Self-adaptive snow melting method and system for photovoltaic snow protection shed tunnel

Abstract

The invention relates to the technical field of intelligent control, and discloses a self-adaptive snow melting method and system for a photovoltaic snow protection shed tunnel, and the method comprises the steps: collecting the multi-mode real-time data of the surface of a photovoltaic panel; calling a digital twin physical model, and calculating a first heating control parameter corresponding to a theoretical energy demand required for melting the current accumulated snow; calling the neural network model to predict and obtain a second heating control parameter; performing dynamic weight fusion on the first heating control parameter and the second heating control parameter by using a weight factor to generate a final heating control parameter; and according to the final heating control parameter, driving a heating unit to perform snow melting operation on the photovoltaic panel, and updating the weight factor and / or the neural network model parameter based on the real-time feedback of the surface temperature of the photovoltaic panel. The objective of the invention is to overcome the cold start problem caused by high dependence on historical data and the defect of limited control precision and energy efficiency caused by lack of a precise physical model in the prior art.

Description

Technical Field

[0001] The present invention relates to the field of intelligent control technology, and in particular to an adaptive snow melting method and system for photovoltaic snow sheds based on digital twins and multimodal perception. Background Art

[0002] Some sections of highways experience sudden snowfalls in cold winters. Snow shelters are often installed to prevent the impact of natural factors such as accumulated snow and ice on road safety and the safety of drivers and passengers. Photovoltaic panels are installed on the exterior of the snow shelters to partially address the energy supply issues for the heating components. However, the efficiency of photovoltaic panels decreases significantly in winter due to snow accumulation. Therefore, the development of efficient and intelligent automatic snow melting technology is crucial to ensure the stable operation of photovoltaic snow shelters.

[0003] In related technical fields, some approaches attempt to achieve intelligent control by using historical data to predict future snow accumulation and adjust heating strategies accordingly. For example, Chinese patent application CN120434841A discloses a self-heating control method for photovoltaic snow shelters in highway tunnels. The method's core approach involves zoning photovoltaic panels, collecting real-time snow depth data, and then, based on historical data (data from multiple preset dates) and target snow removal volumes, obtaining an estimated snow thickness and an expected value. Finally, the heating power is adjusted by comparing the different expected values.

[0004] However, existing snowmelt control systems generally face the problem of a strong dependence on historical data. For a newly deployed system, in the absence of sufficient historical data accumulation (such as a full winter), the accuracy of such prediction models is difficult to guarantee, which is the so-called "cold start" problem. At the same time, interannual climate variability may also cause the representativeness of historical data to decrease, affecting the accuracy of control. In addition, existing control logic is mostly based on single macro-indicators such as snow thickness, and rarely considers the physical properties of snow that have a decisive impact on snowmelt energy consumption, such as density and phase (for example, the energy required to melt 10 cm of fluffy dry snow is very different from that required to melt 10 cm of wet heavy snow). This control method based on indirect indicators and fuzzy comparison has inherent limitations in achieving precise energy matching and maximizing energy efficiency, which may lead to insufficient heating or energy waste.

[0005] Therefore, how to develop an adaptive snow melting method that does not rely on a large amount of historical data, can accurately identify the physical properties of snow, and achieve precise energy control is a technical problem that needs to be solved urgently in this field. Summary of the Invention

[0006] The purpose of the present invention is to overcome the cold start problem caused by the high dependence on historical data in the existing technology and the defects of limited control accuracy and energy efficiency due to the lack of accurate physical models, and to provide a new photovoltaic snow shed tunnel adaptive snow melting method based on digital twin and neural network hybrid intelligence.

[0007] In order to solve the problems existing in the above-mentioned prior art, the technical solutions provided by the present invention include: The photovoltaic snowproof shed cave adaptive snow melting method includes: Collecting multimodal real-time data on the surface of the photovoltaic panel, wherein the multimodal real-time data includes at least snow thickness, photovoltaic panel surface temperature, and environmental parameters; Invoking a digital twin physical model built based on the laws of thermodynamics to calculate, based on the multimodal real-time data, a first heating control parameter corresponding to a theoretical energy demand required to melt the current snow; Invoking a pre-trained neural network model to predict a second heating control parameter based on the multimodal real-time data; Performing dynamic weighted fusion on the first heating control parameter and the second heating control parameter using a weight factor to generate a final heating control parameter; According to the final heating control parameter, the heating unit is driven to perform snow melting operation on the photovoltaic panel, and the weight factor and / or the neural network model parameter are updated based on the real-time feedback of the surface temperature of the photovoltaic panel.

[0008] Preferably, the dynamic weighted fusion satisfies the following formula: ; in, is the weight factor, is the final heating control parameter, and They are the first heating control parameter and the second heating control parameter respectively.

[0009] Preferably, the dynamic adjustment of the weight factor includes: During the cold start period, α was set to ≥ 0.95; During normal operation, α is set to any value in the range of 0.3-0.7 according to the confidence of the neural network prediction; When it is detected that the prediction deviation of the neural network model exceeds a first preset threshold, an abnormal calibration period is entered, and α is set to be greater than or equal to 0.8 until the prediction deviation is less than a second preset threshold.

[0010] Preferably, the pre-trained neural network model is a physical knowledge embedded neural network, and its loss function includes a physical rule residual term composed of thermodynamic equations.

[0011] Preferably, the step of calculating the theoretical energy requirement for melting the current snow further comprises: classifying the physical state of the current snow based on the environmental parameters, and estimating the snow density according to the classification result; The snow mass is calculated based on the estimated snow density and the snow thickness, and the theoretical energy requirement for melting the current snow is calculated.

[0012] Preferably, the theoretical energy requirement for melting the current snow includes: The heat of melting required to heat the snow from its current temperature to its melting point and complete the phase change; Ambient heat loss from photovoltaic panels to the surrounding environment during snow melting.

[0013] Preferably, the step of updating the neural network model parameters further comprises: Set a target snowmelt temperature curve; Continuously comparing the real-time feedback value of the photovoltaic panel surface temperature with the corresponding value on the target snowmelt temperature curve to obtain a temperature error; Based on the temperature error, the neural network model parameters are updated using an online learning algorithm.

[0014] The present invention also provides a photovoltaic snowproof shed and cave adaptive snow melting system, comprising: a multimodal sensing unit configured to collect multimodal real-time data on the surface of the photovoltaic panel, wherein the multimodal real-time data includes at least snow thickness, photovoltaic panel surface temperature, and environmental parameters; A data processing and decision-making unit is communicatively connected to the multimodal perception unit, and the data processing and decision-making unit includes: a physical model module configured to call a digital twin physical model constructed based on the laws of thermodynamics and calculate a first heating control parameter corresponding to a theoretical energy demand required to melt the current snow based on the multimodal real-time data; A neural network module is configured to call a pre-trained neural network model to predict a second heating control parameter based on the multimodal real-time data; a fusion control module configured to dynamically weight the first heating control parameter and the second heating control parameter using a weight factor to generate a final heating control parameter; a heating execution unit, communicatively connected to the data processing and decision-making unit, configured to receive the final heating control parameter and perform a heating and snow-melting operation on the surface of the photovoltaic panel; A feedback updating unit is communicatively connected to the multimodal sensing unit and the data processing and decision-making unit, and is configured to update the weighting factors and / or parameters of the neural network model based on real-time feedback of the surface temperature of the photovoltaic panel.

[0015] Beneficial effects 1. High reliability and precision: The thermodynamic model ensures the system can always make reliable decisions consistent with physical logic, completely resolving the "cold start" problem and providing a stable "safety net" for system decision-making. The continuous learning and fine-tuning of the neural network maximize the system's performance and energy efficiency, achieving control accuracy far exceeding that achievable with a single model.

[0016] 2. Superior Adaptability: Whether addressing the "cold start" challenges of a new deployment or complex field conditions caused by factors such as wind, obstruction, and aging equipment, the hybrid intelligent system of this invention rapidly adapts and continuously optimizes to its optimal state. Its dynamic weighted fusion mechanism ensures both high stability and strong adaptability to ever-changing field conditions.

[0017] 3. Significant Energy Savings: This invention minimizes energy waste through precise calculations and adaptive corrections. By identifying the physical properties of snow (such as density and phase) in real time and accurately calculating energy requirements based on first principles of physics, it achieves optimal energy efficiency while ensuring effective snow melting.

[0018] 4. Robustness and stability: The mutual calibration mechanism of the two models provides an additional layer of fault tolerance for the system, making it more stable in the face of sensor noise or extreme and unexpected weather patterns. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 A schematic flow chart of a method for adaptive snow melting in a photovoltaic snowproof shed cave provided in a preferred embodiment of the present invention; Figure 2 This is a schematic structural diagram of a photovoltaic snow shed and cave adaptive snow melting system provided in a preferred embodiment of the present invention. DETAILED DESCRIPTION

[0020] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0021] Example 1 This embodiment provides an adaptive snow-melting method for photovoltaic snow shelters based on digital twins and multimodal sensing. In a specific application scenario, the method is executed by a central controller deployed on-site at the photovoltaic snow shelter. This central controller is connected to a multimodal sensing unit (including but not limited to an ultrasonic thickness sensor, a patch temperature sensor, a temperature and humidity sensor, an anemometer, etc.) and one or more heating units (for example, electric heating films or heating wires built into the back of the photovoltaic panel).

[0022] like Figure 1 As shown, the photovoltaic snowproof shed cave adaptive snow melting method provided in this embodiment may specifically include the following steps: S1. Collect multimodal real-time data on the surface of photovoltaic panels, where the multimodal real-time data includes at least snow thickness, photovoltaic panel surface temperature, and environmental parameters.

[0023] In this step, data is collected periodically (for example, once every minute) via a multimodal sensing unit. Snow thickness can be measured by ultrasonic or laser ranging sensors deployed at various locations on the photovoltaic panels. Panel surface temperature is measured by an array of thermocouples or thermistors tightly attached to the back of the panels, providing critical feedback for subsequent closed-loop control. Environmental parameters, including at least temperature and humidity, and preferably also wind speed and solar radiation intensity, are collected by a small weather station. This data forms the basis for subsequent dual-model intelligent decision-making.

[0024] S2. Call a digital twin physical model built based on the laws of thermodynamics, and calculate a first heating control parameter corresponding to the theoretical energy demand required to melt the current snow based on the multimodal real-time data.

[0025] The digital twin physical model is pre-built offline based on the inherent physical properties of the photovoltaic panel, such as material properties (for example, the thermal conductivity of the glass cover and the specific heat capacity of the backsheet), geometric dimensions, and parameters such as the rated power and spatial distribution of the built-in heating units. The core function of this model is to provide a theoretical benchmark value that is stable and reliable under all operating conditions.

[0026] Specifically, based on fundamental thermodynamic laws (such as the law of conservation of energy and Fourier's law of heat conduction), this physical model accurately simulates the temporal evolution of the photovoltaic panel surface temperature field under given environmental conditions when a specific power input is applied to the heating unit. More importantly, it can calculate the theoretical total energy (Q_total) required to melt a specific amount of snow. In one embodiment, this total energy can be expressed as: Qtotal = Qmelt + Qloss.

[0027] Among them, Q_melt is the melting heat, that is, the energy required to heat the snow from the current temperature to the melting point and complete phase change; Q_loss is the environmental heat loss, that is, the energy lost by the photovoltaic panel to the surrounding environment due to convection, radiation, etc. during the snow melting process.

[0028] The calculation formula for Q_melt is: Calculate the required heat. Where m is the total mass of snow, T_melt is the melting point of snow and ice (default is 0°C), and c is the specific heat capacity of snow. is the current snow temperature (which can be approximated as the board surface temperature or ambient temperature), is the latent heat of melting of ice (approximately 3.34×10^5 J / kg).

[0029] The calculation of Q_loss leverages the simulation capabilities of the digital twin model. Real-time ambient temperature, wind speed, and other parameters are input into the model. The model then uses a built-in heat transfer model (for example, combining Newton's law of cooling and the Stefan-Boltzmann law) to simulate and calculate heat losses during snowmelt, including heat dissipated from the photovoltaic panels to the surrounding environment through convection, radiation, and other means. The specific calculation method can be reasonably designed by those skilled in the art based on existing technologies and common knowledge.

[0030] Finally, based on the calculated total energy demand Q_total and the system's preset snow melting time target t (for example, the snow melting is expected to be completed within 30 minutes), the first heating control parameter P1 can be determined. It can be a power value (unit: watt) and the calculation formula is P1 = Q_total / t.

[0031] In some preferred embodiments, in order to more accurately calculate the melting heat Q_melt, the method further includes identifying the physical state of snow and estimating its density. The specific steps are as follows: S21. Based on the environmental parameters, the physical state of the current snow is classified, and the snow density is estimated based on the classification results. The central controller runs a built-in snow physical property identification model. The model classifies the current snow into dry snow, wet snow, or ice / freezing rain based on the ambient temperature and humidity data collected in S1. For example, a rule-based classifier can be set: when the ambient temperature is lower than -5°C, it is judged as "dry snow" and a lower density value is assigned to it from the built-in physical parameter library, such as 100 kg / m 3 When the ambient temperature is between -2°C and 1°C and the humidity is greater than 90%, it is considered "wet snow" and assigned a higher density value, such as 300 kg / m 3 When the ambient temperature is continuously below 0°C and the snow is solid, it is considered "ice" and assigned a higher density value, such as 900 kg / m 3 .

[0032] S22. Calculate the mass of snow based on the estimated snow density and snow thickness, and calculate the theoretical energy requirement required to melt the current snow.

[0033] Based on the snow thickness measured by the sensor in S1, the effective area of ​​the photovoltaic panel, and the snow density estimated in S21, the total mass m of snow covering the photovoltaic panel is calculated. This snow mass m is then substituted into the above Qmelt calculation formula to obtain a more accurate melting heat requirement.

[0034] S3. Call a pre-trained neural network model to predict a second heating control parameter based on the multimodal real-time data.

[0035] The role of this neural network model is to supplement the physical model, learning and compensating for complex nonlinear influencing factors (such as local wind field disturbances, snow unevenness, equipment aging, etc.) that are difficult to be accurately described by ideal physical models in a data-driven manner.

[0036] Considering that it is often difficult to obtain a large number of labeled training samples in actual deployment, this embodiment preferably adopts a pre-training method optimized for few-sample learning. In a preferred embodiment, the neural network model is a physics-informed neural network (PINN). The core idea is to add a physical rule residual term composed of thermodynamic equations (such as heat conduction equations) to the loss function of the neural network in addition to the conventional prediction error term (i.e., the difference between the model prediction value and the true value). This is equivalent to forcing the output of the neural network to comply with the basic laws of physics during the training process, thereby greatly reducing the dependence on training data, solving the "cold start" problem, and significantly improving the generalization ability of the model. The loss function L can be expressed as: L = L_data + λ * L_physics, where L_data is the data prediction error, L_physics is the physical rule residual term, and λ is the error weight coefficient.

[0037] This model receives the same multimodal real-time data as the physical model and outputs a second heating control parameter, P2. P2 is a corrected or empirically predicted heating power value based on complex, nonlinear relationships learned from the data that cannot be fully described by the ideal physical model (such as the influence of local wind field disturbances and uneven snow cover).

[0038] S4. Use a weight factor to dynamically weight the first heating control parameter and the second heating control parameter to generate a final heating control parameter.

[0039] This step is designed to organically combine the high reliability of the physical model with the high adaptability of the neural network model. By dynamically adjusting weights, the system can prioritize more reliable model outputs at different operating stages and conditions, achieving a balance between stability and accuracy.

[0040] In some preferred embodiments, the dynamic weighted fusion can be a linear weighted fusion, whose formula is: P_final = α * P1 + (1-α) * P2. Wherein, P_final is the heating control parameter ultimately output to the heating unit, and α is a dynamically changing weight factor, with a value range between [0, 1].

[0041] The dynamic adjustment strategy of the weight factor α is designed to cover the entire life cycle of the system operation, as follows: Cold Start: When the system is first deployed or restarted after a prolonged shutdown, the neural network model may not be reliable because it has not yet undergone sufficient online learning. In this case, the system automatically enters a cold start phase, setting the weight factor α to a high value close to 1, such as 0.95, 0.98, or 1.0. This ensures that the final decision P_final is almost entirely determined by the reliable physical model P1, ensuring stability and security during the initial system startup.

[0042] Normal Operation: As the system continues to operate and online learning proceeds, the neural network model fine-tunes itself by continuously absorbing real-time data, gradually improving its prediction accuracy and confidence. At this point, the system enters its normal operation phase. The weight factor α ranges from 0.3 to 0.7 and is dynamically adjusted based on the confidence level of the neural network's predictions. For example, when the model outputs a high confidence score, indicating high confidence in its predictions, α can be set to a lower value, such as 0.35, to leverage more of the neural network's experience. When confidence is low, α can be set to a higher value, such as 0.65, to rely more on theoretical calculations based on the physical model.

[0043] Abnormal calibration period: During operation, the deviation between P1 and P2 is continuously monitored. If the deviation exceeds a preset first threshold (for example, the absolute value of the deviation is greater than 20% of P1), it indicates that there is a significant difference in the outputs of the two models. This may mean that the neural network has encountered an extreme operating condition that it has never seen before, and its prediction may no longer be reliable. At this time, the system automatically enters the abnormal calibration period and temporarily sets α to a high value of not less than 0.8 (for example, 0.8 or 0.9), forcing the system to return to a more stable physical model-dominated mode to achieve "safe fallback". This state will remain until the prediction deviation of the two models returns to less than a stricter second preset threshold (for example, the deviation is less than 10% of P1), and the system will return to normal operation.

[0044] S5. According to the final heating control parameter, drive the heating unit to melt snow on the photovoltaic panel, and update the weight factor and / or the neural network model parameter based on the real-time feedback of the photovoltaic panel surface temperature.

[0045] This step establishes a complete closed-loop feedback control system. The central controller sends the calculated final heating control parameter P_final (for example, a specific power value) to the driver circuit of the heating unit to drive it.

[0046] At the same time, the system uses the real-time feedback of the plate surface temperature for online learning and optimization. In a preferred embodiment, the updating step of the neural network model parameters specifically includes: S51. Set a target snowmelt temperature curve. This curve defines the ideal heating and snowmelt process. For example, the system expects the panel surface temperature to increase linearly from -5°C to +2°C within 15 minutes and then maintain this temperature for 5 minutes. This curve can be preset or dynamically adjusted based on actual needs and environmental conditions.

[0047] S52: Continuously compare the real-time feedback value of the photovoltaic panel surface temperature with the corresponding value on the target snowmelt temperature curve to obtain a temperature error.

[0048] S53. Based on the temperature error, update the neural network model parameters (such as weights and biases) using an online learning algorithm. For example, a gradient descent-based fine-tuning algorithm or a Kalman filter algorithm may be used. This process enables the neural network to continuously learn from actual control results and continuously adapt to the unique characteristics of the current environment, thereby continuously reducing its subsequent prediction errors. Optionally, the temperature error can also be used to optimize the dynamic adjustment strategy of the weight factor α (such as the confidence-α mapping relationship during normal operation).

[0049] In another embodiment, to achieve more refined control, the photovoltaic panel surface is physically or logically divided into multiple independent control zones. For example, based on snow distribution patterns, the photovoltaic panel surface is divided into top, middle, and bottom zones. Each zone is equipped with an independent sensor set (for measuring snow thickness and surface temperature) and independently controllable heating units. In this case, the adaptive snow melting method described in Example 1 (i.e., steps S1 to S5) is executed independently and in parallel for each zone. The central controller maintains an independent model calculation and decision-making process for each zone, outputting the optimal final heating control parameters for that zone. This system enables highly refined zone-specific control. For example, when only the top zone has heavy snow, the system will drive only the heating units in the top zone at higher power, while the middle and bottom zones may operate at lower power or not at all. This maximizes energy efficiency while ensuring effective snow melting.

[0050] Example 2 like Figure 2 As shown, the photovoltaic snowproof shed and cave adaptive snow melting system provided in this embodiment specifically includes: a multimodal sensing unit configured to collect multimodal real-time data on the surface of the photovoltaic panel, wherein the multimodal real-time data includes at least snow thickness, photovoltaic panel surface temperature, and environmental parameters; A data processing and decision-making unit is communicatively connected to the multimodal perception unit, and the data processing and decision-making unit includes: a physical model module configured to call a digital twin physical model constructed based on the laws of thermodynamics and calculate a first heating control parameter corresponding to a theoretical energy demand required to melt the current snow based on the multimodal real-time data; A neural network module is configured to call a pre-trained neural network model to predict a second heating control parameter based on the multimodal real-time data; a fusion control module configured to dynamically weight the first heating control parameter and the second heating control parameter using a weight factor to generate a final heating control parameter; a heating execution unit, communicatively connected to the data processing and decision-making unit, configured to receive the final heating control parameter and perform a heating and snow-melting operation on the surface of the photovoltaic panel; A feedback updating unit is communicatively connected to the multimodal sensing unit and the data processing and decision-making unit, and is configured to update the weighting factors and / or parameters of the neural network model based on real-time feedback of the surface temperature of the photovoltaic panel.

[0051] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art will appreciate that they may modify the technical solutions described in the foregoing embodiments or substitute equivalents for some of the technical features. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. Photovoltaic snow shed tunnel adaptive snow melting method, characterized in that: include: Collecting multimodal real-time data on the surface of the photovoltaic panel, wherein the multimodal real-time data includes at least snow thickness, photovoltaic panel surface temperature, and environmental parameters; Invoking a digital twin physical model built based on the laws of thermodynamics to calculate, based on the multimodal real-time data, a first heating control parameter corresponding to a theoretical energy demand required to melt the current snow; Invoking a pre-trained neural network model to predict a second heating control parameter based on the multimodal real-time data; Performing dynamic weighted fusion on the first heating control parameter and the second heating control parameter using a weight factor to generate a final heating control parameter; According to the final heating control parameter, the heating unit is driven to perform snow melting operation on the photovoltaic panel, and the weight factor and / or the neural network model parameter are updated based on the real-time feedback of the surface temperature of the photovoltaic panel.

2. The photovoltaic snowproof shed and cave adaptive snow melting method according to claim 1 is characterized in that: The dynamic weighted fusion satisfies the following formula: ; in, is the weight factor, is the final heating control parameter, and They are the first heating control parameter and the second heating control parameter respectively.

3. The photovoltaic snowproof shed cave adaptive snow melting method according to claim 2 is characterized in that: The dynamic adjustment of the weight factor includes: During the cold start period, α was set to ≥ 0.95; During normal operation, α is set to any value in the range of 0.3-0.7 according to the confidence of the neural network prediction; When it is detected that the prediction deviation of the neural network model exceeds a first preset threshold, an abnormal calibration period is entered, and α is set to be greater than or equal to 0.8 until the prediction deviation is less than a second preset threshold.

4. The photovoltaic snowproof shed and cave adaptive snow melting method according to claim 1 is characterized in that: The pre-trained neural network model is a physical knowledge embedded neural network, and its loss function includes a physical rule residual term composed of thermodynamic equations.

5. The photovoltaic snowproof shed cave adaptive snow melting method according to claim 1 is characterized in that: The steps of calculating the theoretical energy requirement for melting the current snowpack further include: classifying the physical state of the current snow based on the environmental parameters, and estimating the snow density according to the classification result; The snow mass is calculated based on the estimated snow density and the snow thickness, and the theoretical energy requirement for melting the current snow is calculated.

6. The photovoltaic snowproof shed and cave adaptive snow melting method according to claim 5 is characterized in that: The theoretical energy requirements for melting the current snow include: The heat of melting required to heat the snow from its current temperature to its melting point and complete the phase change; Ambient heat loss from photovoltaic panels to the surrounding environment during snow melting.

7. The photovoltaic snowproof shed and cave adaptive snow melting method according to claim 1 or 4, characterized in that: The step of updating the neural network model parameters further includes: Set a target snowmelt temperature curve; Continuously comparing the real-time feedback value of the photovoltaic panel surface temperature with the corresponding value on the target snowmelt temperature curve to obtain a temperature error; Based on the temperature error, the neural network model parameters are updated using an online learning algorithm.

8. Photovoltaic snowproof shed and cave adaptive snow melting system, characterized by: include: a multimodal sensing unit configured to collect multimodal real-time data on the surface of the photovoltaic panel, wherein the multimodal real-time data includes at least snow thickness, photovoltaic panel surface temperature, and environmental parameters; A data processing and decision-making unit is communicatively connected to the multimodal perception unit, and the data processing and decision-making unit includes: a physical model module configured to call a digital twin physical model constructed based on the laws of thermodynamics and calculate a first heating control parameter corresponding to a theoretical energy demand required to melt the current snow based on the multimodal real-time data; A neural network module is configured to call a pre-trained neural network model to predict a second heating control parameter based on the multimodal real-time data; a fusion control module configured to dynamically weight the first heating control parameter and the second heating control parameter using a weight factor to generate a final heating control parameter; a heating execution unit, communicatively connected to the data processing and decision-making unit, configured to receive the final heating control parameter and perform a heating and snow-melting operation on the surface of the photovoltaic panel; A feedback updating unit is communicatively connected to the multimodal sensing unit and the data processing and decision-making unit, and is configured to update the weighting factors and / or parameters of the neural network model based on real-time feedback of the surface temperature of the photovoltaic panel.

Images