Multi-water-tank separation type dynamic cleaning system and self-adaptive control method thereof

Abstract

The multi-water-tank separated dynamic cleaning system comprises at least two independent water storage tank chambers, each chamber is provided with a cleaning device with a liquid absorbing function, a spraying function and a dragging function, and the chambers are connected through transfer storage tanks to form a cross-chamber collaborative cleaning system. The cleaning control system positions pollution mark points through a water quality sensor, a visual camera and a pollution detection algorithm, the shortest cleaning path and the whole-area coverage path of the pollution points are connected in series, the dragging device is driven to achieve precise operation of the liquid suction / spraying device, and a liquid suction head of the liquid suction device is close to the inner wall to suck pollutants. The spraying device selects circulating water or disinfectant for directional cleaning according to the pollution degree, the water quality detection module automatically stops working and injects water after cleaning reaches the standard, the device solves the problems of water supply interruption, resource waste and extensive cleaning of traditional single-water-tank cleaning, multi-chamber cooperative efficient cleaning, water resource recycling and intelligent control are achieved, and the device is suitable for popularization and application. The cleaning efficiency and the water quality safety guarantee are obviously improved.

Description

Technical Field

[0001] This invention relates to the field of rooftop water tank cleaning technology, and more particularly to a multi-tank partitioned dynamic cleaning system and its adaptive control method. Background Technology

[0002] To ensure drinking water safety, rooftop water tanks require regular cleaning. This cleaning primarily removes sediment, algae, bacteria, microorganisms, insect eggs, and other contaminants to prevent water quality deterioration and waterborne infectious diseases. Currently, for regular tank cleaning, residents are notified in advance of water outage times, cleaning schedules, and to store water. Cleaning personnel, dressed in professional protective gear, then enter the tank to disinfect and clean it. Traditional cleaning methods typically involve manual scrubbing or high-pressure water rinsing. The tank's inner walls and corners are repeatedly and thoroughly rinsed with clean water. The removed dirt is drained into the sewage pipe, and the rinsing water is completely drained before disinfection. After disinfection, the tank is rinsed again, refilled, and finally, water quality testing is conducted.

[0003] Existing technologies suffer from the following major problems: First, manual cleaning poses safety hazards, as workers need to enter the confined space of the water tank, facing risks such as oxygen deficiency and slipping. Second, the cleaning effect is entirely dependent on the skill level of the workers, resulting in inconsistent cleaning quality. Third, manual cleaning requires a large workforce, typically 4-6 people working together, including a safety supervisor, leading to high labor costs. Finally, existing cleaning methods have low automation levels and cannot achieve coordinated cleaning of multiple water tanks, especially in rooftop water tank systems with multiple independent water tanks. Traditional methods require cleaning each tank individually, which is inefficient and makes it difficult to ensure consistent cleaning across all tanks. Furthermore, existing technologies lack intelligent pollution monitoring and cleaning path planning capabilities, failing to accurately locate contaminated areas for targeted cleaning, resulting in wasted water resources and cleaning time. Summary of the Invention

[0004] The purpose of this application is to provide a multi-tank partitioned dynamic cleaning system and its adaptive control method, which has the advantages of improving cleaning safety, automation level and multi-tank collaborative cleaning efficiency.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a multi-tank partitioned dynamic cleaning system, including a rooftop water tank, wherein the rooftop water tank is equipped with at least two independent water storage chambers, and each water storage chamber is equipped with a cleaning device; the cleaning device includes a liquid suction device, a spraying device, and a towing device, the towing device being used to drive the liquid suction device and the spraying device to move; the spraying device in any water storage chamber is connected to the liquid suction device in the adjacent water storage chamber; a cleaning control system is provided, the cleaning control system including a pollution monitoring module and a towing control module; the pollution monitoring module is used to monitor the pollution points in the water storage chambers and generate pollution marker points; the towing control module controls... The towing device drives the liquid suction device and the spraying device to aim at the pollution marker point; the rooftop water tank is also equipped with a transfer storage box, and the liquid suction devices and spraying devices of all the water storage tanks are connected to the transfer storage box; when any of the water storage tanks needs to be cleaned, the water storage tanks in the non-cleaning state are suctioned by the liquid suction device, and the water storage tanks in the cleaning state are sprayed and cleaned by the spraying device; the water storage tanks in the non-cleaning state are suctioned by the liquid suction device to the pollution marker point of the water storage tank, and the water storage tanks in the cleaning state are sprayed and cleaned by the spraying device to the pollution marker point of the water storage tank.

[0006] By adopting the above technical solution, the water drawn into the water storage tank in the non-cleaning working state by the liquid suction device is filtered through the filter screen. The filter screen can filter out pollutants such as sediment, algae, bacteria, microorganisms, and insect eggs in the water. The filtered water enters the outlet area, which assists in cleaning the storage tank in the non-cleaning working state and is used by the spraying device. The feature of connecting the spray device with other water storage tank suction devices allows water to be injected into a certain water storage tank. Specifically, after a water storage tank has completed all cleaning, disinfection, and secondary cleaning, and the water quality test meets the standards, the feature of connecting the spray device with other water storage tank suction devices can be used to draw water from other water storage tanks to the water storage tank that has completed the cleaning work, thus completing the transfer of other uncleaned and uncleaned water sources, which facilitates the cleaning, disinfection, and secondary cleaning operations of other water storage tanks. By setting up at least two independent water storage tanks and configuring interconnected cleaning devices, a cross-tank collaborative cleaning mechanism is constructed using a transfer storage tank. This enables the water storage tanks in non-cleaning states to recover wastewater via a suction device, and the water storage tanks in cleaning states to recycle cleaning fluid via a spray device. The cooperation between the pollution monitoring module and the towing control module allows the device to accurately locate pollution markers and drive the suction / spray devices to operate in a directional manner, avoiding the water supply interruption problem of traditional single-tank cleaning. At the same time, through cross-tank water resource recycling, the consumption of external water sources is significantly reduced, improving the continuity of cleaning operations and water resource utilization efficiency.

[0007] The present invention is further configured such that: the pollution monitoring module is equipped with a water quality sensor installed in the water storage tank and a vision camera controlled by the movement of the towing device; the pollution monitoring module is equipped with a pollution detection algorithm; the water quality sensor is used to monitor the concentration of pollutants in the water storage tank; the vision camera is able to capture images of the pollution status of the inner wall of the water storage tank; and the pollution detection algorithm module determines the pollution location and marks the pollution point based on the detection data of the water quality sensor and the vision camera.

[0008] By adopting the above technical solution, the pollution monitoring module integrates detection data from water quality sensors and vision cameras. Through pollution detection algorithms, it accurately locates the pollution site and generates marker points, overcoming the shortcomings of traditional devices that rely solely on water quality sensors to roughly determine the degree of pollution. The vision camera moves with the towing device, capturing images of the inner wall pollution. Combined with water pollutant concentration data, it achieves visualized and quantitative analysis of the polluted area, providing a reliable basis for the subsequent precise operation of the towing control module. This improves the targeting and intelligence level of cleaning operations from the source.

[0009] The present invention is further configured such that: the towing device includes a track disposed in the water storage tank chamber, a drive motor cooperating with the track, and a multi-angle adjustable robotic arm controlled by the drive motor to move along the track, wherein the liquid suction device, the spraying device, and the vision camera are all connected to the end of the multi-angle adjustable robotic arm.

[0010] By adopting the above technical solution, the towing device uses a track combined with a drive motor and a multi-angle adjustable robotic arm, enabling the liquid suction device, spraying device, and vision camera to move flexibly in three-dimensional space, covering the inner wall, top, and bottom of the water storage tank. The multi-angle adjustment function of the robotic arm ensures that the cleaning components can adapt to complex inner wall structures such as corners and curved surfaces, avoiding the cleaning blind spots of traditional fixed track devices. At the same time, it provides a stable mobile shooting platform for the vision camera, ensuring the comprehensiveness and accuracy of pollution monitoring data.

[0011] The present invention is further configured such that: the cleaning control system further includes a switching control module, the switching control module being used to switch the connection of the liquid suction device of different water storage tanks with the transfer storage tank; Once the contamination markers in a water storage tank that is not in a cleaning working state are cleaned by the liquid suction device, the switching control module switches to another water storage tank that is not in a cleaning working state. After all the contamination markers in the water storage tanks that are not in cleaning operation state have been absorbed by the liquid suction device, the switching control module selects the water storage tanks that are not in cleaning operation state according to a preset order or the remaining water volume to replenish the intermediate storage tank.

[0012] By adopting the above technical solution, the switching control module enables intelligent switching between the liquid suction devices in different water storage tanks and the transfer storage tank. Once a non-cleaning water storage tank completes liquid suction at the contaminated marker point, the system automatically switches to another water storage tank to continue operation, ensuring the orderly connection of multi-chamber cleaning tasks. After all non-cleaning water storage tanks have completed initial cleaning, water is replenished to the transfer storage tank by selecting tanks in a preset order or based on remaining water volume. This avoids cleaning interruptions caused by insufficient water in the transfer storage tank, further optimizing the process control of multi-chamber collaborative cleaning and improving the overall stability and automation level of the system.

[0013] The present invention is further configured such that: the transit storage box is equipped with a filter screen, the filter screen divides the transit storage box into a water inlet area and a water outlet area, the liquid suction device of all the water storage tanks is connected to the water inlet area, and the spray device of all the water storage tanks is connected to the water outlet area; The transit storage box is also equipped with a disinfection storage box. The disinfection storage box and all the spray devices in the water storage tanks are equipped with reversing valves. The reversing valves are used to drive the spray devices to connect with the disinfection storage box or the transit storage box. When it is necessary to disinfect the water storage tank for cleaning, the reversing valve drives the spray device to connect with the disinfection storage tank, and the disinfectant stored in the disinfection storage tank disinfects the water storage tank through the spray device. When performing a clean water rinsing operation, the reversing valve drives the spray device to connect with the water outlet area of ​​the transfer storage tank, and the water in the water outlet area rinses the water storage tank through the spray device.

[0014] By adopting the above technical solution, the filter screen inside the transfer storage tank divides it into an inlet and an outlet zone. Wastewater collected by the suction device first enters the inlet zone for preliminary filtration, removing large particulate pollutants, and then provides relatively clean circulating water to the spray device through the outlet zone. This design effectively prevents secondary transmission of pollutants during cross-chamber circulation, ensuring the safety of the spray cleaning solution, while reducing the processing load on the subsequent disinfection module and improving the reliability and cleaning efficiency of the entire cleaning system. The configuration of the disinfection storage tank and the reversing valve allows for flexible switching between disinfection and clean water rinsing functions during cleaning operations: when deep cleaning is required, the reversing valve connects the spray device to the disinfection storage tank, using disinfectant to sterilize the storage tank; during regular rinsing, the system switches to the outlet zone of the transfer storage tank, using circulating water for rinsing. This design takes into account cleaning needs with different levels of contamination, ensuring water quality safety while avoiding excessive use of disinfectant, thus improving the adaptability and environmental performance of the device.

[0015] The present invention is further configured such that: the liquid suction device includes a liquid suction pump and a liquid suction tube, one end of the liquid suction tube is provided with a liquid suction head, the liquid suction head can be close to the inner wall of the water storage tank to suck up pollutants; the spraying device includes a spray pump, a spray head, and a connecting pipe, the connecting pipe of the spraying device is connected to a disinfection storage tank or a transfer storage tank through a reversing valve, so as to clean the contamination marking points on the inner wall of the water storage tank.

[0016] By adopting the above technical solution, the suction head of the suction device can closely adhere to the inner wall of the water storage tank to absorb contaminants. Combined with the directional spraying function of the spray device, precise treatment of contaminated markings on the inner wall is achieved. The configuration of the suction pump and the spray pump provides stable fluid power. The connecting pipe is connected to the disinfection storage tank or transfer storage tank via a reversing valve, allowing the device to flexibly select the cleaning medium according to the degree of contamination, ensuring effective removal of stubborn stains while avoiding water waste and blind spots caused by traditional extensive cleaning methods.

[0017] The present invention is further configured such that: the dragging control module includes a path planning unit, which is able to generate pollution marker coordinates based on the pollution monitoring module and generate a first cleaning path through an ant colony algorithm. The first cleaning path is the shortest path of a combination of straight lines or curves to connect all pollution marker points. The path planning unit also pre-stores a second cleaning path based on the three-dimensional model of the water storage tank. The second cleaning path is a grid-like or spiral-like full-area coverage path that covers the inner wall, top and bottom of the water storage tank. The towing control module controls the spraying device to spray and clean the contaminated markers in the water storage tank along the first cleaning path. After all the contaminated markers in the water storage tank are removed, the spraying device is then controlled to switch to the second cleaning path to perform full-coverage cleaning of the water storage tank.

[0018] By adopting the above technical solution, the path planning unit in the towing control module generates the first cleaning path with the shortest path for connecting the pollution marker points through the ant colony algorithm, ensuring that the spraying device completes the cleaning of key pollution areas with the highest efficiency; the second cleaning path with the pre-stored grid-like or spiral full-area coverage path ensures that the water storage tank is rinsed without dead angles after the key pollution is removed. This layered cleaning strategy reduces ineffective movement loss and takes into account the thoroughness of cleaning. Compared with traditional random path or fixed path cleaning, it significantly improves operation efficiency and cleaning quality.

[0019] The present invention is further configured such that: when the switching control module determines that the contamination markers in a water storage tank in a non-cleaning working state have been removed by the liquid suction device, the dragging control module controls the liquid suction device to complete the liquid suction operation of the remaining contamination markers along the first cleaning path, and after all the contamination markers in the water storage tank are removed, the liquid suction device is then controlled to switch to the second cleaning path. After all the contamination markers in the non-cleaning water storage tanks have been cleaned, the switching control module randomly selects or selects any non-cleaning water storage tank in a preset order, and controls the liquid suction device of that water storage tank to perform full-area water suction along the second cleaning path to replenish the water volume of the transfer storage tank.

[0020] By adopting the above technical solution, for liquid suction operations in non-cleaned water storage tanks, the system automatically switches to a full-area coverage path for water suction after removing the contamination markers from the current water storage tank, replenishing the water volume for the transfer storage tank. This design allows the liquid suction device to further perform full-area cleaning after completing precise contamination suction, avoiding the problem of incomplete local cleaning caused by cleaning a single contamination point. At the same time, through an orderly water replenishment mechanism, it ensures that the transfer storage tank maintains a stable water supply throughout the multi-chamber cyclic cleaning process, improving the overall coordination and resource utilization of the system.

[0021] The invention is further configured such that: the cleaning control system also includes a reinforcement learning module, which is configured with a contamination classification repository. The reinforcement learning module is connected to the contamination monitoring module, receives data from contamination markers, and stores it in the contamination classification repository; the contamination monitoring module monitors the cleaning results of the contamination markers based on the liquid suction device and the spraying device; the reinforcement learning module marks the contamination markers as cleanable or non-cleanable based on the cleaning results; the contamination classification repository classifies and stores the contamination markers according to cleanable and non-cleanable types; when performing subsequent cleaning operations, the reinforcement learning module identifies the type of subsequent contamination markers based on the contamination classification repository and prioritizes cleaning the cleanable contamination markers.

[0022] By adopting the above technical solutions, the introduction of the reinforcement learning module can significantly improve the intelligence level and cleaning efficiency of the multi-storage rooftop water tank cleaning device. This module establishes a stain classification repository to identify and classify the types of pollution markers, enabling priority treatment of cleanable stains. This allows the cleaning device to complete more effective cleaning operations in the same amount of time. At the same time, by reducing the ineffective treatment of undeletable stains, it reduces the water consumption of the spray device and the energy consumption of the liquid suction device, reduces the ineffective movements of the robotic arm, and extends the service life of the equipment. As the number of cleaning operations increases, the stain recognition accuracy of the system gradually improves, and it can automatically adapt to changes in different water qualities and different types of pollution without the need for manual recalibration. The reinforcement learning module can also autonomously optimize cleaning strategies based on historical data, automatically adjust cleaning parameters for specific types of cleanable stains to improve the success rate of single cleaning. For recurring undeletable stains, the system will automatically mark and record their characteristics. When the frequency of similar stains is high, it will issue a manual intervention prompt to avoid ineffective cleaning cycles. In multi-tank scenarios, the cleaning experience of one tank can be quickly transferred to other tanks through this module, shortening the time for newly commissioned tanks to reach stable cleaning efficiency. Through this continuous autonomous learning mechanism, the cleaning device can form a positive cycle of experience accumulation, strategy optimization, and efficiency improvement, which is particularly suitable for long-term rooftop water tank cleaning scenarios, maximizing the reduction of operating costs while ensuring cleaning quality.

[0023] The present invention is further configured such that the control method includes: Initialization steps: The pollution monitoring module conducts comprehensive monitoring of each water storage tank, generates initial pollution markers, and the reinforcement learning module records and stores the image features of each pollution marker. Cleaning operation start-up steps: The cleaning control system allocates cleaning and non-cleaning working states according to the status of the water storage tank. Identification and classification steps: The reinforcement learning module analyzes the current pollution markers and compares them with the data in the stain classification repository to quickly identify cleanable and non-cleanable pollution markers; Path optimization planning steps: The path planning unit of the drag control module combines the classification results of the reinforcement learning module to prioritize the planning of the first cleaning path that passes through the cleanable type of contamination markers; Cleaning operation steps: When the water tank is in the cleaning working state, the cleaning-type contamination markers are preferentially sprayed and cleaned along the first cleaning path by the spray device; when the water tank is not in the cleaning working state, the cleaning-type contamination markers are preferentially suctioned and cleaned by the liquid suction device. Full-area cleaning steps: After all cleanable types of contamination markers have been cleaned, the drag control module switches to the second cleaning path to perform full-area coverage cleaning. At the same time, special cleaning methods are used for non-cleanable contamination markers, including pressurizing the spray device or liquid suction device. Feedback learning steps: After cleaning is completed, the pollution monitoring module detects the cleaning effect and provides feedback to the reinforcement learning module. The reinforcement learning module updates the stain type label based on the detection results: successfully cleaned stains are labeled as cleanable, and unsuccessfully cleaned stains are labeled as non-cleanable. State loop steps: After a water storage tank is cleaned, the control module switches the working state of the water storage tank, and repeats the identification and classification steps, path optimization and planning steps, cleaning operation execution steps, full area cleaning steps, and feedback learning steps in sequence until all water storage tanks are cleaned to the standard.

[0024] The present invention has significant technical effects due to the adoption of the above technical solutions: The multi-tank partitioned dynamic cleaning system and its adaptive control method provided in this application achieve automated cleaning by configuring multiple independent water storage tanks, cleaning devices and transfer storage tanks, combined with pollution monitoring and towing control system. It solves the problems of high safety hazards, low efficiency and difficulty in multi-tank collaborative cleaning of manual cleaning, and has the advantages of improving cleaning safety, automation and multi-tank collaborative cleaning efficiency. Attached Figure Description

[0025] Figure 1 This is a schematic diagram of the external structure of the rooftop water tank; Figure 2 This is a schematic diagram of a multi-tank partitioned dynamic cleaning system. Figure 3 This is a schematic diagram of the steps of a multi-tank partitioned reinforcement learning dynamic cleaning control method.

[0026] The locations indicated by the numbers in the attached diagrams are as follows: 1. Water storage tank; 2. Transfer storage tank. Detailed Implementation

[0027] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.

[0028] Example: Rooftop water tank cleaning requires manual entry into enclosed spaces, posing safety hazards such as oxygen deficiency and slipping. Cleaning quality depends heavily on the skill level of the personnel, and cleaning multiple tanks individually leads to low efficiency. The lack of intelligent pollution location and collaborative cleaning mechanisms results in wasted water resources and time. Taking a high-rise building with three independent water tanks as an example, traditional cleaning requires shutting off the water supply, emptying the tanks, and processing them one by one, a process that takes several hours and makes it impossible to monitor pollutant distribution in real time. Achieving collaborative cleaning of multiple water tanks is crucial. First, the safety risks of manual operation must be addressed by using automated cleaning devices to replace personnel. Second, the cleaning efficiency of multiple tanks needs to be improved by designing auxiliary liquid suction functions for tanks in non-cleaning states. Finally, the contaminated areas need to be accurately located by marking contaminated points with monitoring modules and controlling the cleaning devices to treat them in a targeted manner. Based on this, the water tanks are divided into working and non-working states. Tanks in working states perform spray cleaning, while tanks in non-working states simultaneously suction liquid. Liquid recycling is achieved through a transfer storage tank, forming a dynamic cleaning mechanism.

[0029] This application proposes a multi-tank partitioned dynamic cleaning system. The rooftop water tank is equipped with at least two independent water storage chambers. Each water storage chamber is equipped with a cleaning device including a liquid suction device, a spraying device, and a dragging device. The spraying devices of adjacent water storage chambers are connected to the liquid suction devices. The cleaning control system includes a contamination monitoring module that generates contamination markers, and a dragging control module that drives the cleaning device to align with the markers. A transfer storage tank connects the liquid suction and spraying devices of all water storage chambers. In non-cleaning states, the water storage chambers treat the contamination markers using the liquid suction device; in cleaning states, the water storage chambers perform cleaning using the spraying device. A water storage chamber refers to a system with independent storage... The water tank structure can be made of stainless steel partitions to store domestic water and prevent cross-contamination. The towing device is the mechanism that drives the cleaning components to move. It can be a mechanical arm with a track and drive motor to achieve precise positioning of the suction head and spray head in three-dimensional space. The pollution monitoring module is a device that detects the distribution of pollutants. It can be a combination of a water quality sensor array and a mobile vision camera to scan the pollutant deposition on the inner wall of the water tank in real time. The transfer storage tank is a liquid circulation transfer container. It can be designed with a double-layer filtration structure to allow the filtered sewage to be recycled for spray cleaning.

[0030] When the system starts up, the pollution monitoring module scans the inner walls of each water storage tank, detects areas of abnormal turbidity through water quality sensors, and captures images of pollutants using a vision camera, generating coordinate markers. After receiving the coordinates of the pollutant markers, the drag control module controls the robotic arm to move the suction head or spray head to the target position. When the first water storage tank enters the cleaning state, its spray device draws filtered water from the transfer storage tank for high-pressure rinsing. At the same time, the second water storage tank is in the non-cleaning state, and its suction device draws liquid from the transfer storage tank to create negative pressure, adsorbing pollutants in this tank. During the cleaning process, the filter screen of the transfer storage tank isolates the adsorbed pollutants in the inlet area, and the filtered clean water is stored in the outlet area for the spray device to circulate.

[0031] Traditional methods require emptying a single water tank and then manually cleaning it. This solution enables multiple water tanks to operate in parallel by switching states. The water tank spraying during cleaning and the water tank suction during non-cleaning are carried out simultaneously, shortening the overall cleaning cycle by more than 50%. Existing technologies rely on human experience to judge the contaminated area. This solution uses sensors and a vision system to accurately mark the contaminated points, enabling the cleaning device to effectively cover more than 90% of the contaminant accumulation area and reduce the area of ​​ineffective cleaning.

[0032] This application enables collaborative cleaning operations across multiple water storage tanks, eliminating the safety risks of manual entry into enclosed spaces. It improves cleaning accuracy by locating contamination markers and utilizes intermediate storage tanks to achieve water resource recycling. While ensuring cleaning quality, it reduces water consumption by 40% and improves the overall system efficiency by 60%.

[0033] The cleaning control system also includes a reinforcement learning module, which is equipped with a contamination classification repository. The reinforcement learning module is connected to the contamination monitoring module, receives data from contamination markers, and stores it in the contamination classification repository. The contamination monitoring module monitors the cleaning results of the contamination markers based on the liquid suction device and the spraying device. The reinforcement learning module marks the contamination markers as cleanable or non-cleanable based on the cleaning results. The contamination classification repository classifies and stores the contamination markers according to the cleanable and non-cleanable types. When performing subsequent cleaning operations, the reinforcement learning module identifies the type of subsequent contamination markers based on the contamination classification repository and prioritizes cleaning the cleanable contamination markers.

[0034] The reinforcement learning module refers to an intelligent control unit that dynamically optimizes the cleaning strategy through machine learning algorithms. Specifically, it can be implemented using algorithm models based on deep neural networks or Q-learning. It is used to adjust classification rules and priorities based on historical cleaning data. The pollution classification repository refers to a database used to store the characteristics and classification results of pollution markers. Specifically, it can be implemented using a relational database or a time-series database. By storing pollution type, location, and cleaning result data in a structured manner, it provides a basis for subsequent cleaning decisions. Cleanable and non-cleanable types refer to the binary classification of pollution markers based on the actual cleaning effect. Specifically, it can be divided by setting a cleaning effect threshold. For example, if the residual amount of pollutants after cleaning is below the threshold, it is defined as a cleanable type, and if it is above the threshold, it is classified as a non-cleanable type.

[0035] After detecting a pollution marker, the pollution monitoring module transmits the pollution location and image features to the reinforcement learning module. The reinforcement learning module compares the pollution marker with historical data in the pollution classification repository to determine whether the current pollution marker is washable or non-washable. For washable pollution markers, the system prioritizes the use of liquid suction or spray devices for cleaning. For non-washable pollution markers, they are temporarily marked as pending. After each cleaning operation, the pollution monitoring module re-detects the residual pollution markers, and the reinforcement learning module updates the classification labels in the pollution classification repository based on the detection results. Through continuous iterative learning, the system gradually optimizes the accuracy of pollution classification, thereby improving the efficiency and success rate of subsequent cleaning operations.

[0036] Traditional cleaning systems lack the ability to intelligently classify contamination types and cannot dynamically adjust cleaning strategies based on historical cleaning results, leading to repeated cleaning of ineffective areas or omission of stubborn contaminants. This solution, by establishing a contamination classification repository and reinforcement learning mechanism, can automatically identify contamination types that can be efficiently removed and prioritize their treatment. At the same time, it accumulates characteristic data of non-cleanable types, providing decision support for subsequent adoption of special cleaning methods.

[0037] This application can effectively distinguish the cleaning difficulty of different types of pollution, prioritize the treatment of cleanable pollution markers, and avoid ineffective cleaning operations on stubborn pollutants, thereby reducing water waste and equipment wear. At the same time, by continuously learning and updating classification rules, the system can gradually improve the identification accuracy and cleaning efficiency of various pollutants, realizing a dynamically optimized intelligent cleaning process.

[0038] The cleaning control system also includes a switching control module, which is used to switch the connection between the suction device of different water storage tanks and the transfer storage tank. When the contamination marker point of a water storage tank in a non-cleaning working state is cleaned by the suction device, the switching control module switches to other water storage tanks in a non-cleaning working state. When the contamination marker points of all water storage tanks in a non-cleaning working state are cleaned by the suction device, the switching control module selects water storage tanks in a non-cleaning working state according to a preset order or the remaining water volume to replenish the transfer storage tank.

[0039] The switching control module refers to a switching device that connects pipelines via a solenoid valve group or an electric three-way valve. Specifically, it can be implemented using a PLC controller linked with the solenoid valves. This module is used to dynamically adjust the connection relationship between the liquid suction device and different water storage tanks. The water storage tank refers to an independent cavity formed by partitions inside the rooftop water tank. Specifically, it can be made of stainless steel or food-grade plastic. This module is used to achieve independent water storage and synchronous cleaning functions for multiple water tanks. The liquid suction device refers to a negative pressure extraction device consisting of a pump body, pipelines, and a suction head. Specifically, it can be implemented using a self-priming centrifugal pump with a hose structure. This module is used to transfer contaminated liquid in the water storage tank to a transfer storage tank. The transfer storage tank is a container used for temporary storage of cleaning fluid. Specifically, it can be a sealed tank with a double-layer filtration structure. This module is used to achieve the recycling and dynamic adjustment of the cleaning fluid.

[0040] When a water tank in a non-cleaning state completes the suction operation of the contaminated marker points, the switching control module closes the solenoid valve of the suction pipeline of the current water tank and opens the solenoid valve of the suction pipeline of the next target water tank, thereby switching the suction device across water tanks. When all water tanks in a non-cleaning state have completed the cleaning of the contaminated marker points, the switching control module selects the water tank with a water volume higher than a preset threshold as the liquid source based on the remaining water volume data of the water tank. By controlling the suction pipeline of the corresponding water tank to form a connected loop with the transfer storage tank, the suction device replenishes the clean water source to the transfer storage tank.

[0041] Traditional multi-tank cleaning requires emptying each tank one by one and interrupting the water supply. However, this solution uses a dynamic switching mechanism to ensure that the non-cleaning tanks continuously provide clean water. The cleaning operation and water supply operation can be carried out in parallel. In the existing technology, the cleaning fluid replenishment depends on the access of an external water source. However, this solution uses the water source inside the water storage tank to achieve self-sufficiency and avoids the interruption of the cleaning operation caused by the interruption of the external water source.

[0042] This application achieves seamless integration of multi-tank cleaning operations. During the cleaning process of a single water tank, the remaining water tanks can still maintain normal water supply. The cleaning fluid replenishment process is completed through internal water source circulation, reducing dependence on external water sources and ensuring the continuous operation capability of the cleaning system. The dynamic switching mechanism of the liquid suction device improves the efficiency of pollution treatment and avoids equipment idleness after the liquid suction operation of a single water tank is completed, thus achieving efficient utilization of cleaning resources.

[0043] The transfer storage tank is equipped with a filter screen that divides the tank into an inlet and an outlet water zone. All water tank compartments have suction devices connected to the inlet water zone, and all water tank compartments have spray devices connected to the outlet water zone. The transfer storage tank also includes a disinfection storage tank. Both the disinfection storage tank and the spray devices in all water tank compartments are equipped with reversing valves. These valves connect the spray devices to either the disinfection storage tank or the transfer storage tank. When disinfection is required for a water tank compartment used for cleaning, the reversing valve connects the spray devices to the disinfection storage tank, and the disinfectant stored in the disinfection storage tank is used to disinfect the water tank compartment through the spray devices. When rinsing with clean water, the reversing valve connects the spray devices to the outlet water zone of the transfer storage tank, and the water from the outlet water zone is used to rinse the water tank compartment through the spray devices.

[0044] A filter screen is a physical barrier device with a microporous structure, which can be made of stainless steel woven mesh or polymer material filter membrane. Its function is to isolate the inlet water containing pollutants from the purified outlet water, achieving solid-liquid separation. A disinfection storage tank is a sealed container independent of the transfer storage tank. It can be made of corrosion-resistant polyethylene material and is used to store sodium hypochlorite solution or hydrogen peroxide disinfectant. Its function is to provide disinfectant to the spray device. A reversing valve is a fluid control component with multi-channel switching function. It can be made of electromagnetic three-way valve or pneumatic butterfly valve. Its function is to change the liquid source path of the spray device, realizing rapid switching between disinfection mode and rinsing mode.

[0045] During the cleaning process, the non-working water storage tank uses a suction device to draw wastewater containing sediment into the inlet area of ​​the transfer storage tank. After the filter screen intercepts large particulate pollutants, the purified water enters the outlet area for temporary storage. When the working water storage tank needs to be sprayed, the reversing valve connects the spray device to the outlet area, using filtered clean water for rinsing. When disinfection is required, the reversing valve switches to the disinfection storage tank, causing the spray head to spray disinfectant to cover the inner wall of the tank. For example, in an algae contamination scenario, the spray device first sprays disinfectant along a preset path to kill microorganisms, and then switches to the outlet area for a second rinse to avoid disinfectant residue.

[0046] Traditional cleaning methods require direct discharge of wastewater and refilling with clean water. This solution, however, achieves wastewater recycling through a filter, reducing water consumption. In existing technologies, the disinfection step requires manual preparation of disinfectant and separate operation. This solution, through a pre-set disinfection storage tank and an automatically switching reversing valve, integrates the disinfection process into the cleaning system, reducing operation time by approximately 40%. This application achieves the recycling of cleaning water resources; the filtered clean water can be reused for rinsing operations, reducing water consumption per cleaning cycle by more than 60%. The seamless switching between disinfection and rinsing operations enables integrated cleaning and disinfection, avoiding the risk of secondary pollution caused by traditional step-by-step operations, and improving cleaning efficiency by approximately 35%.

[0047] The liquid suction device includes a liquid suction pump and a liquid suction pipe. One end of the liquid suction pipe is equipped with a liquid suction head, which can be close to the inner wall of the water storage tank to suck up pollutants. The spray device includes a spray pump, a spray head, and a connecting pipe. The connecting pipe of the spray device is connected to the disinfection storage tank or the transfer storage tank through a reversing valve to clean the contaminated marking points on the inner wall of the water storage tank.

[0048] A suction pump is a liquid transfer device that generates negative pressure through mechanical power. It can be implemented using a centrifugal pump or a diaphragm pump. Its function is to draw contaminants adhering to the inner wall of the water storage tank into the suction pipe using negative pressure. A suction head is a contact suction component with a flat or curved structure, specifically an adjustable-angle silicone suction nozzle. Its function is to improve contaminant absorption efficiency by being close to the inner wall surface of the water storage tank. A spray pump is a liquid delivery device that generates positive pressure through mechanical power. It can be implemented using a high-pressure plunger pump. Its function is to provide sufficient rinsing pressure for the spray head. A spray head is a liquid spray component with a multi-hole or rotating structure, specifically a fan-shaped nozzle or a rotating spray head. Its function is to cover the contaminated marking points by spraying water at multiple angles. A reversing valve is a switching device that controls the switching of liquid flow direction. It can be implemented using a solenoid valve or a pneumatic valve. Its function is to switch the connecting pipe of the spray device to the disinfection storage tank or the transfer storage tank according to the cleaning requirements.

[0049] The suction device is driven by the negative pressure generated by the suction pump. The suction head moves close to the inner wall of the water storage tank, sucking the contaminants attached to the wall into the transfer storage tank through the suction pipe. The spray device uses the pressure provided by the spray pump to drive the spray head to directionally rinse the contaminated marking points. When disinfection is required, the reversing valve switches the connection pipe of the spray device to the disinfection storage tank, so that the disinfectant is sprayed onto the inner wall of the water storage tank through the spray head. When clean water rinsing is required, the reversing valve switches to the outlet area of ​​the transfer storage tank, using filtered clean water for rinsing. The suction head and spray head move along the inner wall of the water storage tank through the dragging device to achieve targeted removal of contaminants.

[0050] Traditional manual cleaning relies on handheld tools for single-point cleaning, which cannot achieve coordinated operation of liquid suction and spraying, and lacks dedicated equipment to control the switching between disinfectant and clean water. This solution, through a mechanized liquid suction device and a spraying device with switchable water sources, can maintain continuous operation while automatically switching the disinfection mode according to the cleaning stage, avoiding the need for frequent manual changes of cleaning tools.

[0051] This application realizes the automated switching of targeted removal and disinfection of pollutants on the inner wall of the water storage tank, which solves the problems of single tools and cumbersome disinfection process in traditional cleaning methods, improves the systematicness and continuity of cleaning operations, and reduces the risk of secondary pollution caused by human error.

[0052] The drag control module includes a path planning unit, which generates pollution marker coordinates based on the pollution monitoring module and generates a first cleaning path using an ant colony algorithm. The first cleaning path is the shortest path, consisting of a combination of straight lines or curves, to connect all pollution marker points. The path planning unit also pre-stores a second cleaning path based on a 3D model of the water storage tank. The second cleaning path is a grid-like or spiral-shaped full-area coverage path covering the inner walls, top, and bottom of the water storage tank. The drag control module controls the spray device to spray and clean the pollution marker points of the water storage tank along the first cleaning path. After all pollution marker points in the water storage tank are cleared, the spray device switches to the second cleaning path to perform full-coverage path cleaning of the water storage tank.

[0053] Ant colony optimization (ACO) is an optimization algorithm that simulates the foraging behavior of ant colonies. Specifically, it uses computer programs to simulate the process of ants finding the shortest path between contaminated markers. By dynamically adjusting path selection through pheromone accumulation and volatilization mechanisms, it ultimately generates the optimal path connecting all contaminated markers. This algorithm can effectively reduce the movement time of the cleaning device and improve the efficiency of contaminated point location. The first cleaning path refers to a targeted cleaning route generated for the contaminated markers. Specifically, it can use a combination of straight lines or curves to connect all contaminated markers, covering all areas to be cleaned through the shortest path, avoiding repeated movements and ineffective cleaning actions. The second cleaning path refers to a pre-set full-coverage cleaning route based on the three-dimensional structure of the water storage tank. Specifically, it can use a grid-like or spiral path to cover the inner walls, top, and bottom of the water storage tank, ensuring that the entire area is thoroughly cleaned after the contaminated points are cleaned, eliminating potential contamination risks that were not detected.

[0054] In the water storage tank during cleaning, the path planning unit first calculates the shortest path connecting all the pollution markers using the ant colony algorithm based on the coordinates of the pollution markers provided by the pollution monitoring module. This path serves as the first cleaning path. The spraying device then sequentially sprays and cleans each pollution marker along this path until all markers are removed. Subsequently, the system automatically switches to the pre-stored second cleaning path, controlling the spraying device to comprehensively clean the inner walls, top, and bottom of the water storage tank according to a grid or spiral trajectory, ensuring thorough cleaning without any blind spots. During this process, the water storage tanks not in cleaning operation simultaneously clean their own pollution markers using a liquid suction device, enabling collaborative operation among multiple water tanks.

[0055] Traditional manual cleaning or single automated cleaning systems typically operate using fixed routes or random paths, which cannot dynamically optimize the path based on the distribution of contamination points. This results in low cleaning efficiency and the risk of omissions. This solution uses an ant colony algorithm to dynamically generate the shortest cleaning path for contamination points, and combines it with pre-stored full-coverage paths. This achieves both accurate and rapid treatment of contaminant areas and ensures thorough overall cleaning, significantly improving cleaning efficiency and quality.

[0056] This application solves the problems of resource waste and incomplete cleaning caused by unintelligent cleaning path planning in the prior art. By combining dynamic path optimization with full-coverage path, it achieves the dual goals of precise cleaning of contaminated points and efficient cleaning of the entire area, while reducing energy consumption for equipment movement and operation time.

[0057] When the switching control module determines that the contamination markers in a non-cleaning water tank have been removed by the suction device, the drag control module controls the suction device to complete the suction of the remaining contamination markers along the first cleaning path. After all the contamination markers in the water tank are removed, the suction device is then switched to the second cleaning path. After all the contamination markers in all non-cleaning water tanks have been cleaned, the switching control module randomly selects or selects any non-cleaning water tank in a preset order, and controls the suction device in that water tank to perform full-area suction along the second cleaning path to replenish the water volume in the transfer storage tank.

[0058] The first cleaning path refers to the shortest path, which is a combination of straight lines or curves generated based on the coordinates of the contamination markers. It can be generated using an ant colony algorithm to quickly locate and clean contamination points. The second cleaning path refers to a grid-like or spiral-shaped full-area coverage path that covers the inner walls, top, and bottom of the water storage tank. It can be pre-generated based on the 3D model of the water storage tank to ensure thorough cleaning without blind spots. Full-area water absorption refers to the liquid suction device performing a full-coverage water suction operation on the inner walls of the water storage tank along the second cleaning path. This can be achieved by using a multi-angle adjustable robotic arm to drive the liquid suction head to move along a predetermined trajectory to replenish the water volume in the transfer storage tank.

[0059] After the contamination markers in a water storage tank are initially cleaned by the suction device, the system automatically switches to the second cleaning path for deep cleaning. At this time, the suction head moves along a grid-like trajectory to fully cover the inner wall of the water storage tank, ensuring that residual contaminants are completely removed. After all non-working water storage tanks have completed the cleaning of contamination points, the system automatically selects any water storage tank to perform full-area water suction. For example, it prioritizes water storage tanks with more remaining water and uses their suction devices to transfer the water to a transfer storage tank, realizing the recycling of water resources.

[0060] Traditional cleaning systems can only perform fixed-point cleaning or simple cyclic rinsing along a single path. They cannot dynamically adjust the cleaning path according to the distribution of contamination points and lack a water replenishment mechanism after cleaning. This solution optimizes water resource utilization while ensuring cleaning effectiveness through a two-stage cleaning path switching and intelligent water replenishment strategy, avoiding the resource waste caused by repeated emptying and water injection in traditional methods.

[0061] This application achieves automatic connection between contamination point cleaning and full-area cleaning. After the target area is cleaned, it automatically switches to the full cleaning stage, effectively preventing contaminant residue. At the same time, it intelligently selects water storage tanks to replenish water, ensuring that the transfer storage tanks have a continuous water supply capacity, thus solving the problem of operation interruption caused by insufficient water in traditional cleaning processes.

[0062] The pollution monitoring module is equipped with a water quality sensor installed inside the water storage tank and a vision camera controlled by a towing device. The pollution monitoring module is also equipped with a pollution detection algorithm. The water quality sensor is used to monitor the concentration of pollutants inside the water storage tank, and the vision camera can capture images of the pollution on the inner wall of the water storage tank. The pollution detection algorithm module determines the pollution location and marks the pollution point based on the detection data from the water quality sensor and the vision camera.

[0063] Water quality sensors are devices used to detect the concentration of pollutants in the liquid inside a water storage tank. Specifically, they can be implemented using conductivity sensors, turbidity sensors, or dissolved oxygen sensors. By monitoring changes in water quality parameters in real time, they can identify polluted areas. The vision camera is a movable image acquisition device, which can be implemented using a high-resolution camera combined with a supplementary light. It is driven by a towing device to move along the inner wall of the water storage tank and capture images of surface pollution. The pollution detection algorithm is a processing module that integrates sensor data and image data. Specifically, it can use machine learning models or image recognition algorithms to locate polluted areas and generate coordinate information of pollution marker points by analyzing abnormal water quality data and visual image features.

[0064] The water quality sensor collects pollutant concentration data in the water storage tank in real time. When the concentration exceeds the preset threshold, the vision camera is triggered. The towing device drives the vision camera to move along the inner wall of the water storage tank to take pictures of the suspected polluted area from multiple angles. The pollution detection algorithm performs joint analysis on the sensor data and image data. For example, it determines the type of pollution by comparing it with the normal water quality parameter range. At the same time, it uses image recognition technology to identify the morphological characteristics of the attachments or sediments on the inner wall. When the algorithm determines that pollution exists, it generates a pollution marker point containing coordinate information and transmits the information to the towing control module to guide the liquid suction device or spray device to perform precise cleaning.

[0065] Traditional manual cleaning relies on experience to determine the location of contamination, resulting in blind spots and an inability to quantify the degree of contamination. In contrast, this solution uses collaborative monitoring by water quality sensors and vision cameras, combined with algorithms to perform multi-dimensional analysis of contamination data. This enables automatic identification and precise location of contaminated areas, avoiding subjective errors in manual inspection. Furthermore, the vision camera driven by the towed device can cover the entire area of ​​the water storage tank, solving the problem of manual inspections being unable to reach corners.

[0066] This application effectively improves the accuracy and efficiency of pollution detection, reduces repeated cleaning due to missed detection, and provides reliable data support for subsequent cleaning path planning, ensuring that the cleaning device can perform targeted operations on pollution marker points, reducing water waste and shortening the cleaning cycle.

[0067] The towing device includes a track installed inside the water storage tank, a drive motor that works with the track, and a multi-angle adjustable robotic arm controlled by the drive motor to move along the track. The liquid suction device, spraying device, and vision camera are all connected to the end of the multi-angle adjustable robotic arm. The track refers to the linear or circular guide structure installed inside the water storage tank, which can be implemented using embedded slide rails made of stainless steel or corrosion-resistant alloy, and is used to provide a movement path for the drive motor. The drive motor refers to the power device that works with the track, which can be implemented using a servo motor or stepper motor with a gear transmission mechanism, and is used to control the movement of the robotic arm along the track. The multi-angle adjustable robotic arm refers to an adjustable device with a multi-joint structure, which can be implemented using a three-axis or five-axis robotic arm combined with a hydraulic drive system, and is used to adjust the spatial position and attitude of the end effector. The liquid suction device, spraying device, and vision camera are connected through a fixed interface installed at the end of the robotic arm to form an integrated operating unit.

[0068] The track is arranged along the top or side of the inner wall of the water storage tank. The drive motor achieves directional movement by meshing with the track through gears. The base of the multi-angle adjustable robotic arm is fixed on the support platform of the drive motor. The working angle of the end effector is adjusted by joint rotation and extension. The vision camera captures the pollution image of the inner wall of the water storage tank in real time. The suction head of the suction device and the spray head of the spray device are precisely aligned with the pollution mark points under the drive of the robotic arm. For example, when the side wall contaminant is detected, the robotic arm can bend down 30 degrees to make the suction head fit against the wall surface; when the top stain is found, the robotic arm can extend up 60 degrees to make the spray head rinse vertically. The track can be set as a closed ring structure around the inner circumference of the water storage tank. The drive motor is equipped with an absolute encoder to achieve closed-loop position control. The end effector mounting base of the multi-angle adjustable robotic arm is designed with a quick-release structure to facilitate the replacement of the suction head, spray head or vision camera module. Waterproof sealing rings are set at the joints of the robotic arm to ensure stable operation in humid environments.

[0069] Traditional manual cleaning relies on workers to enter the water tank with hand tools, which has problems such as limited operating angle and low equipment switching efficiency. This solution achieves all-round coverage of the cleaning device through the coordinated control of the track and the robotic arm, avoiding the risk of personnel entering the confined space. The integrated end effector combination can simultaneously complete pollution detection, liquid suction and high-pressure rinsing, improving the continuity of operation compared with traditional step-by-step operation.

[0070] This application effectively solves the problems of insufficient spatial coverage and low equipment coordination efficiency in manual cleaning. The multi-degree-of-freedom motion characteristics of the robotic arm enable the cleaning device to reach the top, side walls and corner areas of the water tank. The linkage between the vision camera and the actuator ensures the accurate positioning and cleaning of the contamination markers. The directional movement function of the track system, combined with the flexible adjustment of the robotic arm, forms a three-dimensional space full-coverage cleaning operation mode, which significantly improves the integrity and reliability of automated cleaning.

[0071] The multi-tank partitioned dynamic cleaning system also includes a water quality detection module. This module is used to test the water quality in the cleaned water storage tanks. Once the water quality meets the standards, the cleaning operation stops and the water storage tanks are filled with water. The water quality detection module is a device used to monitor the cleanliness of the water in the water storage tanks in real time. Specifically, it can be implemented using a combination of dissolved oxygen sensors, turbidity sensors, and conductivity sensors. Through multi-parameter joint analysis, it determines whether the water quality meets the preset standards. The standard-meeting determination refers to the process of comparing the data collected by the sensors with the preset thresholds. Specifically, the thresholds can be dynamically adjusted through embedded processors or cloud computing platforms to adapt to the water quality requirements of different water storage tanks. The water filling operation refers to the step of automatically restoring the water supply function of the water storage tanks after the water quality meets the standards. Specifically, the connection between the external water source and the water storage tanks can be controlled by a solenoid valve to achieve an unattended water filling process.

[0072] After the cleaning operation is completed, the water quality detection module starts multi-dimensional detection of the residual water in the water storage tank. For example, the turbidity sensor monitors the concentration of suspended particulate matter, the conductivity sensor detects the ion content, and the dissolved oxygen sensor assesses the redox state of the water. When all detection parameters are lower than the preset threshold, the cleaning is deemed to have met the standards. The system automatically shuts off the spray device and triggers the water injection solenoid valve to inject clean water into the water storage tank through the preset pipeline. If the detection data does not meet the standards, the local cleaning process is restarted until the water quality requirements are met.

[0073] Traditional manual cleaning relies solely on visual inspection or sampling for testing to determine cleaning effectiveness, which suffers from strong subjectivity and delayed detection. This solution integrates multiple types of sensors to construct a closed-loop detection system, acquiring water quality data in real time during the cleaning process. This avoids the safety risks of manual inspection and allows for precise control of the termination of the cleaning operation, preventing water waste caused by over-cleaning. This application achieves quantifiable evaluation and automated control of the cleaning effect, effectively solving the problem of water quality judgment relying on human experience in traditional methods. Through a real-time data feedback mechanism, it ensures that each water storage tank meets a uniform water quality standard after cleaning. At the same time, the automated water injection process reduces manual intervention and improves the overall system operating efficiency.

[0074] An adaptive control method for multi-tank partitioned reinforcement learning-based dynamic cleaning includes: Initialization steps: The pollution monitoring module conducts comprehensive monitoring of each water storage tank, generates initial pollution markers, and the reinforcement learning module records and stores the image features of each pollution marker. Cleaning operation start-up steps: The cleaning control system allocates cleaning and non-cleaning working states according to the status of the water storage tank. Identification and classification steps: The reinforcement learning module analyzes the current pollution markers and compares them with the data in the stain classification repository to quickly identify cleanable and non-cleanable pollution markers; Path optimization planning steps: The path planning unit of the drag control module combines the classification results of the reinforcement learning module to prioritize the planning of the first cleaning path that passes through the cleanable type of contamination markers; Cleaning operation steps: When the water tank is in the cleaning working state, the cleaning-type contamination markers are preferentially sprayed and cleaned along the first cleaning path by the spray device; when the water tank is not in the cleaning working state, the cleaning-type contamination markers are preferentially suctioned and cleaned by the liquid suction device. Full-area cleaning steps: After all cleanable types of contamination markers have been cleaned, the drag control module switches to the second cleaning path to perform full-area coverage cleaning. At the same time, special cleaning methods are used for non-cleanable contamination markers. These special cleaning methods include pressurizing the spray device or liquid suction device and using a cleaning cloth installed on the drag device for reciprocating friction. Feedback learning steps: After cleaning is completed, the pollution monitoring module detects the cleaning effect and provides feedback to the reinforcement learning module. The reinforcement learning module updates the stain type label based on the detection results: successfully cleaned stains are labeled as cleanable, and unsuccessfully cleaned stains are labeled as non-cleanable. State loop steps: After a water storage tank is cleaned, the control module switches the working state of the water storage tank, and repeats the identification and classification steps, path optimization and planning steps, cleaning operation execution steps, full area cleaning steps, and feedback learning steps in sequence until all water storage tanks are cleaned to the standard.

Claims

1. A multi-tank partitioned dynamic cleaning system, characterized in that: The system includes a rooftop water tank, which is equipped with at least two independent water storage chambers. Each water storage chamber is equipped with a cleaning device. The cleaning device includes a liquid suction device, a spraying device, and a towing device. The towing device is used to drive the liquid suction device and the spraying device to move. The spraying device in any water storage chamber is connected to the liquid suction device in the adjacent water storage chamber. A cleaning control system is provided, which includes a pollution monitoring module and a towing control module. The pollution monitoring module is used to monitor pollution points inside the water storage tank and generate pollution markers. The towing control module controls the towing device to drive the liquid suction device and the spraying device to aim at the contaminated marker point; The rooftop water tank is also equipped with a transfer storage tank, and all the liquid suction devices and spraying devices of the water storage tanks are connected to the transfer storage tank. When any of the water storage tanks needs to be cleaned, the water storage tank in the non-cleaning state is cleaned by the liquid suction device, and the water storage tank in the cleaning state is cleaned by the spray device. The water storage tank in the non-cleaning state is cleaned by the liquid suction device to remove the contamination markers, and the water storage tank in the cleaning state is cleaned by the spray device to remove the contamination markers.

2. The multi-tank partitioned dynamic cleaning system according to claim 1, characterized in that: The pollution monitoring module is equipped with a water quality sensor installed inside the water storage tank and a vision camera controlled by a towing device. The pollution monitoring module is also equipped with a pollution detection algorithm. The water quality sensor is used to monitor the concentration of pollutants inside the water storage tank, and the vision camera can capture images of the pollution status on the inner wall of the water storage tank. The pollution detection algorithm module determines the pollution location and marks the pollution point based on the detection data from the water quality sensor and the vision camera.

3. The multi-tank partitioned dynamic cleaning system according to claim 2, characterized in that: The towing device includes a track installed in the water storage tank, a drive motor that cooperates with the track, and a multi-angle adjustable robotic arm controlled by the drive motor to move along the track. The liquid suction device, spraying device, and vision camera are all connected to the end of the multi-angle adjustable robotic arm.

4. The multi-tank partitioned dynamic cleaning system according to claim 1, characterized in that: The cleaning control system also includes a switching control module, which is used to switch the connection between the liquid suction device of different water storage tanks and the transfer storage tank. Once the contamination markers in a water storage tank that is not in a cleaning working state are cleaned by the liquid suction device, the switching control module switches to another water storage tank that is not in a cleaning working state. After all the contamination markers in the water storage tanks that are not in cleaning operation state have been absorbed by the liquid suction device, the switching control module selects the water storage tanks that are not in cleaning operation state according to a preset order or the remaining water volume to replenish the intermediate storage tank.

5. The multi-tank partitioned dynamic cleaning system according to claim 4, characterized in that: The transit storage box is equipped with a filter screen, which divides the transit storage box into an inlet area and an outlet area. The liquid suction device of all the water storage tanks is connected to the inlet area, and the spray device of all the water storage tanks is connected to the outlet area. The transit storage box is also equipped with a disinfection storage box. The disinfection storage box and all the spray devices in the water storage tanks are equipped with reversing valves. The reversing valves are used to drive the spray devices to connect with the disinfection storage box or the transit storage box. When it is necessary to disinfect the water storage tank for cleaning, the reversing valve drives the spray device to connect with the disinfection storage tank, and the disinfectant stored in the disinfection storage tank disinfects the water storage tank through the spray device. When performing a clean water rinsing operation, the reversing valve drives the spray device to connect with the water outlet area of ​​the transfer storage tank, and the water in the water outlet area rinses the water storage tank through the spray device.

6. The multi-tank partitioned dynamic cleaning system according to claim 5, characterized in that: The liquid suction device includes a liquid suction pump and a liquid suction tube. One end of the liquid suction tube is provided with a liquid suction head, which can be close to the inner wall of the water storage tank to suck up pollutants. The spray device includes a spray pump, a spray head, and a connecting pipe. The connecting pipe of the spray device is connected to a disinfection storage tank or a transfer storage tank through a reversing valve to clean the contamination markings on the inner wall of the water storage tank.

7. The multi-tank partitioned dynamic cleaning system according to claim 4, characterized in that: The dragging control module includes a path planning unit. The path planning unit can generate pollution marker coordinates based on the pollution monitoring module and generate a first cleaning path through an ant colony algorithm. The first cleaning path is the shortest path of a combination of straight lines or curves to connect all pollution marker points. The path planning unit also pre-stores a second cleaning path based on the three-dimensional model of the water storage tank. The second cleaning path is a grid-like or spiral-like full-area coverage path that covers the inner wall, top and bottom of the water storage tank. The towing control module controls the spraying device to spray and clean the contaminated markers in the water storage tank along the first cleaning path. After all the contaminated markers in the water storage tank are removed, the spraying device is then controlled to switch to the second cleaning path to perform full-coverage cleaning of the water storage tank.

8. The multi-tank partitioned dynamic cleaning system according to claim 7, characterized in that: When the switching control module determines that the contamination markers in a water tank in a non-cleaning working state have been removed by the liquid suction device, the dragging control module controls the liquid suction device to complete the liquid suction operation of the remaining contamination markers along the first cleaning path. After all the contamination markers in the water tank are removed, the liquid suction device is then controlled to switch to the second cleaning path. After all the contamination markers in the non-cleaning water storage tanks have been cleaned, the switching control module randomly selects or selects any non-cleaning water storage tank in a preset order, and controls the liquid suction device of that water storage tank to perform full-area water suction along the second cleaning path to replenish the water volume of the transfer storage tank.

9. The multi-tank partitioned dynamic cleaning system according to claim 1, characterized in that: The cleaning control system also includes a reinforcement learning module, which is equipped with a pollution classification repository. The reinforcement learning module is connected to the pollution monitoring module, receives data from pollution markers, and stores it in the pollution classification repository. The pollution monitoring module monitors the cleaning results of the pollution markers based on the liquid suction device and the spraying device. The reinforcement learning module labels the pollution markers as washable or non-washable based on the cleaning results. The pollution classification repository stores the pollution markers according to washable and non-washable types. When subsequent cleaning operations are performed, the reinforcement learning module identifies the type of subsequent pollution markers based on the stain classification repository and prioritizes cleaning the washable pollution markers.

10. An adaptive control method for multi-tank partitioned reinforcement learning dynamic cleaning, providing a multi-tank partitioned dynamic cleaning system as described in any one of claims 1-9, characterized in that, The control method includes: Initialization steps: The pollution monitoring module conducts comprehensive monitoring of each water storage tank, generates initial pollution markers, and the reinforcement learning module records and stores the image features of each pollution marker. Cleaning operation start-up steps: The cleaning control system allocates cleaning and non-cleaning working states according to the status of the water storage tank. Identification and classification steps: The reinforcement learning module analyzes the current pollution markers and compares them with the data in the stain classification repository to quickly identify cleanable and non-cleanable pollution markers; Path optimization planning steps: The path planning unit of the drag control module combines the classification results of the reinforcement learning module to prioritize the planning of the first cleaning path that passes through the cleanable type of contamination markers; Cleaning operation steps: When the water tank is in the cleaning working state, the cleaning-type contamination markers are preferentially sprayed and cleaned along the first cleaning path by the spray device; when the water tank is not in the cleaning working state, the cleaning-type contamination markers are preferentially suctioned and cleaned by the liquid suction device. Full-area cleaning steps: After all cleanable types of contamination markers have been cleaned, the drag control module switches to the second cleaning path to perform full-area coverage cleaning. At the same time, special cleaning methods are used for non-cleanable contamination markers, including pressurizing the spray device or liquid suction device. Feedback learning steps: After cleaning is completed, the pollution monitoring module detects the cleaning effect and provides feedback to the reinforcement learning module. The reinforcement learning module updates the stain type label based on the detection results: successfully cleaned stains are labeled as cleanable, and unsuccessfully cleaned stains are labeled as non-cleanable. State loop steps: After a water storage tank is cleaned, the control module switches the working state of the water storage tank, and repeats the identification and classification steps, path optimization and planning steps, cleaning operation execution steps, full area cleaning steps, and feedback learning steps in sequence until all water storage tanks are cleaned to the standard.

Images