Seawater reverse osmosis filtering method based on pressure-conductivity dynamic coupling

Abstract

The invention relates to a seawater filtration technology, and particularly provides a seawater reverse osmosis filtration method based on pressure-conductivity dynamic coupling, which comprises the following steps: A1, seawater in a seawater tank passes through a booster pump and then is conveyed to a reverse osmosis membrane, pure water and concentrated water are output, and the concentrated water returns to the seawater tank through a circulating valve; synchronously collecting upstream and downstream parameters of the reverse osmosis membrane, wherein the parameters comprise pressure and conductivity; a2, extracting pressure-conductivity coupling characteristics by using the parameters, and outputting a pollution type identification result by a pollution type identification model according to the coupling characteristics; a3, generating a booster pump frequency, a circulating valve opening degree and a cleaning instruction based on the identification result; and A4, executing the instruction and feeding back parameters to form closed-loop control. The method has the advantages that the dynamic response speed is increased, and the pollution adaptability is enhanced.

Description

TECHNICAL FIELD

[0001] The present application relates to seawater treatment technology, in particular to a seawater reverse osmosis filtration method based on pressure-conductivity dynamic coupling. BACKGROUND

[0002] As the mainstream means of seawater desalination, the core of reverse osmosis technology is to realize the separation of water and salt through pressure driving, and conductivity directly reflects the desalination effect, and the coordinated control of the two is the key to the efficient operation of the system. The existing technology has the following defects: 1. Parameter coupling redundancy, the existing technology often introduces temperature as the core parameter into the control model, but the influence of temperature on the system can be indirectly reflected through the dynamic relationship between pressure and conductivity. Introducing a temperature sensor alone not only increases the hardware cost, but also reduces the control accuracy due to the cross interference of multiple parameters. In the near-shore sea area where the temperature fluctuation is less than ±5℃, the marginal benefit of the temperature parameter is less than 3%.

[0003] 2. Insufficient pressure feature mining, traditional systems only use static transmembrane pressure difference as a control basis, and do not capture dynamic features such as pressure pulsation frequency and amplitude. The initial stage of membrane fouling (such as colloidal fouling [silt, microbial floc, humic substance]) will cause the pressure pulsation frequency to rise from 0.5Hz to 2Hz, missing the best control opportunity.

[0004] 3. Poor pollution type adaptability, the existing control model uses the same control strategy for biological pollution and scaling pollution, but biological pollution requires reducing pressure and increasing cleaning frequency, while scaling pollution requires increasing flow rate to flush. Strategy confusion can cause the membrane life to be shortened by more than 30%. SUMMARY

[0005] To solve the above problems in the prior art, the present application provides a seawater reverse osmosis filtration method based on pressure-conductivity dynamic coupling.

[0006] The purpose of the present application is achieved by the following technical solutions: A seawater reverse osmosis filtration method based on pressure-conductivity dynamic coupling, the method comprising the following steps: A1. The seawater in the seawater tank is transported to the reverse osmosis membrane through the booster pump, outputting pure water and concentrated water, and the concentrated water returns to the seawater tank through the circulation valve; Synchronously collecting parameters upstream and downstream of the reverse osmosis membrane, the parameters including pressure and conductivity; A2. Extracting pressure-conductivity coupling features using the parameters, and a pollution type identification model outputs a pollution type identification result according to the coupling features; A3. Generating a booster pump frequency, a circulation valve opening degree and a cleaning instruction based on the identification result; A4. Execute instructions and feedback parameters to form a closed-loop control.

[0007] Compared with the prior art, the application has the beneficial effects of: 1. Parameter redundancy elimination: indirectly reflect the influence of temperature and other environments through the dynamic coupling relationship between pressure and conductivity, reduce the number of sensors by 30%, reduce hardware cost and data interference; 2. Dynamic response speed-up: identify membrane pollution in advance using pressure pulsation characteristics, shorten the response time by 2-4 hours compared with traditional transmembrane pressure difference single parameter control; 3. Enhanced pollution adaptability: implement differentiated regulation strategies for different pollution types, extend the service life of membrane modules, and reduce cleaning agent consumption; 4. Extreme working condition stability: under the condition of salt concentration sudden change (±10000 μS / cm) or high turbidity (5-10 NTU), the system has small desalination rate fluctuation, which is much better than the prior art. BRIEF DESCRIPTION OF DRAWINGS

[0008] The disclosure of the application will become more apparent with reference to the accompanying drawings. It is easy for those skilled in the art to understand that these drawings are only used to illustrate the technical solutions of the application, and are not intended to limit the protection scope of the application. In the drawings: Figure 1 is a flowchart of a seawater reverse osmosis filtration method based on pressure-conductivity dynamic coupling; Figure 2 is a schematic diagram of the connection relationship of each component.

[0009] In the drawings, 11 is a seawater tank, 21 is a booster pump, 31 is a reverse osmosis membrane, 41 is a circulating valve, 51 is a first group of sensors, 52 is a second group of sensors, and 53 is a third group of sensors. DETAILED DESCRIPTION

[0010] Figures 1-2 The following description describes alternative specific embodiments of the application to teach those skilled in the art how to implement and reproduce the application. Some conventional aspects have been simplified or omitted in order to teach the technical solutions of the application. Those skilled in the art should understand that variations or substitutions derived from these specific embodiments will be within the scope of the application. Those skilled in the art should understand that the following features can be combined in various ways to form multiple variations of the application. Therefore, the application is not limited to the following alternative specific embodiments, but is only limited by the claims and their equivalents.

[0011] Example 1

[0012] One embodiment of a seawater reverse osmosis filtration method based on pressure-conductivity dynamic coupling is shown in Figure 1 The method comprises the following steps: A1. As Figure 2 As shown, the seawater in the seawater tank 11 is pumped to the reverse osmosis membrane 31 by the booster pump 21, and pure water and concentrated water are output. The concentrated water is returned to the seawater tank 11 through the circulation valve 41.

[0013] The first sensor group 51, the second sensor group 52, and the third sensor group 53 simultaneously collect parameters upstream and downstream of the reverse osmosis membrane 31, including pressure and conductivity.

[0014] The pressure parameters include static and dynamic parameters. Static parameters include upstream pressure P1, downstream concentrate outlet pressure P2, downstream pure water outlet pressure P3, and the transmembrane pressure difference TMP = (P1 + P2) / 2 - P3. The dynamic parameters are obtained by performing a Fourier transform on the transmembrane pressure difference TMP to extract the pressure pulsation frequency f and pulsation amplitude A.

[0015] The conductivity parameters include the upstream inlet conductivity EC. in , downstream pure water outlet permeate conductivity EC out .

[0016] Desalination rate R = (EC) in -EC out ) / EC in ×100%, conductivity gradient ΔEC=EC in -EC out The rate of change is k = ΔEC / Δt, where Δt is time.

[0017] A2. Extract pressure-conductivity coupling features using the aforementioned parameters, specifically by constructing a pressure-conductivity feature matrix. The pressure-conductivity coupling features include: Efficiency characteristics, TMP / ΔEC and P1 / R.

[0018] Dynamic characteristics, f×ΔEC and A / k.

[0019] Health characteristics, N×(1-R), where N is the number of pulses over a certain period of time.

[0020] The pollution type identification model outputs pollution type identification results based on the coupling features, and the pollution types include biological pollution, colloidal pollution, scaling pollution, and normal state.

[0021] The characteristics of biological pollution are that both f and A increase, while ΔEC decreases.

[0022] The characteristics of scaling fouling are: TMP increases, ΔEC remains stable, and k approaches 0.

[0023] Colloidal contamination is characterized by increased N and A levels and ΔEC fluctuations greater than 5%.

[0024] A3. Based on the identification results, generate the frequency of booster pump 21, the opening degree of circulation valve 41, and cleaning instructions.

[0025] The DDPG algorithm is used to construct a state-action-reward control closed loop.

[0026] The state space S contains TMP, f, A, ΔEC, k, and the pollution type.

[0027] Action space A includes the frequency of booster pump 21, the opening degree of circulation valve 41, and the cleaning trigger threshold.

[0028] The reward function H is: H=λ1×R target +λ2×(1-TMP / TMP max )+λ3×(1-|ΔEC-ΔEC opt | / ΔEC opt )-λ4×I clean .

[0029] R target To achieve the target desalination rate, TMP max For the maximum differential pressure that the reverse osmosis membrane 31 can withstand, ΔEC opt For the optimal conductivity difference, I clean The penalty term for the cleaning action is λ1-λ4, which are dynamic weights.

[0030] Under normal conditions, ΔEC opt To achieve the goal, TMP / ΔEC is minimized by adjusting the frequency of booster pump 21.

[0031] To prevent biological contamination, reduce the frequency of the booster pump 21 by 10%-15%, increase the opening of the circulation valve 41 by 20%-30%, and trigger chemical cleaning earlier.

[0032] To prevent scaling and contamination, maintain the frequency of the booster pump 21 and increase the opening of the circulation valve 41, such as by 15%-25%, to suppress crystal deposition through turbulent scouring.

[0033] Colloidal contamination is addressed by momentarily increasing the frequency of the booster pump 21 by 5%-8% to create a pressure shock, then restoring the original frequency and using pulse flow to peel off the colloidal material.

[0034] A4. Execute instructions and provide feedback parameters to form a closed-loop control.

[0035] The sensor array performs core parameter acquisition and calculation every second: Pressure parameters: upstream pressure P1 (accuracy ±0.005MPa), downstream concentrate outlet pressure P2 (±0.005MPa), pure water outlet pressure P3 (±0.005MPa), real-time calculation of TMP and its real-time pulsation frequency f (Fourier transform window is 5 seconds), amplitude A.

[0036] Conductivity parameters: influent ECin (±1μS / cm), product ECout (±1μS / cm), calculate ΔEC, desalination rate R and rate of change k (sliding window 10 seconds).

[0037] In normal closed-loop operation, if pressure and conductivity fluctuations are within the set thresholds, the current control command is maintained; if they exceed the thresholds, the DDPG algorithm is triggered to make an immediate decision (the time for a single decision is ≤0.5 seconds), adjusting the booster pump frequency or the opening of the concentrate circulation valve until the parameters return to the stable range.

[0038] In the cleaning closed loop, after the cleaning command is completed, the system needs to collect the pressure-conductivity dynamic curve within 10 minutes after cleaning and calculate the product water recovery rate: if the recovery rate is ≥90%, the cleaning is deemed effective and the system switches to the normal operation closed loop; if the recovery rate is <70%, a secondary cleaning is triggered (adjusting the cleaning agent type or extending the duration), and the contamination case is added to the training set of the sub-model to optimize the contamination identification accuracy.

[0039] Safety protection closed loop: When P1 exceeds the threshold or ECout exceeds the standard, the safety mode is immediately triggered: the booster pump frequency drops to 20Hz, the circulation valve opening is adjusted to 50%, and an audible and visual alarm is issued at the same time; after the parameters return to the safe range, the system automatically restarts the AI ​​control process and switches from the stable state parameters of the most recent 30 seconds.

[0040] Example 2

[0041] This embodiment 1 illustrates the application of a reverse osmosis filtration method in online liquid flashback reverse osmosis filtration.

[0042] In this application example, the parameters of the selected seawater desalination reverse osmosis membrane 31 are: Desalination rate: 99.8%; Recovery rate: 10%. Maximum pressure: 8.27 MPa; Recommended pressure: 5.5 MPa. Maximum temperature: 45℃; Recommended temperature: 25℃. Maximum influent flow rate: 19.3 m³ / h. 3 / h.

[0043] The pressure sensors in the first group of sensors 51, the second group of sensors 52, and the third group of sensors 53 are RS485 MODBUS pressure transmitters with a maximum range of 60 MPa, a baud rate of 9600, and a sampling frequency of 20 Hz.

[0044] The conductivity sensors in the first group of sensors 51 and the second group of sensors 52 adopt high-conductivity smart electrodes with a measurement range of 0-50000μS / cm, RS485 Modubus communication, baud rate of 9600, and sampling frequency of 20Hz.

[0045] The pollution type identification model was trained using 1000 sets of historical pollution data. The tree depth of the random forest algorithm was set to 15, and the minimum number of samples for node splitting was 5.

[0046] The experience replay pool size for the DDPG algorithm is set to 10. 6 Both the Actor network and the Critic network use a 3-layer fully connected structure (with 128, 64 and 32 neurons respectively).

[0047] like Figure 1 As shown, the reverse osmosis method in this embodiment includes the following steps: A1. The first sensor group 51, the second sensor group 52 and the third sensor group 53 synchronously collect parameters upstream and downstream of the reverse osmosis membrane 31, including pressure and conductivity.

[0048] The inlet water flow rate is set to 10m³. 3 / h, target water production 2 3 / h, the membrane module operating temperature is 15-25℃.

[0049] Pressure parameters are collected at 1-second intervals. These parameters include static and dynamic parameters. Static parameters include upstream pressure P1 (working pressure range 0.8-1.0 MPa, with an alarm triggered if exceeding 1.5 MPa, and an alarm triggered if exceeding 2.0 MPa), downstream concentrate outlet pressure P2 (pressure range 0.5-0.8 MPa), downstream pure water outlet pressure P3 (pressure range 0.1-0.20 MPa), and transmembrane pressure difference TMP = (P1+P2) / 2 - P3 (pressure range 0.5-0.8 MPa). Dynamic parameters are obtained by performing Fourier transform on the transmembrane pressure difference TMP, setting a 5-second Fourier transform window, and extracting the pressure pulsation frequency f (range 0.5-5 Hz) and pulsation amplitude A (normal amplitude ≤ 0.02 MPa, exceeding 0.03 MPa is considered abnormal fluctuation).

[0050] The conductivity parameters include the upstream inlet conductivity EC. in (30000-45000 μS / cm), downstream pure water outlet permeate conductivity EC out (≤100μS / cm standard value).

[0051] Desalination rate R = (EC) in -EC out ) / EC in ×100%, conductivity gradient ΔEC=EC in -EC out (Normal range 29960-44960μS / cm) The rate of change k = ΔEC / Δt, where Δt is the time set to 10s, and the normal fluctuation range of k is ±10μS / cm·min.

[0052] A2. Extract pressure-conductivity coupling features using the aforementioned parameters, specifically by constructing a pressure-conductivity feature matrix. The pressure-conductivity coupling features include: Efficiency characteristics, TMP / ΔEC and P1 / R.

[0053] Dynamic characteristics, f×ΔEC and A / k.

[0054] Health characteristics, N×(1-R), where N is the number of pulses in 10 minutes (normal should be ≤10).

[0055] The pollution type identification model outputs pollution type identification results based on the coupling features, and the pollution types include biological pollution, colloidal pollution, scaling pollution, and normal state.

[0056] The characteristics of biological pollution are: f increases by ≥20%, A increases by ≥20%, and ΔEC decreases by ≥15%.

[0057] The characteristics of scaling fouling are: TMP increase ≥15%, ΔEC fluctuation ≤5%, and k ≤1μS / cm·min.

[0058] Colloidal contamination is characterized by an increase of N ≥30%, an increase of A ≥30%, and a fluctuation of ΔEC greater than 5%.

[0059] A3. Based on the identification results, generate the frequency of booster pump 21, the opening degree of circulation valve 41, and cleaning instructions.

[0060] The DDPG algorithm is used to construct a state-action-reward control closed loop.

[0061] The state space S contains TMP, f, A, ΔEC, k, and the pollution type.

[0062] The action space A includes a booster pump 21 with a frequency adjustment step of 0.2-0.5 Hz / time, a circulation valve 41 with an opening of 1-2% / time, and cleaning trigger thresholds (biological contamination lasting 10 minutes and ΔEC decreasing by ≥15%; scaling contamination lasting 15 minutes and TMP increasing by ≥20%; colloidal contamination lasting 8 minutes and ΔEC fluctuation >8%).

[0063] The reward function H is: H=λ1×R target +λ2×(1-TMP / TMP max )+λ3×(1-|ΔEC-ΔEC opt | / ΔEC opt )-λ4×I clean .

[0064] R target For the target desalination rate (target value ≥ 98%), TMP maxFor the reverse osmosis membrane 31 to withstand the maximum pressure difference (2 MPa), ΔEC opt (≤100μS / cm) represents the optimal conductivity difference, I clean The penalty for the cleaning action is set to 0.8, and λ1-λ4 are the dynamic weights λ1=0.3, λ2=0.4, λ3=0.2, and λ4=0.1, respectively.

[0065] Under normal conditions, ΔEC opt With the target value of ≤100μS / cm, TMP / ΔEC is minimized by adjusting the frequency of booster pump 21.

[0066] To prevent biological contamination, reduce the frequency of the booster pump 21 by 10%-15%, increase the opening of the circulation valve 41 by 20%-30%, and trigger chemical cleaning earlier.

[0067] To prevent scaling and contamination, maintain the frequency of the booster pump 21 and increase the opening of the circulation valve 41, such as by 15%-25%, to suppress crystal deposition through turbulent scouring.

[0068] Colloidal contamination is addressed by momentarily increasing the frequency of the booster pump 21 by 5%-8% to create a pressure shock, then restoring the original frequency and using pulse flow to peel off the colloidal material.

[0069] A4. Execute instructions and provide feedback parameters to form a closed-loop control.

[0070] The sensor array performs core parameter acquisition and calculation every second: During normal closed-loop operation, if the pressure is <1 MPa and the conductivity is <100 μS / cm, the current control command is maintained; if the threshold is exceeded, the DDPG algorithm is triggered to make an immediate decision (the time for a single decision is ≤0.5 seconds), adjusting the booster pump frequency or the opening of the concentrate circulation valve until the parameters return to the stable range.

[0071] The cleaning closed loop requires the system to collect the pressure-conductivity dynamic curve within 10 minutes after cleaning and calculate the product water recovery rate. If the recovery rate is ≥90%, the cleaning is deemed effective and the system switches to the normal operation closed loop. If the recovery rate is <70%, a secondary cleaning is triggered (adjusting the cleaning agent type or extending the duration), and the contamination case is added to the training set of the sub-model to optimize the contamination identification accuracy.

[0072] Safety protection closed loop: When P1 exceeds the threshold of 2 MPa or Ecout exceeds 150 μS / cm, the safety mode is immediately triggered: the booster pump frequency drops to 20 Hz, the circulation valve opening is adjusted to 50%, and an audible and visual alarm is issued at the same time; after the parameters return to the safe range, the system automatically restarts the AI ​​control process and switches from the stable state parameters of the most recent 30 seconds.

[0073] In actual operation, when biological contamination is detected (f=2.1Hz, A=0.3MPa, ΔEC decreases by 15%), the system automatically reduces the frequency of booster pump 21 from 40Hz to 34Hz, increases the opening of concentrate circulation valve 41 from 10% to 25%, and starts sodium hypochlorite cleaning 2 hours in advance. The final membrane flux recovery rate reaches 98%, which is 12% higher than the traditional method.

Claims

1. A seawater reverse osmosis filtration method based on dynamic coupling of pressure and conductivity, characterized in that, The method includes the following steps: A1. The seawater in the seawater tank is pumped to the reverse osmosis membrane by a booster pump, and pure water and concentrated water are output. The concentrated water is returned to the seawater tank through a circulation valve. Parameters, including pressure and conductivity, are collected simultaneously upstream and downstream of the reverse osmosis membrane. A2. Using the parameters, pressure-conductivity coupling features are extracted, and the pollution type identification model outputs pollution type identification results based on the coupling features; A3. Based on the identification results, generate booster pump frequency, circulation valve opening degree, and cleaning command; A4. Execute instructions and provide feedback parameters to form a closed-loop control.

2. The method according to claim 1, characterized in that, Pressure parameters include: Static parameters include upstream pressure P1, downstream concentrate outlet pressure P2, downstream pure water outlet pressure P3, and transmembrane pressure difference TMP = (P1 + P2) / 2 - P3. Dynamic parameters were used to perform Fourier transform on the transmembrane pressure difference TMP, and the pressure pulsation frequency f and pulsation amplitude A were extracted.

3. The method according to claim 2, characterized in that, The conductivity parameters include the upstream inlet conductivity EC. in Downstream pure water outlet permeate conductivity EC out ; Desalination rate R = (EC) in -EC out ) / EC in ×100%, conductivity gradient ΔEC=EC in -EC out The rate of change is k = ΔEC / Δt, where Δt is time.

4. The method according to claim 3, characterized in that, The method for extracting pressure-conductivity coupling characteristics is as follows: Construct the pressure-conductivity characteristic matrix, including: Efficiency characteristics, TMP / ΔEC and P1 / R; Dynamic characteristics, f×ΔEC and A / k; Health characteristics, N×(1-R), where N is the number of pulses over a certain period of time.

5. The method according to claim 4, characterized in that, The types of contamination include biological contamination, colloidal contamination, scaling contamination, and normal conditions; The characteristics of biological pollution are that both f and A increase, while ΔEC decreases; The characteristics of scaling fouling are: TMP increases, ΔEC remains stable, and k approaches 0; Colloidal contamination is characterized by increased N and A levels and ΔEC fluctuations greater than 5%.

6. The method according to claim 5, characterized in that, The recognition model uses a classifier built based on the random forest algorithm.

7. The method according to claim 4, characterized in that, The generation method is as follows: The DDPG algorithm is used to construct a state-action-reward control closed loop; The state space S contains TMP, f, A, ΔEC, k, and the contamination type; Action space A includes booster pump frequency, circulation valve opening degree, and cleaning trigger threshold; The reward function H is: H=λ1×R target +λ2×(1-TMP / TMP max )+λ3×(1-|ΔEC-ΔEC opt | / ΔEC opt )-λ4×I clean ; R target To achieve the target desalination rate, TMP max ΔEC is the maximum pressure differential that the reverse osmosis membrane can withstand. opt For the optimal conductivity difference, I clean The penalty term for the cleaning action is λ1-λ4, which are dynamic weights.

8. The method according to claim 7, characterized in that, The dynamic weights are updated in real time through particle swarm optimization.

9. The method according to claim 7, characterized in that, Under normal conditions, ΔEC opt To achieve the goal, TMP / ΔEC is minimized by adjusting the booster pump frequency; Biological contamination can be addressed by reducing the frequency of the booster pump, increasing the opening of the circulation valve, and triggering chemical cleaning earlier. To combat scaling and contamination, maintain the frequency of the booster pump, increase the opening of the circulation valve, and suppress crystal deposition through turbulent scouring. Colloidal contamination is addressed by momentarily increasing the frequency of the booster pump to create a pressure surge, then restoring the original frequency and using pulsed flow to peel off the colloidal material.

10. The method according to claim 9, characterized in that, To address biological contamination, reduce the booster pump frequency by 10%-15% and increase the circulation valve opening by 20%-30%. Scale buildup and contamination increase the opening degree of the circulation valve by 15%-25%; Colloidal contamination instantly increases the frequency of the booster pump by 5%-8%.

Images