Photovoltaic cleaning robot remote monitoring and fault diagnosis system based on internet of things

Abstract

The application relates to the technical field of robot fault management, and particularly discloses a photovoltaic cleaning robot remote monitoring and fault diagnosis system based on the Internet of Things, which synchronously acquires real-time running current values of driving motors on the left and right sides of a photovoltaic cleaning robot and encoder feedback speed values, calculates an asymmetric index reflecting the resistance difference between the two sides, realizes deep perception and high-sensitivity monitoring of the walking dynamic balance state of the robot, compares an aging coefficient with a material failure benchmark value in a closed loop, and outputs a fault diagnosis report, can accurately predict and indicate the thermal stickiness failure risk of a walking wheel, and fundamentally avoids task interruption, motor burning and even mechanical damage to expensive photovoltaic components caused by wheel body softening and adhesion. Not only is the safe and stable operation of the photovoltaic cleaning robot in a long-term high-irradiation environment ensured, but also the service life of key consumables is prolonged, thereby providing economic value protection for intelligent management and asset preservation of a photovoltaic power station.

Description

Technical Field

[0001] This invention belongs to the field of robot fault management technology, and relates to a remote monitoring and fault diagnosis system for photovoltaic cleaning robots based on the Internet of Things. Background Technology

[0002] As a core piece of equipment for improving the power generation efficiency of photovoltaic power plants and ensuring automated operation and maintenance, the stable operation of photovoltaic cleaning robots plays a decisive role in the energy output of large-scale photovoltaic arrays. In particular, the stability of the wheel-rail contact state of the walking drive system, which serves as the robot's power fulcrum, directly affects whether the robot can achieve accurate inspection and efficient cleaning under complex slope conditions, and is a fundamental physical prerequisite for maintaining the efficient operation of the entire photovoltaic operation and maintenance system. Meanwhile, the thermal environment on the surface of photovoltaic modules, as a byproduct of the photovoltaic cell's photoelectric conversion process, not only affects the service life of the modules but also has a profound physical impact on the robot components that come into direct contact with them. Especially for the walking device, which is under high-frequency friction and load-bearing conditions, the dynamic balance of heat exchange at the contact interface directly affects the motion accuracy and structural integrity of the walking mechanism.

[0003] It is worth noting that the wheels of photovoltaic cleaning robots are typically made of high-polymer elastic materials to balance grip and obstacle avoidance performance. These materials are extremely sensitive to environmental thermal loads. Under high-irradiance conditions in summer, localized high-temperature fields easily form on the surface of dark photovoltaic modules. Long-term or severe thermal wetting not only accelerates the chemical aging of the wheel materials but also causes nonlinear rheological shifts in their physical state. This material performance degradation caused by environmental thermal excitation results in an extremely complex coupling relationship between the mechanical resistance of the walking system and the current response of the drive motor, controlled by environmental variables. Especially for robots operating continuously in non-uniform temperature fields, there is a subtle but cumulative interaction between the localized hot spot effect of the modules and the thermal sensitivity of the walking materials.

[0004] However, current monitoring and management systems for photovoltaic cleaning robots have significant shortcomings in handling such complex coupled failures. Firstly, traditional fault monitoring logic relies excessively on single current thresholds or speed feedback fluctuations, often simplifying complex material-environment coupled failures into common issues like motor overload or mechanical jamming. Maintenance data shows that even when motor control parameters are set within standard ranges, robots may still experience unexplained "motion lag" or "difficulty in starting and peeling off" during specific high-temperature periods. Existing technologies lack the dynamic analytical capability to detect microscopic viscous changes in materials caused by environmental thermal loads, failing to distinguish between sudden obstruction caused by surface foreign objects and deep viscous failure due to thermal softening of the walking materials. This results in existing remote monitoring platforms largely operating in a passive sensing phase, unable to provide early warnings before irreversible thermal damage to the walking wheels, leading to a severe lag in the health management system for photovoltaic robots in complex environments. Summary of the Invention

[0005] In view of the problems existing in the prior art, the present invention provides a remote monitoring and fault diagnosis system for photovoltaic cleaning robots based on the Internet of Things to solve the above-mentioned technical problems.

[0006] To achieve the above and other objectives, the technical solution adopted by the present invention is as follows: The first aspect of this invention provides a remote monitoring and fault diagnosis system for photovoltaic cleaning robots based on the Internet of Things, the system comprising: Side wheel symmetry calculation module: acquires the real-time operating current values ​​and encoder feedback speed values ​​of the drive motors on the left and right sides of the photovoltaic cleaning robot, and calculates the asymmetry index of the resistance difference on the corresponding sides of the photovoltaic cleaning robot based on the real-time operating current values ​​and encoder feedback speed values; compares the asymmetry index with the preset abnormal screening threshold, and when the asymmetry index exceeds the preset abnormal screening threshold, requests the component thermal distribution data and real-time environmental meteorological data associated with the current positioning coordinates of the photovoltaic cleaning robot from the cloud analysis server; Current waveform acquisition module: Based on component thermal distribution data and real-time environmental meteorological data, a thermal environment judgment model is constructed. If the model judgment result indicates that the current environment is in a high thermal excitation state, a thermal relaxation test command containing a predetermined static heat absorption time is generated and sent to the photovoltaic cleaning robot. In response to the thermal relaxation test command, after the photovoltaic cleaning robot completes the static heat absorption action, a pulse step excitation signal of predetermined amplitude is sent to the drive motor, and the transient current response waveform data under the action of the pulse step excitation signal is collected simultaneously. Fault report generation module: Extracts peak stripping current data and phase lag time data at the moment of startup from transient current response waveform data, weights and sums the peak stripping current data and phase lag time data according to preset material rheological weights, calculates the thermal aging coefficient of the walking wheel material corresponding to the photovoltaic cleaning robot, compares the thermal aging coefficient with the preset material failure benchmark value, and generates and outputs a fault diagnosis report indicating that the walking wheel has experienced thermal viscous failure when the thermal aging coefficient exceeds the preset material failure benchmark value.

[0007] The second aspect of this invention provides a method for remote monitoring and fault diagnosis of photovoltaic cleaning robots based on the Internet of Things, including: The real-time operating current values ​​and encoder feedback speed values ​​of the drive motors on the left and right sides of the photovoltaic cleaning robot are obtained, and the asymmetry index of the resistance difference on the two sides of the photovoltaic cleaning robot is calculated based on the real-time operating current values ​​and encoder feedback speed values. The asymmetry index is compared with a preset anomaly screening threshold. When the asymmetry index exceeds the preset anomaly screening threshold, the cloud analysis server is requested to obtain component thermal distribution data and real-time environmental meteorological data associated with the current positioning coordinates of the photovoltaic cleaning robot. A thermal environment determination model is constructed based on component thermal distribution data and real-time environmental meteorological data. If the model determination result indicates that the current environment is in a high thermal excitation state, a thermal relaxation test instruction containing a predetermined static heat absorption time is generated and sent to the photovoltaic cleaning robot. In response to the thermal relaxation test command, after the photovoltaic cleaning robot completes the static heat absorption action, a pulse step excitation signal of predetermined amplitude is sent to the drive motor, and the transient current response waveform data under the action of the pulse step excitation signal is collected simultaneously. Extract the peak stripping current data and phase lag time data at the moment of startup from the transient current response waveform data. Then, weight and sum the peak stripping current data and phase lag time data according to the preset material rheological weights to calculate the thermal aging coefficient of the walking wheel material corresponding to the photovoltaic cleaning robot. The thermal aging coefficient is compared with the preset material failure benchmark value. When the thermal aging coefficient exceeds the preset material failure benchmark value, a fault diagnosis report indicating that the walking wheel has experienced thermal sticking failure is generated and output.

[0008] As described above, the IoT-based remote monitoring and fault diagnosis system for photovoltaic cleaning robots provided by this invention has at least the following beneficial effects: This invention achieves deep perception and high-sensitivity monitoring of the robot's dynamic balance by simultaneously acquiring real-time operating current values ​​of the drive motors on both sides of the photovoltaic cleaning robot and encoder feedback speed values, and calculating an asymmetry index reflecting the difference in resistance between the two sides. By comparing the asymmetry index with a preset anomaly screening threshold in real time, and actively requesting component thermal distribution and real-time meteorological data from the cloud server when the threshold is exceeded, the monitoring of single equipment parameters is elevated to a multi-dimensional spatiotemporal data fusion level based on the Internet of Things. This effectively solves the technical bottleneck of traditional monitoring methods being unable to distinguish between mechanical wear and environmental coupling interference. This process not only analyzes the robot's sub-health state from a dynamic perspective but also achieves precise location of potential fault areas through cloud data links, greatly enhancing the scientific rigor and foresight of the early warning system and providing crucial decision-making basis for subsequent accurate diagnosis.

[0009] This invention constructs a thermosensitive environment judgment model based on acquired heat distribution and meteorological data. After determining that the current environment is in a high thermosensitive excitation state, it generates a thermal relaxation test command containing a predetermined static heat absorption duration and sends it to the robot. The robot is monitored to complete its heat absorption action, and a pulsed step excitation signal is applied. This effectively overcomes the lag problem of passive monitoring modes in identifying latent material failures. By simulating the most unfavorable working condition of the walking wheel contacting high-temperature components, it transforms microscopic material physical changes into quantifiable transient current waveform data. This significantly improves the signal-to-noise ratio of the detection process, ensuring the uniqueness and accuracy of fault feature extraction in complex outdoor environments, and greatly reducing the risk of false alarms and missed alarms caused by sudden environmental changes.

[0010] This invention, by performing a closed-loop comparison between the aging coefficient and the material failure benchmark value and outputting a fault diagnosis report, can accurately predict and indicate the risk of thermal adhesion failure of the walking wheels. This fundamentally avoids task interruptions, motor burnout, and even mechanical damage to expensive photovoltaic modules caused by wheel softening and adhesion. It not only ensures the safe and stable operation of the photovoltaic cleaning robot in long-term high-irradiation environments and extends the service life of key consumables, but also reduces the difficulty of on-site troubleshooting for maintenance personnel through precise fault attribution analysis. Attached Figure Description

[0011] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is a schematic diagram showing the connections of the various modules in the system of the present invention.

[0013] Figure 2 This is a schematic diagram showing the connections between the steps of the method of the present invention. Detailed Implementation

[0014] The following description, in conjunction with the implementation of this invention, is merely an example and illustration of the concept of this invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the inventive concept or exceed the scope defined in these claims, all of which should fall within the protection scope of this invention.

[0015] In traditional fault diagnosis systems for photovoltaic cleaning robots, fixed thresholds and static rules cannot adapt to the dynamic thermo-mechanical coupling effects in complex outdoor environments. When a non-uniform temperature field forms on the surface of photovoltaic modules due to irradiation, the thermoviscous behavior of the wheel material causes nonlinear resistance fluctuations. However, the system lacks environmental perception capabilities and cannot establish a dynamic correlation between resistance differences and thermal conditions. This leads to a mismatch between the initial screening threshold and real-time operating conditions, causing the fault feature screening mechanism to include environmental noise, ultimately affecting the accuracy and timeliness of diagnosis. This invention calculates an asymmetric index by real-time calculation of the current and speed data of the drive motors on both sides, and triggers a cloud data collaboration request only when the index exceeds the limit. This upgrades the system from single threshold judgment to environmental correlation screening, effectively avoiding false alarms caused by temperature fluctuations and improving the specificity of fault identification.

[0016] For example, during midday in summer, the temperature of photovoltaic panels can reach over 70°C. When aging wheels come into contact with this high-temperature surface, their energy storage modulus decreases while their loss modulus increases, leading to a rise in the peak current of the single-sided drive motor and a delay in encoder feedback. Traditional systems still use a preset constant anomaly screening threshold, which may misjudge the characteristics of thermal viscosity as random disturbances and ignore them, or incorrectly classify them as mechanical jamming. The fault indicators output by the multi-source data fusion model lose the crucial thermal sensitivity dimension, making it impossible for the diagnostic logic to identify the specific mode of "thermal viscosity failure." The generated maintenance suggestions may only be to clean foreign objects or adjust the balance, failing to address the root cause of wheel aging. This invention introduces cloud-based thermal distribution and meteorological data to construct a thermally sensitive environment judgment model, which can accurately distinguish between material softening and other mechanical faults under high-temperature conditions. This ensures that subsequent test commands are triggered only under confirmed high thermal sensitivity conditions, thereby eliminating environmental interference and improving the initiation accuracy of the diagnostic process.

[0017] If the above problems are not addressed, misidentification of thermoviscous faults will lead to a misalignment between maintenance measures and actual problems, accelerating wheel failure and potentially causing secondary damage such as robot deviation and component scratches. A rigid feature selection mechanism will hinder the system from capturing the critical point of material transition from an elastic to a viscous state, delaying the optimal time for preventative replacement. The asynchronous processing of environmental and proprioceptive data will also cause phase deviations between thermal excitation and electrical response, reducing the reliability of asymmetric exponential calculations and thermal aging assessments, ultimately creating a negative feedback loop that affects the optimization of the entire robot fleet's operation and maintenance strategy. This invention generates a thermal relaxation test command containing a predetermined static heat absorption duration, allowing the wheels to reach thermal equilibrium on a high-temperature plate surface, simulating the worst-case operating conditions. Subsequently, a pulsed step excitation is applied and transient current waveforms are acquired. This design achieves, for the first time in an operational setting, the controllable excitation and observation of the viscoelastic behavior of materials, providing direct physical evidence for root cause analysis of faults.

[0018] To address the aforementioned issues, this application first considers establishing a dynamic correlation mechanism between resistance differences and environmental thermal state. Traditional systems use fixed thresholds for initial anomaly screening, which leads to the suppression of thermo-viscosity characteristics and an inability to reflect the degree of material aging. To resolve this, this application attempts to couple asymmetric exponents with cloud-based thermal distribution data and real-time meteorological data, dynamically adjusting diagnostic trigger conditions by constructing a thermosensitive environment judgment model. Further analysis reveals that relying solely on operating current and speed data is insufficient to accurately quantify material rheological properties; therefore, an active thermal relaxation test is needed to excite the viscoelastic response of the wheels, and the peak stripping current and phase lag time are extracted from the transient waveform under pulse step excitation. By designing a thermal relaxation test command generation and feature weighted fusion mechanism, the thermosensitive aging coefficient is adaptively calculated according to the actual material state, thereby solving the problems of misjudgment and diagnostic lag. This invention ultimately calculates the thermal aging coefficient by weighted fusion of the stripping current peak value and phase lag time, and compares it with the failure benchmark value. This enables a quantitative assessment of the aging state of the walking wheel material, outputs accurate fault reports, and transforms the maintenance mode from post-remediation to pre-warning. This significantly improves the operational reliability of the robot in high-temperature environments and the overall life-cycle maintenance efficiency.

[0019] After introducing the basic concept of the present invention, the embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Example

[0020] Please see Figure 1 As shown, the IoT-based remote monitoring and fault diagnosis system for photovoltaic cleaning robots includes a side wheel symmetry calculation module, a current waveform acquisition module, and a fault report generation module. The various modules are connected via wired and / or wireless connections to enable data transmission between them; Side wheel symmetry calculation module: acquires the real-time operating current values ​​and encoder feedback speed values ​​of the drive motors on the left and right sides of the photovoltaic cleaning robot, and calculates the asymmetry index of the resistance difference on the corresponding sides of the photovoltaic cleaning robot based on the real-time operating current values ​​and encoder feedback speed values; compares the asymmetry index with the preset abnormal screening threshold, and when the asymmetry index exceeds the preset abnormal screening threshold, requests the component thermal distribution data and real-time environmental meteorological data associated with the current positioning coordinates of the photovoltaic cleaning robot from the cloud analysis server; Current waveform acquisition module: Based on component thermal distribution data and real-time environmental meteorological data, a thermal environment judgment model is constructed. If the model judgment result indicates that the current environment is in a high thermal excitation state, a thermal relaxation test command containing a predetermined static heat absorption time is generated and sent to the photovoltaic cleaning robot. In response to the thermal relaxation test command, after the photovoltaic cleaning robot completes the static heat absorption action, a pulse step excitation signal of predetermined amplitude is sent to the drive motor, and the transient current response waveform data under the action of the pulse step excitation signal is collected simultaneously. Fault report generation module: Extracts peak stripping current data and phase lag time data at the moment of startup from transient current response waveform data, weights and sums the peak stripping current data and phase lag time data according to preset material rheological weights, calculates the thermal aging coefficient of the walking wheel material corresponding to the photovoltaic cleaning robot, compares the thermal aging coefficient with the preset material failure benchmark value, and generates and outputs a fault diagnosis report indicating that the walking wheel has experienced thermal viscous failure when the thermal aging coefficient exceeds the preset material failure benchmark value. Example

[0021] like Figure 2 As shown, a remote monitoring and fault diagnosis method for photovoltaic cleaning robots based on the Internet of Things (IoT) is described, which includes: The real-time operating current values ​​and encoder feedback speed values ​​of the drive motors on the left and right sides of the photovoltaic cleaning robot are obtained, and the asymmetry index of the resistance difference on the two sides of the photovoltaic cleaning robot is calculated based on the real-time operating current values ​​and encoder feedback speed values.

[0022] Preferably, the asymmetric index of the resistance difference on both sides of the photovoltaic cleaning robot is calculated, including: The system simultaneously collects the real-time operating current values ​​of the left and right drive motors of the photovoltaic cleaning robot and the corresponding encoder feedback speed values, and locks the data when it confirms that the encoder feedback speed values ​​on both sides are within the preset effective straight-line travel range. For each drive motor, the locked real-time operating current value is divided by its corresponding encoder feedback speed value to calculate the unit speed load value on the left and the unit speed load value on the right respectively. Calculate the ratio between the unit speed load value on the left and the unit speed load value on the right, and determine the degree of deviation between this ratio and the preset ideal balance reference value as the asymmetry index.

[0023] Preferably, data locking is performed when the encoder feedback speed values ​​on both the left and right sides are confirmed to be within the preset effective straight-line driving range, specifically including: The absolute value of the difference between the feedback speed values ​​of the left encoder and the right encoder is calculated in real time, and this absolute value of the difference is defined as the speed synchronization deviation value. Determine whether the speed synchronization deviation value is less than the preset straight-line driving tolerance threshold, and at the same time monitor whether the encoder feedback speed values ​​on both the left and right sides are greater than the preset minimum stable operating speed threshold. When the speed synchronization deviation value is less than the preset straight-line driving tolerance threshold and the encoder feedback speed value is greater than the preset minimum stable operating speed threshold, it is determined that the photovoltaic cleaning robot is currently in the preset effective straight-line driving range, and a data locking operation is triggered for the real-time operating current value and the encoder feedback speed value.

[0024] In one specific embodiment, the real-time operating current value of the left drive motor is synchronously acquired at a preset high-frequency sampling rate by using Hall encoders located at the ends of the robot's left and right drive wheel axles and current sampling resistors in the motor driver. and encoder feedback speed value And the real-time operating current value of the right-side drive motor. and encoder feedback speed value ;in and All units are amperes; and All units are meters per second; After acquiring the raw data, the system immediately performs a validity screening based on motion state, aiming to eliminate non-viscous load fluctuations caused by steering, starting, or slippage. This screening logic is based on a preset valid straight-line driving interval determination algorithm, and the specific calculation process is as follows: the system calculates the absolute value of the speed difference between the left and right sides in real time. This value is defined as the speed synchronization deviation value. It is only valid if the simultaneous conditions are met. Only then is the data deemed valid and locked. The preset straight-line driving tolerance threshold is recommended to be between 0.05 m / s and 0.1 m / s. The logic for setting this threshold is based on the inherent assembly clearance of the robot's mechanical structure and the steady-state error of the closed-loop control, and is used to tolerate small speed fluctuations that are not of a deviation nature. The minimum stable operating speed threshold is preset, and it is recommended to set it to 0.15m / s to 0.2m / s. The necessity of setting this lower limit is that when the motor is in the low-speed start-up stage, the cogging torque and static friction will cause nonlinear high-frequency noise in the current signal, which cannot truly reflect the road surface viscous resistance. Therefore, it is necessary to filter out the low-speed data.

[0025] After data locking is completed, the system calculates the unit speed load value for each motor. The calculation formula is based on the balance principle of motor output power and mechanical load power, and is derived as follows: .in This represents the unit speed load value on the left or right side, with the physical dimension being ampere-seconds per meter. The design of this formula is based on the fact that the current value alone fluctuates linearly with the robot's speed and cannot be directly used as a resistance indicator. By dividing the current by the speed, the electrical signal is essentially converted into a normalized physical quantity analogous to "energy consumption per unit distance traveled," thereby eliminating the interference of speed on resistance assessment and accurately characterizing the frictional resistance characteristics between the wheels and the photovoltaic module surface.

[0026] Subsequently, the system calculates the asymmetric exponent based on the normalized load value. Its core calculation formula is: In the formula, This is an asymmetric index that characterizes the difference in resistance on both sides, and is a dimensionless percentage value; This is the preset ideal balance reference value, with a default value of 1.0. (Introduced here...) The logic behind the parameters is that, considering the structural design of some photovoltaic cleaning robots, their center of gravity is not originally centered, and the ideal load ratio of the left and right motors at the factory may not be 1:1. This is addressed through calibration. This formula can offset the inherent gravitational asymmetry of the mechanical structure, ensuring that the calculated index fully reflects the abnormal resistance differences caused by the external environment. The advantage of using a ratio form is that when the robot is climbing, the current of both motors increases synchronously due to the gravitational component, and the difference fluctuates with the slope. The ratio form can automatically cancel out the influence of the gravitational component shared by both sides through division, making the index highly robust and consistent on photovoltaic modules with different slopes.

[0027] The asymmetry index is compared with a preset anomaly screening threshold. When the asymmetry index exceeds the preset anomaly screening threshold, the cloud analysis server is requested to obtain component thermal distribution data and real-time environmental meteorological data associated with the current positioning coordinates of the photovoltaic cleaning robot.

[0028] Real-time environmental meteorological data includes solar irradiance values ​​and ambient temperature values; The current location coordinates of the photovoltaic cleaning robot are input into the pre-built digital twin database of photovoltaic module array for spatial indexing, and the unique identifier of the target photovoltaic module covered by the photovoltaic cleaning robot at the current moment is obtained. Based on the unique identifier of the target photovoltaic module, the corresponding historical infrared thermal imaging records are retrieved from the digital twin database. The local temperature difference distribution feature value of the target photovoltaic module under historical high temperature conditions is extracted and used as the module thermal distribution data.

[0029] In one specific embodiment, threshold gating logic is first executed to gate the asymmetric exponent. Compared with the preset anomaly screening threshold Numerical comparisons are performed. Among these, a preset initial anomaly screening threshold is used. The preferred setting is within the range of 15% to 20%. This range is designed based on extensive field test data, which shows that resistance differences caused by mechanical assembly errors or minor road surface unevenness typically range from 0% to 10%. When the asymmetry index exceeds 15%, it highly likely that the single-sided wheel is experiencing continuous abnormal viscous force or structural jamming. Once determined The robot sends its current high-precision GPS positioning coordinates to a cloud analysis server via a 5G IoT module. The data request packet. In response, the cloud server first calls the interface that is connected in real-time with the photovoltaic power station's weather station to obtain real-time environmental meteorological data, specifically including real-time solar irradiance values. (Unit: Watts per square meter) and ambient temperature value .

[0030] Next, the robot's current location coordinates are... Input a pre-built digital twin database of photovoltaic module arrays. This database is built based on UAV orthophotos and CAD drawings and contains geometric boundary information of all modules. The ray casting method is used to determine which polygon area the coordinate point P falls into, so as to uniquely match the unique identifier of the target photovoltaic module covered by the photovoltaic cleaning robot at the current moment.

[0031] Based on the unique identifier, the system retrieves historical infrared thermal imaging records of the component from the historical operation and maintenance database during past summer high-temperature periods and performs feature mining to generate component thermal distribution data.

[0032] Local temperature difference distribution characteristic value Its calculation formula is designed as a statistically weighted average temperature difference: In the formula, N represents the total number of times a hot spot has been detected in the component in historical records; The highest temperature of the local hotspot in the i-th record; This represents the average surface temperature of the entire component during this recording. This is the time decay weighting coefficient; the closer a record is to the current time, the greater its weight. The calculation formula involves not just looking at the temperature once, but quantifying how much hotter the component is "habitually" than its surroundings at historically high temperatures, thus characterizing its inherent localized heating intensity.

[0033] A thermal environment determination model is constructed based on component thermal distribution data and real-time environmental meteorological data. If the model determination result indicates that the current environment is in a high thermal excitation state, a thermal relaxation test instruction containing a predetermined static heat absorption time is generated and sent to the photovoltaic cleaning robot.

[0034] Preferably, generating a thermal relaxation test instruction containing a predetermined static heat absorption duration includes: Real-time environmental meteorological data is used as the basic environmental thermal variable, and component thermal distribution data is used as the local thermal correction factor. The real-time component surface temperature prediction value at the current position of the photovoltaic cleaning robot is calculated through the thermal sensitive environment judgment model. The real-time component surface temperature prediction value is compared with the preset rheological conversion temperature threshold of the wheel material. When the real-time component surface temperature prediction value is greater than the preset rheological conversion temperature threshold, it is confirmed that the judgment condition of the high thermosensitive excitation state is met, indicating that the current contact surface temperature is sufficient to induce the material to produce a change in viscoelastic properties.

[0035] Based on the difference between the real-time predicted surface temperature of the component and the preset rheological conversion temperature threshold, the corresponding predetermined static heat absorption time is matched in the preset thermal relaxation time mapping table, and the predetermined static heat absorption time is encapsulated in the thermal relaxation test instruction and sent to the photovoltaic cleaning robot.

[0036] Preferably, the real-time predicted value of the module surface temperature at the current location of the photovoltaic cleaning robot is calculated, including: The ambient temperature and solar irradiance values ​​are extracted from real-time environmental meteorological data. The product of the solar irradiance value and the preset photothermal conversion coefficient is calculated, and the product is added to the ambient temperature value to obtain the theoretical benchmark average temperature reflecting the photovoltaic module under fault-free conditions. Local temperature difference distribution feature values ​​are extracted from the component thermal distribution data, and these feature values ​​are defined as the temperature increment correction term of the target photovoltaic module in a specific hot spot area relative to the overall panel surface. The real-time predicted surface temperature of the module is obtained by summing the theoretical reference average temperature and the local temperature difference distribution characteristic value. The real-time predicted surface temperature of the module is equal to the theoretical reference average temperature plus the local temperature difference distribution characteristic value.

[0037] Preferably, the construction logic of the thermal environment determination model is as follows: Establish a reference sample set containing historical operation records of photovoltaic power plants. For each historical record, extract the corresponding historical ambient temperature value, historical solar irradiance value and historical module backsheet temperature monitoring value as model input samples. The difference obtained by subtracting the historical ambient temperature value from the historical module backsheet temperature monitoring value is defined as the photothermal temperature rise effect value. The historical solar irradiance values ​​in the reference sample set are divided into multiple continuous irradiance intensity intervals according to their numerical values. Within each irradiance intensity interval, the preset central trend value with the most concentrated distribution of photothermal temperature rise effect values ​​is selected. The ratio of the preset central trend value to the corresponding historical solar irradiance value is calculated and set as the benchmark photothermal conversion coefficient target value corresponding to the irradiance intensity interval. A linear regression function structure with solar irradiance values ​​as independent variables and photothermal conversion coefficient as dependent variable is constructed. The least squares method is used to iteratively fit the mapping relationship between historical solar irradiance values ​​and the benchmark photothermal conversion coefficient target value. By minimizing the mean square error between the function output value and the benchmark photothermal conversion coefficient target value, the environmental heat conduction slope and the basic thermal resistance intercept in the function structure are solved, thereby generating a thermally sensitive environment judgment model.

[0038] In one specific embodiment, the system first extracts a massive amount of full-cycle operation records from the historical operation and maintenance database of the photovoltaic power plant to establish a reference sample set. For each timestamped record in the sample set, the system extracts three core physical quantities: historical ambient temperature values. Historical solar irradiance values and the temperature of the photovoltaic module backsheet at that time The system defines the photothermal temperature rise effect value. The formula for calculating the pure temperature rise of the component under specific irradiation is as follows: .

[0039] Simultaneously, a binning statistical method was adopted: According to example The step size is divided into M consecutive irradiance intensity intervals. Within each interval, kernel density estimation is used to screen out... The value with the highest probability distribution is used as the preset central trend value. And calculate the target value of the baseline photothermal conversion coefficient for this interval. The unit is Subsequently, a system was constructed with solar irradiance G as the independent variable and the photothermal conversion coefficient as the input variable. Linear regression model with dependent variable Using the least squares method to... The slope of environmental heat conduction is obtained by iterative fitting of point pairs. and basic thermal resistance intercept ,in Characterizing the thermal saturation nonlinear decay rate induced by enhanced irradiation, Characterizes the inherent temperature rise efficiency under low irradiation. This represents the representative irradiance value in the historical irradiance binning statistics. During the model training phase, the system uses historical solar irradiance... The system is divided into multiple intervals with a fixed step size, and the average irradiance or center value of each interval is taken as the threshold. ; When the cloud server receives a request from the robot, it first sends the real-time solar irradiance values. Input the trained model above to dynamically calculate the photothermal conversion coefficient under the current operating conditions. Next, the theoretical baseline average temperature is calculated based on the law of conservation of energy. The calculation formula is: This value reflects the "healthy temperature" that the component surface should have when there are no internal circuit defects. This indicates the real-time ambient temperature, which is the ambient temperature value obtained from the meteorological sensor or data interface at the current moment.

[0040] At this point, the system introduces a local thermal correction factor from the digital twin data, namely the local temperature difference distribution characteristic value. The system performs the final superposition operation: Thus, the real-time predicted value of the component surface temperature is obtained. .

[0041] It should be added that the characteristic values ​​of local temperature difference distribution It is obtained through analysis and spatial modeling of historical infrared thermal imaging data of photovoltaic modules. Its core calculation logic is as follows: Retrieve historical infrared image sequences for the specific component area. Compare the measured temperature of each pixel in the infrared image with the theoretical health temperature model of the component under the same operating conditions (same irradiation and ambient temperature) to generate a residual thermal map. In the residual thermal map, identify localized overheating areas caused by cell cracks, microcracks, or solder ribbon aging. Typically, statistical values ​​(such as average temperature difference) of the temperature deviation of these abnormal areas relative to the surrounding normal areas are taken.

[0042] Finally, the predicted values With preset rheological conversion temperature threshold Compare them. These are physical constants determined based on dynamic thermomechanical analysis and testing of the wheel material, representing the critical point where the material's storage modulus begins to decrease significantly and viscous characteristics begin to dominate. If The system has confirmed that it has entered a high-thermosensitive excitation state. The system calculation exceeds the amplitude limit. The predetermined static heat absorption time is determined by consulting a preset thermal relaxation time mapping table. The mapping table follows a non-linear growth principle; for example, when... hour, =10s (lightly Microsoft-optimized, quick testing); when hour, =30s (significant softening, sufficient heat absorption is required to reach steady state); when hour, =60s (deep viscous flow state, requiring a long period of stillness to simulate the worst-case scenario). Finally, the system generates a stream containing this... The thermal relaxation test command for the parameters is sent to the robot.

[0043] In response to the thermal relaxation test command, after the photovoltaic cleaning robot completes the static heat absorption action, a pulse step excitation signal of predetermined amplitude is sent to the drive motor, and the transient current response waveform data under the action of the pulse step excitation signal is collected simultaneously.

[0044] Preferably, the logic for determining whether the photovoltaic cleaning robot has completed the static heat absorption action is as follows: The thermal relaxation test command is analyzed, and the predetermined static heat absorption time contained therein is extracted as the time domain judgment benchmark. The encoder feedback speed values ​​of the drive motors on the left and right sides are monitored in real time. When the encoder feedback speed value is detected to be continuously lower than the preset zero speed judgment threshold, the current system time is recorded as the static start time. The real-time static cumulative duration is calculated based on the current system real-time time and the static start time. The real-time static cumulative duration is compared with the predetermined static heat absorption duration. At the same time, it is monitored whether the cumulative displacement deviation value of the drive motor is kept within the preset mechanical lock-up tolerance range during this comparison period, so as to eliminate the interference of thermal contact surface offset caused by gravity sliding. When the cumulative real-time static duration reaches the predetermined static heat absorption duration, and the cumulative displacement deviation value does not exceed the preset mechanical locking tolerance range, it is confirmed that the photovoltaic cleaning robot has met the time and space stability conditions required for the local thermal softening of the walking wheel material, and a static heat absorption action completion signal is generated.

[0045] Preferably, the logic for determining the predetermined amplitude of the pulse step excitation signal in the pulse step excitation signal sent to the drive motor is as follows: The average real-time holding current of the drive motor of the photovoltaic cleaning robot during the static heat absorption action was statistically analyzed. This average real-time holding current was defined as the gravity compensation benchmark value to quantitatively characterize the static holding torque required for the robot to overcome the gravity component of the current slope. The real-time component surface temperature prediction value is retrieved, and the corresponding test incremental current value is obtained based on the preset temperature-viscosity mapping table. This test incremental current value is used to characterize the additional torque margin required to overcome the estimated adhesion force of the wheel material at different temperatures. The gravity compensation reference value and the test incremental current value are superimposed and summed, and the resulting sum is determined as the predetermined amplitude of the pulse step excitation signal.

[0046] In one specific embodiment, the predetermined static heat absorption time is obtained. The system then starts a microsecond-level timer. During this period, the system polls the feedback speed of the left and right motor encoders in real time at a period of 10ms. To eliminate the effects of sensor background noise and micro-vibration, the system sets a zero-speed detection threshold. The threshold is set based on the dead-zone characteristics of the motor servo system; only speeds below this threshold are physically considered "intended to stop." The system records speeds continuously below this threshold. The first frame time is the static start time. And continuously calculate the cumulative duration of real-time stillness. . Indicates the current moment.

[0047] Meanwhile, to prevent the robot from slipping slightly off the tilted photovoltaic panel due to gravity, a process that is barely perceptible to the naked eye, the system simultaneously monitors the cumulative displacement deviation. The calculation formula is as follows: This involves integrating the infinitesimal velocity during the period of rest. The system will... Compared with the preset mechanical locking tolerance range Compare them. A setting of 2mm to 5mm is recommended. The physical meaning of this parameter lies in encompassing the mechanical tooth backlash of the gearbox and the elastic deformation limit of the rubber wheel. If... This means that a macroscopic displacement has occurred, the thermal contact surface has shifted, and the system will reset the timing or report an error; only when the following conditions are met... The system generates a "stationary heat absorption action completion signal" only when both conditions are met.

[0048] Next, the system executes the dynamic determination logic for the amplitude of the pulse step excitation signal. This logic abandons the traditional fixed amplitude test method and instead adopts an adaptive superposition strategy of "gravity reference + viscous increment" to ensure that the excitation signal can overcome the current gravity load and has sufficient margin to detect the viscous state of the material.

[0049] The first step is to calculate the gravity compensation benchmark value. System backtracking to During this period of static heat absorption, the set of real-time current data output by the drive motor to maintain its position. And calculate their arithmetic mean: The necessity of using the mean value here is to filter out the high-frequency current ripple caused by PWM chopping. Physically, this is directly equivalent to the gravitational component of the downward motion of the robot at its current slope. ,in It is the motor torque constant, which is the baseline current to prevent the wheel from "falling off".

[0050] Physically, this is directly equivalent to the gravitational component of the downward motion of the robot at its current slope. The specific understanding is explained as follows: Imagine a solar-powered cleaning robot positioned on an inclined solar panel, say at a 20-degree angle. Gravity pulls the robot downwards. This downward pull creates a sliding torque on the axles. If the motors don't work, the robot will slide down like a slide. To remain stationary, the motors must output a holding torque that is opposite in direction and exactly equal in magnitude. When stationary, the motor torque equals the sliding torque caused by gravity. This is a case of equilibrium of two forces in physics.

[0051] The second step is to determine the incremental test current value. The system calls the real-time component surface temperature prediction value. This is input into a pre-defined temperature-viscosity mapping table. This mapping table is constructed based on materials rheology experiments and characterizes the additional triggering torque required at different temperatures to induce instantaneous peeling of the softened, adhered material of the wheel. The corresponding functional relationship is set as follows: Generally, this function exhibits an S-shaped growth trend: in the low-temperature region... The temperature is relatively small, but as the temperature rises into the glass transition range, Significantly increased to match the dramatic increase in viscous drag. Final composite amplitude. The calculation formula is: The unit is ampere.

[0052] The peak stripping current and phase lag time data at the moment of startup are extracted from the transient current response waveform data. These data are then weighted and summed according to preset material rheological weights to calculate the thermal aging coefficient of the walking wheel material corresponding to the photovoltaic cleaning robot. If the thermal aging coefficient still exceeds the standard after the first thermal relaxation test, the static heat absorption time is extended by 20%, and the pulse step excitation amplitude is reduced by 15%, and iterative tests are performed.

[0053] Preferably, the thermal aging coefficient of the walking wheel material corresponding to the photovoltaic cleaning robot is calculated, including: For transient current response waveform data, the maximum current amplitude within a preset monitoring time window after the application of the pulse step excitation signal is identified and defined as the stripped current peak data. Acquire encoder speed feedback waveform data that is recorded synchronously with transient current response waveform data, calculate the time difference between the rising edge trigger moment of the current waveform and the motion response moment of the speed waveform, and define it as phase lag time data; The viscosity strength factor is obtained by comparing the peak stripping current data with the preset reference peak current, and the flow hysteresis factor is obtained by comparing the phase lag time data with the preset reference response time. The viscosity strength factor and the flow hysteresis factor are then weighted and summed according to the preset material rheological weights to calculate the thermal aging coefficient of the walking wheel material corresponding to the photovoltaic cleaning robot.

[0054] The thermal aging coefficient is compared with the preset material failure benchmark value. When the thermal aging coefficient exceeds the preset material failure benchmark value, a fault diagnosis report indicating that the walking wheel has experienced thermal sticking failure is generated and output.

[0055] Time-domain feature extraction is performed based on the transient current response waveform data I(t) acquired through high-speed sampling and the encoder speed feedback waveform data v(t). The system sets a preset monitoring time window. Within this window, the system uses a peak search algorithm to identify the absolute maximum value of the current waveform, defining it as the extracted current peak data. The physical meaning of this parameter is that it represents the limiting torque current that the drive motor must output to overcome the "suction cup effect" caused by high temperature between the wheels and the component surface, as well as the static friction of the material itself.

[0056] Meanwhile, the system analyzes the hysteresis characteristics of the velocity waveform relative to the current waveform. Due to stress relaxation and creep phenomena in viscoelastic materials, there is a time difference between the application of current (force) and the generation of velocity (displacement). The system detects the moment when the rising edge of the current waveform reaches a preset trigger threshold. And the moment when the velocity waveform leaves the zero velocity region. Calculate the difference between the two. , defined as phase lag time data. This data directly reflects the delay in the untangling of polymer chain segments inside the material under stress. The longer the time, the closer the material is to a high-viscosity fluid state.

[0057] The thermal aging coefficient was then calculated. The calculation formula is constructed as follows: .

[0058] In the formula, The preset reference current peak value is based on the typical peak value of this robot model when subjected to the same pulse test on a standard photovoltaic panel at room temperature. This ratio item This constitutes the viscosity strength factor, used to quantify "how tightly it sticks"; The preset reference response time is determined based on the inherent delay of mechanical transmission at room temperature. This difference term... This constitutes the flow hysteresis factor, used to quantify "how slowly it moves"; This represents the viscous strength weight, a dimensionless parameter, with a preferred default value of 0.6. It is the flow lag weight, and its physical dimension is the reciprocal of a second.

[0059] Finally, the system will calculate the results. The material failure threshold is compared with a preset material failure benchmark. The material failure benchmark is usually set between 1.5 and 1.8. This threshold is a critical point determined based on a large number of destructive aging experiments, indicating that the material properties have undergone irreversible rheological changes.

[0060] It should be noted that the interval and threshold sizes are set for ease of comparison. The size of the threshold depends on the amount of sample data and the base number set by those skilled in the art for each set of sample data, as long as it does not affect the proportional relationship between the parameter and the quantized value. Furthermore, the above formulas are all dimensionless calculations, and the formulas are derived from software simulations using a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0061] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0062] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0063] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for remote monitoring and fault diagnosis of photovoltaic cleaning robots based on the Internet of Things, characterized in that, include: The system acquires the real-time operating current values ​​and encoder feedback speed values ​​of the drive motors on both sides of the photovoltaic cleaning robot, and calculates the asymmetry index of the resistance difference between the two sides of the photovoltaic cleaning robot based on the real-time operating current values ​​and encoder feedback speed values. Specifically, this includes: The system simultaneously collects the real-time operating current values ​​of the left and right drive motors of the photovoltaic cleaning robot and the corresponding encoder feedback speed values, and locks the data when it confirms that the encoder feedback speed values ​​on both sides are within the preset effective straight-line travel range. For each drive motor, the locked real-time operating current value is divided by its corresponding encoder feedback speed value to calculate the unit speed load value on the left and the unit speed load value on the right respectively. Calculate the ratio between the unit speed load value on the left and the unit speed load value on the right, and determine the degree of deviation between this ratio and the preset ideal balance reference value as the asymmetry index; The asymmetry index is compared with a preset anomaly screening threshold. When the asymmetry index exceeds the preset anomaly screening threshold, the cloud analysis server is requested to obtain component thermal distribution data and real-time environmental meteorological data associated with the current positioning coordinates of the photovoltaic cleaning robot. A thermal environment determination model is constructed based on component thermal distribution data and real-time environmental meteorological data. If the model determination result indicates that the current environment is in a high thermal excitation state, a thermal relaxation test instruction containing a predetermined static heat absorption time is generated and sent to the photovoltaic cleaning robot. In response to the thermal relaxation test command, after the photovoltaic cleaning robot completes the static heat absorption action, a pulse step excitation signal of predetermined amplitude is sent to the drive motor, and the transient current response waveform data under the action of the pulse step excitation signal is collected simultaneously. The peak stripping current and phase lag time data at the moment of startup are extracted from the transient current response waveform data. These data are then weighted and summed according to preset material rheological weights to calculate the thermal aging coefficient of the walking wheel material corresponding to the photovoltaic cleaning robot, including: For transient current response waveform data, the maximum current amplitude within a preset monitoring time window after the application of the pulse step excitation signal is identified and defined as the stripped current peak data. Acquire encoder speed feedback waveform data that is recorded synchronously with transient current response waveform data, calculate the time difference between the rising edge trigger moment of the current waveform and the motion response moment of the speed waveform, and define it as phase lag time data; The viscosity strength factor is obtained by comparing the peak stripping current data with the preset reference peak current, and the flow hysteresis factor is obtained by comparing the phase lag time data with the preset reference response time. The viscosity strength factor and the flow hysteresis factor are weighted and summed according to the preset material rheological weights to calculate the thermal aging coefficient of the walking wheel material corresponding to the photovoltaic cleaning robot. The thermal aging coefficient is compared with the preset material failure benchmark value. When the thermal aging coefficient exceeds the preset material failure benchmark value, a fault diagnosis report indicating that the walking wheel has experienced thermal sticking failure is generated and output.

2. The method for remote monitoring and fault diagnosis of photovoltaic cleaning robots based on the Internet of Things according to claim 1, characterized in that, Once the encoder feedback speed values ​​on both the left and right sides are confirmed to be within the preset effective straight-line driving range, data locking is performed, specifically including: The absolute value of the difference between the feedback speed values ​​of the left encoder and the right encoder is calculated in real time, and this absolute value of the difference is defined as the speed synchronization deviation value. Determine whether the speed synchronization deviation value is less than the preset straight-line driving tolerance threshold, and at the same time monitor whether the encoder feedback speed values ​​on both the left and right sides are greater than the preset minimum stable operating speed threshold. When the speed synchronization deviation value is less than the preset straight-line driving tolerance threshold and the encoder feedback speed value is greater than the preset minimum stable operating speed threshold, it is determined that the photovoltaic cleaning robot is currently in the preset effective straight-line driving range, and a data locking operation is triggered for the real-time operating current value and the encoder feedback speed value.

3. The method for remote monitoring and fault diagnosis of photovoltaic cleaning robots based on the Internet of Things according to claim 1, characterized in that, Generate thermal relaxation test instructions containing a predetermined static heat absorption duration, including: Real-time environmental meteorological data is used as the basic environmental thermal variable, and component thermal distribution data is used as the local thermal correction factor. The real-time component surface temperature prediction value at the current position of the photovoltaic cleaning robot is calculated through the thermal sensitive environment judgment model. The real-time component surface temperature prediction value is compared with the preset rheological conversion temperature threshold of the wheel material. When the real-time component surface temperature prediction value is greater than the preset rheological conversion temperature threshold, the judgment condition of the high thermosensitive excitation state is confirmed. Based on the difference between the real-time predicted surface temperature of the component and the preset rheological conversion temperature threshold, the corresponding predetermined static heat absorption time is matched in the preset thermal relaxation time mapping table, and the predetermined static heat absorption time is encapsulated in the thermal relaxation test instruction and sent to the photovoltaic cleaning robot.

4. The method for remote monitoring and fault diagnosis of photovoltaic cleaning robots based on the Internet of Things according to claim 3, characterized in that, The real-time predicted surface temperature of the photovoltaic modules at the current location of the photovoltaic cleaning robot is calculated, including: The ambient temperature and solar irradiance values ​​are extracted from real-time environmental meteorological data. The product of the solar irradiance value and the preset photothermal conversion coefficient is calculated, and the product is added to the ambient temperature value to obtain the theoretical benchmark average temperature reflecting the photovoltaic module under fault-free conditions. Local temperature difference distribution feature values ​​are extracted from the component thermal distribution data, and these feature values ​​are defined as the temperature increment correction term of the target photovoltaic module in a specific hot spot area relative to the overall panel surface. The real-time component surface temperature prediction value is obtained by summing the theoretical reference average temperature and the local temperature difference distribution characteristic value. The real-time component surface temperature prediction value is equal to the theoretical reference average temperature plus the local temperature difference distribution characteristic value.

5. The method for remote monitoring and fault diagnosis of photovoltaic cleaning robots based on the Internet of Things according to claim 3, characterized in that, The construction logic of the thermal environment determination model is as follows: Establish a reference sample set containing historical operation records of photovoltaic power plants. For each historical record, extract the corresponding historical ambient temperature value, historical solar irradiance value and historical module backsheet temperature monitoring value as model input samples. The difference obtained by subtracting the historical ambient temperature value from the historical module backsheet temperature monitoring value is defined as the photothermal temperature rise effect value. The historical solar irradiance values ​​in the reference sample set are divided into multiple continuous irradiance intensity intervals according to their numerical values. Within each irradiance intensity interval, the preset central trend value with the most concentrated distribution of photothermal temperature rise effect values ​​is selected. The ratio of the preset central trend value to the corresponding historical solar irradiance value is calculated and set as the benchmark photothermal conversion coefficient target value corresponding to the irradiance intensity interval. A linear regression function structure with solar irradiance values ​​as independent variables and photothermal conversion coefficient as dependent variable is constructed. The least squares method is used to iteratively fit the mapping relationship between historical solar irradiance values ​​and the benchmark photothermal conversion coefficient target value. By minimizing the mean square error between the function output value and the benchmark photothermal conversion coefficient target value, the environmental heat conduction slope and the basic thermal resistance intercept in the function structure are solved, thereby generating a thermally sensitive environment judgment model.

6. The method for remote monitoring and fault diagnosis of photovoltaic cleaning robots based on the Internet of Things according to claim 1, characterized in that, The logic for determining whether a photovoltaic cleaning robot has completed a stationary heat absorption action is as follows: The thermal relaxation test command is parsed, and the predetermined static heat absorption time contained therein is extracted as the time domain judgment benchmark. The encoder feedback speed values ​​of the drive motors on the left and right sides are monitored in real time. When the encoder feedback speed value is detected to be continuously lower than the preset zero speed judgment threshold, the current system time is recorded as the static start time. The real-time static cumulative duration is calculated based on the current system real-time time and the static start time, and then compared with the predetermined static heat absorption duration. At the same time, it is monitored whether the cumulative displacement deviation value of the drive motor remains within the preset mechanical lock-up tolerance range during this comparison period. When the cumulative real-time static duration reaches the predetermined static heat absorption duration, and the cumulative displacement deviation value does not exceed the preset mechanical locking tolerance range, it is confirmed that the photovoltaic cleaning robot has met the time and space stability conditions required for the local thermal softening of the walking wheel material, and a static heat absorption action completion signal is generated.

7. The method for remote monitoring and fault diagnosis of photovoltaic cleaning robots based on the Internet of Things according to claim 1, characterized in that, The logic for determining the predetermined amplitude of the pulse step excitation signal sent to the drive motor is as follows: The average real-time holding current of the drive motor of the photovoltaic cleaning robot during the static heat absorption action was statistically analyzed, and this average real-time holding current was defined as the gravity compensation benchmark value. The real-time component surface temperature prediction value is retrieved, and the corresponding test incremental current value is obtained based on the preset temperature-viscosity mapping table. This test incremental current value is used to characterize the additional torque margin required to overcome the estimated adhesion force of the wheel material at different temperatures. The gravity compensation reference value and the test incremental current value are superimposed and summed, and the resulting sum is determined as the predetermined amplitude of the pulse step excitation signal.

8. A remote monitoring and fault diagnosis system for photovoltaic cleaning robots based on the Internet of Things, characterized in that, It is implemented based on the IoT-based remote monitoring and fault diagnosis method for photovoltaic cleaning robots as described in any one of claims 1-7, including: Side wheel symmetry calculation module: acquires the real-time operating current values ​​and encoder feedback speed values ​​of the drive motors on the left and right sides of the photovoltaic cleaning robot, and calculates the asymmetry index of the resistance difference on the corresponding sides of the photovoltaic cleaning robot based on the real-time operating current values ​​and encoder feedback speed values; compares the asymmetry index with the preset abnormal screening threshold, and when the asymmetry index exceeds the preset abnormal screening threshold, requests the component thermal distribution data and real-time environmental meteorological data associated with the current positioning coordinates of the photovoltaic cleaning robot from the cloud analysis server; Current waveform acquisition module: Based on component thermal distribution data and real-time environmental meteorological data, a thermal environment judgment model is constructed. If the model judgment result indicates that the current environment is in a high thermal excitation state, a thermal relaxation test command containing a predetermined static heat absorption time is generated and sent to the photovoltaic cleaning robot. In response to the thermal relaxation test command, after the photovoltaic cleaning robot completes the static heat absorption action, a pulse step excitation signal of predetermined amplitude is sent to the drive motor, and the transient current response waveform data under the action of the pulse step excitation signal is collected simultaneously. Fault report generation module: Extracts peak stripping current data and phase lag time data at the moment of startup from transient current response waveform data, weights and sums the peak stripping current data and phase lag time data according to preset material rheological weights, calculates the thermal aging coefficient of the walking wheel material corresponding to the photovoltaic cleaning robot, compares the thermal aging coefficient with the preset material failure benchmark value, and generates and outputs a fault diagnosis report indicating that the walking wheel has experienced thermal viscous failure when the thermal aging coefficient exceeds the preset material failure benchmark value.

Images