A photovoltaic module cooperative tracking control system and method for active protection against extreme weather and intelligent operation and maintenance

Abstract

The application relates to the technical field of solar power generation, in particular to a photovoltaic module cooperative tracking control system and method for active protection and intelligent operation and maintenance of extreme weather, which comprises an environment sensing and data acquisition module, a central processing and intelligent decision module, a driving execution and communication interaction module and an energy management and power distribution module. Through multi-source data fusion, a least square support vector machine model optimized by an improved snake optimizer is used to accurately predict the dust deposition degree and power loss of the module, the extreme weather risk is evaluated in real time, according to the prediction and evaluation results, intelligent decision and execution of high-precision sun tracking, active safety protection or dust deposition management auxiliary action are carried out. Through the construction of the integrated cooperative control platform of "sensing-decision-execution", the system can carry out unified intelligent decision based on the real-time fused multi-source data, and realizes the intelligent unification and dynamic optimization of power generation income, equipment safety and operation and maintenance efficiency.

Description

Technical Field

[0001] This invention relates to the field of solar power generation technology, specifically a photovoltaic module collaborative tracking control system and method for active protection and intelligent operation and maintenance in the face of extreme weather. Background Technology

[0002] As one of the core forms of clean energy, the long-term reliable operation and power generation efficiency optimization of photovoltaic power generation systems have always been the focus of industry attention. Currently, photovoltaic tracking systems have evolved from fixed brackets to single-axis or dual-axis systems that can automatically track the sun's position, significantly improving power generation. However, facing complex and ever-changing outdoor operating environments, especially against the backdrop of frequent extreme weather events and increasingly prominent component contamination, existing technologies have significant shortcomings in terms of systemicity, intelligence, and proactivity.

[0003] Firstly, regarding system architecture and functional coordination, existing solutions are mostly simple superpositions or independent operation of single functional modules. Common tracking controllers drive the rotation of components solely based on solar position algorithms or light sensor signals to maximize power generation. Protection against extreme weather such as strong winds and hail often relies on a separate mechanical limit device or a simple wind speed switch, which performs an emergency leveling action when the measured wind speed exceeds a threshold. For contamination issues affecting power generation efficiency, such as dust and snow accumulation on the component surface, regular manual inspections or the deployment of independent detection devices are typically used. These functional modules lack data sharing and have separate decision-making logic, failing to form an intelligent whole with information fusion and collaborative optimization, and thus unable to cope with complex operating conditions involving multiple conflicting objectives.

[0004] Secondly, in terms of extreme weather protection, existing technologies generally lack initiative and predictability, remaining in a state of "passive endurance" or "delayed response." Most tracking systems' safety strategies are based on on-site measured meteorological data, a model that carries risks such as delayed response and a lack of simplistic strategies.

[0005] Finally, regarding the monitoring and management of efficiency degradation factors such as component dust accumulation, the current mainstream method is to indirectly determine performance loss by comparing theoretical power generation with actual power generation. However, power loss is affected by multiple factors, including irradiance fluctuations, temperature changes, and component degradation itself. Relying solely on power difference is insufficient to accurately identify and quantify the share of loss caused by dust accumulation, resulting in a high false alarm rate. Visual inspection solutions mostly remain at the stage of qualitatively judging "with or without dust," are greatly affected by lighting conditions, and fail to establish a precise mathematical model between image features and power generation loss rate. Monitoring results are severely disconnected from operational actions. The system typically only issues general alerts and cannot provide intelligent decision support such as when to clean or what the cleaning priority should be, nor can it automatically adjust component orientation to facilitate cleaning operations, failing to form a closed-loop operation and maintenance system of "monitoring-prediction-decision-assisted execution."

[0006] Therefore, the current photovoltaic tracking technology field urgently needs an integrated control platform and method that can build a multi-source information deep fusion, power generation tracking, safety protection and intelligent operation and maintenance collaborative decision-making, transform passive response into active defense, integrate external early warning information, realize the execution of graded and differentiated safety strategies based on risk prediction, and realize closed-loop intelligent operation and maintenance from prediction to governance assistance. Summary of the Invention

[0007] To address the above technical problems, this invention provides an integrated control platform and method capable of constructing a multi-source information deep fusion, power generation tracking, security protection, and intelligent operation and maintenance collaborative decision-making, in order to solve the problems of functional isolation, passive response to extreme weather, and serious disconnect between dust accumulation monitoring results and operation and maintenance in existing photovoltaic module control systems.

[0008] To address the aforementioned technical problems, the present invention provides a photovoltaic module collaborative tracking and control system for active protection and intelligent operation and maintenance in the context of extreme weather, specifically comprising the following modules:

[0009] The environmental sensing and data acquisition module is used to comprehensively collect environmental and component status data. It consists of a multispectral environmental sensor array, a component status monitoring unit, and a meteorological data interface.

[0010] The multispectral environmental sensor array includes at least a high-precision radiometer, a temperature and humidity sensor, and an ultrasonic / mechanical anemometer to obtain accurate solar irradiance, ambient temperature and humidity, instantaneous wind speed, and wind direction angle.

[0011] The component condition monitoring unit includes vibration acceleration sensors installed on key stress-bearing parts of the tracking bracket, such as the spindle and connecting rod, to monitor mechanical vibration and structural health; PT100 or thermocouple temperature sensors attached to the back panel of the component to monitor the operating temperature of the component; and a vision sensor facing the surface of the component to acquire images of the component surface to analyze stains, dust accumulation, or physical damage.

[0012] The meteorological data interface connects to the Internet via Ethernet and 4G / 5G modules, receiving and parsing short-term weather forecasts and disaster warnings from regional meteorological observatories or dedicated meteorological services in real time.

[0013] The central processing and intelligent decision-making module, acting as the system's "brain," is responsible for data fusion, model computation, and instruction generation. Its core is an industrial-grade embedded computing platform, integrating a programmable logic controller and a microprocessor.

[0014] Programmable logic controllers (PLCs) are responsible for routine I / O control, task scheduling, driver instruction issuance, and system self-testing.

[0015] The microprocessor is used to run complex AI algorithm models, specifically embedding two core models: an ash accumulation prediction model and a safety control model. The ash accumulation prediction model is constructed using a least-squares support vector machine optimized with an improved snake optimizer, and is used to predict the degree of ash accumulation and the corresponding power generation loss rate based on historical and real-time data. The safety control model is used to comprehensively assess the risk level of extreme weather by integrating meteorological data and vibration data, and generate corresponding component protection posture commands.

[0016] The drive execution and communication interaction module is responsible for accurately executing actions and realizing human-machine interaction. It consists of a high-precision dual-axis gimbal drive mechanism, a local human-machine interface, and a remote communication unit. Specifically:

[0017] The high-precision dual-axis gimbal drive mechanism consists of a precision geared motor, a worm gear transmission pair with self-locking function, a high-precision encoder, and a relay drive circuit, which drives the photovoltaic module to rotate and position precisely in a two-dimensional space of azimuth and elevation angles.

[0018] The local human-machine interface uses an industrial touch screen to display system status, early warning information, and power generation data in real time through a graphical interface, and supports parameter settings;

[0019] The remote communication unit uploads critical data to the cloud monitoring platform via Ethernet or wireless network and receives remote commands.

[0020] The energy management and power distribution module adopts an intelligent switching power architecture, supporting dual power supply from mains input and the built-in lithium battery pack. Internally, it includes multi-stage AC / DC and DC / DC power conversion and protection circuits to provide a matched and stable power supply for sensors, controllers, drive motors, etc. in the system, ensuring that the system can rely on the battery to complete critical protection actions and maintain core monitoring when the mains power is interrupted.

[0021] Furthermore, the system also includes a local manual control unit, which is independent of the automatic control system. This unit includes physical controls such as an emergency stop button, an automatic / manual mode switch, and an azimuth / altitude jog button, giving maintenance personnel the highest priority direct control rights and ensuring equipment safety in special circumstances.

[0022] A collaborative tracking control method for photovoltaic modules for active protection and intelligent operation and maintenance in extreme weather is also provided, including the following steps:

[0023] S1, Multi-source data acquisition and fusion: Simultaneously acquire component output power, ambient irradiance, component temperature, ambient humidity, wind speed and direction, vibration acceleration, and component surface images; preprocess the data, including moving average filtering for noise reduction, image illumination compensation, and power temperature coefficient correction, to establish an accurate "power-irradiance" benchmark model.

[0024] S2, Intelligent prediction of ash accumulation based on ISO-LSSVM: The preprocessed feature dataset is input into the ash accumulation prediction model; the model uses an improved snake optimizer to globally optimize the penalty factor γ and kernel function width σ of LSSVM; the ISO algorithm introduces the Levy flight strategy in the early stage of iteration to enhance the global exploration capability, and adopts the Gaussian mixture mutation strategy in the later stage of iteration to improve the local search accuracy; the optimized model outputs the ash accumulation degree index and power loss rate prediction values.

[0025] S3, Multi-level Intelligent Decision-Making and Command Generation: The central processing module processes dust accumulation prediction results and safety risk assessment results in parallel. Under normal, risk-free conditions, it calculates the sun's position based on astronomical algorithms and generates the optimal tracking angle command. When the safety control model determines that there is a risk of extreme weather such as strong winds or hail, it immediately interrupts tracking, generates and prioritizes the execution of corresponding protective attitude commands, adjusting to the minimum windward angle and vertical state. When the dust accumulation prediction model output exceeds a preset threshold, it generates dust accumulation warning information and can instruct the gimbal to adjust the components to a specific tilt angle, such as ±45°, that is convenient for manual cleaning.

[0026] S4, Drive Execution and Feedback: The drive execution module receives instructions and controls the pan-tilt unit to move precisely to the target angle; after execution, the system re-collects data to evaluate the effectiveness of the protective or cleaning auxiliary actions.

[0027] S5, Online Model Optimization: The effect data after each decision execution is used as feedback to dynamically update the training set of the ISO-LSSVM model, realizing online adaptive fine-tuning of model parameters and continuously improving prediction accuracy.

[0028] Compared with the prior art, the present invention has the following advantages:

[0029] 1. This invention fundamentally changes the traditional system's single-function and fragmented decision-making. By constructing an integrated collaborative control platform of "perception-decision-execution," the system can make unified intelligent decisions based on real-time fused multi-source data. It accurately tracks data to maximize power generation when there is no risk; prioritizes safety protection during disaster warnings; and triggers maintenance assistance when dust accumulation is severe. This multi-objective collaboration and dynamic priority management capability enables the system to automatically switch core objectives under different operating conditions, transforming it from equipment solely pursuing power generation into an intelligent asset capable of autonomously balancing safety, power generation, and maintenance costs to achieve optimal comprehensive benefits throughout its entire lifecycle. This realizes the intelligent unification and dynamic optimization of power generation revenue, equipment safety, and maintenance efficiency.

[0030] 2. This invention introduces short-term weather warning information as decision input, enabling the system to perform protective actions in advance before a disaster occurs. At the same time, it combines a graded and differentiated protection strategy to provide the optimal protection posture for different disasters, rather than the traditional uniform flat operation. With the help of a mechanically self-locking drive mechanism and a high-priority hardware emergency stop circuit, a proactive safety protection closed loop of "early warning drive, strategy optimization, and reliable execution" is formed, which significantly reduces the risk of component damage, support deformation, or even collapse caused by extreme weather.

[0031] 3. This invention proposes a prediction model based on an improved snake optimizer to optimize a least-squares support vector machine. This model effectively improves the global convergence and stability of parameter optimization by introducing a Lévy flyby and Gaussian mixture mutation strategy. By fusing dynamically calculated theoretical cleaning power with component surface image texture features, the model can accurately isolate environmental interference and quantify the power loss caused by dust accumulation with high precision. Based on this, the system establishes a hierarchical early warning and cleaning assistance mechanism, forming a closed-loop operation and maintenance system from "monitoring" to "execution," upgrading the operation and maintenance mode from periodic, extensive "planned maintenance" to precise and efficient "predictive maintenance."

[0032] 4. This invention adopts a heterogeneous computing architecture of PLC+AI processor, which balances control reliability and algorithm complexity; the dual-path intelligent power supply design of mains power + lithium battery pack ensures that core monitoring and critical protection functions remain available even in extreme weather conditions where mains power is interrupted; the highest priority hardware manual control channel, independent of software, ensures that maintenance personnel can obtain final control in any abnormal situation; through fully automatic intelligent early warning and decision-making, the system significantly reduces reliance on manual inspections and lowers the labor costs of daily maintenance. Attached Figure Description

[0033] Figure 1 This is a flowchart of the system control method.

[0034] Figure 2 This is a schematic diagram of the electrical principle of the gimbal drive and control module.

[0035] Figure 3 Architecture diagram for the driver execution and communication interaction module.

[0036] Figure 4 Optimize the LSSVM flowchart for ISO.

[0037] Figure 5 This is a diagram of the control system structure. Detailed Implementation

[0038] The present invention will be further described below with reference to the accompanying drawings.

[0039] Example 1:

[0040] like Figure 2 , 3 As shown in Figures 4 and 5, a photovoltaic module collaborative tracking and control system for active protection and intelligent operation and maintenance in extreme weather conditions is mainly composed of the following core modules in its physical structure:

[0041] 1. Environmental Perception and Data Acquisition Module: This module acts as the system's sensory nerve endings, responsible for high-precision and high-reliability data acquisition. Specifically, it includes:

[0042] ① Multispectral environmental sensor array:

[0043] The radiometer uses a Class II standard total irradiance meter, such as the Kipp & Zonn CMP11, with a spectral range of 285 to 2800 nm. It is installed on a dedicated, unobstructed column next to the module array, parallel to the module plane. The total irradiance value is output via analog (4-20 mA) or digital interface (RS-485, Modbus protocol), with a sampling frequency of 1 Hz.

[0044] The temperature and humidity sensor uses an integrated digital sensor, such as the Sensirion SHT85, with an accuracy of ±1.5%RH and ±0.1°C. It is installed inside a ventilated radiation shield and outputs data through an RS-485 interface.

[0045] The anemometer uses an ultrasonic sensor, such as the Vaisala WMT52, which has no mechanical rotating parts, requires little maintenance, and can measure wind speed from 0-60 m / s and wind direction from 0-360°, outputting a three-dimensional wind speed vector; it uploads data at a frequency of 1 Hz via an RS-485 interface and provides statistical values ​​such as 10-minute average wind speed and gust wind speed.

[0046] ② Component status monitoring unit:

[0047] The vibration accelerometer uses an IEPE-type triaxial accelerometer, such as the PCB Piezotronics 356A16, with a range of ±50g and a frequency range of 0.5-3000Hz. It is double-fixed to the bearing housing of the tracker's main rotating shaft using high-strength adhesive and bolts to monitor torsional vibration of the main drive structure. Another accelerometer is mounted on the crossbeam of the support in the middle of the component array to monitor lateral bending vibration. Sensor signals are connected to the data acquisition card of the central processing module via shielded cables.

[0048] The component backplane temperature sensor uses a PT100 platinum resistance thermometer with Class A accuracy. It is mounted on the center of a representative component backplane and covered with an insulation layer to reduce environmental interference. It is connected to the PLC's dedicated RTD input module via a four-wire connection to eliminate lead resistance errors.

[0049] The vision sensor employs a 5-megapixel global shutter industrial camera, such as the Hikvision MV-CH050-10UM, equipped with anti-glare glass and a narrow bandpass filter, with a center wavelength of 650nm and a bandwidth of ±10nm, to enhance the contrast between dust and background. The camera is mounted on an independent column at a certain angle to the component array, automatically capturing a set of component surface images every 30 minutes and transmitting the images to the central processor via Ethernet.

[0050] ③ Meteorological Data Interface: The system PLC connects to the power station's local area network via a built-in Ethernet port and is configured with a Python script as a background service. This script accesses the API interface of the China Meteorological Administration or commercial meteorological service providers such as Moji Weather and Xinzhi Weather every 5-10 minutes to obtain gridded short-term weather forecasts centered on the power station's latitude and longitude. It analyzes key information including precipitation type (rain, hail), precipitation intensity, wind speed and direction, and temperature for the next 0-2 hours, and writes the warning information (such as a blue gale warning and an orange hail warning) into the PLC's shared memory area in a structured JSON format.

[0051] 2. Central Processing and Intelligent Decision-Making Module: This module is the brain of the system, employing a heterogeneous computing architecture of "industrial PLC + embedded AI computing unit". Its specific components are:

[0052] ① Supporting hardware setup:

[0053] The programmable logic controller (PLC) selected is the Siemens S7-1200 series (CPU 1215C), which features digital / analog I / O modules, a high-speed counter module for receiving encoder feedback, and a PROFINET / Ethernet communication port. The PLC is responsible for polling control of all underlying hardware (reading all sensor data), system state machine management, issuing drive commands, and HMI communication with the touch screen.

[0054] The microprocessor, or AI computing unit, uses an NVIDIA Jetson Nano or a higher-performance Jetson Xavier NX module. This module communicates with the PLC via USB or Ethernet and is specifically responsible for running AI algorithm models that require high computing power. It runs the Ubuntu operating system and deploys the PyTorch or Scikit-learn framework.

[0055] ② Software for the core algorithm model:

[0056] Data preprocessing and feature engineering:

[0057] Time-series data filtering: The power-irradiance data streams read from the irradiance meter and electricity meter via Modbus are filtered by moving average using a sliding window of length 10 to eliminate noise caused by instantaneous cloud cover.

[0058] Image preprocessing: The RGB images acquired by the vision sensor are first converted to grayscale images, and then the Standard Normal Transform (SNV) is applied. SNV is calculated by subtracting the mean of all pixels in the grayscale vector of each pixel in the image, and then dividing by its standard deviation. This effectively compensates for the effects of non-uniform illumination under different time periods and lighting angles. Subsequently, texture features such as contrast, correlation, and entropy of the Gray-Level Co-occurrence Matrix (GLCM) are extracted from the preprocessed image as grayscale representations.

[0059] Power normalization: Using data from the module backplane temperature sensor, the measured DC power is corrected to the value under standard test conditions based on the temperature coefficient in the module specification (e.g., -0.34% / °C).

[0060] Training and deployment of the ISO-LSSVM dust accumulation prediction model:

[0061] Model input: Constructing feature vectors

[0062] .

[0063] Wherein, G_corr is the corrected effective irradiance, in W / m², referring to the actual effective solar irradiance projected onto the inclined surface of the photovoltaic module (i.e., under the current tilt and azimuth angles), after necessary physical correction; T_module is the module backsheet operating temperature, in °C, the temperature at the center point of the module backsheet directly measured by a temperature sensor such as PT100; RH is the ambient relative humidity, in %, representing the atmospheric relative humidity at the installation location; P_clean is the theoretical cleaning power, in W or kW, referring to the theoretical DC power that should be output under the current real-time environmental conditions (G_corr, T_module), assuming the module surface is absolutely clean; Texture_Contrast is the image texture contrast, one of the gray-level co-occurrence matrix (GLCM) features extracted from the module surface image, reflecting the severity of local gray-level changes in the image, i.e., the depth of the texture's "grooves"; Texture_Entropy is the image texture entropy, also a GLCM feature, measuring the randomness or disorder of the image texture. A higher entropy value indicates a greater amount of information contained in the image, and a more complex and disordered texture; "..." indicates other potential features, including but not limited to the real-time measured output power of components, wind speed, etc.

[0064] Model output: The target value Y is " ".

[0065] Where P_actual is the actual measured power, which refers to the real-time output power directly measured by the DC meter of the photovoltaic string or the DC side sensor of the inverter at a specific moment; P_clean is the theoretical clean power, which refers to the power that should theoretically be output under the same environmental conditions as P_actual (i.e., at the same moment, with the same irradiance and the same component temperature), assuming that the component surface is absolutely clean and free of any contamination.

[0066] Improved Snake Optimizer (ISO) Implementation: Implement the Snake Optimizer (SO) baseline code in Python and make the following key improvements:

[0067] Levy flight strategy: In the first 40% of the algorithm iterations, when simulating the "exploration" behavior of the snake swarm, the Levy flight step size is added to the individual position update formula. The Levy step size is generated by the Mantegna algorithm, and its random step size has a heavy-tailed distribution, which can encourage individuals to make long-distance jumps, enhance global search capabilities, and escape local optima.

[0068] Gaussian Mixture Mutation Strategy: In the last 60% of the iterations, when simulating "exploitation" behavior, a Gaussian mixture mutation is applied to the position vector of the current optimal solution. That is, with a certain probability, a Gaussian distributed random perturbation with a mean of 0 and dynamically decaying variance is superimposed on the optimal solution. This is equivalent to fine-tuning the area near the optimal solution, improving convergence accuracy.

[0069] LSSVM model training: Load historical datasets using Scikit-learn compatible libraries (such as liac-arff) on the Jetson module; the fitness function of the ISO algorithm is defined as the reciprocal of the root mean square error (RMSE) of the LSSVM model on the validation set. ISO finds an optimal combination of hyperparameters, including the penalty factor γ and the radial basis function kernel width σ, through iterative optimization.

[0070] Online prediction: The trained ISO-LSSVM model is saved as a serialized file. During system operation, the Jetson module receives the real-time feature vector X_new from the PLC every 30 minutes, loads the model, and predicts the current dust accumulation loss rate Y_pred. Y_pred is compared with a preset threshold (e.g., Level 1 warning: 5%, Level 2 warning: 10%), and the status code and predicted value are sent back to the PLC via Socket communication.

[0071] Safety control model: This is a rule-based expert system embedded in the ladder logic of the PLC. Its main rules include:

[0072] Strong wind protection rules: IF The received wind speed warning level ≥ blue OR real-time 10-minute average wind speed > 17m / s THEN Safety sign position. When the safety sign is true, the PLC immediately terminates the solar tracking program and calls the "strong wind protection subroutine". This subroutine calculates and controls the pan-tilt unit to rotate the component plane to be parallel to the wind direction (i.e., the minimum windward angle, usually 0°-10°) based on the wind direction angle, and activates the electromagnetic brake of the drive mechanism to lock it.

[0073] Hail Protection Rules: If the received warning information includes "Hail" OR radar reflectivity factor > 45 dBZ, then the highest priority safety flag is set. This flag has the highest interrupt priority, and the system forces components to be adjusted to a vertical position of 85°-90° to minimize the impact area of ​​the hail.

[0074] Vibration anomaly rule: FFT analysis of the IF vibration acceleration sensor shows that the amplitude continuously exceeds the limit at the natural frequency of the support, triggering a "mechanical fault warning" and limiting the tracking speed.

[0075] 3. Driver execution and communication interaction module

[0076] ① Dual-axis gimbal drive mechanism:

[0077] Motors and Drives: Both the azimuth and altitude axes utilize AC servo motors paired with precision planetary gear reducers with large reduction ratios (e.g., 1:100). A worm gear pair serves as the final stage of transmission, with a self-locking angle of less than 5°, ensuring that the component's posture is automatically maintained during power outages or braking.

[0078] Position feedback: A 23-bit multi-turn absolute encoder is integrated at the rear of the motor. At the same time, a high-precision rotary transformer is installed at the worm gear output as redundant position feedback. The data from the two are compared in the PLC to ensure position reliability.

[0079] Control Loop: The PLC sends position commands to the servo drive via the PROFINET bus. The drive internally performs closed-loop control of position, speed, and current. The tracking algorithm in the PLC program calculates the target azimuth angle α and elevation angle β, and generates a smooth motion trajectory command by planning a trapezoidal or S-curve.

[0080] ② Local Human-Machine Interface (HMI): Employs a 10-inch industrial touchscreen, communicating with the PLC via Ethernet. The HMI screen design includes:

[0081] Main monitoring screen: Real-time display of component angle, irradiance, power, wind speed, and system status (e.g., automatic / manual / fault).

[0082] Warning information screen: The screen displays information such as the content, level, and time of dust accumulation warnings, weather warnings, and vibration warnings in a scrolling manner.

[0083] Data trend screen: You can query historical power, irradiance, and ash loss rate curves.

[0084] Parameter settings screen: This screen is for advanced users to set parameters for the tracking algorithm, warning thresholds, communication parameters, etc. Password is required for security reasons.

[0085] ③ Remote Communication Unit: The PLC connects to the power plant's industrial ring network switch via its Ethernet port. An MQTT client runs on the PLC and / or Jetson module, publishing critical operational data (status, alarms, power generation, forecast results) in JSON format to the power plant's central monitoring cloud platform. Simultaneously, by subscribing to command topics issued by the cloud platform, it can receive remote commands such as "lock," "return," and "cleaning mode."

[0086] 4. Energy Management and Power Distribution Module

[0087] ① Hardware components:

[0088] Mains power input: Connect to a single-phase AC220V / 50Hz power supply, which is connected to the module via a 10A circuit breaker.

[0089] Lithium battery pack: Uses lithium iron phosphate batteries with a capacity of 48V / 100Ah and a built-in battery management system.

[0090] Intelligent switching circuit: The core consists of an automatic transfer switch and a set of priority control logic. Under normal conditions, the entire system is powered by mains electricity through a high-efficiency AC / DC power supply (output DC48V), which simultaneously charges the lithium battery pack.

[0091] Multi-stage power conversion: The DC48V main voltage is converted into various voltages required by the system through a series of DC / DC conversion modules: one DC / DC converter is converted to DC24V to power the PLC, sensors, and relays; another is converted to DC12V to power the camera, Jetson module, etc.; the servo driver is directly driven by DC48V or AC220V.

[0092] ② Control Logic:

[0093] When a mains power failure is detected, the ATS seamlessly switches to battery power within 3-10ms.

[0094] In battery-powered mode, the PLC automatically enters "energy-saving mode": solar tracking is suspended, maintaining only core sensor data acquisition, safety monitoring, and necessary communication. If an extreme weather warning is triggered at this time, the system still has enough energy to drive the motor to perform protective actions.

[0095] When the battery level drops below 20%, the system will report a critical "low battery" alarm via the remote communication unit.

[0096] 5. Local manual control unit: This unit is a hardware loop completely independent of the PLC automatic control program, ensuring the highest reliability.

[0097] Hardware components: An IP65-rated emergency operation box containing:

[0098] Emergency stop button: Directly connected in series in the main power contactor coil circuit of the drive motor. When pressed, it physically cuts off the power supply to all motors.

[0099] Key switch: Used to switch between "automatic" and "manual" modes. In "manual" mode, all drive commands output by the PLC are isolated by hardware relays.

[0100] Self-resetting knob: controls "azimuth angle forward / reverse" and "altitude angle forward / reverse". The knob signal is directly connected to a separate small manual controller, which directly outputs PWM signals to the enable and direction terminals of the servo driver to achieve low-speed jogging.

[0101] Priority: Regardless of the PLC's state, the emergency stop button has the absolute highest hardware priority. The manual mode switch has the next highest priority, but it is higher than any PLC software instruction.

[0102] Example 2:

[0103] like Figure 1 The control method of a photovoltaic module collaborative tracking control system for active protection and intelligent operation and maintenance in response to extreme weather conditions, shown below, takes a complete process from normal tracking to hail warning and then to dust accumulation warning as an example, and specifically includes the following steps:

[0104] Step S1: System Initialization and Data Acquisition. The system is powered on, and the PLC completes its hardware self-test. The Jetson module starts and loads the pre-trained ISO-LSSVM model. The environmental perception module begins operation: the radiometer and anemometer sample at a frequency of 1Hz; the vision camera prepares for its first image capture.

[0105] Step S2, Normal Solar Tracking and Parallel Prediction. In the absence of warnings, the PLC runs a solar tracking subroutine. This program, based on precise GPS time and the power station's latitude and longitude, uses a photovoltaic solar energy algorithm or a more precise solar positioning algorithm to calculate the sun's azimuth and altitude angles in real time, taking into account the mechanical limitations of the support structure, to calculate the optimal target angle for the components. Simultaneously, the AI ​​computing unit executes a dust accumulation prediction subroutine every 30 minutes: it acquires the average irradiance, temperature, and power over the past 30 minutes, as well as the latest processed image texture features, inputs them into the ISO-LSSVM model, and obtains the current predicted dust accumulation loss rate of 3% (below the 5% threshold). The result is then fed back to the PLC as "normal." The safety control model continuously monitors wind speed and the meteorological interface.

[0106] Step S3: Extreme Weather Warning Interruption and Active Protection. The meteorological data interface script receives an orange alert from the meteorological bureau indicating that severe hail is expected within the next 30 minutes and writes {"alert":"hail","level":"orange"} to the PLC. The PLC's safety control model immediately triggers the highest priority interrupt.

[0107] At this moment, the PLC immediately pauses the current solar tracking task. Simultaneously, it generates instructions to call the "hail protection subroutine" and generate target angle instructions: azimuth angle to zero (reference zero point), elevation angle adjusted to 88°, which is close to vertical. The PLC drives the component to a safe position and activates the brake locking after it is in place.

[0108] The HMI screen background flashes red, displaying "Hail Warning! Components have entered vertical protection posture." Simultaneously, the warning status and action logs are reported to the cloud platform via MQTT.

[0109] Step S4: Warning Cancellation and System Recovery. The meteorological interface confirms the hail warning has been cancelled, and the safety control model resets the flag. The PLC control component slowly returns from a vertical attitude to the current solar position (recalculated based on astronomical algorithms) and resumes automatic tracking mode.

[0110] Step S5: Dust Accumulation Early Warning Trigger and Assisted Governance Decision. After a week without rain, three days later, the AI ​​computing unit's prediction value continued to rise. When the ISO-LSSVM model output a dust accumulation loss rate of 11.5%, exceeding the level two early warning threshold, the PLC received a "severe dust accumulation" signal. Since there was no weather warning at the moment, the PLC decided to execute the dust accumulation governance auxiliary strategy and generated a "cleaning assistance" instruction. The target elevation angle was set to 45°, which is convenient for the water gun of the cleaning truck to spray. The azimuth angle was set to a specific angle according to the cleaning truck's path planning.

[0111] Once the module is smoothly adjusted to a 45° tilt angle, a yellow warning box pops up on the HMI screen: "Heavy dust accumulation detected, cleaning recommended. The module has been adjusted to the cleaning angle." At the same time, a work order containing the power plant number, string location, and recommended cleaning time is automatically pushed to the cloud-based operation and maintenance management system and the mobile app of the local operation and maintenance personnel.

[0112] Step S6: Effect Evaluation and Online Model Optimization. After completing the cleaning, the maintenance personnel click "Cleaning Complete" on the HMI to confirm. Over the following few sunny days, the system collects "Power-Irradiance" data pairs after cleaning, calculates the performance comparison before and after cleaning, and confirms that the dust loss rate has decreased to approximately 1%.

[0113] The background service on the Jetson module takes the complete data cycle from "dust accumulation" to "cleaning" as a new sample, including feature data and the final "clean" label, and adds it to the end of the historical dataset. Every Sunday morning, the system automatically triggers incremental training: using the dataset containing the new samples, it reruns the ISO algorithm to fine-tune the LSSVM parameters, generating an updated model file. This feedback mechanism allows the model to adapt to the natural decline in component efficiency and the slow changes in local dust characteristics, achieving continuous improvement in prediction accuracy.

[0114] Step S7, Manual Intervention. When inspection personnel discover abnormal noise from a tracker and wish to fix it for inspection, they should turn the key switch on the local control box to "Manual" mode. Automatic control will then be disabled. Carefully adjust the component to a level position convenient for maintenance by jogging the knob, and then proceed with the work. After the work is completed, switch back to "Automatic" mode, and the system will automatically resume operation on the next business day.

Claims

1. A photovoltaic module collaborative tracking and control system for active protection and intelligent operation and maintenance in response to extreme weather conditions, characterized in that, include: The environmental sensing and data acquisition module is used to collect environmental parameters, component status data, and external weather warning information; The central processing and intelligent decision-making module is connected to the environmental perception and data acquisition module. It is used to fuse multi-source data and run the dust accumulation prediction model and safety control model to generate control commands. The ash accumulation prediction model is a model built based on an improved snake optimizer optimized least squares support vector machine; The drive execution and communication interaction module is connected to the central processing and intelligent decision-making module, and is used to drive the photovoltaic module to rotate according to control commands, and realize human-machine interaction and remote communication. The energy management and power distribution module is used to provide stable power to all modules of the system and has the ability to switch between mains power and battery power.

2. The photovoltaic module collaborative tracking control system for active protection and intelligent operation and maintenance in extreme weather as described in claim 1, characterized in that: The environmental sensing and data acquisition module includes: a multispectral environmental sensor group, including at least a radiometer, a temperature and humidity sensor, and a wind speed and direction meter; The component status monitoring unit includes at least a vibration acceleration sensor mounted on a support, a temperature sensor mounted on the back panel of the component, and a vision sensor for capturing images of the component surface. Meteorological data interface, used to access the Internet to obtain short-term weather forecasts and disaster warnings.

3. The photovoltaic module collaborative tracking control system for active protection and intelligent operation and maintenance in response to extreme weather conditions as described in claim 2, characterized in that: The central processing and intelligent decision-making module includes a programmable logic controller and a microprocessor; The programmable logic controller is used for system scheduling and instruction issuance; The microprocessor is used to run the ash accumulation prediction model and the safety control model; The improved snake optimizer incorporates the Levy flight strategy and the Gaussian mixture mutation strategy during the iteration process.

4. The photovoltaic module collaborative tracking control system for active protection and intelligent operation and maintenance in extreme weather as described in claim 3, characterized in that: The drive execution and communication interaction module includes: a dual-axis gimbal drive mechanism, which adopts worm gear transmission and has a mechanical self-locking function; The local human-machine interface uses an industrial touchscreen. The remote communication unit supports wired or wireless network access.

5. The photovoltaic module collaborative tracking control system for active protection and intelligent operation and maintenance in extreme weather as described in claim 4, characterized in that: It also includes a local manual control unit, which is independent of the automatic control loop and has an emergency stop button and manual operation controls at the equipment site, and has the highest operation priority.

6. A control method applied to the photovoltaic module collaborative tracking control system for active protection and intelligent operation and maintenance in extreme weather as described in any one of claims 1-5, characterized in that, Includes the following steps: Collect operating parameters, environmental parameters, and surface image data of photovoltaic modules, and perform preprocessing. The preprocessed data is input into the least squares support vector machine ash accumulation prediction model optimized by the improved snake optimizer to obtain the predicted values ​​of ash accumulation degree and power loss rate. Based on the predicted values ​​and real-time meteorological safety risk assessment results, corresponding control commands are generated and executed. The control commands include one or more of the following: solar tracking commands, extreme weather protection attitude commands, and dust accumulation control auxiliary commands. After executing the instructions, data is collected again to evaluate the effect, and the effect data is used to optimize the ash accumulation prediction model online.

7. The control method for the photovoltaic module collaborative tracking control system for active protection and intelligent operation and maintenance in extreme weather as described in claim 6, characterized in that: The generation and execution of the corresponding control commands specifically includes: executing solar tracking commands to maximize power generation when there is no weather risk; When an extreme weather warning is received, the protective posture command is executed first, adjusting the components to a preset safe angle; When the predicted dust accumulation value exceeds the preset threshold, the dust accumulation management auxiliary command is executed, the warning information is output, and the component can be optionally adjusted to a tilt angle that is easy to clean.

8. The control method for the photovoltaic module collaborative tracking control system for active protection and intelligent operation and maintenance in extreme weather as described in claim 6, characterized in that: The process of optimizing the least squares support vector machine using the improved snake optimizer includes: using the preprocessed data as training samples to initialize the snake optimizer parameters; In the early stages of the iteration, the Levy flight strategy was used to update individual positions for global exploration; In the later stages of the iteration, a Gaussian mixture mutation strategy is used to update the individual positions for local fine-grained search; The final least squares support vector machine prediction model is constructed using the optimal combination of penalty factor and kernel function parameters obtained from the search.

9. The control method for the photovoltaic module collaborative tracking control system for active protection and intelligent operation and maintenance in extreme weather as described in claim 6, characterized in that: The power and irradiance time-series data are filtered by moving average, the component surface image is transformed by standard normal variable to eliminate uneven illumination, and the component output power is corrected by temperature coefficient.

10. The control method for the photovoltaic module collaborative tracking control system for active protection and intelligent operation and maintenance in extreme weather as described in claim 6, characterized in that: It provides a local manual control channel independent of the automatic control process, receives manual operation commands, and directly drives the actuators.

Images