the performance of this system, the applicant conducted a series of experiments, including tracking accuracy testing, adaptability testing, and reliability testing.

[0160] The system was tested under various weather conditions. Tests were conducted in multiple scenarios, including real-world operation in sunny, cloudy, cloud-moving, and partially obscured conditions. The following steps were followed during the testing process: An automatic solar photovoltaic tracking system was deployed in an experimental site with controlled lighting conditions to ensure that the surrounding environment represented a typical application scenario. A high-precision solar position tracker was used as a reference, with an accuracy of within 0.01° in measuring the real-time position of the sun. The actual angle output of the device was recorded synchronously as a benchmark for the system's tracking accuracy.

[0161] Test steps: Initial calibration: Before the experiment begins, adjust the photovoltaic panel to the initial position of the system, i.e., both the horizontal axis and the pitch axis are 0°.

[0162] Weather recording: Real-time recording of weather data during the test, including environmental parameters such as light intensity, cloud cover, wind speed, temperature, and humidity.

[0163] System operation: Under different weather conditions, the solar photovoltaic automatic tracking system is operated, and the fusion of sensors and algorithms is used for dynamic prediction and feedback correction.

[0164] Data acquisition: The system records the horizontal and pitch axis angles output by the system, as well as the actual angle of the reference device and the true position of the sun, every 10 seconds through the data recording module.

[0165] Multiple weather tests: Under clear weather conditions: with stable and unobstructed lighting, test whether the tracking response of the test system is consistent with that of the reference instrument, and record the accuracy error.

[0166] Cloudy conditions: Test the system's response speed and accuracy when intermittently obscured by clouds, including the effect of changes in illumination uniformity on the weight adjustment of the tracking algorithm.

[0167] Partial shading: Obstacles are artificially introduced around the photovoltaic panels to simulate partial shadow scenarios and test the tracking and compensation capabilities of the feedback correction module for the shift in the center of gravity of the illumination.

[0168] High wind environment: A high wind speed environment is simulated by using a wind turbine to verify the response and adjustment accuracy of the abnormal protection module under different wind speeds.

[0169] Repeat testing and result validation: To ensure the accuracy of the experimental results, the test was repeated at least three times for each weather condition, with the same time period selected for the test, and the results were compared and analyzed to verify the consistency and stability of the system.

[0170] Tracking accuracy test Test results show that the tracking accuracy of this system is better than 0.1°, which is about 50% higher than that of traditional systems. The tracking accuracy is evaluated using solar position measurement error, which is less than 0.1°.

[0171] Tracking accuracy evaluation formula: (42) in, For tracking error; and This represents the actual position of the sun. and This is the system output angle.

[0172] This formula is used to evaluate the tracking accuracy of the system and provides a basis for performance optimization.

[0173] Adaptability test The system's adaptability was tested under conditions of partial occlusion and severe weather. Test results show that the system can quickly respond to changes in illumination and partial occlusion, and its tracking performance is stable. The adaptability test employed experiments involving sudden changes in illumination and partial occlusion; the system could respond to illumination changes within 10 ms, with a tracking error of less than 0.2°.

[0174] Reliability testing Long-term reliability testing was conducted on the system for 1000 hours. Test results showed that the system's fault-free operating time was greater than 99.9%, indicating high reliability. Reliability testing used Mean Time Between Failures (MTBF) for evaluation, and the MTBF was greater than 10,000 hours.

[0175] in conclusion The solar photovoltaic automatic tracking system proposed in this invention, based on a dual-axis dynamic prediction-feedback fusion algorithm, achieves high-precision and highly adaptable automatic tracking of photovoltaic panels by integrating dynamic prediction of solar trajectory with real-time feedback of multiple parameters.

[0176] The system adopts a modular design, including a data acquisition module, an algorithm processing module, an execution drive module, and a power supply module. Data is collected through four high-precision light sensors and various environmental and system status sensors, and transmitted to the processor via a CAN bus. The algorithm processing module runs a dual-axis dynamic prediction-feedback fusion algorithm, integrating an improved solar trajectory dynamic prediction model with multi-parameter feedback correction results to dynamically adjust weights and generate optimal angle commands. The execution drive module achieves precise dual-axis positioning using a brushless DC motor and a high-precision encoder. The power supply module ensures the system's energy self-sufficiency. Experiments show that the system achieves a tracking accuracy better than 0.1°, exhibits strong adaptability and high reliability, and can significantly improve the efficiency of solar photovoltaic utilization.

Claims

1. A method for automatic solar photovoltaic tracking, characterized in that, Includes the following steps: Step S1: Predict the position of the sun at a future time T; Step S2: Real-time data collection of light intensity, photovoltaic output power, photovoltaic panel temperature, and wind speed; Calculate the uniformity of illumination based on the real-time collected light intensity; Step S3: Generate an illumination matrix using bilinear interpolation based on the real-time collected illumination. The illumination centroid is obtained by convolution calculation of the illumination matrix; Step S4: Calculate the prediction weight and feedback weight based on the illumination uniformity; obtain the final horizontal axis angle signal and the final pitch axis angle signal based on the horizontal axis prediction angle, pitch axis prediction angle, horizontal axis feedback correction amount, pitch axis feedback correction amount, prediction weight, and feedback weight.

2. The automatic solar photovoltaic tracking method according to claim 1, characterized in that, Step S3 includes: Step S3-1: Set the small angle variable And calculate the power gradient with respect to angle based on the measured output power of the photovoltaic panel: in, For the output power of the photovoltaic panel, This represents the real-time angle of the horizontal axis. This represents the real-time angle of the pitch axis. The angle is a preset small variable; For positive fine-tuning of the horizontal axis The actual output power after that, For pitch axis positive fine adjustment The measured output power afterwards; Step S3-2: Generate the illumination matrix using bilinear interpolation: in, This is the lighting matrix; The weights are bilinear interpolation weights, which are dynamically adjusted based on the distance between the light sensor position and the grid points. Let be the light intensity of the k-th light sensor; Step S3-3: Calculate the illumination centroid through convolution: in, The x-coordinate of the centroid of illumination; The vertical coordinate of the centroid of illumination; Step S3-4: In the case of partial occlusion, calculate the lighting centroid correction angle used for angle correction: in, For the horizontal axis illumination centroid correction angle; For the pitch axis illumination centroid correction angle; This is the horizontal axis centroidal correction factor; This is the pitch axis center of gravity correction factor; This is the geometric center of the photovoltaic panel.

3. The automatic solar photovoltaic tracking method according to claim 2, characterized in that, In step S3-2, the bilinear interpolation weights are: in, For grid point coordinates, , , , The coordinates are the positions of the four light sensors.

4. The automatic solar photovoltaic tracking method according to claim 2, characterized in that, Step S4 includes: Step S4-1: Calculate the uniformity of illumination: Where U represents the uniformity of illumination; Let the light intensity be the light intensity of the k-th light sensor; The average light intensity obtained from the four light sensors; Step S4-2: Calculate the prediction weight and feedback weight based on the illumination uniformity U: in, To predict weights; For feedback weights; Step S4-3: Obtain the prediction result based on real-time environmental parameters: in, This is the final angle signal of the horizontal axis; This is the final angle signal of the pitch axis; These are the predicted angles for the horizontal and pitch axes, respectively. This is the horizontal axis feedback correction amount; This is the pitch axis feedback correction amount.

5. The automatic solar photovoltaic tracking method according to claim 1, characterized in that, Step S1 includes: Step S1-1: Calculate the solar declination angle and hour angle: = in, The solar declination angle; The hour angle; n It is accumulated over a year; t True solar time; Step S1-2: Calculate the solar altitude angle and azimuth angle using the local latitude and longitude: in, To predict the declination angle, To predict the hour angle, These are time-varying coefficients; Solar altitude angle and solar azimuth Calculate using the following formulas respectively: in, The local geographical latitude; Step S1-3: Calculate the angle prediction correction amount caused by cloud cover: in, The correction amount for angle prediction caused by cloud cover; kc The cloud layer influence coefficient; The rate of change of light intensity; This refers to the speed of cloud movement. Cloud thickness; The revised prediction angle is: in, These are the predicted angles for the horizontal axis and the pitch axis, respectively.

6. The automatic solar photovoltaic tracking method according to claim 5, characterized in that, In step S1-2, the time-varying coefficients are updated in real time using the least squares method: in, i represents the historical declination angle sample value; ti represents the corresponding time; and N represents the number of sample points involved in the fitting.

7. A solar photovoltaic automatic tracking system, characterized in that, include: The data acquisition module includes four high-precision light sensors, one wind speed sensor, one voltage / current sensor, and one temperature sensor. The four high-precision light sensors are respectively arranged at the four corners of the photovoltaic panel to collect the light intensity at each point in real time. The voltage / current sensor is used to obtain the current and voltage of the photovoltaic panel. The wind speed sensor is used to obtain the wind speed of the environment where the photovoltaic panel is located. The temperature sensor is used to obtain the temperature of the photovoltaic panel. The trajectory prediction module uses an improved declination angle and hour angle model, combined with seasonal correction factors, regional offset and cloud influence correction, to achieve dynamic prediction of the sun's position; The algorithm processing module obtains the final horizontal axis angle signal and the final pitch axis angle signal based on the data collected by the data acquisition module and by executing the steps of the solar photovoltaic automatic tracking method according to any one of claims 1-6. The execution drive module drives the dual-axis brushless DC motor to adjust the horizontal and pitch axis angles based on the final horizontal axis angle signal and the final pitch axis angle signal obtained by the algorithm processing module.

8. The solar photovoltaic automatic tracking system according to claim 7, characterized in that, The execution drive module uses a dual-axis brushless DC motor and a high-precision encoder, combined with PID closed-loop control to achieve precise angle execution and error compensation.

9. The solar photovoltaic automatic tracking system according to claim 7, characterized in that, The trajectory prediction module uses a cubic curve dynamic model, which is trained based on historical solar position and cloud data, and the model coefficients are updated continuously to predict the solar position in the next 10 minutes.

10. A storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps of the automatic solar photovoltaic tracking method as described in any one of claims 1-6.

Images