Photovoltaic steam self-cleaning system based on multi-source sensing and waste heat cooperation and control method

Abstract

The invention belongs to the technical field of photovoltaic power generation system operation and maintenance, and discloses a photovoltaic steam self-cleaning system based on multi-source sensing and waste heat cooperation. Comprising an inverter efficiency monitoring module, a visual AI stain identification module, an intelligent decision controller, a multi-heat-source cooperative system, a steam generation and low-pressure safety system, a thermal shock protection water mixing device, an automatic spray gun and a cleaning execution unit. According to the system, through inverter efficiency monitoring and visual AI fusion recognition, stains and shadows are accurately distinguished in combination with the attenuation rate, and the cleaning decision accuracy is improved; multi-parameter safety interlocking such as temperature, humidity and wind speed is integrated, and the risk of severe weather operation is eradicated; a multi-heat-source cooperative system is built by innovatively utilizing heating and ventilation, energy storage and waste heat of an inverter, and the dependence of traditional electric heating is thoroughly eliminated through MPC algorithm scheduling; water mixing cooling and closed-loop temperature control effectively limit temperature rise of the panel, thermal shock is avoided fundamentally, and the service life of the assembly is prolonged.

Description

Technical Field

[0001] This invention belongs to the field of photovoltaic power generation system operation and maintenance technology, specifically a photovoltaic steam self-cleaning system and control method based on multi-source sensing and waste heat synergy. Background Technology

[0002] With the widespread deployment of distributed photovoltaic (PV) power stations, the problem of reduced power generation efficiency caused by the accumulation of dirt such as dust, bird droppings, and snow on the surface of PV modules is becoming increasingly prominent. Regular cleaning is a key measure to ensure the power generation benefits of the system. Currently, the industry mainly offers the following cleaning solutions, but each has its own limitations: 1. Traditional timed or manual cleaning mode: This model typically involves cleaning based on a fixed schedule (such as weekly or monthly) or relying on decisions made after manual on-site inspections. Main shortcomings: ① Low level of intelligence: It cannot accurately perceive the actual severity of stains and environmental impact, which can easily lead to "over-cleaning" or "under-cleaning". The former results in waste of water and electricity resources, while the latter results in loss of power generation; ② Lack of safety guarantee: It may still be forced to operate in severe weather conditions such as strong winds, low temperatures, and high humidity, which poses safety risks such as support swaying, pipe freezing and cracking, and personnel working at height; ③ High operation and maintenance costs: Manual cleaning is expensive and inefficient. 2. Automated cleaning solution based on a single sensor: Existing technologies (such as CN114897918B and CN117593675B) mainly identify stains through visual images or trigger cleaning by monitoring power attenuation through a single power generation efficiency sensor. Main shortcomings: ① Poor recognition accuracy: Single visual recognition is prone to misjudgment under backlight and shadow conditions; a single efficiency sensor cannot distinguish between power fluctuations caused by momentary shadow occlusion and power loss caused by long-term dirt accumulation, leading to false triggering or missed triggering; ② Lack of environmental safety interlock: The cleaning trigger logic does not integrate key environmental parameters such as temperature, wind speed, and humidity, and cannot automatically disable in extreme weather conditions, resulting in a high risk of equipment damage; ③ Limited cleaning methods: Mostly mechanical brushes or high-pressure water guns are used, which have limited cleaning effect and may scratch components, and are not coordinated with the building energy system; 3. Existing solutions for steam cleaning systems: Electric heating steam system: It directly uses the electricity generated by the photovoltaic system to heat and produce steam. It has huge energy consumption, usually consuming 25%-40% of the system's own power generation, which seriously offsets the power generation gain brought by cleaning. It is not economical and does not conform to the zero carbon concept. Engine waste heat drive solution (e.g., CN104096691B): Utilizes waste heat from industrial engines to generate steam; main shortcomings: ① Limited application scenarios: Only applicable to industrial power plants equipped with large engines, unable to adapt to the mainstream distributed photovoltaic scenarios such as industrial and commercial rooftops and park buildings; ② Lack of intelligent control: Mostly uses timed or manual triggering, failing to solve the problems of accurate stain identification and safe triggering; ③ Lack of thermal shock protection: Does not consider the risk of "thermal shock" caused by the huge temperature difference (ΔT can reach over 120℃) between high-temperature steam and low-temperature photovoltaic glass panels, which may cause the glass to crack; 4. Existing solutions for building waste heat recovery: Existing technologies (such as CN208475598U) mainly focus on the heat recovery of air conditioning condensate for internal building circulation or flushing, but they do not establish effective synergy with photovoltaic cleaning systems, resulting in an incomplete heat utilization chain and seasonal limitations (such as the absence of heat source when the HVAC system is shut down in winter). 5. Existing solutions for thermal management of energy storage systems: The waste heat generated during the operation of energy storage systems is usually discharged directly into the environment through air cooling or liquid cooling, resulting in energy waste; existing technologies have not combined this stable low-grade heat source with the needs of photovoltaic cleaning, thus failing to achieve the cascade utilization of energy. In summary, existing photovoltaic cleaning technologies generally suffer from the following common problems: inaccurate stain identification leading to decision-making errors; neglecting environmental safety conditions during cleaning, resulting in high risks; high energy consumption during the cleaning process, especially the poor economic efficiency of electric heating steam solutions; isolated operation of various energy systems (photovoltaics, energy storage, and HVAC), with waste heat not being utilized in a coordinated manner; and a lack of effective thermal shock protection mechanisms, which may damage expensive photovoltaic modules. Therefore, there is an urgent need for a highly efficient, energy-saving, and reliable photovoltaic self-cleaning solution that integrates intelligent sensing, safe decision-making, energy coordination, and safety protection. Summary of the Invention

[0003] The purpose of this invention is to provide a photovoltaic steam self-cleaning system and control method based on multi-source sensing and waste heat synergy, so as to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a photovoltaic steam self-cleaning system based on multi-source sensing and waste heat synergy, comprising: an inverter efficiency monitoring module, a visual AI stain recognition module, an intelligent decision controller, a multi-heat source synergy system, a steam generation and low-pressure safety system, a thermal shock protection mixing device, and an automatic spray gun and cleaning execution unit; the inverter efficiency monitoring module reads inverter data in real time through a Modbus TCP interface, calculates theoretical power generation, efficiency decay, and decay time derivative, and distinguishes between shadows and stains; the visual AI module acquires component images and detects stains using a YOLOv9 / v12 model, outputting coverage and a heat map; the intelligent decision controller triggers cleaning through multi-condition AND logic; the multi-heat source synergy system generates steam using HVAC condensation heat, energy storage heat dissipation, and inverter heat dissipation; the thermal shock protection device avoids thermal damage through mixing and closed-loop control; and the automatic spray gun cleans precisely according to the heat map.

[0005] Preferably, the theoretical power generation calculation formula of the inverter efficiency monitoring module is as follows: In the formula Photovoltaic rated peak power, Horizontal irradiance, With a temperature coefficient of 0.0045, To measure the component temperature, This is the correction factor for the incident angle of the inclined surface, with a value range of [0.85, 1.0].

[0006] Preferably, the visual AI stain recognition module is equipped with a camera with IP67 protection rating, ≥2MP resolution and FOV≥90°, which captures one frame of image every 30 seconds; after grayscale and normalization preprocessing, it is input into the model, with mAP@0.5≥95% and inference latency <200ms / frame; the stain coverage rate C is calculated with an accuracy of ±5%, and is divided into three levels: normal, moderate and severe according to C<10%, 10%-30% and ≥30%.

[0007] Preferably, the multi-condition AND logic of the intelligent decision controller includes: AI stain coverage C>10%; efficiency decay percentage DP%>5%; decay time derivative dDP / dt<1.0; ambient temperature 0℃≤T≤35℃, wind speed V≤8m / s, relative humidity RH≤85%; time since last cleaning >4 hours; when all conditions are met simultaneously, a cleaning trigger signal is output; otherwise, a negative signal is returned or the device is added to the waiting queue.

[0008] Preferably, the multi-heat source collaborative system includes three heat sources: HVAC condensing heat Q_HVAC with a power of 15-25kW, energy storage heat dissipation Q_ESS with a power of 3-6kW, and inverter heat dissipation Q_INV with a power of 1-4kW. The multi-heat source collaborative system uses the MPC algorithm for heat source scheduling, with a prediction window of 10 minutes. The optimal mixing ratio is solved through quadratic programming, and the optimization objective is to minimize heat waste and achieve a heat exchange efficiency of ≥85%.

[0009] Preferably, the steam generated by the steam generation and low-pressure safety system has the following parameters: pressure 0.1-0.2 MPa, temperature 100-120℃, and flow rate 5-10 kg / h. The steam generation and low-pressure safety system is equipped with a pressure gauge, a safety valve with a set pressure of 0.25 MPa, and a quality monitoring system. It automatically releases pressure when overpressure occurs to ensure compliance with low-pressure safety standards.

[0010] Preferably, the thermal shock protection mixing device uses an electric three-way mixing valve to mix steam above 100°C with coolant at around 25°C. After mixing, the steam temperature is ≤60°C, and the temperature difference between the steam and the cold photovoltaic panel is ≤70°C. The closed-loop control uses first-order derivative adjustment, with the panel temperature rise rate ≤5°C / minute and the absolute temperature rise ≤20°C. If either limit is exceeded, the cleaning will stop immediately and the cooling mode will be activated.

[0011] Preferably, the automatic spray gun and cleaning execution unit have an automatic scanning mechanism, which prioritizes cleaning high-coverage areas based on the 8×8 grid heat map of stain distribution; dynamically adjusts the number of spray guns turned on by 30%-70% according to the temperature rise rate, with a cleaning cycle of 8-12 minutes, and automatically switches to cooling mode after completion.

[0012] A photovoltaic steam control method based on multi-source sensing and waste heat synergy, the specific steps of which are as follows: Step 1: Parameter Acquisition: Acquire inverter, environmental, and heat source data every minute, and acquire component images every 30 seconds; Step 2, Data Processing: Calculate efficiency decay and derivative, distinguish between shadows and stains, generate stain coverage and heat map, and report all data to AIHUB database; Step 3, Decision Judgment: Determine whether the five conditions described in claim 4 are met. If they are met, trigger cleaning; otherwise, return to the parameter acquisition stage and wait for the next cycle. Step 4: Heat source scheduling: Determine the heat source mixing ratio using the MPC algorithm to generate qualified steam; Step 5: Cleaning Execution: Record the initial temperature of the photovoltaic panel. The spray gun is activated to inject mixed steam; closed-loop control is executed once per second: the current temperature on the panel is detected, and if it exceeds... If the temperature reaches +20℃, immediately stop cleaning; calculate the temperature rise rate, and if it exceeds 5℃ / minute, reduce the number of spray guns by 30%; check the steam temperature, and if it exceeds 60℃, increase the coolant mixing ratio; Step 6, Cooling: After cleaning, turn off the steam supply and turn on the coolant circulation pump; when the panel temperature drops to within 5°C of the ambient temperature and the coolant temperature is <30°C, turn off the cooling pump, record the cleaning timestamp and power generation recovery effect, and report to AIHUB. Step 7, Waiting for the loop: Set the next cleaning interval, return to the parameter acquisition stage and enter the next loop.

[0013] Preferably, the steam generation in step four requires heating the preheated mixture from multiple heat sources for 5-10 minutes, and the heating process is achieved through segmented temperature control.

[0014] The beneficial effects of this invention are as follows: By employing inverter efficiency monitoring and YOLOv9 / v12 visual AI multi-source fusion recognition, combined with attenuation rate judgment, the system accurately distinguishes between stains and shadows, significantly improving the accuracy of cleaning decisions and avoiding false triggers and missed triggers. It introduces multi-parameter safety interlock logic based on temperature, wind speed, and humidity to eliminate risky operations in severe weather, ensuring the safety of equipment and personnel. It innovatively integrates HVAC condensation heat, energy storage, and inverter heat dissipation to construct a multi-heat source collaborative system. Through MPC algorithm optimization scheduling, it completely eliminates the dependence of traditional electric heating on photovoltaic power, significantly improving energy utilization efficiency. By utilizing mixed-water cooling and closed-loop temperature rise control, it limits the panel temperature rise rate and absolute temperature difference, fundamentally solving the problem of thermal shock cracking and extending module lifespan. This solution is suitable for distributed scenarios such as commercial and industrial rooftops and park buildings, and can be seamlessly integrated with existing energy management systems. It combines intelligence, safety, energy saving, and compatibility, comprehensively optimizing the operation and maintenance effects and benefits of photovoltaic systems. Attached Figure Description

[0015] Figure 1 This is a system block diagram of the present invention; Figure 2 This is a flowchart of the method of the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] like Figure 1 and Figure 2As shown, this embodiment of the invention provides a photovoltaic steam self-cleaning system based on multi-source sensing and waste heat synergy, including: an inverter efficiency monitoring module, a visual AI stain recognition module, an intelligent decision controller, a multi-heat source synergy system, a steam generation and low-pressure safety system, a thermal shock protection mixing device, an automatic spray gun, and a cleaning execution unit; the inverter efficiency monitoring module reads inverter data in real time through a Modbus TCP interface, calculates theoretical power generation, efficiency decay, and decay time derivative, and distinguishes between shadows and stains; the visual AI module collects component images and detects stains using a YOLOv9 / v12 model, outputting coverage and a heat map; the intelligent decision controller triggers cleaning through multi-condition AND logic; the multi-heat source synergy system generates steam using HVAC condensation heat, energy storage heat dissipation, and inverter heat dissipation; the thermal shock protection device avoids thermal damage through mixing and closed-loop control; and the automatic spray gun cleans precisely according to the heat map.

[0018] Combining power generation efficiency change rate with AI visual recognition, it effectively distinguishes between stains and shadows, resulting in high accuracy in cleaning decisions. It introduces multi-parameter environmental safety interlock logic to prevent risky operations in adverse weather conditions, ensuring the safety of equipment and personnel. It innovatively utilizes waste heat from existing HVAC and energy storage systems within the building as a steam heat source, significantly reducing or even eliminating the direct consumption of photovoltaic power during the cleaning process and improving the overall energy efficiency of the system. Through mixed-water cooling and closed-loop temperature rise control, it fundamentally solves the problem of photovoltaic glass thermal shock cracking that may occur during steam cleaning, extending the lifespan of the modules.

[0019] The theoretical power generation calculation formula for the inverter efficiency monitoring module is as follows: In the formula Photovoltaic rated peak power, Horizontal irradiance, With a temperature coefficient of 0.0045, To measure the component temperature, This is the correction factor for the incident angle of the inclined surface, with a value range of [0.85, 1.0].

[0020] By integrating photovoltaic rated parameters, real-time irradiance, module temperature, and angle correction coefficients to match actual operating conditions, the efficiency decay and time derivative analysis based on the calculation results can effectively distinguish between shadows and stains, solving the problem of misjudgment by a single sensor. Accurate data provides a reliable basis for intelligent decision-making, avoids ineffective cleaning, and ensures the scientific nature of the cleaning triggering timing, helping to improve the power generation efficiency and operation and maintenance economy of photovoltaic systems.

[0021] The visual AI stain recognition module is equipped with a camera with IP67 protection rating, ≥2MP resolution and FOV≥90°, which captures one frame of image every 30 seconds. After grayscale and normalization preprocessing, the image is input into the model, with mAP@0.5≥95% and inference latency <200ms / frame. The stain coverage rate C calculation accuracy is ±5%, and it is divided into three levels: normal, moderate and severe according to C<10%, 10%-30% and ≥30%.

[0022] The IP67 protective camera is suitable for complex outdoor environments. Its high resolution and wide field of view ensure full coverage of component surfaces. Image preprocessing combined with a high-performance YOLO model enables high-accuracy and low-latency stain detection. Precise coverage calculation and grade classification provide a quantitative basis for cleaning decisions, avoiding mis-cleaning or missed cleaning due to fuzzy recognition. At the same time, the 30-second acquisition interval balances real-time performance and resource conservation.

[0023] The multi-condition AND logic of the intelligent decision controller includes: AI stain coverage C>10%; efficiency decay percentage DP%>5%; decay time derivative dDP / dt<1.0; ambient temperature 0℃≤T≤35℃, wind speed V≤8m / s, relative humidity RH≤85%; time since last cleaning >4 hours; when all conditions are met simultaneously, a cleaning trigger signal is output; otherwise, a negative signal is returned or the device is added to the waiting queue.

[0024] The multi-condition AND logic forms a strict cleaning trigger threshold, controlling multiple dimensions such as the degree of dirt, power generation impact, attenuation type, environmental safety, and operation frequency. This avoids ineffective cleaning triggered by insufficient dirt, non-dirt obstruction, or extreme weather, reducing equipment wear and energy consumption. It also prevents resource waste caused by excessive cleaning, ensuring stable component operation. The scientific combination of conditions makes cleaning decisions more precise, ensuring that the operation is initiated only when "cleaning is needed, can be cleaned, and should be cleaned," balancing cleaning effectiveness with system safety and economy.

[0025] The multi-heat source collaborative system includes three heat sources: HVAC condensing heat Q_HVAC with a power of 15-25kW, energy storage heat dissipation Q_ESS with a power of 3-6kW, and inverter heat dissipation Q_INV with a power of 1-4kW. The multi-heat source collaborative system uses the MPC algorithm for heat source scheduling, with a prediction window of 10 minutes. It solves the optimal mixing ratio through quadratic programming, with the optimization objective being to minimize heat waste and achieve a heat exchange efficiency of ≥85%.

[0026] By integrating three types of waste heat resources, the heat source supply channels are broadened, ensuring the stability and continuity of steam generation and avoiding the limitations of relying on a single heat source. The MPC algorithm, combined with a 10-minute prediction window and quadratic programming solution, achieves dynamic optimal allocation of heat sources, maximizing the utilization of waste heat and minimizing heat waste. A high heat exchange efficiency of ≥85% ensures energy utilization, replaces the traditional electric heating mode, consumes zero additional electricity, and significantly reduces the energy consumption cost of photovoltaic systems.

[0027] The steam generated by the steam generation and low-pressure safety system has the following parameters: pressure 0.1-0.2MPa, temperature 100-120℃, and flow rate 5-10kg / h. The steam generation and low-pressure safety system is equipped with a pressure gauge, a safety valve with a set pressure of 0.25MPa, and a quality monitoring system. It automatically releases pressure when overpressure occurs to ensure compliance with low-pressure safety standards.

[0028] The 0.1-0.2MPa pressure meets cleaning and decontamination requirements without requiring special equipment qualifications, lowering the deployment threshold; the 100-120℃ temperature and 5-10kg / h flow rate are suitable for photovoltaic module cleaning scenarios, balancing cleaning power with water and energy conservation; the pressure gauge, safety valve, and quality monitoring system form a triple safety protection, with automatic pressure relief in case of overpressure to eliminate safety hazards and ensure compliant system operation; the low-pressure design reduces equipment wear, stable parameters ensure consistent cleaning results, and at the same time reduces operation and maintenance risks and costs, improving system safety, reliability, and practicality.

[0029] Among them, the thermal shock protection mixing device uses an electric three-way mixing valve to mix steam above 100°C with coolant at around 25°C. After mixing, the steam temperature is ≤60°C and the temperature difference with the cold photovoltaic panel is ≤70°C. The closed-loop control uses first-order derivative regulation, the panel temperature rise rate is ≤5°C / minute, and the absolute temperature rise is ≤20°C. If either limit is exceeded, the cleaning will stop immediately and the cooling mode will be activated.

[0030] The electric three-way mixing valve quickly mixes high-temperature steam and coolant, controlling the mixing temperature to ≤60℃, significantly reducing the temperature difference with the cold panel and avoiding the risk of thermal shock from the source; the closed-loop control with first-order derivative adjustment controls thermal stress changes by limiting the rate of temperature rise and the absolute temperature rise, keeping the panel breakage rate below 0.01%; the emergency mechanism of stopping when the threshold is exceeded + cooling mode further strengthens the safety defense, ensuring the structural integrity of the photovoltaic module without affecting the cleaning effect, extending the service life of the equipment, and improving the stability and safety of the system operation.

[0031] The automatic spray gun and cleaning execution unit have an automatic scanning mechanism. Based on the 8×8 grid heat map of stain distribution, it prioritizes cleaning areas with high coverage. It dynamically adjusts the number of spray guns opened by 30%-70% according to the temperature rise rate. The cleaning cycle is 8-12 minutes, and it automatically switches to cooling mode after completion.

[0032] Based on the priority cleaning mechanism of the stain heat map, it focuses on high-pollution areas for precise operation, avoids ineffective spraying, and improves cleaning efficiency and cleanliness; the number of spray guns is dynamically adjusted according to the temperature rise rate, which can not only meet the needs of thermal shock protection, but also rationally allocate steam resources; the optimized cleaning cycle of 8-12 minutes balances the cleaning effect and energy consumption, and automatically switches the cooling mode to ensure safe cooling of components.

[0033] A photovoltaic steam control method based on multi-source sensing and waste heat synergy, the specific steps of which are as follows: Step 1: Parameter Acquisition: Acquire inverter, environmental, and heat source data every minute, and acquire component images every 30 seconds; Step 2, Data Processing: Calculate efficiency decay and derivative, distinguish between shadows and stains, generate stain coverage and heat map, and report all data to AIHUB database; Step 3, Decision Judgment: Determine whether the five conditions of claim 4 are met. If they are met, trigger cleaning; otherwise, return to the parameter acquisition stage and wait for the next cycle. Step 4: Heat source scheduling: Determine the heat source mixing ratio using the MPC algorithm to generate qualified steam; Step 5: Cleaning Execution: Record the initial temperature of the photovoltaic panel. The spray gun is activated to inject mixed steam; closed-loop control is executed once per second: the current temperature on the panel is detected, and if it exceeds... If the temperature reaches +20℃, immediately stop cleaning; calculate the temperature rise rate, and if it exceeds 5℃ / minute, reduce the number of spray guns by 30%; check the steam temperature, and if it exceeds 60℃, increase the coolant mixing ratio; Step 6, Cooling: After cleaning, turn off the steam supply and turn on the coolant circulation pump; when the panel temperature drops to within 5°C of the ambient temperature and the coolant temperature is <30°C, turn off the cooling pump, record the cleaning timestamp and power generation recovery effect, and report to AIHUB. Step 7, Waiting for the loop: Set the next cleaning interval, return to the parameter acquisition stage and enter the next loop.

[0034] A complete closed loop of "collection-processing-decision-execution-cooling-cycle" is formed, with rigorous logic and close alignment with actual operation and maintenance needs; parameter collection takes into account both real-time performance and resource optimization, providing accurate data support for subsequent stages; data processing enables precise differentiation between stains and shadows, and reporting to the database facilitates traceability and analysis; decision-making strictly follows multiple condition thresholds to avoid ineffective cleaning; heat source scheduling achieves efficient utilization of waste heat through the MPC algorithm; second-by-second closed-loop control of cleaning execution builds a solid thermal shock protection line; cooling ensures safe cooling of components; cyclic settings avoid frequent operations; the entire process is automated, requiring no manual intervention, which not only improves cleaning efficiency and effectiveness but also reduces energy consumption, equipment wear and tear, and operation and maintenance costs, ensuring the long-term stable and efficient operation of the photovoltaic system.

[0035] In step four, steam generation requires heating the preheated mixture from multiple heat sources for 5-10 minutes, and the heating process is achieved through segmented temperature control.

[0036] The 5-10 minute heating time is adapted to the characteristics of multiple heat sources and waste heat supply, ensuring that the mixture is fully heated to the target steam parameters; segmented temperature control can accurately control the heating rate, avoid local overheating or uneven heating, ensure stable steam quality, and reduce energy waste.

[0037] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0038] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A photovoltaic steam self-cleaning system based on multi-source sensing and waste heat synergy, characterized in that, include: The system comprises an inverter efficiency monitoring module, a visual AI stain recognition module, an intelligent decision controller, a multi-heat source collaborative system, a steam generation and low-pressure safety system, a thermal shock protection mixing device, and an automatic spray gun and cleaning execution unit. The inverter efficiency monitoring module reads inverter data in real time via a Modbus TCP interface, calculates theoretical power generation, efficiency degradation, and the time derivative of degradation, and distinguishes between shadows and stains. The visual AI module acquires component images and detects stains using a YOLOv9 / v12 model, outputting coverage and a heat map. The intelligent decision controller triggers cleaning through multi-condition AND logic. The multi-heat source collaborative system generates steam using HVAC condensation heat, energy storage heat dissipation, and inverter heat dissipation. The thermal shock protection device avoids thermal damage through water mixing and closed-loop control. The automatic spray gun cleans precisely according to the heat map.

2. The photovoltaic steam self-cleaning system based on multi-source sensing and waste heat synergy according to claim 1, characterized in that: The theoretical power generation calculation formula for the inverter efficiency monitoring module is as follows: In the formula Photovoltaic rated peak power, Horizontal irradiance, With a temperature coefficient of 0.0045, To measure the component temperature, This is the correction factor for the incident angle of the inclined surface, with a value range of [0.85, 1.0].

3. The photovoltaic steam self-cleaning system based on multi-source sensing and waste heat synergy according to claim 1, characterized in that: The visual AI stain recognition module is equipped with a camera with IP67 protection rating, ≥2MP resolution and FOV≥90°, which captures one frame of image every 30 seconds. After grayscale and normalization preprocessing, the image is input into the model, with mAP@0.5≥95% and inference latency <200ms / frame. The stain coverage rate C is calculated with an accuracy of ±5%, and is divided into three levels: normal, moderate and severe according to C<10%, 10%-30% and ≥30%.

4. The photovoltaic steam self-cleaning system based on multi-source sensing and waste heat synergy according to claim 1, characterized in that: The multi-condition AND logic of the intelligent decision controller includes: AI stain coverage C>10%; efficiency decay percentage DP%>5%; decay time derivative dDP / dt<1.0; ambient temperature 0℃≤T≤35℃, wind speed V≤8m / s, relative humidity RH≤85%; time since last cleaning >4 hours; when all conditions are met simultaneously, a cleaning trigger signal is output; otherwise, a negative signal is returned or the device is added to the waiting queue.

5. A photovoltaic steam self-cleaning system based on multi-source sensing and waste heat synergy as described in claim 1, characterized in that: The multi-heat source collaborative system includes three heat sources: HVAC condensing heat Q_HVAC with a power of 15-25kW, energy storage heat dissipation Q_ESS with a power of 3-6kW, and inverter heat dissipation Q_INV with a power of 1-4kW. The multi-heat source collaborative system uses the MPC algorithm for heat source scheduling, with a prediction window of 10 minutes. It solves the optimal mixing ratio through quadratic programming, with the optimization objective being to minimize heat waste and achieve a heat exchange efficiency of ≥85%.

6. The photovoltaic steam self-cleaning system based on multi-source sensing and waste heat synergy according to claim 1, characterized in that: The steam generated by the steam generation and low-pressure safety system has the following parameters: pressure 0.1-0.2MPa, temperature 100-120℃, and flow rate 5-10kg / h. The steam generation and low-pressure safety system is equipped with a pressure gauge, a safety valve with a set pressure of 0.25MPa, and a quality monitoring system. It automatically releases pressure when overpressure occurs to ensure compliance with low-pressure safety standards.

7. A photovoltaic steam self-cleaning system based on multi-source sensing and waste heat synergy as described in claim 1, characterized in that: The thermal shock protection mixing device uses an electric three-way mixing valve to mix steam above 100°C with coolant at around 25°C. After mixing, the steam temperature is ≤60°C, and the temperature difference between the steam and the cold photovoltaic panel is ≤70°C. The closed-loop control uses first-order derivative regulation, with the panel temperature rise rate ≤5°C / minute and the absolute temperature rise ≤20°C. If either limit is exceeded, the cleaning will stop immediately and the cooling mode will be activated.

8. A photovoltaic steam self-cleaning system based on multi-source sensing and waste heat synergy as described in claim 1, characterized in that: The automatic spray gun and cleaning execution unit have an automatic scanning mechanism. Based on the 8×8 grid heat map of stain distribution, it prioritizes cleaning areas with high coverage. It dynamically adjusts the number of spray guns turned on by 30%-70% according to the temperature rise rate. The cleaning cycle is 8-12 minutes. After completion, it automatically switches to cooling mode.

9. A photovoltaic steam control method based on multi-source sensing and waste heat synergy, characterized in that, The specific steps are as follows: Step 1: Parameter Acquisition: Acquire inverter, environmental, and heat source data every minute, and acquire component images every 30 seconds; Step 2, Data Processing: Calculate efficiency decay and derivative, distinguish between shadows and stains, generate stain coverage and heat map, and report all data to AIHUB database; Step 3, Decision Judgment: Determine whether the five conditions described in claim 4 are met. If they are met, trigger cleaning; otherwise, return to the parameter acquisition stage and wait for the next cycle. Step 4: Heat source scheduling: Determine the heat source mixing ratio using the MPC algorithm to generate qualified steam; Step 5: Cleaning Execution: Record the initial temperature of the photovoltaic panel. The spray gun is activated to inject mixed steam; closed-loop control is executed once per second: the current temperature on the panel is detected, and if it exceeds... If the temperature reaches +20℃, immediately stop cleaning; calculate the temperature rise rate, and if it exceeds 5℃ / minute, reduce the number of spray guns by 30%; check the steam temperature, and if it exceeds 60℃, increase the coolant mixing ratio; Step 6, Cooling: After cleaning, turn off the steam supply and turn on the coolant circulation pump; when the panel temperature drops to within 5°C of the ambient temperature and the coolant temperature is <30°C, turn off the cooling pump, record the cleaning timestamp and power generation recovery effect, and report to AIHUB. Step 7, Waiting for the loop: Set the next cleaning interval, return to the parameter acquisition stage and enter the next loop.

10. A photovoltaic steam control method based on multi-source sensing and waste heat synergy according to claim 9, characterized in that: The steam generation in step four requires heating the preheated mixture from multiple heat sources for 5-10 minutes, and the heating process is achieved through segmented temperature control.

Images