Intelligent mineralization treatment process for reverse osmosis produced water in the pearl river estuary

Abstract

The application discloses an intelligent mineralization treatment process for reverse osmosis water of brackish water in the Pearl River Estuary, comprising the following steps: collecting water quality parameters of brackish water at a water intake of the Pearl River Estuary and constructing a water quality fluctuation prediction model in a salt tide period; obtaining reverse osmosis water after pretreatment and reverse osmosis desalination of the brackish water; mixing the reverse osmosis water with carbon dioxide in a pipeline static mixer after pressure boosting to form a carbonic acid solution, and monitoring the actual dissolution amount of the carbon dioxide at the outlet of the static mixer in real time; when the measured dissolution amount is lower than a theoretical calculation value, automatically executing a compensation measure; sending the carbonic acid solution from the bottom into a fixed bed mineralization reactor filled with limestone filter material to perform a mineralization reaction, multiple online pH sensors are arranged at different heights in the reactor, and the operation state of the reactor is judged and automatically regulated according to the gradient distribution of the pH values at the heights; and monitoring the Langley index of the effluent after mineralization in real time and dynamically feeding back and adjusting. The application realizes intelligent and accurate regulation and control of the whole process of mineralization treatment of reverse osmosis water.

Description

Technical Field

[0001] This invention relates to the post-treatment of brackish water desalination, specifically to an intelligent mineralization treatment process for reverse osmosis permeate from brackish water in the Pearl River Estuary, which is particularly suitable for precise mineralization control of reverse osmosis permeate under conditions of drastic fluctuations in water quality during the saltwater intrusion period in the Pearl River Estuary. Background Technology

[0002] The Pearl River Estuary is a typical area of ​​saltwater intrusion in my country. Affected by reduced upstream water flow and the combined effects of tides, varying degrees of saltwater intrusion occur annually during the dry season from October to April of the following year. During saltwater intrusion, the salinity of the slightly brackish water at the Pearl River Estuary intake can rise sharply from the usual 0.5‰ to over 15‰, with water quality parameters such as total dissolved solids and turbidity also fluctuating dramatically. Multiple water intakes in Zhuhai City have experienced varying degrees of salinity increases during saltwater intrusion, seriously threatening the safety of the municipal water supply. Nearly 90% of the city's raw water relies on the Xijiang River as a single source, and this "relying on the river for water" model faces severe challenges in the face of increasingly frequent saltwater intrusion.

[0003] Reverse osmosis (RO) seawater desalination technology is an important way to solve the freshwater shortage in coastal areas. However, while RO membranes efficiently remove salt from water, they also remove large amounts of beneficial minerals such as calcium and magnesium, resulting in severely low hardness, alkalinity, and pH levels in the RO permeate. The permeate obtained from treating the slightly brackish water of the Pearl River Estuary using RO typically has a calcium ion content of only 3-5 mg / L, a total hardness of only 13-17 mg / L (as CaCO3), a total alkalinity of only 10-15 mg / L (as CaCO3), and a Langerile Index (LSI) as low as -3.0 to -4.0. This low-hardness, low-alkalinity desalinated water is highly corrosive. Directly entering the municipal water supply network will dissolve the original calcium carbonate protective layer on the inner wall of the pipes, causing corrosion of iron pipes and resulting in "yellow water" or "red water," seriously affecting residents' normal water use.

[0004] Therefore, reverse osmosis permeate must undergo mineralization treatment before entering the water transmission network. This involves adding minerals to increase the water's hardness and alkalinity, stabilizing its chemical properties to meet the requirements of the network and drinking water supply. Currently, the main mineralization methods for seawater desalination permeate include direct chemical addition, carbon dioxide dissolution of limestone, and carbon dioxide-lime slurry. Among these, the carbon dioxide dissolution of limestone method has become the most commonly used mineralization method in large-scale seawater desalination projects due to its advantages such as wide availability of raw materials, low operating costs, and no introduction of additional chemical substances.

[0005] There are existing reports on the carbon dioxide dissolution limestone mineralization process. For example, Chinese patent CN102826689A discloses a post-treatment process and equipment for seawater desalination, which involves adding CO2 to desalinated water and then passing it through a mineralization tank with a calcium carbonate-filled bed for mineralization. Chinese patent CN102897945A discloses a method for mineralizing desalinated water, which involves adding limestone to desalinated water and then introducing CO2 followed by stirring and reaction. Chinese patent CN106608692A discloses a highly efficient and controllable seawater desalination process, which uses an absorption tower to mix CO2 with desalinated water before introducing it into a reaction vessel for mineralization. Chinese patent CN114249472A discloses a method and apparatus for remineralizing desalinated water using a limestone fixed-bed reactor, in which acidified desalinated water is passed through a limestone-filled fixed-bed reactor for limestone dissolution and mineralization.

[0006] However, the aforementioned existing technologies all have the following technical defects: First, there is a lack of real-time sensing and compensation mechanisms for the actual amount of CO2 dissolved. Current processes generally employ an "open-loop control" model for CO2 dosing—adding a set amount of CO2 is considered sufficient for dissolution, without online monitoring of the actual dissolved amount. However, the solubility of CO2 in water is affected by various factors such as temperature, pressure, and the ionic composition of the water. After reverse osmosis treatment, the ionic composition of the product water from the slightly brackish water of the Pearl River Estuary changes with fluctuations in the salinity of the raw water, directly affecting the CO2 dissolution efficiency. When sudden changes in water quality lead to a decrease in CO2 dissolution efficiency, existing processes cannot detect and compensate for this in a timely manner, resulting in insufficient acidity of the carbonic acid solution entering the mineralization reactor and a deterioration in the mineralization effect.

[0007] Secondly, mineralization reactors are considered "black boxes," lacking online monitoring methods for their internal operating status. Current technology only monitors the influent and effluent water quality parameters, remaining completely unaware of the reaction processes and distribution within the reactor. In reality, the reaction progress varies at different heights of the limestone filter media layer—the bottom, in contact with fresh carbonate solution, experiences the most vigorous reaction; at the top, most of the carbonate has been consumed, and the reaction tends to slow down.

[0008] Third, the mineralization control strategy is simplistic and lacks multi-parameter coordinated control based on the internal state of the reactor. Existing technologies mainly rely on feedback regulation based on influent and effluent water quality parameters, resulting in a delayed response and an inability to precisely intervene in local anomalies within the reactor. Summary of the Invention

[0009] The purpose of this invention is to overcome the shortcomings of existing technologies and provide an intelligent mineralization treatment process for reverse osmosis permeate from the Pearl River Estuary. By monitoring the pH gradient distribution inside the reactor, it can reflect the spatial distribution of filter media activity and whether abnormal operating conditions such as water flow short-circuiting or localized caking exist. Specifically, it includes the following steps: S1. Water Quality Monitoring and Prediction Procedures Water quality parameters of slightly saline water at the Pearl River Estuary intake were collected, including salinity, temperature, and total dissolved solids. A water quality fluctuation prediction model for saline tide periods was constructed based on the time series data of these water quality parameters.

[0010] Preferably, the method for constructing the water quality fluctuation prediction model during the saline tide period includes: collecting daily variation data of salinity, temperature, and total dissolved solids during the historical same period of the Pearl River Estuary saline tide period; establishing prediction sub-models for each parameter using a Long Short-Term Memory Neural Network (LSTM); and obtaining a comprehensive water quality prediction result by weighted fusion of the prediction results of each sub-model, with a prediction time window of 4-24 hours. The model parameters are updated online every 24 hours using the latest collected measured data.

[0011] Saltwater intrusion in the Pearl River Estuary exhibits significant tidal periodicity and seasonality. By using LSTM neural networks to learn from historical data, the periodic characteristics of saltwater intrusion and the correlation between them and water quality parameters can be effectively captured, enabling accurate prediction of water quality changes in the next 4-24 hours and providing a time window for the pre-adjustment of subsequent mineralization process parameters.

[0012] S2. Reverse osmosis desalination step The slightly brackish water is pretreated and then desalinated by reverse osmosis to obtain reverse osmosis permeate. The pretreatment includes coagulation sedimentation, media filtration, and ultrafiltration. The reverse osmosis desalination uses a seawater desalination reverse osmosis membrane, and the recovery rate of the reverse osmosis system is 45%-60%.

[0013] S3. Pressurized carbonation and real-time monitoring and compensation steps for CO2 dissolution The reverse osmosis permeate is pressurized to 0.3-0.6 MPa by a booster pump and then thoroughly mixed with gaseous carbon dioxide in a static mixer in the pipeline to form a carbonic acid solution. The purpose of pressurization is to increase the solubility of CO2 in water—according to Henry's Law, the solubility of a gas in water is directly proportional to pressure. Pressurizing the reverse osmosis permeate to 0.3-0.6 MPa can significantly increase the amount of CO2 that can be dissolved.

[0014] An online carbon dioxide concentration sensor is installed at the outlet of the static mixer in the pipeline to monitor the actual dissolved carbon dioxide in the water in real time. When the actual dissolved amount is lower than 90% of the theoretically calculated dissolved amount, at least one of the following compensation measures is automatically implemented: ① Increase the outlet pressure of the booster pump. According to Henry's Law, increasing the pressure can directly increase the equilibrium solubility of CO2. The increase should be 0.05-0.15 MPa.

[0015] ② Increase the turbulence intensity within the static mixer in the pipeline. By adjusting the angle of the guide vanes inside the mixer, the degree of fluid turbulence is increased, the gas-liquid mass transfer coefficient is improved, and the Reynolds number is increased by 20%-50%, thereby accelerating the CO2 dissolution process.

[0016] When the actual dissolved amount is less than 85% of the theoretically calculated dissolved amount, the above-mentioned compensation measures ① and ② are simultaneously implemented to ensure that the dissolved CO2 amount is restored to the target level. The theoretically calculated dissolved amount is based on Henry's Law, and the calculation formula is C* = H·P, where H is the Henry's constant for carbon dioxide in water (related to temperature), and P is the partial pressure of carbon dioxide at the outlet of the booster pump.

[0017] This step introduces online real-time monitoring of the actual CO2 dissolution amount in the seawater desalination mineralization process and establishes a closed-loop compensation mechanism of "dissolution amount deviation → multi-level compensation measures". This mechanism completely changes the erroneous assumption of "dosage amount = dissolution amount" in existing technologies, realizing the leap from "open-loop control" to "closed-loop control" of CO2 dosing.

[0018] S4. Upflow fixed-bed mineralization reaction and pH gradient monitoring steps The carbonic acid solution is fed from the bottom into a fixed-bed mineralization reactor filled with limestone filter media at a flow rate of 0.10-0.30 m / min, and the mineralization reaction is carried out by an upflow through the limestone filter media layer. Carbonic acid reacts with calcium carbonate in the limestone to form calcium bicarbonate, and the empty bed contact time of the mineralization reaction is 4-10 minutes.

[0019] The limestone filter media has a particle size of 1-5 mm, a calcium carbonate mass percentage of ≥95%, a filter media layer height of 1.2-2.5 m, and a diameter-to-height ratio of 1:3 to 1:6 for the fixed-bed mineralization reactor.

[0020] At least three online pH sensors are installed at different heights along the water flow direction within the fixed-bed mineralization reactor to acquire pH values ​​at each height in real time and calculate the pH gradient distribution. Preferably, the three online pH sensors are respectively located 0.2m above the bottom of the limestone filter media layer, in the middle of the filter media layer, and 0.2m below the top of the filter media layer. The pH gradient distribution includes the pH difference between adjacent sensors and the total difference between the bottom sensor and the top sensor.

[0021] The core advantage of using the upward flow method is that as the water flows from the bottom upward, it comes into full contact with the limestone filter media. At the same time, the upward water flow can maintain a certain expansion rate of the filter media layer, avoid filter media caking, and improve reaction efficiency.

[0022] The core chemical reaction of mineralization is: CaCO3(s) + CO2(aq) + H2O ⇌ Ca2+ (aq) + 2HCO3⁻(aq) This reaction is reversible, and its direction is affected by CO2 concentration, temperature, and pH. The pH is lowest (4.5-5.8) when the carbonated solution enters from the bottom, and gradually increases after contact with limestone, reaching 7.0-8.0 when the water exits from the top. This process of pH increase along the way directly reflects the limestone dissolution reaction.

[0023] S5. Operational status diagnosis and control steps based on pH gradient distribution The operating status of the fixed-bed mineralization reactor is determined based on the pH gradient distribution obtained in step S4. ① When the pH difference between the bottom and top of the filter media is less than 0.3, the limestone filter media is considered to have insufficient activity—indicating that the pH hardly increases after the carbonate solution flows through the filter media layer, and the limestone dissolution reaction does not occur effectively. Possible reasons include the filter media surface being covered by contaminants and losing its activity, or a large amount of CaCO3 in the filter media being consumed. At this point, a warning signal is issued to replenish the filter media or switch to a standby reactor.

[0024] ② When the pH difference between two adjacent online pH sensors is greater than 0.8, it is determined that there is a water flow short circuit or local caking in the filter media layer—indicating that the carbonic acid solution is unevenly distributed in the filter media layer, with most of the water flow passing through a certain local area, causing over-reaction in that area (a sharp increase in pH), while the filter media in other areas does not effectively participate in the reaction. At this time, a local backwashing procedure is initiated: the inlet valve of the fixed-bed mineralization reactor is closed, and the backwashing inlet valve and backwashing air inlet valve are opened. The mineralized effluent is used as the backwashing water source, and compressed air is used as the backwashing air source to perform combined air-water backwashing on the filter media layer. The backwashing time is 5-15 minutes.

[0025] This step establishes a multi-dimensional pH monitoring network inside the mineralization reactor. By inferring the reactor's operating status through the spatial distribution characteristics of the pH gradient, it achieves "internal visualization" and precise diagnosis of the mineralization reactor. In existing technologies, mineralization reactors are considered "black boxes"—only the influent and effluent are monitored, without understanding what is happening inside. This invention, through the "window" of pH gradient distribution, achieves real-time perception of the reactor's internal operating status and precise location of abnormal conditions.

[0026] S6. Feedback control steps based on effluent LSI The mineralized effluent from step S4 is sent to a post-mineralization tank, and the Langerile Index (LSI) of the post-mineralized effluent is monitored in real time. LSI is a core indicator for evaluating the chemical stability of water quality, calculated as LSI = pH - pHs, where pHs is the saturation pH value. A negative LSI indicates a tendency for corrosion, while a positive LSI indicates a tendency for scaling.

[0027] The amount of carbon dioxide added in step S3 and / or the flow rate of the carbonated solution in step S4 are dynamically adjusted based on the actual LSI measurement value. When the LSI is below -0.50 (the water quality is corrosive), increase the amount of carbon dioxide added (to increase the acidity of the carbonic acid solution and accelerate the dissolution of limestone) and decrease the feed rate of the carbonic acid solution (to prolong the hydraulic residence time and increase the degree of reaction). When the LSI is higher than +0.50 (the water quality has a tendency to scale), reduce the amount of carbon dioxide added and increase the flow rate of the carbonate solution to shorten the hydraulic residence time and reduce the degree of reaction; so that the LSI of the effluent after mineralization is controlled between -0.50 and +0.50.

[0028] S7. CO2 Recycling Steps A portion of the effluent from the mineralized water tank is pressurized and then returned to the front end of the static mixer in step S3. It is then mixed with the reverse osmosis permeate before carbon dioxide is added. The reflux ratio is controlled between 10% and 30%. The reflux water contains unreacted dissolved CO2 and bicarbonate ions. Mixing it with fresh reverse osmosis permeate before CO2 addition effectively improves the overall CO2 utilization rate.

[0029] Compared with the prior art, the present invention has the following advantages using the above method: (1) In the prior art, CO2 addition is "open-loop control"—adding the set amount is considered sufficient for complete dissolution, without verifying the actual amount dissolved. In this invention, a CO2 concentration sensor is installed at the outlet of the static mixer in the pipeline to monitor the actual amount dissolved in real time. When the measured value is lower than the theoretical value, a three-level compensation measure of "pressure increase-turbulence increase-surfactant addition" is automatically implemented. This mechanism completely changes the traditional assumption that "addition amount = dissolution amount", realizes true closed-loop control of CO2 addition, and improves the utilization rate of CO2 and the stability of mineralization effect.

[0030] (2) Existing technologies only monitor the quality of influent and effluent, treating the reactor as a "black box". This invention sets up multiple online pH sensors at different heights in the reactor, and judges the filter media activity and diagnoses abnormal operating conditions such as water flow short circuit and local caking by the spatial distribution characteristics of pH gradient.

[0031] (3) The monitoring information at three levels (CO2 dissolution at the molecular level, pH gradient distribution at the reactor level, and LSI at the effluent water quality level) is organically integrated to form a full-process intelligent control network from micro to macro and from local to overall. The monitoring information at the three levels complements and corroborates each other, jointly ensuring the stability and reliability of the mineralization effect.

[0032] (4) The process parameters (pressurization pressure 0.3-0.6MPa, CO2 dissolution deviation threshold 90% / 85%, pH gradient judgment threshold 0.3 / 0.8, LSI control range -0.50~+0.50, etc.) designed specifically for the slightly brackish water of the Pearl River Estuary have been specially optimized for the salinity of 0.5‰-18‰, large TDS fluctuation range, and periodic salt tide of the slightly brackish water of the Pearl River Estuary. Attached Figure Description

[0033] Figure 1 A schematic diagram of the intelligent mineralization treatment process for reverse osmosis permeate from the slightly brackish water in the Pearl River Estuary; Figure 2 A schematic diagram showing the arrangement of pH sensors and pH gradient distribution inside a fixed-bed mineralization reactor; Figure 3 Logic block diagram for real-time monitoring and multi-level compensation control of CO2 dissolution; Figure 4 This is a flowchart illustrating the operational status diagnosis and control decision-making process based on pH gradient distribution in this invention. Figure 5 This is a comparison curve of LSI in the mineralized effluent from the example and the conventional process after 30 days of continuous operation during the saline tide period. Detailed Implementation

[0034] The present invention will be further described in detail below with reference to the embodiments. The embodiments are only intended to provide a clearer understanding of the technical features, objectives and effects of the present invention.

[0035] The present invention discloses an intelligent mineralization treatment process for reverse osmosis permeate from the slightly brackish water of the Pearl River Estuary, comprising the following steps: S1. Collect water quality parameters of slightly saline water at the Pearl River Estuary intake and construct a water quality fluctuation prediction model during the saline tide period. The water quality parameters include salinity, temperature and total dissolved solids. S2. The slightly brackish water is pretreated and then desalinated by reverse osmosis to obtain reverse osmosis permeate; S3. After the reverse osmosis permeate is pressurized to 0.3-0.6 MPa by a booster pump, it is thoroughly mixed with gaseous carbon dioxide in a pipeline static mixer to form a carbonic acid solution. An online carbon dioxide concentration sensor is installed at the outlet of the pipeline static mixer to monitor the actual amount of carbon dioxide dissolved in the water in real time. When the actual amount of dissolved carbon dioxide is lower than 90% of the theoretically calculated amount, at least one of the following compensation measures is automatically executed: ① Increase the outlet pressure of the booster pump; ② Increase the turbulence intensity in the pipeline static mixer. S4. The carbonic acid solution is fed from the bottom into a fixed-bed mineralization reactor filled with limestone filter media at a flow rate of 0.10-0.30 m / min. The mineralization reaction is carried out by passing through the limestone filter media layer in an upflow manner. At least three online pH sensors are installed at different heights along the water flow direction in the fixed-bed mineralization reactor to obtain the pH value at each height in real time and calculate the pH gradient distribution. S5. Determine the operating status of the fixed-bed mineralization reactor based on the pH gradient distribution obtained in step S4: when the difference between the bottom pH value and the top pH value is less than 0.3, it is determined that the limestone filter media is not active enough, and an early warning signal is issued to supplement the filter media or switch the reactor; when the pH difference between two adjacent online pH sensors is greater than 0.8, it is determined that there is a water flow short circuit or local caking in the filter media layer, and a local backwashing procedure is initiated. S6. The mineralized effluent from step S4 is sent into the mineralized water tank, and the Langerile Index (LSI) of the mineralized effluent is monitored in real time. The amount of carbon dioxide added in step S3 and / or the flow rate of the carbonated solution in step S4 are dynamically adjusted according to the measured LSI value so that the LSI of the mineralized effluent is controlled between -0.50 and +0.50.

[0036] In some embodiments, the method for constructing the water quality fluctuation prediction model during the saltwater intrusion period in step S1 includes: collecting daily variation data of salinity, temperature and total dissolved solids during the historical same period of the Pearl River Estuary saltwater intrusion period, establishing prediction sub-models for each parameter using a long short-term memory neural network, and obtaining a comprehensive water quality prediction result by weighted fusion of the prediction results of each sub-model, with a prediction time window of 4-24 hours.

[0037] In some embodiments, the theoretical dissolution amount in step S3 is calculated based on Henry's Law, and the calculation formula is C* = H·P, where H is the Henry's constant for carbon dioxide in water and P is the partial pressure of carbon dioxide at the outlet of the booster pump; when the actual dissolution amount is less than 85% of the theoretically calculated dissolution amount, the compensation measures ① and ② are executed simultaneously.

[0038] In some embodiments, the increase in the outlet pressure of the booster pump in compensation measure ① is 0.05-0.15 MPa; the method to increase the turbulence intensity in the static mixer in the pipeline in compensation measure ② is to adjust the angle of the guide vanes inside the mixer to increase the Reynolds number by 20%-50%.

[0039] In some embodiments, in step S4, three online pH sensors are installed at different heights along the water flow direction inside the fixed-bed mineralization reactor, respectively located 0.2m above the bottom of the limestone filter layer, in the middle of the filter layer, and 0.2m below the top of the filter layer; the pH gradient distribution includes the pH difference between adjacent sensors and the total difference between the bottom sensor and the top sensor.

[0040] In some embodiments, the method of initiating the local backwashing procedure in step S5 includes: closing the water inlet valve of the fixed bed mineralization reactor, opening the backwash water inlet valve and the backwash air inlet valve, using the mineralized effluent as the backwash water source and compressed air as the backwash air source, and performing air-water combined backwashing on the filter media layer for a backwashing time of 5-15 minutes.

[0041] In some embodiments, the limestone filter media has a particle size of 1-5 mm, a calcium carbonate mass percentage of ≥95%, a filter media layer height of 1.2-2.5 m, and a diameter-to-height ratio of 1:3 to 1:6 for the fixed-bed mineralization reactor.

[0042] In some embodiments, the method of dynamically adjusting the carbon dioxide dosage and / or the carbonate solution feed rate according to the measured LSI value in step S6 includes: when the LSI is below -0.50, increasing the carbon dioxide dosage and decreasing the carbonate solution feed rate; when the LSI is above +0.50, decreasing the carbon dioxide dosage and increasing the carbonate solution feed rate.

[0043] In some embodiments, the pretreatment includes coagulation sedimentation, media filtration and ultrafiltration, and the reverse osmosis desalination process uses a seawater desalination reverse osmosis membrane, with a recovery rate of 45%-60% for the reverse osmosis system.

[0044] In some embodiments, the present invention further includes step S7: a portion of the effluent from the mineralized water tank is pressurized and then returned to the front end of the static mixer in the pipeline in step S3, where it is mixed with the reverse osmosis permeate before carbon dioxide is added, with the return ratio controlled at 10%-30%. Example

[0045] This embodiment uses the water intake of a pumping station in the Pearl River Estuary as the application scenario. The pilot plant has a processing capacity of 5 m³ / h (reverse osmosis permeate), and the test period is from October 2025 to March 2026, covering the main saltwater intrusion period in the Pearl River Estuary.

[0046] (I) Detailed Implementation of Step S1 Online water quality monitoring instruments, including salinity meters, thermometers, and TDS meters, are installed at the water intake point to continuously collect raw water quality parameters at a sampling frequency of once per hour.

[0047] Diurnal variation data of salinity, temperature, and TDS during the Pearl River Estuary's saltwater intrusion period (October to March of the following year, 2020-2024) were collected to construct a training dataset. Long Short-Term Memory (LSTM) neural networks were used to establish prediction sub-models for each parameter. The input to each sub-model was the water quality parameter time series of the past 72 hours, and the output was the predicted value for the next 24 hours. The prediction results of each sub-model were weighted and fused to obtain a comprehensive water quality prediction result. The weight coefficients were determined according to the degree of influence of each parameter on the mineralization effect (salinity weight 0.50, temperature weight 0.30, TDS weight 0.20).

[0048] The model updates its parameters online every 24 hours using the latest collected measured data. In this embodiment, the model's average relative error in predicting salinity for the next 12 hours is 8.7%, which meets the requirements for engineering applications.

[0049] (II) Detailed Implementation of Step S2 The slightly saline water from the intake is pumped up by a submersible pump and then enters a coagulation sedimentation tank, a multi-media filter and an ultrafiltration system for pretreatment before entering the reverse osmosis system for desalination.

[0050] The coagulation and sedimentation system uses polyaluminum chloride (PAC) as the coagulant, with a dosage of 5-15 mg / L (adjusted in real-time according to the influent turbidity), and polyacrylamide (PAM) as the coagulant aid, with a dosage of 0.1-0.5 mg / L. The multi-media filter uses anthracite-quartz sand dual-layer filter media, with a designed filtration rate of 8 m / h. The ultrafiltration system uses a PVDF hollow fiber ultrafiltration membrane, with a designed flux of 50 L / m²·h and a recovery rate ≥90%.

[0051] The reverse osmosis system uses seawater desalination reverse osmosis membranes, with 6 membrane modules arranged in series, designed to produce 5 m³ / h of water. The high-pressure pump outlet pressure is automatically adjusted according to the influent salinity, ranging from 4.5 to 6.5 MPa, and the system recovery rate is 50%-60%.

[0052] The water quality parameters of reverse osmosis permeate are: TDS 80-350mg / L (varying with feed water salinity), calcium ions 3-8mg / L, total hardness (as CaCO3) 13-35mg / L, total alkalinity (as CaCO3) 10-30mg / L, pH 6.5-7.2, LSI -2.5 to -4.0.

[0053] (III) Detailed Implementation of Step S3 The reverse osmosis permeate is pressurized to 0.45 MPa by a booster pump and then enters the pipeline static mixer (DN50, with built-in spiral guide vanes). Food-grade carbon dioxide is depressurized to 0.5 MPa from a cylinder via a pressure reducing valve, and then precisely metered by a mass flow controller (MFC) before entering the static mixer to thoroughly mix with the pressurized reverse osmosis permeate to form a carbonic acid solution. The baseline CO2 dosage is 12 g / m³ of permeate, and feedforward adjustment is performed based on the predicted salinity value from step S1—for every 1‰ increase in the predicted salinity value, the CO2 dosage increases by 8%.

[0054] An online CO2 concentration sensor (based on infrared absorption principle, range 0-1000 mg / L, accuracy ±2%FS) is installed at the outlet of the pipeline static mixer to monitor the actual amount of CO2 dissolved in the water in real time.

[0055] The theoretical solubility calculation is based on Henry's Law. Taking a water temperature of 25℃ as an example, the Henry's constant H for CO2 is approximately 3.4 × 10⁻⁶. -2 Given a concentration of mol / (L·atm), a booster pump outlet pressure of 0.45 MPa (approximately 4.44 atm), and a CO2 partial pressure of approximately 0.4 MPa (approximately 3.95 atm), the theoretical solubility C* = 3.4 × 10⁻⁶ mol / (L·atm). -2 × 3.95 ≈0.134mol / L≈5.9g / L. The actual CO2 dosage in the produced water is 12g / m³ = 0.012g / L, which is far below the saturation solubility, so theoretically it should dissolve completely.

[0056] When the measured CO2 dissolution is lower than 90% of the theoretically calculated dissolution (i.e., lower than 0.0108 g / L), the control system automatically implements compensation measures. In this embodiment, during a saltwater intrusion, the raw water salinity increased from 2‰ to 12‰ within 6 hours, causing a change in the ionic composition of the reverse osmosis permeate, and the measured CO2 dissolution decreased to 82% of the theoretical value. The control system automatically implemented the following compensation measures: ① Increase the outlet pressure of the booster pump by 0.10 MPa (from 0.45 MPa to 0.55 MPa); ② Adjust the angle of the guide vanes inside the static mixer (from 30° to 45°) to increase the Reynolds number from about 8000 to about 11000.

[0057] After the two compensation measures were implemented, the measured CO2 solubility recovered to 94% of the theoretical value, and the pH value of the carbonic acid solution decreased from 5.8 to 5.2, meeting the requirements of the mineralization reaction.

[0058] (iv) Detailed implementation of step S4 The carbonic acid solution enters the fixed-bed mineralization reactor from the bottom at a flow rate of 0.20 m / min. The mineralization reactor is a cylindrical steel structure with an inner diameter of 0.6 m and a total height of 3.0 m, resulting in a diameter-to-height ratio of 1:5.

[0059] The reactor has a water distribution plate at the bottom (35% opening rate, 5mm pore size), above which is a support layer (gravel, 5-10mm particle size, 0.2m high), and above that is a limestone filter media layer. The limestone filter media is food-grade calcite-type calcium carbonate with a CaCO3 content ≥96%, a particle size of 2-4mm, a filter media layer height of 2.0m, and an effective volume of approximately 0.57m³.

[0060] The carbonic acid solution flows upward through the limestone filter bed, with an empty bed contact time of approximately 6.0 minutes. The mineralized effluent is collected by the overflow weir at the top of the reactor before flowing out.

[0061] Three online pH sensors are installed along the water flow direction within the limestone filter bed of the fixed-bed mineralization reactor: pH1: 0.2m above the bottom of the filter media (inlet end) pH2: Middle of the filter media (1.0m from the bottom) pH3: 0.2m below the top of the filter media (outlet end) Data from the three pH sensors is transmitted to the central control system in real time at a sampling frequency of once per minute.

[0062] (V) Detailed Implementation of Step S5 The central control system receives data from three pH sensors in real time and calculates the pH gradient distribution. Under normal operating conditions, the typical pH distribution is as follows: pH1≈5.0-5.5 (bottom, when the carbonic acid solution has just entered the filter media), pH2≈6.0-6.8 (middle, when the limestone dissolution reaction is underway), and pH3≈7.2-7.8 (top, when the reaction is nearing completion). The pH difference between the bottom and top, ΔpH_total, is approximately 1.5-2.8, calculated as pH3 - pH1.

[0063] The control system executes the following diagnostic and control logic: Operating Condition 1 (Normal Operation): ΔpH_total is within the range of 1.5-2.8, and the pH difference between adjacent sensors is within the range of 0.5-0.8. The system maintains normal operation.

[0064] Condition 2 (Insufficient Filter Media Activity): When ΔpH_total < 0.3, it indicates that the pH hardly increases after the carbonated solution flows through the entire filter media layer, and the limestone dissolution reaction basically does not occur. At this time, the control system issues a warning signal: "Insufficient limestone filter media activity, it is recommended to add fresh filter media or switch to the standby reactor." In this example, this condition occurred after approximately 120 days of continuous operation. The operator switched to the standby reactor according to the warning signal and replaced the filter media in the original reactor.

[0065] Condition 3 (Water Flow Short Circuit or Local Cloggedness): When the difference between two adjacent pH sensors is > 0.8, it indicates that the carbonic acid solution is severely unevenly distributed in the filter media layer. For example, a monitoring data showed pH1=5.2, pH2=6.5 (difference 1.3>0.8), and pH3=7.0 (pH2 and pH3 difference 0.5), indicating a water flow short circuit in the lower part of the filter media layer—most of the carbonic acid solution rapidly passes through a certain local area, causing a sharp increase in pH in that area, while other areas of the filter media do not have sufficient contact. At this time, the control system automatically starts a local backwashing program: close the inlet water valve, open the backwash inlet water valve and the backwash air inlet valve, using mineralized effluent as the backwash water source with a backwash intensity of 10L / (m²·s) and compressed air as the backwash air source with a pressure of 0.3MPa and an air washing intensity of 15L / (m²·s), performing a combined air-water backwash for 10 minutes. After the backwash is completed, operation resumes, and the pH gradient distribution returns to normal.

[0066] (vi) Detailed implementation of step S6 The mineralized effluent enters a post-mineralization water tank (effective volume 10m³). An online LSI calculator is installed on the effluent pipeline of the post-mineralization water tank to automatically calculate the LSI by monitoring pH, temperature, total hardness, total alkalinity, and TDS.

[0067] The central controller reads LSI data every 10 minutes and executes the following feedback adjustment logic: When LSI < -0.50 (corrosion tendency), the controller issues the following commands: ① Increase the MFC setpoint, increasing the CO2 dosage by 8%; ② Reduce the influent pump frequency, decreasing the carbonate solution flow rate from 0.20 m / min to 0.15 m / min (extending the hydraulic residence time). After 30 minutes, the LSI is retested. If it remains below -0.50, adjustments continue until the LSI rises back to the target range (-0.30 to +0.30).

[0068] When LSI > +0.50 (prone to scaling), the controller issues the following commands: ① Reduce the MFC setpoint, decreasing the CO2 dosage by 8%; ② Increase the influent pump frequency, increasing the carbonate solution flow rate from 0.20 m / min to 0.25 m / min (shortening the hydraulic retention time). LSI is retested after 30 minutes until it returns to the target range.

[0069] In this embodiment, the LSI of the effluent after mineralization was stably controlled between -0.25 and +0.25, achieving a compliance rate of 94.7%.

[0070] Detailed Implementation of Step S7 A return pipe is led out from the outlet pipe of the mineralized water tank, pressurized by a return booster pump, and then connected to the carbonation inlet pipe at the front end of the static mixer. The return ratio is controlled at 15%-20%. The return water contains unreacted dissolved CO2 and bicarbonate ions. After mixing with fresh reverse osmosis permeate, CO2 is added, effectively improving the overall CO2 utilization rate. Calculations show that the CO2 utilization rate in this embodiment is increased from approximately 65% ​​in conventional processes to 82%. Example

[0071] This embodiment is basically the same as Embodiment 1, except that in step S1, the water quality fluctuation prediction model during the saline tide period uses the ARIMA time series model instead of the LSTM neural network, which is suitable for application scenarios with limited computing resources.

[0072] The order (p, d, q) of the ARIMA model is determined using the Akaike Information Criterion (AIC). Taking salinity prediction as an example, the optimal model selected by the AIC criterion is ARIMA(2,1,1). Independent ARIMA models are established for each water quality parameter, and the prediction results are weighted and fused to obtain a comprehensive prediction result. Although the prediction accuracy of the ARIMA model is slightly lower than that of the LSTM model (the root mean square error RMSE is about 10%-15% higher), it has lower computational cost, stronger model interpretability, and is suitable for embedded control systems. Example

[0073] This embodiment is basically the same as Embodiment 1, except that the operating parameters of the mineralization reactor in step S4 are adjusted and compensated according to the inlet water temperature.

[0074] The water temperature in the Pearl River Estuary can drop to 10-15℃ in winter and reach as high as 28-32℃ in summer. Temperature has a significant impact on the equilibrium constant and reaction rate of the calcium carbonate dissolution reaction. In this embodiment, when the influent temperature is below 18℃, the upstream flow rate of the carbonate solution is appropriately reduced (from 0.20 m / min to 0.12 m / min), and the empty bed contact time is extended from 6.0 minutes to 8.0 minutes to compensate for the decrease in reaction rate under low temperature conditions. When the influent temperature is above 28℃, the upstream flow rate is appropriately increased (from 0.20 m / min to 0.28 m / min), and the empty bed contact time is shortened from 6.0 minutes to 4.5 minutes to avoid over-mineralization.

[0075] The temperature compensation coefficient was calibrated through previous experiments: for every 5°C decrease in temperature, the empty bed contact time increases by 20%; for every 5°C increase in temperature, the empty bed contact time decreases by 15%.

[0076] The process of Embodiment 1 of the present invention was compared with two control processes under the same conditions. The test period was from October 2025 to March 2026 (a total of 150 days, covering the entire saline intrusion period). The comparison results are as follows:

[0077] Experimental data show that: First, compared with control process A (operating with fixed parameters), the LSI compliance rate of the mineralized effluent increased from 62.5% to 94.7%, an improvement of more than 50%. This fully demonstrates the significant advantages of the multi-level intelligent control strategy based on water quality prediction, CO2 dissolution monitoring, and pH gradient diagnosis in coping with water quality fluctuations during saline intrusion periods.

[0078] Second, compared to control process B (which only uses influent and effluent feedback control, without CO2 dissolution monitoring and pH gradient diagnosis), the LSI compliance rate of this invention is still 16.4 percentage points higher. This indicates that relying solely on influent and effluent feedback control has a significant lag—when water quality changes abruptly, feedback control can only begin to adjust after the effluent quality deteriorates. In contrast, the real-time CO2 dissolution monitoring of this invention can detect anomalies and compensate immediately at the carbonation solution stage, and the pH gradient diagnosis can detect anomalies inside the reactor and intervene precisely.

[0079] Third, the CO2 utilization rate of this invention reaches 82%, which is 24 percentage points higher than that of control process A and 15 percentage points higher than that of control process B. This is mainly attributed to two innovative designs of this invention: ① real-time monitoring and multi-stage compensation of CO2 dissolution improve the initial CO2 dissolution rate; ② partial mineralized effluent recirculation realizes the recycling of undissolved CO2. The CO2 saving effect produced by the synergistic effect of these two designs is far greater than the simple sum of their individual uses.

[0080] Fourth, the replacement cycle of the limestone filter media in this invention is extended to 10 months, which is 2.5 times that of the control process A. This is due to: ① precise control of CO2 dosage and monitoring of dissolution volume to avoid excessive consumption of filter media by excessive acidification; ② pH gradient diagnosis can promptly detect and deal with abnormal conditions such as local caking, preventing the local deterioration of filter media performance from spreading to the whole; ③ the upflow mode keeps the filter media layer moderately expanded, reducing the accumulation of pollutants; ④ regular air-water combined backwashing effectively restores the activity of the filter media.

[0081] Fifth, the response time of this invention to abnormal operating conditions is shortened from 30-60 minutes in control process B to 5-10 minutes. This is thanks to the pH gradient monitoring network, which can detect abnormalities (such as local pH changes caused by water flow short circuits) in real time inside the reactor, without having to wait for the effluent water quality to deteriorate before responding passively.

[0082] The above embodiments fully illustrate the feasibility and technical advantages of the present invention. Those skilled in the art can make various modifications and variations based on the principles of this invention, and all such modifications should be included within the scope of protection of this invention.

Claims

1. A smart mineralization treatment process for reverse osmosis permeate from the slightly brackish water of the Pearl River Estuary, characterized in that, Includes the following steps: S1. Collect water quality parameters of slightly saline water at the Pearl River Estuary intake and construct a water quality fluctuation prediction model during the saline tide period. The water quality parameters include salinity, temperature and total dissolved solids. S2. The slightly brackish water is pretreated and then desalinated by reverse osmosis to obtain reverse osmosis permeate; S3. The reverse osmosis permeate is pressurized to 0.3-0.6 MPa by a booster pump and then fully mixed with gaseous carbon dioxide in a pipeline static mixer to form a carbonic acid solution. An online carbon dioxide concentration sensor is installed at the outlet of the pipeline static mixer to monitor the actual amount of carbon dioxide dissolved in the water in real time. When the actual amount of dissolved carbon dioxide is lower than 90% of the theoretically calculated amount, at least one of the following compensation measures is automatically executed: ① Increase the outlet pressure of the booster pump; ② Increase the turbulence intensity in the pipeline static mixer. S4. The carbonated solution is fed from the bottom into a fixed-bed mineralization reactor filled with limestone filter media at a flow rate of 0.10-0.30 m / min. The mineralization reaction is carried out by passing through the limestone filter media layer in an upflow manner. At least three online pH sensors are installed at different heights along the water flow direction in the fixed-bed mineralization reactor to obtain the pH value at each height in real time and calculate the pH gradient distribution. S5. Determine the operating status of the fixed-bed mineralization reactor based on the pH gradient distribution obtained in step S4: when the difference between the bottom pH value and the top pH value is less than 0.3, it is determined that the limestone filter media is not active enough, and an early warning signal is issued to supplement the filter media or switch the reactor; when the pH difference between two adjacent online pH sensors is greater than 0.8, it is determined that there is a water flow short circuit or local caking in the filter media layer, and a local backwashing procedure is initiated. S6. The mineralized effluent from step S4 is sent into the mineralized water tank, and the Langerile Index (LSI) of the mineralized effluent is monitored in real time. The amount of carbon dioxide added in step S3 and / or the flow rate of the carbonated solution in step S4 are dynamically adjusted according to the measured LSI value so that the LSI of the mineralized effluent is controlled between -0.50 and +0.

50.

2. The intelligent mineralization treatment process for reverse osmosis permeate from the slightly brackish water of the Pearl River Estuary according to claim 1, characterized in that, The method for constructing the water quality fluctuation prediction model during the saltwater intrusion period described in step S1 includes: collecting daily variation data of salinity, temperature and total dissolved solids during the historical same period of the Pearl River Estuary saltwater intrusion period; establishing prediction sub-models for each parameter using a long short-term memory neural network; and obtaining a comprehensive water quality prediction result by weighted fusion of the prediction results of each sub-model, with a prediction time window of 4-24 hours.

3. The intelligent mineralization treatment process for reverse osmosis permeate from the slightly brackish water of the Pearl River Estuary according to claim 1, characterized in that, The theoretical dissolution amount in step S3 is calculated based on Henry's Law, and the calculation formula is C* = H·P, where H is the Henry's constant for carbon dioxide in water and P is the partial pressure of carbon dioxide at the outlet of the booster pump; when the actual dissolution amount is less than 85% of the theoretically calculated dissolution amount, the compensation measures ① and ② are executed simultaneously.

4. The intelligent mineralization treatment process for reverse osmosis permeate from the slightly brackish water of the Pearl River Estuary according to claim 3, characterized in that, The compensation measure ① increases the outlet pressure of the booster pump by 0.05-0.15 MPa; the compensation measure ② increases the turbulence intensity in the static mixer by adjusting the angle of the guide vanes inside the mixer to increase the Reynolds number by 20%-50%.

5. The intelligent mineralization treatment process for reverse osmosis permeate from the slightly brackish water of the Pearl River Estuary according to claim 1, characterized in that, In step S4, three online pH sensors are installed at different heights along the water flow direction inside the fixed-bed mineralization reactor, respectively located 0.2m above the bottom of the limestone filter layer, in the middle of the filter layer, and 0.2m below the top of the filter layer; the pH gradient distribution includes the pH difference between adjacent sensors and the total difference between the bottom sensor and the top sensor.

6. The intelligent mineralization treatment process for reverse osmosis permeate from the slightly brackish water of the Pearl River Estuary according to claim 5, characterized in that, The method for initiating the local backwashing procedure in step S5 includes: closing the water inlet valve of the fixed bed mineralization reactor, opening the backwash water inlet valve and the backwash air inlet valve, using the mineralized effluent as the backwash water source and compressed air as the backwash air source, and performing air-water combined backwashing on the filter media layer for a backwashing time of 5-15 minutes.

7. The intelligent mineralization treatment process for reverse osmosis permeate from the slightly brackish water of the Pearl River Estuary according to claim 1, characterized in that, The limestone filter media has a particle size of 1-5 mm, a calcium carbonate mass percentage of ≥95%, a filter media layer height of 1.2-2.5 m, and a diameter-to-height ratio of 1:3 to 1:6 for the fixed-bed mineralization reactor.

8. The intelligent mineralization treatment process for reverse osmosis permeate from the slightly brackish water of the Pearl River Estuary according to claim 1, characterized in that, The method for dynamically adjusting the carbon dioxide dosage and / or the carbonate solution feed rate based on the measured LSI value in step S6 includes: when the LSI is below -0.50, increasing the carbon dioxide dosage and decreasing the carbonate solution feed rate; when the LSI is above +0.50, decreasing the carbon dioxide dosage and increasing the carbonate solution feed rate.

9. The intelligent mineralization treatment process for reverse osmosis permeate from the slightly brackish water of the Pearl River Estuary according to claim 1, characterized in that, The pretreatment includes coagulation sedimentation, media filtration and ultrafiltration. The reverse osmosis desalination process uses a seawater desalination reverse osmosis membrane, and the recovery rate of the reverse osmosis system is 45%-60%.

10. The intelligent mineralization treatment process for reverse osmosis permeate from the slightly brackish water of the Pearl River Estuary according to claim 1, characterized in that, It also includes step S7: a portion of the effluent from the mineralized water tank is deducted, pressurized, and then returned to the front end of the static mixer in the pipeline in step S3, where it is mixed with the reverse osmosis permeate before carbon dioxide is added, with the return ratio controlled at 10%-30%.

Images