Intelligent cleaning operation control method for reverse osmosis membrane

Abstract

The invention discloses a reverse osmosis membrane intelligent cleaning operation control method, which belongs to the field of environmental engineering and intelligent control, and comprises the following steps: obtaining organic matter concentration, inorganic matter related particle information and microorganism concentration, and forming an original data set; extracting characteristic parameters of data of each sensor, fusing the characteristic parameters into a multi-dimensional characteristic vector, inputting the multi-dimensional characteristic vector into a convolutional neural network model, outputting a pollution type and a concentration value, and identifying the pollution type and the concentration; data are synchronously collected through multiple sensors, multi-dimensional feature vectors are constructed through preprocessing and feature extraction, pollution types and concentrations are accurately recognized through a convolutional neural network, an optimal cleaning scheme is automatically matched according to the pollution types and concentrations, and meanwhile a closed-loop optimization mechanism is formed by combining real-time monitoring feedback. The problem of blind cleaning caused by the fact that pollution types cannot be distinguished is solved, damage to the membrane due to improper cleaning agents is avoided, the service life of the membrane is prolonged, efficient and stable operation of a membrane system can be guaranteed, the water quality stability of produced water is improved, and the maintenance cost is reduced.

Description

Technical Field

[0001] This invention belongs to the field of environmental engineering and intelligent control technology, specifically relating to an intelligent cleaning and operation control method for reverse osmosis membranes. Background Technology

[0002] In the fields of industrial water treatment and environmental protection, reverse osmosis (RO) membrane systems are widely used as core equipment in industrial water purification, wastewater treatment, and water resource recycling. However, membrane fouling has become a key bottleneck restricting its efficient and stable operation: traditional detection methods mainly rely on macroscopic indicators such as pressure difference and permeate flow rate, or adopt timed chemical cleaning strategies. These methods have problems such as the inability to accurately identify the type of fouling (organic / inorganic / biological fouling) and the lack of targeted cleaning solutions, leading to over-cleaning or under-cleaning, accelerating membrane aging, and significantly increasing operation and maintenance costs. Summary of the Invention

[0003] The purpose of this invention is to provide an intelligent cleaning and operation control method for reverse osmosis membranes, in order to solve the problem that traditional reverse osmosis membrane systems rely on macroscopic indicators such as pressure difference and permeate flow rate or timed cleaning strategies for fouling detection, which cannot accurately identify the types of organic, inorganic, and biological fouling, and the cleaning schemes lack specificity, leading to over-cleaning or under-cleaning, which affects the stable operation and service life of the membrane.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a reverse osmosis membrane intelligent cleaning and operation control method, comprising the following steps:

[0005] Step S1: Obtain organic matter concentration, inorganic particulate information, and microbial concentration to form the original dataset;

[0006] Step S2: After preprocessing the data from step S1, extract the feature parameters of each sensor data and fuse them into a multi-dimensional feature vector. Input the vector into the convolutional neural network model and output the pollution type and concentration value to identify the pollution type and concentration.

[0007] Step S3: Based on the type and concentration of pollution, and according to the light, moderate and heavy pollution conditions, match the corresponding cleaning agent concentration, pH value and temperature parameters for organic matter, inorganic matter and biological pollution respectively, and match the corresponding cleaning solution for different pollution conditions;

[0008] Step S4: Convert the cleaning plan into instructions for execution, monitor the cleaning effect in real time, adjust the plan and repeat the cleaning if the standard is not met, and record the data to optimize the model and rules to ensure that the cleaning meets the standard and continuously improve the system performance.

[0009] Furthermore, in step S1, spectral data in the wavelength range of 200-800nm ​​are collected by an ultraviolet-visible spectrometer to obtain the concentration of organic matter, the particle size distribution and quantity of colloidal particles in water are monitored by an online particle counter to obtain information on inorganic particles, and the ATP fluorescence intensity of microorganisms in water is detected by an ATP fluorescence detector to obtain the concentration of microorganisms. Moreover, multiple sensors collect data synchronously at a frequency of once per minute.

[0010] Furthermore, in step S2, the feature parameter extraction uses principal component analysis to extract the organic features of the spectral data, extract the average particle size, distribution range and concentration of the particle counter data, and extract the mean and standard deviation of ATP fluorescence intensity.

[0011] Furthermore, the convolutional neural network model in step S2 includes an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer;

[0012] A convolutional neural network is used to build a membrane pollution classification model. The input feature vector is processed through convolutional extraction, pooling dimensionality reduction, and fully connected integration. The output includes organic, inorganic, and biological pollution types and their corresponding pollution concentration values: organic pollution index c. org Inorganic pollution index c inorg Or biological pollution index c bio .

[0013] Furthermore, in step S3, the concentration of the cleaning agent and the cleaning conditions for moderate contamination are determined by linear interpolation based on the parameters of light and heavy contamination, according to the quantitative index of contamination degree. The calculation formula is as follows:

[0014]

[0015] Furthermore, in step S3, organic pollution occurs:

[0016] Light pollution c org ≤0.2cm -1 hour:

[0017] Cleaning agent concentration: 0.1% NaOH + 0.025% Na-SDS;

[0018] Cleaning conditions: pH = 11; temperature = 25℃;

[0019] Severe pollution c org ≥0.5cm -1 hour:

[0020] Cleaning agent concentration: 0.15% NaOH + 0.0375% Na-SDS; Cleaning conditions: pH = 12; Temperature = 30℃;

[0021] Moderate pollution 0.2cm -1<c org <0.5cm -1 At that time, the concentration of the cleaning agent was:

[0022]

[0023] Cleaning conditions:

[0024]

[0025] Furthermore, inorganic contamination occurs in step S3:

[0026] Light pollution c inorg When ≤100 cells / mL,

[0027] Cleaning agent concentration: 0.2% HCl

[0028] Cleaning conditions: pH = 4; Temperature = 25℃;

[0029] Severe pollution c inorg When ≥500 cells / mL,

[0030] Cleaning agent concentration: 0.32% HCl;

[0031] Cleaning conditions: pH=2; temperature=35℃;

[0032] Moderate contamination: 100 cells / mL < c inorg When the number of particles is less than 500 per mL, the cleaning agent concentration is:

[0033]

[0034] Cleaning conditions:

[0035]

[0036] Furthermore, biological contamination in step S3:

[0037] Light pollution c bio When ≤100 RLU:

[0038] Cleaning agent concentration: 0.1% NaOH + 0.025% Na-SDS;

[0039] Cleaning conditions: pH = 11; temperature = 25℃;

[0040] Severe pollution c bio When ≥300 RLU:

[0041] Cleaning agent concentration: 0.15% NaOH + 0.0375% Na-SDS;

[0042] Cleaning conditions: pH = 12; temperature = 30℃;

[0043] Moderate pollution 100 RLU < c bio When <300 RLU:

[0044] Cleaning agent concentration:

[0045]

[0046] Cleaning conditions:

[0047]

[0048] Furthermore, during the cleaning process in step S4, the membrane element is first rinsed with RO product water at a pressure of 0.2-0.3 MPa for 15-20 minutes, and then the cleaning solution is injected for low-pressure circulation, with the circulation pressure controlled at 0.1-0.2 MPa.

[0049] Compared with the prior art, the beneficial effects of the present invention are:

[0050] This intelligent reverse osmosis membrane cleaning and operation control method collects data through multiple sensors, constructs multi-dimensional feature vectors through preprocessing and feature extraction, and then uses convolutional neural networks to accurately identify the type and concentration of fouling. Based on this, it automatically matches the optimal cleaning scheme. At the same time, it combines real-time monitoring feedback to form a closed-loop optimization mechanism, which avoids the problem of blind cleaning caused by the inability to distinguish the type of fouling, avoids damage to the membrane caused by improper cleaning agents, extends the service life of the membrane, ensures the efficient and stable operation of the membrane system, improves the stability of the produced water quality, and reduces maintenance costs. Attached Figure Description

[0051] Figure 1 This is a schematic diagram of the intelligent cleaning and operation control method for reverse osmosis membranes.

[0052] Figure 2 This is a schematic diagram of the pollution type identification process. Detailed Implementation

[0053] The present invention will be further described below with reference to embodiments.

[0054] The following embodiments are used to illustrate the present invention, but should not be used to limit the scope of protection of the present invention. The conditions in the embodiments can be further adjusted according to specific conditions, and simple improvements to the method of the present invention under the premise of the concept of the present invention are all within the scope of protection claimed by the present invention.

[0055] Please see Figure 1 and Figure 2 This invention provides an intelligent cleaning and operation control method for reverse osmosis membranes, applied to the fouling detection and precise control of water treatment reverse osmosis (RO) membrane systems, to achieve efficient and stable operation of the RO membrane system, comprising the following steps:

[0056] Step S1: Multi-sensor data acquisition;

[0057] Multiple sensors were used to collect data, including an ultraviolet-visible spectrometer (UV-Vis), an online particle counter, and an ATP fluorescence detector.

[0058] ① Ultraviolet-Vis spectrometer (UV-Vis):

[0059] Detection principle: Conjugated double bonds, aromatic rings, and other functional groups in organic molecules produce characteristic absorptions in the ultraviolet-visible spectral region (200-800 nm). According to the Lambert-Beer law: A = ε·c·l

[0060] Where A is the absorbance, ε is the molar absorptivity (related to the properties of the organic matter), c is the concentration of the organic matter, and l is the optical path length (sample cell thickness). The concentration of the organic matter c can be deduced by measuring the absorbance A.

[0061] An ultraviolet-visible spectrometer is deployed on the inlet side of the membrane to collect spectral data in the wavelength range of 200-800 nm in real time. By utilizing the absorption characteristics of organic matter in the water to light of specific wavelengths, the concentration index c of organic matter is obtained. org (e.g., UV254 value);

[0062] ② Online particle counter:

[0063] Detection principle: Based on the principle of laser scattering, when the sample flows through the detection area, the laser beam irradiates the particles and generates scattered light. The photodetector receives the scattered light, and the signal intensity is proportional to the particle surface area. After calibration, it is converted into particle size and quantity, and the particle concentration of each particle size channel is output in real time.

[0064] An online particle counter is deployed on the inlet side of the membrane to monitor the particle size distribution and quantity of colloidal particles in the water in real time, thereby reflecting the inorganic pollution situation;

[0065] ③ The ATP fluorescence detector detects the ATP fluorescence intensity of microorganisms in water in real time, thereby obtaining information on microbial concentration.

[0066] The multi-sensor acquisition module synchronously collects data at a set frequency (once per minute) and outputs a raw dataset containing information on organic matter, inorganic matter, and microbial contamination.

[0067] Step S2: Data processing and contamination type identification;

[0068] The data from step S1 is transferred to step S2. Data processing and pollution type identification preprocess the data from step S1 (e.g., denoising, normalization), followed by principal component analysis (PCA) to extract the principal components of organic pollution characteristics from the spectral data. The average particle size, distribution range, and concentration from the particle counter data are also extracted, along with the mean and standard deviation of ATP fluorescence intensity. Finally, these feature parameters are fused to construct a multidimensional feature vector, which serves as the input data for pollution type identification.

[0069] A membrane fouling type classification model was established using a Convolutional Neural Network (CNN). The model consists of an input layer, convolutional layers, pooling layers, fully connected layers, and an output layer. A constructed multidimensional feature vector is input into the input layer. Local features are extracted from the data using convolutional kernels in the convolutional layers. The pooling layer reduces the dimensionality of the features output from the convolutional layers to decrease computation. The fully connected layer integrates the features output from the pooling layers. Finally, the output layer outputs the classification probabilities for three types of fouling: organic, inorganic, and biological. During the model training phase, a large number of membrane influent samples with different fouling types were collected. The fouling type of the samples was accurately determined using laboratory analysis methods and used as training data for supervised training of the CNN model, achieving a model accuracy of ≥90% in identifying membrane fouling types. The model derives the fouling type and outputs the corresponding fouling solubility value: c. org (Organic pollution index) or c inorg (Inorganic pollution index) or c bio (Biocontamination indicators)

[0070] Step S3: Dynamic matching and decision-making for cleaning solutions;

[0071] Based on the pollution type identification results, the corresponding cleaning solution will be automatically matched:

[0072] 1. Organic pollution:

[0073] In standard reverse osmosis membrane experiments, by controlling c org After 24 hours of operation, the membrane flux data showed that when c org ≤0.2cm -1 No obvious contaminant deposition was observed on the membrane surface when the thickness was 0.2 cm. -1 <c org <0.5cm -1 At this point, the membrane flux decay rate increases, and the permeate conductivity begins to rise. When c org ≥0.5cm -1 However, after a long period of operation, it was found that the membrane pores were severely blocked by organic matter.

[0074] Based on the dissociation characteristics of organic matter, it is known that at pH = 11, OH... - =10 -3At a concentration of mol / L, this concentration effectively disrupts the hydrogen bonds between organic matter and the membrane surface. At this point, the NaOH concentration is 0.1%, making 0.1% the safe initial concentration for reverse osmosis membrane tolerance. When the NaOH concentration is ≥0.1% at 25℃, the surfactant system reaches its critical micelle concentration after being combined with 0.025% Na-SDS, significantly improving the solubilization effect on organic matter. When the NaOH concentration exceeds 0.15% and the temperature exceeds 30℃, tests show that the hydrolysis rate of amide bonds in the reverse osmosis membrane increases at a constant rate, leading to a decrease in membrane desalination rate; therefore, 0.15% is the maximum concentration. When the NaOH to Na-SDS concentration ratio is 4:1, the optimal balance between cleaning efficiency and membrane safety is achieved. Organic matter concentration c org With UV 254 The values ​​show a linear relationship, therefore the increase in the concentration of the cleaning agent is proportional to the degree of contamination.

[0075] ① Light pollution c org ≤0.2cm -1 hour,

[0076] Cleaning agent concentration: 0.1% NaOH + 0.025% Na-SDS;

[0077] Cleaning conditions: pH = 11; temperature = 25℃;

[0078] ② Severe pollution c org ≥0.5cm -1 hour,

[0079] Cleaning agent concentration: 0.15% NaOH + 0.0375% Na-SDS;

[0080] Cleaning conditions: pH = 12; temperature = 30℃;

[0081] ③ Moderate pollution 0.2cm -1 <c org <0.5cm -1 hour,

[0082] Cleaning agent concentration:

[0083]

[0084] Cleaning conditions:

[0085]

[0086] 2. Inorganic pollution:

[0087] Observations using an online particle counter revealed that when c inorg When the concentration is >100 cells / mL, an inorganic colloidal filter cake layer begins to form on the membrane surface, and the pore blockage rate increases daily; inorgWhen the density is <100 particles / mL, the particles pass through the membrane pores in a monodisperse state, and the deposition rate is <0.1 μg / cm³. 2 c inorg When the filter cake density is 500 cells / mL, the filter cake thickness on the membrane surface exceeds 20 μm, the porosity decreases from 35% to 20%, and the flux recovery rate after washing is <80%. inorg When the number of crystals per mL is greater than 500, the membrane pores become severely clogged with crystals, and chemical cleaning is insufficient to restore its performance.

[0088] Experimental data analysis revealed that when c inorg When the concentration of the cleaning agent (HCl) is ≤100 particles / mL, the concentration is 0.2%, the pH is 4, and the temperature is 25℃, the dissolution rate of CaCO3 reaches 10 mg / (cm³). 2 •h), the cleaning effect is better. When c inorg When the concentration of the cleaning agent is ≥500 particles / mL, the HCl concentration is 0.32%, the pH is 2, and the temperature is 35℃, the CaCO3 dissolution rate reaches 15 mg / (cm³). 2 •h) provides better cleaning results.

[0089] ① Light pollution c inorg When ≤100 cells / mL,

[0090] Cleaning agent concentration: 0.2% HCl;

[0091] Cleaning conditions: pH=4; temperature=25℃;

[0092] ② Severe pollution c inorg When ≥500 cells / mL,

[0093] Cleaning agent concentration: 0.32% HCl;

[0094] Cleaning conditions: pH=2; temperature=35℃;

[0095] ③ Moderate pollution: 100 cells / mL < c inorg When <500 cells / mL,

[0096] Cleaning agent concentration:

[0097]

[0098] Cleaning conditions:

[0099]

[0100] 3. Biological contamination:

[0101] Observation using an ATP fluorescence detector revealed that when c bio When the biofilm thickness exceeds 100 RLU, microbial colonies begin to appear on the membrane surface, and the biofilm thickness increases daily;bio At <100 RLU, microorganisms are in a single-celled free state and have not formed a structured biofilm. bio At 300 RLU, the biofilm thickness exceeds 50 μm, forming a multilayered bacterial floc structure, and the flux recovery rate after washing is <80%; bio When the value is greater than 300 RLU, the membrane pores are severely blocked by biofilm, and chemical cleaning is insufficient to restore performance.

[0102] Experimental data analysis revealed that when c bio For concentrations ≤100 RLU, the best cleaning effect is achieved when using a cleaning agent of 0.1% NaOH + 0.025% Na-SDS, pH = 11, and temperature = 25℃. When c bio For a concentration of ≥300 RLU, the best cleaning effect is achieved when using 0.15% NaOH + 0.0375% Na-SDS, pH = 12, and temperature = 30℃.

[0103] ① Light pollution c bio When ≤100RLU,

[0104] Cleaning agent concentration: 0.1% NaOH + 0.025% Na-SDS;

[0105] Cleaning conditions: pH = 11; temperature = 25℃;

[0106] ② Severe pollution c bio When ≥300 RLU,

[0107] Cleaning agent concentration: 0.15% NaOH + 0.0375% Na-SDS;

[0108] Cleaning conditions: pH = 12; temperature = 30℃;

[0109] ③ Moderate pollution 100 RLU < c bio When <300 RLU,

[0110] Cleaning agent concentration:

[0111]

[0112] Cleaning conditions:

[0113]

[0114] Step S4: Cleaning execution and closure optimization;

[0115] The matched cleaning scheme and cleaning indicators are output as specific control commands and sent to the reverse osmosis system's operating station, including chemical cleaning devices and dosing devices, to initiate the corresponding cleaning program. During the cleaning process, parameters such as the flow rate and temperature of the cleaning solution, as well as operating indicators of the membrane system such as permeate flow rate and influent / effluent conductivity, are monitored in real time, and this data is fed back to the data processing unit. The cleaning effect is evaluated by comparing the changes in the membrane system's operating parameters before and after cleaning. If the membrane flux recovery rate after cleaning does not reach the expected target (e.g., not reaching 80%) or some operating indicators still do not return to normal, the fouling situation is re-analyzed based on the feedback data, and the cleaning scheme is adjusted (e.g., extending the cleaning time, adjusting the cleaning agent concentration, etc.), and the cleaning program is executed again until the membrane system's operating performance returns to normal. Simultaneously, relevant data and cleaning effects from each cleaning process are recorded to optimize and improve the fouling type identification model and cleaning scheme matching rules, continuously improving the system's control accuracy and adaptability.

[0116] Example: A demineralized water treatment workshop in an industrial park uses a membrane system to treat raw water. During system operation, reverse osmosis membrane fouling frequently occurs, leading to decreased membrane flux, unstable product water quality, and unsatisfactory results from traditional cleaning methods.

[0117] The reverse osmosis membrane intelligent cleaning operation control method of the present invention:

[0118] A UV-Vis spectrometer, an online particle counter, and an ATP fluorescence detector are installed on the feed side of the reverse osmosis membrane. The UV-Vis spectrometer monitors the UV254 value of the feed water in real time, the online particle counter detects the particle size distribution and concentration of colloidal particles in the water, and the ATP fluorescence detector acquires the ATP fluorescence intensity of microorganisms. Each sensor collects data once per minute and transmits it to the data processing unit.

[0119] The data processing unit performs digital processing on the collected raw data. It extracts the main component characteristics reflecting organic pollution from the spectral data, extracts characteristic parameters such as the average particle size and concentration of colloidal particles from the particle counter data, and extracts statistical features such as fluorescence intensity from the ATP fluorescence detector data to construct a multidimensional feature vector.

[0120] The multidimensional feature vectors are input into a trained CNN model, and the model output is displayed as organic pollution. org =0.45cm -1 .

[0121] Based on the pollution type identification results, an organic pollution cleaning plan is automatically matched and calculated to be 0.14% NaOH + 0.035% Na-SDS, pH 12, and temperature 29℃. Before cleaning with sodium dodecyl sulfonate (Na-SDS) solution, the membrane elements are first rinsed with RO product water. Prepare 0.035% Na-SDS and 0.14% NaOH, controlling the solution temperature at 29℃ and the pH at 12. Circulate the cleaning solution under low pressure, approximately 0.1 MPa, not exceeding 0.2 MPa. Discharge the first 10%–15% of the cleaning solution, as this portion contains a higher concentration of pollutants. Within permissible limits, higher water temperatures are better, contributing to better cleaning results. After cleaning, drain all cleaning solution and completely empty the cleaning tank. Use pretreated product water or RO product water to thoroughly remove and rinse away all residual cleaning solution from the pressure vessel and pipelines.

[0122] The control execution module translates the cleaning plan into control commands, activating the corresponding chemical cleaning and dosing devices to execute the cleaning procedure. During the cleaning process, the feedback monitoring module monitors parameters such as cleaning fluid flow rate, pressure, temperature, membrane system permeate flow rate, and influent and effluent conductivity in real time. After cleaning, the membrane flux reaches the expected cleaning effect. The cleaning data is recorded for further optimization of the fouling type identification model and cleaning plan matching rules.

[0123] The above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the implementation of the present invention. For those skilled in the art, other variations or modifications can be made based on the above description, and these obvious variations or modifications derived from the spirit of the present invention are still within the protection scope of the present invention.

Claims

1. A method for intelligent cleaning and operation control of reverse osmosis membranes, characterized in that, Includes the following steps: Step S1: Obtain information on organic matter concentration, inorganic particulate matter concentration, and microbial concentration to form the original dataset; Step S2: After preprocessing the data from step S1, extract the feature parameters of each sensor data and fuse them into a multi-dimensional feature vector. Input the vector into the convolutional neural network model and output the pollution type and concentration value to identify the pollution type and concentration. Step S3: Based on the type and concentration of pollution, and according to the light, moderate and heavy pollution conditions, match the corresponding cleaning agent concentration, pH value and temperature parameters for organic matter, inorganic matter and biological pollution respectively, and match the corresponding cleaning solution for different pollution conditions; Step S4: Convert the cleaning plan into instructions for execution, monitor the cleaning effect in real time, adjust the plan and repeat the cleaning if the standard is not met, and record the data to optimize the model and rules to ensure that the cleaning meets the standard and continuously improve the system performance.

2. The intelligent cleaning and operation control method for reverse osmosis membranes according to claim 1, characterized in that: In step S1, spectral data in the wavelength range of 200-800nm ​​are collected by an ultraviolet-visible spectrometer to obtain the concentration of organic matter. The particle size distribution and quantity of colloidal particles in the water are monitored by an online particle counter to obtain information on inorganic particles. The ATP fluorescence intensity of microorganisms in the water is detected by an ATP fluorescence detector to obtain the concentration of microorganisms. Data is collected synchronously by multiple sensors at a frequency of once per minute.

3. The intelligent cleaning and operation control method for reverse osmosis membranes according to claim 1, characterized in that: In step S2, the feature parameter extraction uses principal component analysis to extract the organic features of the spectral data, extract the average particle size, distribution range and concentration of the particle counter data, and extract the mean and standard deviation of ATP fluorescence intensity.

4. The intelligent cleaning and operation control method for reverse osmosis membranes according to claim 1, characterized in that: The convolutional neural network model in step S2 includes an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer. A convolutional neural network is used to build a membrane pollution classification model. The input feature vector is processed through convolutional extraction, pooling dimensionality reduction, and fully connected integration. The output includes organic, inorganic, and biological pollution types and their corresponding pollution concentration values: organic pollution index c. org Inorganic pollution index c inorg Or biological pollution index c bio .

5. The intelligent cleaning and operation control method for reverse osmosis membranes according to claim 4, characterized in that, In step S3, the concentration of the cleaning agent and the cleaning conditions for moderate contamination are determined by linear interpolation based on the parameters of light and heavy contamination, using a quantitative index of contamination degree. The calculation formula is as follows:

6. The intelligent cleaning and operation control method for reverse osmosis membranes according to claim 5, characterized in that, Organic pollution in step S3: Light pollution c org ≤0.2cm -1 hour: Cleaning agent concentration: 0.1% NaOH + 0.025% Na-SDS; Cleaning conditions: pH = 11; temperature = 25℃; Severe pollution c org ≥0.5cm -1 hour: Cleaning agent concentration: 0.15% NaOH + 0.0375% Na-SDS; Cleaning conditions: pH = 12; temperature = 30℃; Moderate pollution 0.2cm -1 <c org <0.5cm -1 hour, Cleaning agent concentration: Cleaning conditions:

7. The intelligent cleaning and operation control method for reverse osmosis membranes according to claim 5, characterized in that, Inorganic contamination in step S3: Light pollution c inorg When ≤100 cells / mL, Cleaning agent concentration: 0.2% HCl Cleaning conditions: pH=4; temperature=25℃; Severe pollution c inorg When ≥500 cells / mL, Cleaning agent concentration: 0.32% HCl; Cleaning conditions: pH=2; temperature=35℃; Moderate contamination: 100 cells / mL < c inorg When <500 cells / mL, Cleaning agent concentration: Cleaning conditions:

8. The intelligent cleaning and operation control method for reverse osmosis membranes according to claim 5, characterized in that, Biological contamination in step S3: Light pollution c bio When ≤100 RLU: Cleaning agent concentration: 0.1% NaOH + 0.025% Na-SDS; Cleaning conditions: pH = 11; temperature = 25℃; Severe pollution c bio When ≥300 RLU: Cleaning agent concentration: 0.15% NaOH + 0.0375% Na-SDS; Cleaning conditions: pH = 12; temperature = 30℃; Moderate pollution 100 RLU < c bio When <300 RLU: Cleaning agent concentration: Cleaning conditions:

9. The intelligent cleaning and operation control method for reverse osmosis membranes according to claim 1, characterized in that: When cleaning is performed in step S4, the membrane element is first rinsed with RO product water at a pressure of 0.2-0.3 MPa for 15-20 minutes, and then the cleaning solution is injected for low-pressure circulation, with the circulation pressure controlled at 0.1-0.2 MPa.

Images