Ultrafiltration membrane water treatment control method

Abstract

The invention discloses an ultrafiltration membrane water treatment control method. The method comprises the following steps: collecting inlet water quality parameters and membrane module operation parameters in real time; dynamically predicting a membrane pollution trend by using an online trained machine learning model based on the acquired parameters, and adaptively adjusting a transmembrane pressure difference control threshold according to a prediction result and historical operation data; executing membrane filtration operation according to the adjusted transmembrane pressure difference control threshold value, and monitoring a membrane flux change rate and an energy consumption change rate in real time; triggering a multi-stage backwashing program based on the membrane flux change rate, and synchronously optimizing the pressure, period and interval of backwashing in combination with the prediction model; and after backwashing is completed, membrane integrity detection is executed. According to the method, the machine learning model fusing the support vector machine and the long-short-term memory network is constructed, the membrane flux change rate is dynamically predicted in combination with real-time monitoring data, the trans-membrane pressure difference control threshold value and the backwashing strategy are intelligently adjusted accordingly, and collaborative optimization of the system operation efficiency and the membrane service life is achieved.

Description

Technical Field

[0001] This invention relates to the technical field of membrane separation water treatment, and more particularly to a method for controlling ultrafiltration membrane water treatment. Background Technology

[0002] Ultrafiltration membrane water treatment systems are widely used in industrial wastewater treatment, reclaimed water reuse, and advanced drinking water purification. Their operational performance largely depends on the stability of membrane flux and the control of membrane fouling. Traditional membrane system control strategies rely on fixed transmembrane pressure thresholds and periodic backwashing. However, in environments with drastic water quality fluctuations or high pollution loads, this type of control mode cannot effectively respond to the dynamic characteristics of membrane flux decay, leading to increased membrane fouling, higher operating energy consumption, and even system failure.

[0003] In recent years, although some studies have introduced sensor data for control optimization, most of them are based on simple linear prediction or empirical rules and lack the ability to deeply model and intelligently adjust time series data.

[0004] Therefore, there is an urgent need for a new method based on real-time prediction and adaptive differential pressure control to improve the operational stability and intelligence level of membrane systems under complex water quality conditions. Summary of the Invention

[0005] The purpose of this section is to outline some aspects of embodiments of the present invention and to briefly describe some preferred embodiments. Simplifications or omissions may be made in this section, as well as in the abstract and title of this application, to avoid obscuring the purpose of these documents; however, such simplifications or omissions should not be construed as limiting the scope of the invention.

[0006] In view of the problems existing in the current ultrafiltration membrane water treatment control methods, the present invention is proposed.

[0007] Therefore, the purpose of this invention is to provide an ultrafiltration membrane water treatment control method, which constructs a machine learning model that integrates support vector machine and long short-term memory network, dynamically predicts the membrane flux change rate by combining real-time monitoring data, and intelligently adjusts the transmembrane pressure difference control threshold and backwashing strategy accordingly, thereby achieving synergistic optimization of system operating efficiency and membrane life.

[0008] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a method for controlling ultrafiltration membrane water treatment, comprising the following steps: Real-time acquisition of influent water quality parameters and membrane module operating parameters; Based on the collected parameters, the membrane fouling trend is dynamically predicted using an online-trained machine learning model, and the transmembrane pressure differential control threshold is adaptively adjusted according to the prediction results and historical operating data. Membrane filtration is performed based on the adjusted transmembrane pressure differential control threshold, and the membrane flux change rate and energy consumption change rate are monitored in real time. A multi-stage backwashing procedure is triggered based on the membrane flux change rate, and the backwashing pressure, cycle and interval are simultaneously optimized by combining a predictive model. After backwashing is completed, membrane integrity is tested, and the test results and operational data are fed back to the machine learning model for dynamic optimization of subsequent processing procedures.

[0009] As a preferred embodiment of the ultrafiltration membrane water treatment control method of the present invention, the machine learning model is a hybrid model that integrates support vector machine and long short-term memory network. The model input includes the influent water quality and transmembrane pressure difference data of the most recent 12 hours, and the output is the predicted value of membrane flux decay rate for the next hour.

[0010] As a preferred embodiment of the ultrafiltration membrane water treatment control method of the present invention, the specific implementation of the adaptive adjustment of the transmembrane pressure difference control threshold includes: S1. When the influent turbidity is ≤10 NTU and the predicted flux decay rate is ≤5% / h, the transmembrane pressure difference threshold is set to 0.17 MPa. S2. When 10 NTU < influent turbidity ≤ 30 NTU or 5% / h < predicted attenuation rate ≤ 10% / h, the transmembrane pressure difference threshold will be adaptively maintained in the range of 0.12 to 0.17 MPa. S3. When the influent turbidity is >30 NTU or the predicted decay rate is >10% / h, the pretreatment enhancement program will be automatically started and the threshold will be reduced to 0.10 MPa.

[0011] As a preferred embodiment of the ultrafiltration membrane water treatment control method of the present invention, the multi-stage backwashing procedure includes: Phase 1: Maintain a back pressure of 0.25 MPa for 30 seconds; The second stage: backwashing is performed in a pulse manner with a pulse pressure of 0.35 MPa, alternating between high pressure and low pressure of 0.18 MPa for 5s / 10s, for a total of N cycles, where N is determined by the degree of contamination predicted by the machine learning model; Third stage: Flush with a stable pressure of 0.18MPa for 60 seconds.

[0012] In a preferred embodiment of the ultrafiltration membrane water treatment control method of the present invention, the number N of the pulse phase cycle is determined as follows: When the model predicts a flux decay rate of >15% / h over the next 1 hour, N=8; When 10% / h < predicted attenuation rate ≤ 15% / h, N = 5; When the predicted attenuation rate is ≤10% / h, N=3.

[0013] In a preferred embodiment of the ultrafiltration membrane water treatment control method of the present invention, the membrane integrity detection includes: Inject 0.1 MPa of compressed air into the product water side and monitor the rate of pressure drop; When the pressure drop rate is greater than 0.02 MPa / min, the membrane is determined to be damaged and the damage location procedure is initiated. When 0.01 MPa / min < pressure drop rate ≤ 0.02 MPa / min, it is marked as a potential risk and subsequent monitoring should be strengthened; When the pressure drop rate is ≤0.01MPa / min, the membrane module is confirmed to be intact. The damage location procedure includes: Close the inlet valves of each membrane module in sequence and monitor the pressure changes on the product water side; If the pressure returns to the normal range after a certain group of valves is closed, then that group is identified as the damaged group. The damaged membrane module is automatically switched to the standby membrane module and the operating parameters are updated synchronously.

[0014] As a preferred embodiment of the ultrafiltration membrane water treatment control method of the present invention, it further includes chemical cleaning optimization control: Chemical cleaning is triggered when the cumulative running time reaches 72 hours or the transmembrane pressure difference exceeds the dynamic threshold by 30%. The cleaning process begins with circulating an alkaline cleaning solution with pH=11 for 40 minutes under the following conditions: temperature 35±2℃, flow rate 1.5m / s. If the flux recovery rate after cleaning is <85%, switch to a pH=2 acidic cleaning solution for a second treatment of 20 min. After chemical cleaning is completed, flux calibration is performed and the replicated operating parameters are written into the model adaptation library.

[0015] An ultrafiltration membrane water treatment control system includes: The data acquisition module is used to collect influent water quality parameters and membrane module operating parameters in real time; The prediction and decision module integrates an online-trained SVM–LSTM hybrid prediction model, which is used to predict the future membrane flux decay rate based on the data provided by the acquisition module and output the transmembrane pressure difference control threshold. The execution control module is used to control the operation of membrane filtration based on the threshold output by the prediction and decision module, and to monitor the membrane flux change rate and energy consumption change rate in real time. The backwashing control module is used to automatically adjust the backwashing pressure, cycle and interval according to a multi-stage backwashing strategy when the execution control module detects that the membrane flux change rate has reached the trigger condition. The integrity detection module is used to perform membrane integrity detection on the membrane module after backwashing and feeds back the detection results and flux recovery rate to the prediction and decision module to achieve online adaptive optimization.

[0016] The beneficial effects of this invention are as follows: This invention constructs a machine learning model that integrates support vector machines and long short-term memory networks, combines real-time monitoring data to dynamically predict the rate of change of membrane flux, and intelligently adjusts the transmembrane pressure difference control threshold and backwashing strategy accordingly, thereby achieving synergistic optimization of system operating efficiency and membrane lifetime. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 This is a flowchart illustrating the overall steps of the ultrafiltration membrane water treatment control method of the present invention. Figure 2 This is a flowchart of the membrane integrity detection process in the ultrafiltration membrane water treatment control method of the present invention. Detailed Implementation

[0018] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0019] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0020] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0021] Secondly, the present invention is described in detail with reference to the schematic diagrams. When detailing the embodiments of the present invention, for ease of explanation, the cross-sectional views illustrating the device structure may be partially enlarged, not according to the usual scale. Furthermore, the schematic diagrams are merely examples and should not limit the scope of protection of the present invention. In addition, actual fabrication should include three-dimensional spatial dimensions of length, width, and depth.

[0022] Reference Figures 1-2 A method for controlling water treatment using an ultrafiltration membrane is provided, comprising the following steps: Real-time acquisition of influent water quality parameters and membrane module operating parameters; Based on the collected parameters, the membrane fouling trend is dynamically predicted using an online-trained machine learning model, and the transmembrane pressure differential control threshold is adaptively adjusted according to the prediction results and historical operating data. Membrane filtration is performed based on the adjusted transmembrane pressure differential control threshold, and the membrane flux change rate and energy consumption change rate are monitored in real time. A multi-stage backwashing procedure is triggered based on the membrane flux change rate, and the backwashing pressure, cycle and interval are simultaneously optimized by combining a predictive model. After backwashing is completed, membrane integrity is tested, and the test results and operational data are fed back to the machine learning model for dynamic optimization of subsequent processing procedures.

[0023] The machine learning model is a hybrid model that integrates support vector machines and long short-term memory networks. The model input includes the influent water quality and transmembrane pressure difference data for the most recent 12 hours, and the output is the predicted value of the membrane flux decay rate for the next hour.

[0024] Specifically, the online-trained machine learning model is a hybrid prediction model that integrates Support Vector Machine (SVM) and Long Short-Term Memory Network (LSTM), and includes the following modules and steps: 1. Data Preprocessing Module: The collected raw influent water quality parameters and membrane module operating parameters are subjected to noise reduction and normalization processing. A sliding window approach was used to construct training samples based on feature vectors collected at 5-minute intervals within the last 12 hours, including turbidity, COD, pH, conductivity, transmembrane pressure difference, membrane flux, and system energy consumption. Perform a difference transformation on the time series for each window to highlight the flux decline trend and stabilize the series variance; 2. SVM sub-model: Support vector regression (SVR) with radial basis function (RBF) kernel was used to make a preliminary prediction of short-term (10-30 minutes) membrane flux decay rate; SVM hyperparameter settings: Penalty parameter C = 1.0, kernel width ; The optimal C and ☐ were determined using grid search and 5-fold cross-validation. ; Output short-time prediction results ; 3. LSTM sub-model: A three-layer LSTM network is constructed. The first layer has an input dimension of 7 (corresponding to the preprocessed features) and 128 hidden units. The second and third layers each have 64 hidden units. A fully connected regression layer is connected after the three-layer LSTM to map the hidden state of the last time step to the flux decay rate sequence prediction within the next hour (12 five-minute steps); LSTM hyperparameter settings: learning rate 0.001, batch size 64, number of training epochs 50; Mean squared error (MSE) is used as the loss function, and Adam is the optimizer. 4. Integration and Output: The prediction results of SVM and LSTM are fused in a weighted manner: ; Among them, the fusion coefficient It can be adjusted according to online performance; Output the final predicted value As a prediction of membrane flux decay rate in the next hour, it is used for threshold adaptive adjustment; 5. Online update mechanism: After each backwash, compare the actual flux decline rate before and after the latest backwash with the model prediction. Store them together in the training set; Every 24 hours or when more than 500 new samples are added cumulatively, the SVM and LSTM sub-models are fine-tuned online to ensure the model's adaptability to changes in system state and raw water quality.

[0025] Furthermore, the specific implementation of adaptively adjusting the transmembrane pressure difference control threshold includes: S1. When the influent turbidity is ≤10 NTU and the predicted flux decay rate is ≤5% / h, the transmembrane pressure difference threshold is set to 0.17 MPa. S2. When 10 NTU < influent turbidity ≤ 30 NTU or 5% / h < predicted attenuation rate ≤ 10% / h, the transmembrane pressure difference threshold will be adaptively maintained in the range of 0.12 to 0.17 MPa. S3. When the influent turbidity is >30 NTU or the predicted decay rate is >10% / h, the pretreatment enhancement program will be automatically started and the threshold will be reduced to 0.10 MPa.

[0026] The multi-stage backwashing procedure includes: Phase 1: Maintain a back pressure of 0.25 MPa for 30 seconds; The second stage: backwashing is performed in a pulse manner with a pulse pressure of 0.35 MPa, alternating between high pressure and low pressure of 0.18 MPa for 5s / 10s, for a total of N cycles, where N is determined by the degree of contamination predicted by the machine learning model; Third stage: Flush with a stable pressure of 0.18MPa for 60 seconds.

[0027] Specifically, the number of pulse phase cycles N is determined as follows: When the model predicts a flux decay rate of >15% / h over the next 1 hour, N=8; When 10% / h < predicted attenuation rate ≤ 15% / h, N = 5; When the predicted attenuation rate is ≤10% / h, N=3.

[0028] Membrane integrity testing includes: Inject 0.1 MPa of compressed air into the product water side and monitor the rate of pressure drop; When the pressure drop rate is greater than 0.02 MPa / min, the membrane is determined to be damaged and the damage location procedure is initiated. When 0.01 MPa / min < pressure drop rate ≤ 0.02 MPa / min, it is marked as a potential risk and subsequent monitoring should be strengthened; When the pressure drop rate is ≤0.01MPa / min, the membrane module is confirmed to be intact. The damage location procedure includes: Close the inlet valves of each membrane module in sequence and monitor the pressure changes on the product water side; If the pressure returns to the normal range after a certain group of valves is closed, then that group is identified as the damaged group. The damaged membrane module is automatically switched to the standby membrane module and the operating parameters are updated synchronously.

[0029] Specifically, this also includes optimized control of chemical cleaning: Chemical cleaning is triggered when the cumulative running time reaches 72 hours or the transmembrane pressure difference exceeds the dynamic threshold by 30%. The cleaning process begins with circulating an alkaline cleaning solution with pH=11 for 40 minutes under the following conditions: temperature 35±2℃, flow rate 1.5m / s. If the flux recovery rate after cleaning is <85%, switch to a pH=2 acidic cleaning solution for a second treatment of 20 min. After chemical cleaning is completed, flux calibration is performed and the replicated operating parameters are written into the model adaptation library.

[0030] Specific application scenarios:

[0031] Example 1:

[0032] The circulating water reuse system of a steel plant uses high-turbidity iron-containing industrial wastewater (turbidity 30-100 NTU) as raw water, and the pollution fluctuates greatly.

[0033] The control process is as follows: 1. Data Acquisition and Processing Data such as turbidity, COD, transmembrane pressure difference, and membrane flux are collected every 5 minutes; a sliding window is used to construct time-series features and input them into the SVM+LSTM model. 2. Prediction and Regulation The model predicts that the flux will drop to 15% within 40 minutes. The system automatically reduces the transmembrane pressure differential threshold from 0.15 MPa to 0.12 MPa and schedules backwashing in advance. 3. Backwashing optimization With a COD of up to 85 mg / L, the pulse backwashing is set to 8 cycles, each cycle including 5 seconds of high pressure + 10 seconds of low pressure, which greatly improves the membrane recovery rate. 4. Results Compared with the original fixed strategy, the membrane fouling rate was reduced by 22%, the cleaning frequency was reduced by 13%, and the system energy consumption was reduced by 10%.

[0034] Example 2: Rural drinking water ultrafiltration stations in groundwater purification projects A rural drinking water project in northern China receives groundwater as its influent. The water quality changes slowly but is accompanied by low concentrations of organic matter over a long period.

[0035] The control process is as follows: 1. Initial stage The groundwater turbidity fluctuates between 5 and 15 NTU, and the system dynamically maintains the differential pressure threshold between 0.13 and 0.15 MPa according to the rules. 2. Pollution early warning The model predicts that the flux will decrease slowly (<8% / h) in the next 2 hours. Maintain the original operating parameters and extend the backwashing cycle to 90 minutes. 3. Membrane integrity testing If the pressure drop rate is abnormal (0.025 MPa / min) after a backwash, the system will automatically determine that the membrane module is damaged, isolate the section of the membrane module, and transfer it to the standby module. 4. Cleaning Management The system indicates that the 72-hour cleaning time limit has been met. Perform alkaline cleaning and flux recovery assessment to ensure long-term stable operation of the system.

[0036] The present invention also includes an ultrafiltration membrane water treatment control system, comprising: The data acquisition module is used to collect influent water quality parameters and membrane module operating parameters in real time; The prediction and decision module integrates an online-trained SVM–LSTM hybrid prediction model, which is used to predict the future membrane flux decay rate based on the data provided by the acquisition module and output the transmembrane pressure difference control threshold. The execution control module is used to control the operation of membrane filtration based on the threshold output by the prediction and decision module, and to monitor the membrane flux change rate and energy consumption change rate in real time. The backwashing control module is used to automatically adjust the backwashing pressure, cycle and interval according to a multi-stage backwashing strategy when the execution control module detects that the membrane flux change rate has reached the trigger condition. The integrity detection module is used to perform membrane integrity detection on the membrane module after backwashing and feeds back the detection results and flux recovery rate to the prediction and decision module to achieve online adaptive optimization.

[0037] Specifically, the data acquisition module includes: turbidity sensor, COD sensor, pH sensor, conductivity sensor; differential pressure sensor, flow meter, and electricity meter; each sensor uploads data to the prediction and decision-making module at a 5-minute interval.

[0038] The prediction and decision-making module specifically includes: Preprocessing submodule: performs noise reduction, differencing, and normalization on the raw data; SVM submodule: Based on radial basis kernel function, it realizes short-term flux decay rate prediction of 10–30 minutes; LSTM submodule: A three-layer LSTM network is used for 1-hour decay rate sequence prediction; Fusion Unit: Fusion of SVM and LSTM outputs according to preset weights to generate final threshold suggestions.

[0039] Furthermore, the execution control module is used to: receive the transmembrane pressure difference threshold issued by the prediction and decision module; control the membrane pump and valves to maintain the membrane system operating within the pressure difference range; and collect and upload operating parameters in real time for subsequent decision-making.

[0040] The backwash control module includes: Primary loosening unit: Maintain a back pressure of 0.25 MPa for 30 seconds; Pulse flushing unit: alternately outputs high and low pressures of 0.35MPa / 0.18MPa, with the number of cycles N determined by the prediction module instruction; Stable flushing unit: outputs a stable pressure of 0.18MPa for 60 seconds.

[0041] Specifically, the backwash control module also includes: The number of cycles N is calculated based on the influent COD concentration and the predicted decay rate: When the decay rate is >15% / h or COD is >50mg / L, N=8; When 10% / h < attenuation rate ≤ 15% / h or 20–50 mg / L, N = 5; When the decay rate is ≤10% / h or COD <20mg / L, N=3.

[0042] The integrity detection module includes: An air injection unit is used to inject 0.1 MPa compressed air into the product water side; The pressure drop monitoring unit is used to monitor the rate of pressure drop in real time and perform risk assessment. A pressure greater than 0.02 MPa / min indicates damage. 0.01-0.02 MPa / min indicates potential risks; ≤0.01MPa / min confirmed to be in good working order.

[0043] Additionally, a chemical cleaning management module is included, which is automatically triggered when the cumulative operation reaches 72 hours or the transmembrane pressure difference exceeds the dynamic threshold by 30%. Alkaline cleaning: pH=11 cleaning solution circulates for 40 min (35±2℃, 1.5 m / s). Acidic cleaning: When the flux recovery rate is <85%, use a pH=2 cleaning solution for 20 minutes; The flux recovery rate and cleaning parameters are then fed back to the prediction and decision-making module to update the adaptive library.

[0044] The present invention also provides a computer device applicable to an ultrafiltration membrane water treatment control method and system, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize an ultrafiltration membrane water treatment control method and system as described in the above embodiments.

[0045] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0046] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the ultrafiltration membrane water treatment control method and system as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0047] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for controlling ultrafiltration membrane water treatment, characterized in that, Includes the following steps: Real-time acquisition of influent water quality parameters and membrane module operating parameters; Based on the collected parameters, the membrane fouling trend is dynamically predicted using an online-trained machine learning model, and the transmembrane pressure differential control threshold is adaptively adjusted according to the prediction results and historical operating data. Membrane filtration is performed based on the adjusted transmembrane pressure differential control threshold, and the membrane flux change rate and energy consumption change rate are monitored in real time. A multi-stage backwashing procedure is triggered based on the membrane flux change rate, and the backwashing pressure, cycle and interval are simultaneously optimized by combining a predictive model. After backwashing is completed, membrane integrity is tested, and the test results and operational data are fed back to the machine learning model for dynamic optimization of subsequent processing procedures.

2. The ultrafiltration membrane water treatment control method according to claim 1, characterized in that: The machine learning model is a hybrid model that integrates support vector machine and long short-term memory network. The model input includes the influent water quality and transmembrane pressure difference data of the most recent 12 hours, and the output is the predicted value of membrane flux decay rate for the next hour.

3. The ultrafiltration membrane water treatment control method according to claim 2, characterized in that: The specific implementation of the adaptive adjustment of the transmembrane pressure difference control threshold includes: S1. When the influent turbidity is ≤10 NTU and the predicted flux decay rate is ≤5% / h, the transmembrane pressure difference threshold is set to 0.17 MPa. S2. When 10 NTU < influent turbidity ≤ 30 NTU or 5% / h < predicted attenuation rate ≤ 10% / h, the transmembrane pressure difference threshold will be adaptively maintained in the range of 0.12 to 0.17 MPa. S3. When the influent turbidity is >30 NTU or the predicted decay rate is >10% / h, the pretreatment enhancement program will be automatically started and the threshold will be reduced to 0.10 MPa.

4. The ultrafiltration membrane water treatment control method according to claim 3, characterized in that: The multi-stage backwashing procedure includes: Phase 1: Maintain a back pressure of 0.25 MPa for 30 seconds; The second stage: backwashing is performed in a pulse manner with a pulse pressure of 0.35 MPa, alternating between high pressure and low pressure of 0.18 MPa for 5s / 10s, for a total of N cycles, where N is determined by the degree of contamination predicted by the machine learning model; Third stage: Flush with a stable pressure of 0.18MPa for 60 seconds.

5. The ultrafiltration membrane water treatment control method according to claim 4, characterized in that: The number of pulse phase cycles N is determined as follows: When the model predicts a flux decay rate of >15% / h over the next 1 hour, N=8; When 10% / h < predicted attenuation rate ≤ 15% / h, N = 5; When the predicted attenuation rate is ≤10% / h, N=3.

6. The ultrafiltration membrane water treatment control method according to claim 5, characterized in that: The membrane integrity detection includes: Inject 0.1 MPa of compressed air into the product water side and monitor the rate of pressure drop; When the pressure drop rate is greater than 0.02 MPa / min, the membrane is determined to be damaged and the damage location procedure is initiated. When 0.01 MPa / min < pressure drop rate ≤ 0.02 MPa / min, it is marked as a potential risk and subsequent monitoring should be strengthened; When the pressure drop rate is ≤0.01MPa / min, the membrane module is confirmed to be intact. The damage location procedure includes: Close the inlet valves of each membrane module in sequence and monitor the pressure changes on the product water side; If the pressure returns to the normal range after a certain group of valves is closed, then that group is identified as the damaged group. The damaged membrane module is automatically switched to the standby membrane module and the operating parameters are updated synchronously.

7. The ultrafiltration membrane water treatment control method according to claim 1, characterized in that: It also includes optimized control of chemical cleaning: Chemical cleaning is triggered when the cumulative running time reaches 72 hours or the transmembrane pressure difference exceeds the dynamic threshold by 30%. The cleaning process begins with circulating an alkaline cleaning solution with pH=11 for 40 minutes under the following conditions: temperature 35±2℃, flow rate 1.5m / s. If the flux recovery rate after cleaning is <85%, switch to a pH=2 acidic cleaning solution for a second treatment of 20 min. After chemical cleaning is completed, flux calibration is performed and the replicated operating parameters are written into the model adaptation library.

8. A water treatment control system for ultrafiltration membranes, characterized in that, include: The data acquisition module is used to collect influent water quality parameters and membrane module operating parameters in real time; The prediction and decision module integrates an online-trained SVM–LSTM hybrid prediction model, which is used to predict the future membrane flux decay rate based on the data provided by the acquisition module and output the transmembrane pressure difference control threshold. The execution control module is used to control the operation of membrane filtration based on the threshold output by the prediction and decision module, and to monitor the membrane flux change rate and energy consumption change rate in real time. The backwashing control module is used to automatically adjust the backwashing pressure, cycle and interval according to a multi-stage backwashing strategy when the execution control module detects that the membrane flux change rate has reached the trigger condition. The integrity detection module is used to perform membrane integrity detection on the membrane module after backwashing and feeds back the detection results and flux recovery rate to the prediction and decision module to achieve online adaptive optimization.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the ultrafiltration membrane water treatment control method according to any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the ultrafiltration membrane water treatment control method according to any one of claims 1-7.

Images