Outdoor cabinet intelligent self-cleaning system adapting to complex environment

Abstract

The invention relates to the technical field of outdoor electrical equipment protection, and discloses an outdoor cabinet intelligent self-cleaning system adaptive to a complex environment, the self-cleaning system is composed of a cabinet shell, a dustproof module, a self-cleaning module, a dust collection groove, a detection module and a control module, the cabinet shell provides a high-strength and corrosion-resistant protection structure, and the dust collection groove is provided with a dust collection groove. Air circulation is realized through the air inlet and the air outlet; the dustproof module is used for intercepting external dust; the self-cleaning module is used for removing dust accumulated on a filter screen, the dust collecting groove is used for collecting dust generated in the cleaning process, the detection module is used for monitoring the environment and the state of the filter screen in real time and collecting key data, and the control module is used for data processing, intelligent decision making, control signal generation and dynamic cleaning strategy adjustment. Through cooperative work of the cabinet shell, the dustproof module, the self-cleaning module, the dust collection groove, the detection module and the control module, the system can accurately sense environment changes, self-adaptively trigger cleaning operation and efficiently reduce manual intervention.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of outdoor electrical equipment protection, in particular to an intelligent self-cleaning system for outdoor cabinets suitable for complex environments. BACKGROUND

[0002] Outdoor cabinets are widely used in communication, power and transportation fields to protect internal electronic equipment from environmental impact. However, in desert, urban or industrial environments, dust accumulation can cause filter screen blockage, reduce heat dissipation efficiency, and even cause equipment overheating, performance degradation or failure. In the prior art, some cabinets use fixed filter screens to intercept dust, but require regular manual cleaning, which is high in maintenance cost and low in efficiency; a few cabinets are equipped with automatic cleaning devices, but mostly rely on a single indicator (such as timing or pressure difference), lack adaptability to complex environments (such as humidity and dust concentration changes), and have fixed cleaning strategies that cannot be optimized according to geographical location differences.

[0003] In recent years, the development of sensor technology and machine learning has made it possible for intelligent environmental monitoring. However, existing solutions have not fully combined multi-sensor data fusion and online learning technology, and cannot dynamically adapt to dust characteristics in different environments. Therefore, there is an urgent need for a system that integrates dust prevention, self-cleaning and intelligent optimization to improve the reliability and maintenance efficiency of outdoor cabinets. SUMMARY

[0004] (I) Technical problems solved To solve the problems of high maintenance cost of traditional fixed filter screens, poor adaptability of single indicator triggered cleaning, and inability to dynamically adapt to dust characteristics in different geographical environments, leading to reduced heat dissipation efficiency and increased risk of equipment failure, the present application provides an intelligent self-cleaning system for outdoor cabinets suitable for complex environments, which has the advantages of multi-sensor data fusion perception, dynamic adaptive cleaning strategy, low power consumption and efficient dust removal, and remote visualized control.

[0005] (II) Technical solutions To achieve the above-mentioned purpose, the present application provides the following technical solutions: an intelligent self-cleaning system for outdoor cabinets suitable for complex environments, the self-cleaning system is composed of a cabinet shell, a dust prevention module, a self-cleaning module, a dust collection tank, a detection module and a control module; The cabinet shell is fixed to the cabinet vent, used to accommodate the dust prevention module, the self-cleaning module, the dust collection tank and the detection module, and provides a stable installation foundation and protective shell for the entire system; The dust prevention module includes multiple layers of filter screens and a sealing structure, the filter screens are composed of primary efficiency filter screens, and the sealing structure uses a silicone seal strip to intercept external dust, and ensures that air only passes through the filter screens into the cabinet, achieving effective filtration of air entering the cabinet; The self-cleaning module includes a vibration device and a back-blowing device, which are used to periodically shake off the dust accumulated on the filter screen by the vibration motor and blow off the residual dust accumulated on the filter screen by reversing the generated reverse airflow using the cooling fan, thereby achieving self-cleaning of the filter screen. The dust collection groove is located at the bottom of the shell and is used to collect the dust shaken off by the vibration device and blown off by the back-blowing device, and the dust is discharged outside the cabinet through the bottom gap of the air outlet louver during the operation of the device or through detachable cleaning. The detection module includes a multi-sensor array, specifically a differential pressure sensor, a dust concentration sensor, a temperature sensor, and a humidity sensor, which are used to monitor the pressure difference before and after the filter screen, the particle concentration at the air inlet, the ambient temperature, and the ambient humidity in real time. The control module is based on a microcontroller and integrates dust level detection algorithms and machine learning optimization functions, which are used to receive sensor data transmitted by the detection module, process the data through Kalman filter fusion, environmental correction, data smoothing, and outlier removal, calculate the dust level score, and send the calculated dust level score and confidence to S2.4 in real time to determine whether the self-cleaning phase and confidence are needed, and generate control signals according to the score level and confidence to drive the self-cleaning module to work, while optimizing the relevant parameters of dust level detection through online machine learning algorithms to dynamically adapt to different geographical environments. The cabinet shell, dust prevention module, self-cleaning module, dust collection groove, detection module, and control module are connected through mechanical and electrical connections. The dust prevention module, self-cleaning module, dust collection groove, and detection module are installed and fixed in the corresponding positions inside the cabinet shell. The vibration motor is fixed to the filter screen frame, the sensors of the detection module are connected to the feedback control interface of the control module through lines, and the control module is connected to the vibration motor and cooling fan through lines.

[0006] Preferably, the cabinet shell is fixed to the cabinet air inlet and is made of high-strength, corrosion-resistant materials. It contains the dust prevention module, self-cleaning module, and detection module inside, and has an air inlet and an air outlet outside, with a protection level of IP65.

[0007] Preferably, the dust prevention module includes multiple layers of filter screens and a sealing structure, which are used to intercept external dust. The filter screens are composed of G4 grade primary filter screens with a particle size interception of ≥5µm, and the sealing structure uses a silicone seal strip.

[0008] Preferably, the self-cleaning module includes a vibration device and a back-blowing device. The vibration device includes a vibration motor. The back-blowing device reverses the built-in cooling fan of the cabinet to generate a reverse airflow to blow off the dust and periodically shake off the dust. The self-cleaning module removes the dust accumulated on the filter screen through vibration and fan reversal.

[0009] Preferably, the dust collection trough is located at the bottom of the outer casing and is used to collect dust that is shaken off and blown away. It can be disassembled for cleaning or discharged outside the cabinet through the gap at the bottom of the air outlet louvers when the equipment is running.

[0010] Preferably, the detection module monitors the environment and filter status in real time through a multi-sensor array, specifically including a differential pressure sensor, a dust concentration sensor, a temperature sensor, and a humidity sensor.

[0011] Preferably, the differential pressure sensor has a measurement range of 0-100Pa and an accuracy of ±2Pa, and detects the pressure difference before and after the filter to reflect the degree of blockage. The dust concentration sensor has a measurement range of 0-1000 µg / m³. 3 Accuracy ±10µg / m 3 Monitor the concentration of particulate matter at the air inlet; The temperature sensor has a measurement range of -20℃ to 60℃, an accuracy of ±0.5℃, and calibrates pressure data. The humidity sensor has a measurement range of 0% to 100%RH, an accuracy of ±3%, and corrects for dust concentration data.

[0012] Preferably, the control module is based on a microcontroller, integrates dust level detection algorithms and machine learning optimization, and is responsible for data processing and control signal generation.

[0013] Preferably, the dust level detection algorithm flow is as follows: S1.1 Inputs: Pressure difference (P), dust concentration (C), temperature (T), humidity (H); S1.2 Preprocessing: Kalman filtering is used to fuse sensor data to reduce noise in the pressure difference, dust concentration, temperature and humidity data in S1.1; S1.3. Perform environmental correction calculations on the pressure difference, dust concentration, temperature, and humidity data after preprocessing in S1.2. The correction formula is as follows:

[0014] In the formula, To correct the pressure, The pressure after filtration. This indicates the temperature after filtration. When T=40℃, =0.936 , When H = 80%RH, =0.94 ; S1.4 Data Smoothing: The corrected pressure difference and the corrected dust concentration will be smoothed using a 5-point moving average filter to remove outliers other than 3σ. S1.5, Scoring: (1) Calculation of standardized scores:

[0015]

[0016] In the formula, represents the pressure standardized score, represents the dust standardized score, is the corrected pressure, represents the theoretical maximum value of the pressure index, represents the corrected dust concentration value, represents the theoretical maximum value of the dust concentration; (2) Dynamic threshold: , Based on the median and interquartile range of historical data; (3) The total score is represented by , , the initial , ; (4) Grade division is performed; (5) Dynamic adjustment of threshold: reduce 9-10% in high dust environment and increase 9-10% in low dust environment; (6) The calculation formula of confidence is: Confidence=

[0017] In the formula, , is the standard deviation and mean of the last 5 score values, is the standardized standard deviation of environmental factors, and 0.2 is the weight of environmental variability. The calculation of confidence can ensure the reliability of decision-making.

[0018] Preferably, the machine learning optimization process: S2.1, online learning: using small batch gradient descent, optimizing and , minimizing the prediction error, the optimized and are used for dust grade scoring in the next detection cycle, realizing closed-loop iteration; S2.2, the loss function is: , M=10 The loss value calculated by the loss function is updated every 9-10 samples, which is directly fed back to S2.1 to drive the gradient descent direction; S2.3, geographical adaptation: through historical data analysis, identify environmental characteristics, dynamically adjust weights and thresholds, and write the adjusted thresholds into the control logic of S2.4 in real time to ensure that the same region uses new thresholds for subsequent sampling; S2.4, control logic: (1) Low / Medium: Record data and continue monitoring; (2) High: Start vibration + fan reverse; (3) Serious: Start cleaning + alarm through CAN; (4) Low confidence: Delay 1-2 minutes for re-sampling; After the control action is completed, the system automatically records the action effect as a new sample back to S2.1 for online learning in the next round; S2.5, working principle: (1) Dust prevention stage: external air enters the cabinet through the filter screen, dust is intercepted, and clean air is used for equipment heat dissipation. The interception efficiency directly affects the differential pressure sensor reading, which becomes the trigger basis for S2.4 control logic; (2) Detection stage: the control module collects sensor data every 5-6 minutes, and after Kalman filtering, environmental correction and machine learning optimization, calculates the dust level score. The calculated dust level score and confidence are sent to S2.4 in real time to determine whether to enter the self-cleaning stage; (3) Self-cleaning stage: when the level is high or serious, start the vibration motor to shake off the dust, and the fan is reversed to blow off the residual dust into the dust collection tank. After cleaning, the differential pressure drops data is used as a positive sample back to S2.1 for updating the gradient descent direction; (4) Optimization stage: adjust the weights through online learning to adapt to the geographical environment. The adjustment result immediately affects the next detection cycle to form a complete closed loop of detection, decision, cleaning and learning.

[0019] Compared with the prior art, the present application provides an outdoor cabinet intelligent self-cleaning system that adapts to complex environments, which has the following beneficial effects: The present application forms a cooperative self-cleaning mechanism by integrating a vibration device and a reverse blowing device, combined with the efficient dust collection and processing structure of the dust collection tank, completely changing the traditional fixed filter screen mode relying on manual maintenance, thereby greatly reducing the operation and maintenance cost of the self-cleaning system, and avoiding the problem of reduced heat dissipation efficiency due to delayed manual cleaning, and ultimately improving the continuous and stable operation ability of the outdoor cabinet equipment.

[0020] The present application realizes accurate determination of the filter screen state and environmental characteristics by fusing the pressure difference, dust concentration, temperature and humidity sensors in the multi-sensor array, and simultaneously matching the Kalman filter data preprocessing and dynamic threshold scoring algorithm, so as to solve the defects of poor adaptability of single index triggering cleaning in the traditional self-cleaning technology, ensure that the cleaning action is started only when necessary, not only guarantee the timeliness of system cleaning, but also avoid the waste of energy consumption caused by invalid operation, and finally improve the intelligence and economy of system operation.

[0021] The machine learning optimization (small batch gradient dynamic weight adjustment) and geographical adaptation logic integrated by the control module can identify the dust characteristics of different geographical environments (such as high-dust desert and low-dust city) according to historical data, automatically adjust the cleaning strategy and threshold parameters, thereby effectively solving the problem of increased device failure risk caused by the inability of traditional systems to dynamically adapt to complex environments, enabling the system to maintain high-efficiency dust removal performance in various outdoor scenarios, and being applicable to less supervised deserts and coastal areas. BRIEF DESCRIPTION OF DRAWINGS

[0022] Figure 1 It is a system layout schematic diagram of the present application; Figure 2 It is an enlarged schematic diagram of the self-cleaning module of the present application; Figure 3 It is a sealing method diagram of the self-cleaning module of the present application; Figure 4 It is a sealing method diagram of the cabinet shell of the present application; Figure 5 It is a control module workflow diagram of the present application; 1, cabinet shell; 2, differential pressure sensor; 3, dust prevention module; 4, dust concentration sensor; 5, temperature and humidity sensor; 6, control module; 7, self-cleaning module; 11, cabinet shell door plate assembly; 12, cabinet door plate and cabinet sealing strip; 31, dust prevention module mounting; 32, dust prevention module and cabinet sealing strip; 33, dust prevention louver air duct assembly; 34, dustproof cotton; 35, cooling fan / self-cleaning fan; 71, self-cleaning module vibration device; 72, self-cleaning module dust collection groove. DETAILED DESCRIPTION

[0023] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0024] Please refer to Figure 1 - Figure 5An outdoor cabinet intelligent self-cleaning system suitable for complex environments, the self-cleaning system is composed of a cabinet shell, a dust prevention module, a self-cleaning module, a dust collection tank, a detection module and a control module; The cabinet shell 1 is fixed to the cabinet air vent and is made of high-strength and corrosion-resistant material. The outside is provided with an air inlet and an air outlet, and the protection level reaches IP65. It is used to accommodate the dust prevention module 3, the self-cleaning module 77, the dust collection tank 72 and the detection module, and provides a stable installation foundation and a protective shell for the entire system, ensuring that the internal modules are not directly affected by the external harsh environment. The dust prevention module 3 includes multiple layers of filter screens and a sealing structure. The filter screens are composed of primary efficiency filter screens (G4 level, intercepting particle size ≥5µm), and the sealing structure uses a silica gel sealing strip 32 to intercept external dust. The sealing structure ensures that air only enters the cabinet through the filter screens, effectively filtering the air entering the cabinet and reducing dust entering the cabinet interior. The self-cleaning module 7 includes a vibration device and a back-blowing device 71. The vibration device contains a vibration motor (power 5W, frequency 50Hz, amplitude 1mm). The back-blowing device utilizes the built-in cooling fan 35 of the cabinet to reverse, periodically shakes off the dust accumulated on the filter screens through the vibration motor, and simultaneously blows off the residual dust on the filter screens using the reverse airflow generated by the reverse cooling fan 35, achieving self-cleaning of the filter screens. The dust collection tank 72 is located at the bottom of the shell with an inclination angle of 15° and a capacity of 0.5L. It is used to collect the dust shaken off by the vibration device and blown off by the back-blowing device 71. The dust can be removed outside the cabinet through a detachable cleaning or through the gap at the bottom of the air outlet louver 33 during equipment operation. The detection module includes a multi-sensor array, specifically a differential pressure sensor 2 (range 0-100Pa, accuracy ±2Pa), a dust concentration sensor 4 (range 0-1000µg / m 3 , accuracy ±10µg / m 3 ), a temperature sensor (range -20℃ to 60℃, accuracy ±0.5℃) and a humidity sensor (range 0-100%RH, accuracy ±3%) 5, which are used to monitor the pressure difference before and after the filter screens (reflecting the degree of blockage), the particle concentration at the air inlet, the environmental temperature (correcting the pressure data) and the environmental humidity (correcting the dust concentration data) in real time. The control module 6 is based on a microcontroller and integrates dust level detection algorithms and machine learning optimization functions. It receives sensor data transmitted by the detection module, processes the data through Kalman filter fusion, environmental correction, data smoothing and outlier removal, calculates the dust level score, and sends the calculated dust level score and confidence to S2.4 in real time to determine whether to enter the self-cleaning phase and the confidence. According to the score level and confidence, a control signal is generated to drive the self-cleaning module 77 to work. At the same time, the online machine learning algorithm (small batch gradient descent) is used to optimize the related parameters of dust level detection, dynamically adapting to different geographical environments. The cabinet shell, dust prevention module, self-cleaning module, dust collection tank, detection module and control module are connected through mechanical connection and electrical connection. The dust prevention module, self-cleaning module, dust collection tank and detection module are all fixedly installed in the corresponding positions inside the cabinet shell, wherein the vibration motor is fixed to the filter screen frame, each sensor of the detection module is connected to the feedback control interface of the control module through a line, and the control module is connected to the vibration motor and the cooling fan through a line to realize data transmission and control signal transmission. In the working process, external air enters from the air inlet of the cabinet shell, passes through the filter screen of the dust prevention module and enters the cabinet to cool the internal equipment, and the dust is intercepted by the filter screen; the detection module collects sensor data every 5 minutes and transmits the data to the control module; the control module processes the received data, calculates the dust level score, and sends the calculated dust level score and confidence to S2.4 in real time to determine whether to enter the self-cleaning stage and the confidence, when the level is high and the confidence is ≥0.7, a control signal is generated to drive the vibration motor to work for 30 seconds to shake off the dust, and then drive the cooling fan to reverse for 60 seconds to blow away the residual dust, which falls into the dust collection tank; when the level is serious and the confidence is ≥0.7, the cleaning is started at the same time and an alarm is given through CAN; the control module also continuously adjusts the related parameters through online learning to adapt to different geographical environments and ensure stable and efficient operation of the system.

[0025] Specifically, the cabinet shell is fixed to the cabinet air vent and is made of high-strength and corrosion-resistant material, and internally accommodates the dust prevention module, the self-cleaning module and the detection module, and externally provided with an air inlet and an air outlet, with an IP65 protection level.

[0026] The advantages are: by using high-strength corrosion-resistant materials, setting IP65 protection level and air inlet and outlet structure, the system can run stably in complex outdoor environment and provide reliable protection for internal modules.

[0027] Specifically, the dust prevention module includes multiple layers of filter screens and a sealing structure for intercepting external dust, and the filter screens are composed of G4 primary filter screens with an interception particle size of ≥5µm, and the sealing structure adopts a 32 silica gel sealing strip to ensure that air only enters the cabinet through the filter screens.

[0028] The advantages are: by combining multiple layers of G4 primary filter screens (intercepting ≥5µm particle size) with a silica gel sealing structure, the system can efficiently intercept dust and avoid air leakage, thereby ensuring a clean cooling environment for the internal equipment of the cabinet.

[0029] Specifically, the self-cleaning module includes a vibration device and a back-blowing device 71, the vibration device 71 includes a vibration motor (power 5W, frequency 50Hz, amplitude 1mm); the back-blowing device 71 reverses by the built-in heat dissipation fan 35 in the cabinet, generates a reverse airflow to blow off the accumulated dust, regularly shakes off the accumulated dust, and the self-cleaning module 7 removes the accumulated dust on the filter screen through vibration and fan reversal.

[0030] The advantage is that the vibration in the self-cleaning module and the fan reversal are combined to remove the accumulated dust on the filter screen, form a self-cleaning mechanism, can make the system efficiently remove the accumulated dust, and reduce manual maintenance.

[0031] Specifically, the dust collection groove is located at the bottom of the shell, the inclination angle of the dust collection groove is 15°, and the capacity is 0.5L, which is used for collecting the dust shaken off and blown off, can be disassembled and cleaned, or discharged outside the cabinet through the gap at the bottom of the air outlet louver 33 during the operation of the equipment.

[0032] The advantage is that the dust collection groove collects the dust shaken off and blown off, so that the system can centrally process the dust, is convenient for cleaning and maintenance, the inclination angle design and the detachable structure facilitate dust discharge and cleaning, and reduce the influence of dust accumulation on the equipment.

[0033] Specifically, the detection module monitors the environment and the filter screen state in real time through a multi-sensor array, provides an accurate trigger signal for the self-cleaning module, and ensures the timeliness and effectiveness of cleaning, and specifically includes a differential pressure sensor, a dust concentration sensor, a temperature sensor and a humidity sensor.

[0034] Specifically, the measurement range of the differential pressure sensor is 0-100Pa, the accuracy is ±2Pa, the pressure difference before and after the filter screen is detected, and the blockage degree is reflected; The measurement range of the dust concentration sensor is 0-1000µg / m 3 , the accuracy is ±10µg / m 3 , and the particulate matter concentration at the air inlet is monitored; The measurement range of the temperature sensor is -20℃ to 60℃, the accuracy is ±0.5℃, and the pressure data is corrected; The measurement range of the humidity sensor is 0%~100%RH, the accuracy is ±3%, and the dust concentration data is corrected.

[0035] The advantage is that the multi-sensor is fused, the pressure difference in the differential pressure sensor, the dust concentration in the dust concentration sensor, the temperature in the temperature sensor and the humidity in the humidity sensor are combined, and the reliability of the data in the system is improved through Kalman filtering.

[0036] Specifically, the control module is based on a microcontroller, integrates a dust level detection algorithm and machine learning optimization, is responsible for data processing and control signal generation, and through the dust level detection algorithm and machine learning optimization, the system can dynamically adjust the cleaning strategy and adapt to the dust characteristics in different geographical environments.

[0037] Specifically, the dust level detection algorithm process is as follows: S1.1, input: pressure difference (P), dust concentration (C), temperature (T), humidity (H); S1.2, preprocessing: using Kalman filtering to fuse sensor data, reducing the noise of pressure difference, dust concentration, temperature and humidity data in S1.1; S1.3, the pressure difference, dust concentration, temperature and humidity data after S1.2 preprocessing are calculated by environment correction, and the correction formula is:

[0038] In the formula, is the corrected pressure (unit: Pa), is the filtered pressure (unit: Pa), represents the filtered temperature (unit: ℃), when T=40℃, =0.936 , when H=80%RH, =0.94 ; S1.4, data smoothing: smoothing the corrected pressure difference and the corrected dust concentration, 5-point moving average filtering (moving average filtering can smooth the signal, and the average value of one-dimensional signal in a certain window is calculated to reduce the influence of noise data), and the abnormal value outside 3σ is eliminated (data abnormal value processing, reducing noise data interference, improving analysis accuracy); S1.5, scoring: (1) calculation of normalized score:

[0039]

[0040] In the formula, represents the pressure normalized score, represents the dust normalized score, is the corrected pressure, represents the theoretical maximum value of the pressure index, represents the corrected dust concentration value, represents the theoretical maximum value of the dust concentration; (2) dynamic threshold: , Calculated based on historical data, median, and interquartile range; (3) The total score is determined by express, ,initial , ; (4) Classify into levels:

[0041] (5) Dynamically adjust threshold: reduce by 10% in high dust environment and increase by 10% in low dust environment. The dynamic threshold is adaptively adjusted based on historical data, which helps to enhance environmental adaptability. (6) The formula for calculating the confidence level is: Confidence=

[0042] In the formula, , The standard deviation and mean of the last 5 scores. The standard deviation of environmental factors is 0.2, which is the weight of environmental variability. The calculation of confidence level can ensure the reliability of decision-making, thereby avoiding misoperation and improving system stability.

[0043] The advantages are: by learning online and dynamically adjusting weights, the system's adaptability and reliability are further improved, and the risk of misoperation is reduced.

[0044] Specifically, the machine learning optimization process: S2.1, Online Learning: Optimization using mini-batch gradient descent (batch size 10) , Minimize prediction error, optimized and The dust level score is immediately used for the next testing cycle, achieving closed-loop iteration. S2.2, The loss function is: M=10 The loss value calculated by this loss function is updated every 9-10 samples and directly backpropagated to S2.1 to drive the gradient descent direction; S2.3, Geographic Adaptation: Through historical data analysis (100 samples), environmental characteristics (such as high-dust deserts and low-dust cities) are identified, and weights and thresholds are dynamically adjusted. The adjusted thresholds are immediately written into the control logic of S2.4 to ensure that subsequent sampling in the same area immediately adopts the new thresholds. S2.4, Control Logic: (1) Low / Medium: Record data and continue monitoring; (2) High (confidence ≥ 0.7): Start vibration (30 seconds) + fan reverse (60 seconds); (3) Severe (confidence ≥ 0.7): Start cleaning + alarm through CAN; (4) Low confidence: Delay 1 minute for re-sampling; After the control action is completed, the system automatically records the action effect (pressure difference change, dust concentration drop) as a new sample backflow S2.1 for online learning in the next round; S2.5, working principle: (1) Dust prevention stage: external air enters the cabinet through the filter screen, and dust is intercepted. The clean air is used for equipment heat dissipation. The interception efficiency directly affects the pressure difference sensor reading, which becomes the trigger basis for the control logic of S2.4; (2) Detection stage: the control module collects sensor data every 5 minutes, which is filtered by Kalman filter, corrected by environment, and optimized by machine learning. The dust grade score is calculated, and the calculated dust grade score and confidence are sent to S2.4 in real time to determine whether it needs to enter the self-cleaning stage; (3) Self-cleaning stage: when the grade is high or severe, the vibration motor is started to shake off the dust, and the fan is reversed to blow off the residual dust into the dust collection tank. After cleaning, the pressure difference drop data is used as a positive sample backflow S2.1 to update the gradient descent direction; (4) Optimization stage: adjust the weight through online learning to adapt to the geographical environment. The adjustment result immediately affects the next detection cycle to form a complete closed loop of detection, decision, cleaning and learning, ensuring that the system can run with high precision and low energy consumption in complex environments such as deserts, coasts and cities, and ensure detection accuracy.

[0045] Advantages: In machine learning optimization, the weight is dynamically adjusted through online small batch gradient descent, which can adapt the system to different geographical environments.

[0046] This invention's self-cleaning system comprises a cabinet shell, a dustproof module, a self-cleaning module, a dust collection tank, a detection module, and a control module. The cabinet shell provides a high-strength, corrosion-resistant protective structure, is fixed to the cabinet's ventilation opening, and houses the internal modules, allowing airflow through inlets and outlets. The dustproof module intercepts external dust, ensuring clean air enters the cabinet through multiple layers of G4-grade pre-filters and a silicone sealing structure. The self-cleaning module removes dust from the filters, using a vibration device to shake them off and a reverse fan to generate a reverse airflow that blows away the dust, forming a self-cleaning mechanism. The dust collection tank collects dust generated during the cleaning process, featuring a 15° tilt design and a detachable design. The disassembly structure enables convenient handling or automatic discharge of accumulated dust; the detection module is used to monitor the environment and filter status in real time, collecting key data through a multi-sensor array (differential pressure, dust concentration, temperature, and humidity sensors); the control module is used for data processing and intelligent decision-making, based on a microcontroller-integrated dust level detection algorithm and machine learning optimization, generating control signals and dynamically adjusting the cleaning strategy. Through the combined and coordinated work of the above components, the system of this invention can accurately perceive environmental changes, adaptively trigger cleaning operations, and efficiently reduce manual intervention. It has the advantages of multi-sensor data fusion perception, dynamic adaptive cleaning strategy, low-power and high-efficiency dust removal, and remote visual management and control.

[0047] The system of this invention is applicable to various types of outdoor cabinets, especially in desert and coastal areas with less regulation, which can greatly improve equipment operating efficiency and reduce maintenance difficulty.

[0048] Example 1 like Figure 1 As shown ( Figure 1 This is a schematic diagram of the system layout, showing the layout of the outer shell, dustproof module, self-cleaning module, detection module, and control module. The present invention provides an outdoor cabinet dustproof and self-cleaning system, including an outer shell 1, a dustproof module 2, detection modules 3 / 4 / 5, a control module 6, and a vibration module 7. The outer shell 1 is made of SGCC (hot-dip galvanized cold-rolled steel sheet) and is fixed to the cabinet ventilation opening, housing all modules inside. The dustproof module 2 includes a primary G4 grade filter 34, which is fixed behind the air inlet. like Figure 2 As shown ( Figure 2 This is an enlarged schematic diagram of the self-cleaning module, showing the vibration motor and fan. The self-cleaning module includes a vibration motor 36 (power 15 W) and a cooling fan 35 (model 18060). The vibration motor 36 is fixed to the filter frame by an L-shaped bracket and transmits vibration with an amplitude of 1 mm. When the fan 35 reverses, it generates a reverse airflow, blowing away the accumulated dust to the dust collection tank 37, which is located at the bottom of the entire dustproof module for easy collection. The detection module includes a differential pressure sensor 3, a dust concentration sensor 4, a temperature sensor, and a humidity sensor 5, which are respectively connected to the feedback control interface of the control module 6; The control module 6 employs a microcontroller, running a dust level detection algorithm ( Figure 3 ) ; As Figure 3 shown (the figure shows the self-cleaning module sealing method), the working process of the control module 6 is as follows: (1) Initialize Kalman filter, load historical data (100 samples), set initial weights ( , ) ; (2) Read sensor data every 5 minutes, fuse through Kalman filter, get , , , ; (3) Environmental correction: calculate , ; (4) Dynamic threshold: calculate , , adjust threshold (30, 60, 90) ; (5) Score: calculate , , score ; (6) Machine learning: perform small batch gradient descent every 10 samples, update , ; (7) Confidence: calculate Confidence, if ≥0.7, perform high / severe level action; (8) Test results: in a desert environment with a median of 600 µg / m 3 , the system detects P=60Pa, C=600µg / m 3 , T= 30℃, H=70%, calculates Score=72.25, the level is high, the confidence is 0.92, triggers cleaning, the pressure drops to 20Pa, verifying the self-cleaning effect.

[0049] Example 2 In an urban environment with a median of 150 µg / m 3 , the threshold is adjusted by 10% (33, 66, 99), the system detects P=40Pa, C=200µg / m 3 , calculates Score=45, the level is medium, no cleaning is needed, optimizing energy consumption; (1) Machine learning adjustment ( =0.7, =0.3), adapting to low dust environment; (2) According to the system flow setting of the application, the vibration motor power of the self-cleaning module is 5W, the frequency is 50Hz, the air volume of the back blowing fan is 50m 3 / h, and the cleaning period is vibration for 30 seconds plus back blowing for 60 seconds. (3) According to the system flow setting of the application, the control module calculates the confidence degree, and only when the confidence degree is greater than or equal to 0.7, the cleaning or alarm action is executed.

[0050] Summary: The above embodiment 1 in the high dust desert environment, the system through multi-sensor data fusion and dust grade detection algorithm, accurately identifies the high blockage state of the filter screen, and triggers the vibration and back blowing cooperative cleaning, so that the pressure is reduced from 60Pa to 20Pa, which directly shows the high efficiency of the application of the self-cleaning mechanism in the system; Through machine learning optimization, the characteristics of high dust environment are dynamically adapted to verify the stable dust removal capacity of the system in extreme environment; while embodiment 2 in the low dust urban environment, the system automatically adjusts the threshold by 10%, and adjusts the weight through machine learning to adapt to the low dust characteristics, and does not trigger the cleaning action for the medium pollution state, so as to realize the energy consumption optimization; At the same time, through the confidence degree determination logic, invalid operation is avoided, to verify the intelligent decision and energy saving performance of the system in the low dust environment, the above two embodiments jointly verify that the system can dynamically adjust the cleaning strategy according to the dust characteristics of different geographical environments, both have high dust removal capacity in high dust environment and energy saving performance in low dust environment, which fully embodies the practical effect of multi-sensor fusion sensing, machine learning self-adaptive optimization and cooperative self-cleaning mechanism in the system.

[0051] Figure 4 It is a cabinet shell sealing mode diagram, which shows the cabinet shell sealing mode. Figure 5 It is a control module work flow diagram, which shows sensor data processing, dust grade detection and machine learning optimization logic.

[0052] Although the embodiments of the application have been shown and described, it can be understood by those of ordinary skill in the art that various changes, modifications, replacements and variations can be made to these embodiments without departing from the principles and spirits of the application, and the scope of the application is defined by the appended claims and their equivalents.

Claims

1. An outdoor cabinet intelligent self-cleaning system suitable for complex environments, characterized in that, The self-cleaning system is composed of a cabinet shell, a dustproof module, a self-cleaning module, a dust collection tank, a detection module and a control module. The cabinet shell (1) is fixed to the cabinet air vent and is used to accommodate the dustproof module (3), the self-cleaning module, the dust collection tank (72) and the detection module, and provides a stable installation foundation and a protective shell for the entire system. The dustproof module (3) includes multiple layers of filter screens and a sealing structure, the filter screens are composed of primary efficiency filter screens, and the sealing structure adopts a silica gel sealing strip (32) to intercept external dust and ensure that air only passes through the filter screens to enter the cabinet, thereby effectively filtering the air entering the cabinet. The self-cleaning module includes a vibration device and a back-blowing device (71) to periodically shake off dust accumulated on the filter screens through a vibration motor, and to blow off residual dust accumulated on the filter screens through a reverse airflow generated by a cooling fan (35) in reverse, thereby achieving self-cleaning of the filter screens. The dust collection tank (72) is located at the bottom of the shell and is used to collect dust shaken off by the vibration device and blown off by the back-blowing device (71), and the dust is discharged outside the cabinet through a bottom gap of the air outlet louver (33) during device operation or through detachable cleaning. The detection module includes a multi-sensor array, specifically a differential pressure sensor (2), a dust concentration sensor (4), a temperature sensor and a humidity sensor (5), to monitor the pressure difference before and after the filter screens, the particle concentration at the air inlet, the environmental temperature and the environmental humidity in real time. The control module (6) is based on a microcontroller and integrates a dust level detection algorithm and a machine learning optimization function, is used to receive sensor data transmitted by the detection module, and after processing such as Kalman filter fusion, environmental correction, data smoothing and outlier rejection, calculates a dust level score and a confidence level, determines whether to enter a self-cleaning phase and the confidence level, and generates a control signal according to the score level and the confidence level to drive the self-cleaning module (7) to work, while optimizing related parameters of dust level detection through an online machine learning algorithm to dynamically adapt to different geographical environments. The cabinet shell, the dustproof module, the self-cleaning module, the dust collection tank, the detection module and the control module are connected through mechanical and electrical connections, the dustproof module, the self-cleaning module, the dust collection tank and the detection module are all installed and fixed at corresponding positions inside the cabinet shell, the vibration motor is fixed to the filter screen frame, each sensor of the detection module is connected to the feedback control interface of the control module through a line, and the control module is connected to the vibration motor and the cooling fan through a line.

2. The intelligent self-cleaning system for outdoor cabinet adapting to complex environment according to claim 1, characterized in that: The cabinet shell is fixed to the cabinet air vent, is made of high-strength and corrosion-resistant materials, accommodates the dustproof module, the self-cleaning module and the detection module inside, and is provided with an air inlet and an air outlet outside, with a protection level of IP65.

3. The intelligent self-cleaning system for outdoor cabinet adapting to complex environment according to claim 1, characterized in that: The dustproof module includes multiple layers of filter screens and a sealing structure to intercept external dust, the filter screens are composed of primary efficiency filter screens with a particle size interception of ≥5µm and a G4 level, and the sealing structure adopts a silica gel sealing strip (32).

4. The intelligent self-cleaning system for outdoor cabinet adapting to complex environment according to claim 1, characterized in that: The self-cleaning module includes a vibration device and a back-blowing device (71), the vibration device (71) includes a vibration motor, the back-blowing device (71) is reversed by a cabinet built-in cooling fan (35), reverse airflow is generated to blow off accumulated dust, and accumulated dust is shaken off regularly; the self-cleaning module (7) removes accumulated dust on the filter screen through vibration and fan reversal.

5. The intelligent self-cleaning system for outdoor cabinet adapting to complex environment according to claim 1, characterized in that: The dust collection groove is located at the bottom of the shell and is used for collecting the dust shaken off and blown off, and can be removed and cleaned or discharged outside the cabinet through the gap between the bottom of the air outlet louver during equipment operation.

6. The intelligent self-cleaning system for outdoor cabinet adapting to complex environment according to claim 1, characterized in that: The detection module monitors the environment and the filter screen state in real time through a multi-sensor array, specifically including a differential pressure sensor, a dust concentration sensor, a temperature sensor and a humidity sensor.

7. The intelligent self-cleaning system for outdoor cabinet adapting to complex environment according to claim 6, characterized in that: The measurement range of the differential pressure sensor is 0-100 Pa, the accuracy is ±2 Pa, the pressure difference before and after the filter screen is detected, and the blockage degree is reflected; The dust concentration sensor has a measurement range of 0-1000 pg / m 3 , accuracy ± 10 pg / m 3 , monitors the particle concentration at the air inlet The measurement range of the temperature sensor is -20℃ to 60℃, the accuracy is ±0.5℃, and the pressure data is corrected; The measurement range of the humidity sensor is 0% to 100% RH, the accuracy is ±3%, and the dust concentration data is corrected. 8.The intelligent self-cleaning system of outdoor cabinet adapting to complex environment according to claim 1, characterized in that: The control module is based on a microcontroller, integrates a dust level detection algorithm and machine learning optimization, and is responsible for data processing and control signal generation.

9. The intelligent self-cleaning system for outdoor cabinet adapting to complex environment according to claim 8, characterized in that: The dust level detection algorithm process is as follows: S1.1, input: differential pressure (P), dust concentration (C), temperature (T), humidity (H); S1.2, preprocessing: use Kalman filtering to fuse sensor data to reduce the noise of the differential pressure, dust concentration, temperature and humidity data in S1.1; S1.3, the differential pressure, dust concentration, temperature and humidity data after S1.2 preprocessing are calculated by environment correction, and the correction formula is: In the formula, Pc = 0.95 Pf = 0.95 Tf = 40°C = 0.936 , Hc = 80% RH = 0.94 ; S1.4, data smoothing: smooth the corrected differential pressure and corrected dust concentration, 5-point moving average filtering, and remove abnormal values outside 3σ; S1.5, scoring: (1) calculation of standardized score: In the formula, represents a pressure standardized score, represents a dust standardized score, is a corrected pressure, represents a theoretical maximum value of a pressure index, represents a corrected dust concentration value, represents a theoretical maximum value of a dust concentration; (2) Dynamic threshold: , Based on median and interquartile range of historical data (3) The total score is represented by , , initial , ; (4) grade division; (5) dynamic adjustment of threshold: reduce 9%-10% in high-dust environment and increase 9%-10% in low-dust environment; (6) the calculation formula of confidence is: Confidence = In the formula, , The standard deviation and the average of the last 5 score values are, The standardized standard deviation of the environmental factors, and 0.2 is the weight of the environmental variability. The calculation of the confidence can ensure the reliability of the decision.

10. The intelligent self-cleaning system for outdoor cabinet adapting to complex environment according to claim 8, characterized in that: The machine learning optimization process is as follows: S2.1, online learning: using mini-batch gradient descent, optimize with minimizing prediction error, optimized with dirt level score for next detection cycle, enabling closed loop iteration; S2.2, the loss function is: , M=10 The loss value calculated by the loss function is updated every 9-10 samples, directly fed back to S2.1, and driven in the gradient descent direction; S2.3, geographical adaptation: identify environmental characteristics through historical data analysis, dynamically adjust weights and thresholds, and write the adjusted thresholds into the control logic of S2.4 in real time to ensure that new thresholds are immediately used for subsequent sampling in the same region; S2.4, control logic: (1) low / medium: record data and continue monitoring; (2) high: start vibration + fan reversal; (3) severe: start cleaning + alarm through CAN; (4) low confidence: delay 1-2 minutes to resample; After the control action is completed, the system automatically records the action effect as a new sample and returns to S2.1 for online learning in the next round; S2.5, working principle: (1) Dust prevention stage: External air enters the cabinet through the filter screen, and dust is intercepted. Clean air is used for equipment heat dissipation. The interception efficiency directly affects the differential pressure sensor reading, which becomes the trigger basis for S2.4 control logic; (2) Detection stage: The control module collects sensor data every 5-6 minutes, and after Kalman filtering, environmental correction and machine learning optimization, calculates the dust grade score. The calculated dust grade score and confidence are sent to S2.4 in real time to determine whether it needs to enter the self-cleaning stage; (3) Self-cleaning stage: When the level is high or serious, start the vibration motor to shake off the dust, and the fan is reversed to blow off the residual dust into the dust collection tank. After cleaning, the differential pressure drop data is used as a positive sample to return to S2.1 to update the gradient descent direction; (4) Optimization stage: Adjust the weight through online learning to adapt to the geographical environment. The adjustment result immediately affects the next detection cycle to form a complete closed loop of detection, decision, cleaning and learning.

Images