Electrochemical desalination system based on lye circulation reuse and automatic blowdown filtration

Abstract

The application discloses an electrochemical desalination system based on lye circulation reuse and automatic blowdown filtration, relates to the technical field of water treatment, and comprises an electrolysis module, a lye treatment module, an acid liquid reflux module and an automatic control module, wherein the automatic control module is provided with an LSTM time series prediction model, an XGBoost regression model and a reinforcement learning model. The application realizes efficient reuse of lye supernatant and filtered water through a double reflux circuit, combines the scaling function of a lye collecting tank with a dynamic reuse strategy, and achieves the effect of descaling. Through real-time data analysis and prediction, the dynamic adjustment of blowdown cycle, stirring parameters, backwashing timing and reflux ratio is realized. Through multi-objective global optimization, the energy consumption per ton of water is stabilized at 1.0-1.1 kWh, and the precise control of stirring and backwashing reduces energy consumption and water resource waste.

Description

Technical Field

[0001] This application relates to the field of water treatment technology, and in particular to an electrochemical desalination system based on alkaline solution recycling and automatic sewage filtration. Background Technology

[0002] Industrial circulating water is a critical supporting system in industrial production, and its water quality directly affects equipment operating efficiency and service life. Hard ions (Ca) in the water... 2+ Mg 2+ ) and chloride (Cl - The accumulation of scale and corrosion can easily lead to decreased heat exchange efficiency, increased maintenance costs, and even safety accidents. Electrochemical desalination technology has become the mainstream treatment technology for such scenarios because it does not require the addition of chemical agents and can simultaneously remove hardening and chlorine.

[0003] Existing electrochemical desalination systems have significant drawbacks. At the hardware level, most systems employ a simple electrolytic desalination structure, directly discharging the highly alkaline water generated in the cathode chamber, resulting in water waste. At the control level, control units often use fixed thresholds and preset timing sequences, leading to a severe one-size-fits-all problem. Summary of the Invention

[0004] This application provides an electrochemical desalination system based on alkali recycling and automatic sewage filtration to solve the problems of direct alkali discharge and reliance on fixed thresholds and preset timing for control in the prior art.

[0005] This application provides an electrochemical desalination system based on alkali recycling and automatic sewage discharge filtration, including: The electrolysis module includes an electrolytic cell with a double-membrane, three-chamber structure. The cell comprises, in sequence, an anode plate, an anode chamber, an anion exchange membrane, a raw water chamber, a cation exchange membrane, a cathode chamber, and a cathode plate. Water discharged from the industrial circulation equipment is buffered in a circulating water tank and then transported to the raw water chamber. After electrolysis by the anode and cathode plates and screening by the anion and cation exchange membranes, Mg... 2+ and Ca 2+ Ions accumulate in the cathode chamber and form an alkaline solution, Cl - Ions migrate to the anode chamber and are removed by oxidation; The alkali treatment module includes an alkali collection tank, an automatic sewage discharge device, an automatic backwash filter, and a sedimentation tank connected in sequence. The alkali collection tank is connected to the bottom of the cathode chamber and is used to collect the alkali and sediment discharged from the cathode chamber. The automatic sewage discharge device is connected to the sewage outlet of the alkali collection tank for sewage discharge control. The automatic backwash filter is connected to the output end of the automatic sewage discharge device for filtering and backwashing the sediment in the alkali. The filtered alkali is discharged into the sedimentation tank, and the supernatant of the alkali collection tank and the sedimentation tank is transported back to the cathode chamber. The acid reflux module is connected to the anode chamber and is used to discharge the acid in the anode chamber into the circulating water tank. The automatic control module is equipped with an LSTM time-series prediction model, an XGBoost regression model, and a reinforcement learning model. The LSTM time-series prediction model is used for adaptive sewage discharge and stirring control, the XGBoost regression model is used for intelligent backwashing control, and the reinforcement learning model is used for synergistic optimization of alkali reuse and acid reflux and multi-objective global optimization control.

[0006] The electrochemical desalination system based on alkali recycling and automatic backwash filtration in this application has the following advantages: 1. A dual-recirculation loop enables efficient reuse of the alkaline supernatant and filtered water, achieving a water recovery rate of over 90%. Simultaneously, the scaling function of the alkaline collection tank, combined with a dynamic reuse strategy, achieves effective descaling.

[0007] 2. Through real-time data analysis and prediction, the system can dynamically adjust the sewage discharge cycle, mixing parameters, backwashing timing, and return ratio to adapt to fluctuations in raw water quality, flow rate, and temperature, thus avoiding problems such as sedimentation and blockage, excessive backwashing, and neutralization imbalance.

[0008] 3. Through multi-objective global optimization, the energy consumption per ton of water is stabilized at 1.0-1.1 kWh, and the precise control of stirring and backwashing reduces energy consumption and water waste.

[0009] 4. Through the automated process of data acquisition, model decision-making, PLC execution, and optimization iteration, there is no need for manual parameter setting or process intervention. Only 1-2 model calibrations are required per year, which greatly reduces operating costs and management difficulty. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a schematic diagram of the composition of an electrochemical desalination system based on alkaline solution recycling and automatic sewage discharge filtration provided in the embodiments of this application.

[0012] Explanation of reference numerals: 100, raw water chamber; 200, cathode chamber; 210, cation exchange membrane; 220, cathode plate; 300, anode chamber; 310, anion exchange membrane; 320, anode plate. Detailed Implementation

[0013] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0014] Figure 1 A diagram illustrating the composition of an electrochemical desalination system based on alkali recycling and automatic backwash filtration, provided in an embodiment of this application. This embodiment of the application provides an electrochemical desalination system based on alkali recycling and automatic backwash filtration, comprising: The electrolysis module includes an electrolytic cell with a double-membrane, three-chamber structure. The cell comprises an anode plate 320, an anode chamber 300, an anion exchange membrane 310, a raw water chamber 100, a cation exchange membrane 210, a cathode chamber 200, and a cathode plate 220 arranged sequentially. Water discharged from the industrial circulation equipment is buffered in a circulating water tank and then transported to the raw water chamber 100. After electrolysis by the anode plate 320 and cathode plate 220, and screening by the anion exchange membrane 310 and cation exchange membrane 210, Mg... 2+ and Ca 2+ Ions accumulate in cathode chamber 200 and form an alkaline solution, Cl - Ions migrate to the anode chamber 300 and are removed by oxidation; The alkali treatment module includes an alkali collection tank, an automatic sewage discharge device, an automatic backwash filter, and a sedimentation tank connected in sequence. The alkali collection tank is connected to the bottom of the cathode chamber 200 and is used to collect the alkali and sediment discharged from the cathode chamber 200. The automatic sewage discharge device is connected to the sewage outlet of the alkali collection tank and is used for sewage discharge control. The automatic backwash filter is connected to the output end of the automatic sewage discharge device and is used to filter and backwash the sediment in the alkali. The filtered alkali is discharged into the sedimentation tank, and the supernatant of the alkali collection tank and the sedimentation tank is transported back to the cathode chamber 210. The acid reflux module is connected to the anode chamber and is used to discharge the acid in the anode chamber into the circulating water tank. The automatic control module is equipped with an LSTM time-series prediction model, an XGBoost regression model, and a reinforcement learning model. The LSTM time-series prediction model is used for adaptive sewage discharge and stirring control, the XGBoost regression model is used for intelligent backwashing control, and the reinforcement learning model is used for synergistic optimization of alkali reuse and acid reflux and multi-objective global optimization control.

[0015] For example, the electrolytic cell is made of PP material and has dimensions of 1500mm × 800mm × 600mm. The anode plate 320 is a Ti / IrO2-Ta2O5 coated electrode with dimensions of 1400mm × 700mm × 2mm and a mesh structure with a pore size of 5-10mm. The cathode plate 220 is made of 316L stainless steel and its dimensions match those of the anode plate 320. Both the anion exchange membrane 310 and the cation exchange membrane 210 are selected from the DuPont Nafion series, and their dimensions are slightly larger than the dimensions of the raw water chamber, leaving a 50mm allowance on each side of the exchange membrane. The distance between the two electrodes is 8mm, and the distance between the exchange membrane and the corresponding electrode is 2mm.

[0016] Furthermore, the bottom of the cathode chamber 200 is provided with an inverted conical discharge port, the inlet of the alkali collection tank is connected to the discharge port, and the bottom of the alkali collection tank is provided with a stirring unit, which is used to stir the precipitate in the alkali collection tank to a suspended state.

[0017] The alkali solution collection tank is made of PP material, with dimensions of 1200mm × 800mm × 1000mm and an effective volume of 0.8m³. 3 The bottom also features an inverted conical drain outlet with a 60° cone angle and a DN50 diameter. The mixing unit uses an alkali-resistant polyurethane impeller with a 500mm diameter. The variable frequency motor driving the impeller has a power of 0.75kW and an adjustable speed of 50-150r / min. The automatic drain device uses a DN50 stainless steel solenoid valve and a 1.5kW drain pump. The automatic backwash filter is made of stainless steel with a filtration accuracy of 3μm and a rated flow rate of 6m³ / min. 3 The backwashing pressure is 0.2 MPa, and the backwashing time is 3-5 minutes. The sedimentation tank is a PP inclined tube sedimentation tank with an inclination angle of 60° and an effective volume of 5 m³ / h. 3 The settling time is ≥1 hour. The reflux pump for delivering the supernatant is an alkali-resistant metering pump with a flow rate of 0-1 m³ / h. 3 The recirculation valve connected to the recirculation pump is a fluororubber-sealed solenoid valve with adjustable speed (h).

[0018] The acid reflux module includes an acidic water reflux pipeline and a flow regulating valve. The acidic water reflux pipeline is made of PP material with a diameter of DN40, and the flow regulating valve is made of stainless steel with an adjustment range of 0.5-1m. 3 / h.

[0019] Furthermore, a turbidity sensor and a raw water hardness monitor are installed in the raw water chamber, a level sensor and a turbidity sensor are installed in the alkali collection tank, a level sensor is installed in the sedimentation tank, pressure sensors are installed at the inlet and outlet of the automatic backwash filter, and an effluent hardness monitor is installed on the pipeline that transports the supernatant back to the cathode chamber 210. The raw water hardness monitor, level sensor, turbidity sensor, pressure sensor and effluent hardness monitor are all electrically connected to the automatic control module.

[0020] Specifically, the PLC controller in the automatic control module is a Siemens S7-200 series. The level sensors, with a range of 0-2m, are installed on the side walls of the sedimentation tank and the alkali collection tank, 1.8m from the bottom. The turbidity sensor, with a range of 0-20 NTU, is installed at the supernatant outlet of the alkali collection tank. The pressure sensor has a range of 0-1MPa. The touchscreen is a 10-inch industrial touchscreen used to display operating parameters and control status in real time. The stirring unit is electrically connected to the PLC controller, supporting automatic start / stop and speed adjustment.

[0021] Based on the above hardware structure, the working process of the electrochemical desalination system in this application will be described below.

[0022] Step 1: Data Acquisition and Preprocessing.

[0023] 1.1 Data Acquisition.

[0024] Collect the following key data at a sampling frequency of 1 time / second for 6 consecutive months to build a historical dataset: Raw water characteristic parameters: Ca 2+ Mg 2+ Concentration, Cl - Concentration, flow rate, temperature, and suspended solids concentration, with the suspended solids concentration calculated from the turbidity collected by the turbidity sensor.

[0025] Electrolysis module operating parameters: current density, pH value of each chamber, voltage of anode and cathode, and flow rate of raw water chamber.

[0026] Parameters for the alkali treatment module: liquid level and turbidity of the alkali collection tank, power of the stirring motor, inlet and outlet pressure difference of the automatic backwash filter, and liquid level of the sedimentation tank.

[0027] Actuator status parameters: start / stop status of sewage pump, on / off status of reuse valve, flow rate of return pump, and opening degree of flow regulating valve.

[0028] Treatment effect parameters: Ca in the effluent 2+ Mg 2+ Concentration, Cl - Concentration, energy consumption per ton of water, and water resource recovery rate.

[0029] 1.2 Data preprocessing.

[0030] Outlier removal: Using the 3σ criterion, extreme data caused by sensor malfunctions are removed, such as sudden changes in turbidity exceeding 20 NTU or instantaneous pressure difference reaching 0.5 MPa.

[0031] Data completion: Missing data (such as data from short-term sensor offline events) is completed using linear interpolation to ensure the continuity of the dataset; Data standardization: Parameters of different magnitudes are normalized, such as current density and flow rate, and converted into values ​​in the [0,1] interval. The formula is as follows: ,in x This is the original data. x min The minimum value of the original data. x max The maximum value of the original data. x norm This is the normalized data.

[0032] Data partitioning: The preprocessed dataset is divided into training set, validation set and test set in a ratio of 7:2:1.

[0033] After the above collection and preprocessing, a high-quality, standardized historical dataset will be obtained, containing approximately 15.768 million valid data points, covering different operating conditions such as raw water quality fluctuations and equipment operating load changes, providing reliable data support for subsequent model training.

[0034] Step 2, model training.

[0035] Based on the preprocessed dataset, three core models were trained in the edge computing gateway to adapt to different control functions. The model training environment was Python 3.9, the deep learning framework was TensorFlow 2.8, and the machine learning library was Scikit-learn 1.0.2.

[0036] 2.1 Structure and training of LSTM time series prediction model.

[0037] In the embodiments of this application, the LSTM time series prediction model sequentially includes an input layer, an embedded layer, an LSTM layer, a fully connected layer, and an output layer. The ion concentration of the raw water, the liquid level and turbidity of the alkaline solution collection tank, the pH value of the cathode chamber, and the electrolysis running time are input into the LSTM time series prediction model. The predicted precipitation accumulation rate and the time to reach the discharge threshold are output. When the time to reach the discharge threshold is within 12 hours, the stirring speed is controlled according to the precipitation accumulation rate. After stirring is completed, the automatic discharge device is turned on to transport the alkaline solution containing suspended precipitates to the automatic backwash filter.

[0038] Specifically, the LSTM has two layers: the first layer has 64 neurons, and the second layer has 32 neurons. The input features of this model include the Ca2+ of the raw water. 2+ Mg 2+ The concentration, liquid level in the alkali collection tank, turbidity, pH value of cathode chamber 220 and electrolysis running time, output precipitation accumulation rate (unit mg / h) and time required for precipitation to reach the discharge threshold (unit h), wherein the discharge threshold is set to 10% of the volume of the alkali collection tank.

[0039] The loss function uses mean squared error (MSE), and the formula is as follows: ,in Loss Represents the loss function. y i This refers to the actual accumulation time of sedimentation. The predicted value output by the model. N This represents the number of samples.

[0040] In the training parameters, the learning rate is 0.001, the number of iterations is 100, the batch size is 32, the Adam optimizer is used, and a Dropout layer is added to prevent overfitting with a dropout rate of 0.2.

[0041] The training process of the model is as follows: Initialize model parameters and input the training set into the model in batches; Calculate the loss function between the predicted and actual values, and update the model weights through backpropagation; Every 10 rounds, the model performance is evaluated using a validation set. If the validation set loss increases for 3 consecutive rounds, training is stopped and the optimal model is saved. Use a test set to verify the model's accuracy, requiring a prediction error ≤ 5%.

[0042] 2.2 Structure and training of the XGBoost regression model.

[0043] In the embodiments of this application, the pressure difference between the inlet and outlet of the automatic backwash filter, the suspended solids concentration of the raw water, and the discharge flow rate of the automatic discharge device are input into the XGBoost regression model, and the predicted trend of pressure difference change within the next 1 hour is output. If the pressure difference will reach 0.1 MPa within the next 1 hour, the backwashing procedure is started in advance. If the filter element contamination degree calculated based on the pressure difference reaches 30% or more after the discharge is completed, the backwashing process is also started.

[0044] Specifically, the model employs a gradient boosting tree structure with 100 decision trees at a depth of 6 layers and a minimum of 5 samples per leaf node. The model's input features include: the concentration of suspended solids in the raw water, the flow rate of the wastewater pump, the return flow rate of the filtered water, the current differential pressure of the automatic backwash filter, and the electrolysis runtime. The outputs are the filter cartridge fouling rate (in MPa / h) and whether the differential pressure will reach the backwash threshold within the next hour, which is set to 0.1 MPa.

[0045] The loss function uses the squared error loss.

[0046] In the training parameters, the learning rate is 0.05, and the regularization parameters are λ=0.1 and α=0.01 to prevent the model from overfitting.

[0047] The training process of the model is as follows: Initialize the weak classifier and decision tree, and calculate the initial loss; A new decision tree is constructed based on the gradient descent direction and integrated into the model; Repeat the above steps until the preset number of trees is reached, then adjust the parameters using the validation set. The test set validates the model, requiring a contamination rate prediction error of ≤3% and a backwash threshold prediction accuracy of ≥95%.

[0048] 2.3 Structure and training of reinforcement learning model (DQN).

[0049] In the embodiments of this application, the reinforcement learning model adopts a deep Q-network (DQN). The state space of the reinforcement learning model includes the ion concentration of the raw water, the pH value of the cathode chamber 200, the turbidity and liquid level of the alkaline collection tank, the pH value of the circulating water tank, and the flow rate of the raw water chamber. The action space includes the alkaline return flow rate, the acid return ratio, the current density, and the flow velocity of the raw water chamber.

[0050] Specifically, the model has the following structure: input layer → hidden layer → output layer. The input layer has 12 neurons, corresponding to the dimension of the state space. There are two hidden layers, with the first layer having 128 neurons and the second layer having 64 neurons. The activation function is ReLU. The output layer has 8 neurons, corresponding to the dimension of the discretized action space.

[0051] The formula for the reward function is: Reward = w 1× E + w 2× W + w 3× C in, E To eliminate hard efficiency, The target is ≥90%, and when the target is achieved... E=1, not reached. E = 1. W For water resource recovery rate, The target is ≥90%, and when the target is achieved... W =1, decrease by 0.1 for every 1% decrease. C To optimize energy consumption per ton of water, When energy consumption is ≤1.1kWh C =1, decrease by 0.2 for every 0.1kWh increase. Weight w 1 = 0.4 w 2 = 0.3 w 3 = 0.3, ensuring multi-objective balance.

[0052] In the training parameters, the learning rate is 0.0001, the experience replay buffer size is 10000, the target network update cycle is 100 steps, and the initial value of the exploration rate is 0.9, which gradually decays to 0.1 with iteration.

[0053] The training process of the model is as follows: Initialize the agent and the environment. The agent randomly selects an action to execute and records the state-action-reward-next state data, storing it in the experience replay buffer. Randomly sample batches of data from the experience replay buffer, with a batch size of 64, train the value network, and minimize the loss between the predicted Q-value and the target Q-value. The target network parameters are updated regularly, and the iteration is repeated for 50,000 steps. The model was validated using a test set, requiring that the hardness efficiency be stable at ≥90%, the water resource recovery rate be ≥88%, and the energy consumption per ton of water be ≤1.1kWh.

[0054] After the above processing, three models with satisfactory performance are obtained. The model files are stored on the edge computing gateway, supporting real-time access and online iteration.

[0055] Step 3, adaptive sewage discharge and stirring control.

[0056] 3.1 Triggering conditions.

[0057] When the system is running in real time, the edge computing gateway calls the LSTM time series prediction model every 5 minutes, inputting the current Ca of the raw water. 2+ Mg 2+ The concentration, alkali collection tank level / turbidity, cathode chamber 200 pH value, and electrolysis operation time are used to predict the precipitation accumulation rate and the time to reach the discharge threshold.

[0058] 3.2 Control process.

[0059] 3.2.1 If the time T for sedimentation to reach the discharge threshold is less than or equal to 12 hours, then proceed to the discharge preparation stage; if 12 hours < T < 36 hours, maintain the current state and re-evaluate every 30 minutes; if T ≥ 36 hours, re-evaluate every hour.

[0060] 3.2.2 When entering the wastewater discharge preparation stage, the gateway calculates the optimal stirring parameters based on the sedimentation accumulation rate: If the precipitation accumulation rate is ≥20mg / h: stir speed 120-150r / min, stir time 8-10min; If 10 mg / h < precipitation accumulation rate < 20 mg / h: stirring speed 80-120 r / min, stirring time 6-8 min; If the precipitation accumulation rate is ≤10mg / h: stir speed 50-80r / min, stirring time 5-6min.

[0061] The gateway sends mixing parameter commands to the PLC controller, which then drives the mixing system to start.

[0062] 3.2.3 After the mixing process is completed, the gateway sends a command to the PLC controller: close the reuse valve to cut off the alkali reuse loop, and simultaneously start the automatic sewage discharge device. Wastewater containing suspended sediment enters the automatic backwash filter through the DN50 sewage discharge pipeline. The sewage discharge time is determined by the sedimentation amount. When the sedimentation amount predicted by the model is ≥10% of the tank volume, the sewage discharge time is 30 minutes; when the sedimentation amount is 5%-10% of the tank volume, the sewage discharge time is 20 minutes; when the sedimentation amount is <5% of the tank volume, the sewage discharge time is 15 minutes.

[0063] This step achieves automated control of sedimentation prediction, stirring adaptation, and precise sewage discharge. The sewage discharge cycle can be dynamically adjusted from 12 to 36 hours, stirring energy consumption is reduced by 18%, and sewage discharge thoroughness is improved by 30%, laying the foundation for subsequent filtration and reuse.

[0064] Step 4, intelligent backwash control.

[0065] 4.1 Real-time monitoring and prediction.

[0066] During the sewage discharge process, the edge computing gateway collects the pressure difference between the inlet and outlet of the automatic backwash filter, the suspended solids concentration of the raw water, and the flow rate of the sewage pump every minute. It then inputs these data into the XGBoost regression model to predict the filter cartridge fouling rate and the pressure difference change trend within the next hour.

[0067] 4.2 Decision on the timing of backwashing.

[0068] If the model predicts that the pressure difference will reach 0.1 MPa within the next hour, the backwashing procedure will be started 30 minutes in advance.

[0069] If the filter element contamination rate calculated by the model is less than 30% after the sewage discharge is completed (contamination rate = current pressure difference / 0.1MPa × 100%), the filter will enter standby mode and backwashing will not be initiated. If the filter element is ≥30% contaminated, start the backwashing procedure immediately.

[0070] 4.3 Adaptive adjustment of backwashing parameters.

[0071] The gateway invokes a reinforcement learning model to dynamically adjust the backwashing duration and pressure, aiming for a backwashing differential pressure recovery rate ≥95% and minimal backwashing water consumption. Contamination level 30%-50%: backwashing time 2-3 minutes, backwashing pressure 0.15 MPa; Contamination level 50%-80%: backwashing time 3-4 minutes, backwashing pressure 0.2 MPa; Contamination level ≥80%: backwashing time 4-5 minutes, backwashing pressure 0.25 MPa.

[0072] The gateway sends backwashing parameter commands to the PLC controller, which then controls the automatic backwashing filter to perform combined air and water backwashing. The backwashing wastewater is discharged through the drain pipe.

[0073] 4.4 Standby and status feedback after backflushing.

[0074] After backwashing is complete, the automatic backwash filter enters standby mode. The gateway continuously monitors the differential pressure recovery. If the differential pressure recovery rate after backwashing is <95%, the backwashing program is restarted, i.e., a second backwash. If the recovery rate is ≥95%, the backwashing parameters and effects are recorded for model iterative optimization.

[0075] The backwashing process described above enables intelligent adaptation of backwashing timing and parameters, reduces the energy consumption of the automatic backwash filter by 10%, and avoids problems of excessive or incomplete backwashing, ensuring stable filtered water quality.

[0076] Step 5: Optimize the synergistic use of alkali and acid reflux.

[0077] 5.1 Optimization of alkali solution reuse.

[0078] 5.1.1 After the sewage discharge and backwashing are completed, the gateway collects the turbidity and liquid level of the supernatant in the alkaline collection tank and the pH value of the cathode chamber 200 in real time to determine the basic conditions for reuse, namely, turbidity ≤ 5 NTU and liquid level ≥ 70% of tank volume.

[0079] 5.1.2 The gateway invokes the reinforcement learning model, inputting the Ca of the raw water. 2+ Mg 2+Concentration, pH value of cathode chamber 200, turbidity of alkali collection tank, and output of optimal alkali reflux flow rate: If the pH of the cathode chamber 200 is <11: the reflux flow rate is 0.8-1.0 m³ / h. 3 / h, increases OH - Concentration accelerates scaling; If pH ≤ 12 for cathode chamber 200: reflux flow rate is 0.6-0.8 m³ / h. 3 / h, maintaining scaling efficiency; If the pH of cathode chamber 200 is >12: the reflux flow rate is 0.5-0.6 m³ / h. 3 / h, to avoid secondary scaling.

[0080] The gateway sends the return flow command to the PLC controller, which then controls the return pump to start. The supernatant from the alkali collection tank merges with the water return pipeline from the sedimentation tank through the supernatant reuse pipeline and is then connected to the cathode chamber 200.

[0081] 5.2 Optimization of acid reflux.

[0082] 5.2.1 The gateway collects the pH value of the circulating water tank, the pH value of the acidic water in the anode chamber 300, and the flow rate of the raw water chamber in real time to determine the neutralization effect.

[0083] 5.2.2 The gateway invokes the reinforcement learning model and outputs the optimal acid reflux ratio: If the pH of the circulating water tank is >8.5: the reflux ratio is 20%-25%, and the input of acidic water is increased; If the pH of the circulating water tank is 7.5 ≤ 8.5: the reflux ratio is 10%-15% to maintain neutralization balance; If the pH of the circulating water tank is <7.5: the reflux ratio should be 8%-10% to reduce the input of acidic water.

[0084] The gateway sends the return ratio command to the PLC controller, which controls the opening of the flow regulating valve. The acidic water in the anode chamber 300 is connected to the circulating water tank through the acidic water return pipeline to achieve self-neutralization.

[0085] After the above treatment, the alkali reuse and acid reflux are adapted to dynamic working conditions, the hardening removal efficiency is stable at ≥90%, the water resource recovery rate is increased to over 90%, and the neutralization cost is zero.

[0086] Step 6: Multi-objective global optimization and model iteration.

[0087] 6.1 Real-time performance evaluation.

[0088] The gateway collects the system's core performance indicators every hour: hard material removal efficiency, energy consumption per ton of water, water resource recovery rate, and filter cartridge lifespan. These are compared with preset targets, namely, hard material removal efficiency ≥90%, energy consumption per ton of water ≤1.1kWh, water resource recovery rate ≥90%, and filter cartridge lifespan ≥800h, and a performance score is calculated.

[0089] 6.2 Global parameter optimization.

[0090] If the performance score is less than 90 points (out of 100), the gateway invokes the reinforcement learning model to dynamically adjust the core parameter combination of the electrolysis module and the alkali treatment module. For example, if the raw water hardness increases (300 mg / L → 400 mg / L) and the hardness removal efficiency drops to 85%, the model output optimization parameter is a current density of 38 mA / cm². 2 +Intermediate freshwater chamber flow velocity: 0.25 m / s +Alkali reflux flow rate: 1.0 m³ / s 3 / h, improving hard removal efficiency to 92%.

[0091] For example, if the energy consumption per ton of water increases to 1.2 kWh, the model output optimization parameter is a current density of 32 mA / cm². 2 + The flow rate in the intermediate freshwater chamber is 0.18 m / s + the acid reflux ratio is 12%, reducing energy consumption to 1.05 kWh.

[0092] The gateway sends the optimized parameters to the PLC controller, enabling the coordinated adjustment of parameters across the entire system.

[0093] 6.3 Online model iteration.

[0094] The gateway automatically extracts the previous 24 hours' worth of operational data every day at midnight, incrementally trains the LSTM, XGBoost, and DQN models, updates the model weights, and improves the model's adaptability to new operating conditions. If the amount of new data reaches 20% of the original training set, a full retraining is performed to ensure continuous optimization of model performance.

[0095] After the above treatment, the overall performance of the system is continuously improved, the energy consumption per ton of water is stabilized at 1.0-1.1 kWh, the water resource recovery rate is ≥90%, the hard efficiency fluctuation is ≤2%, and the optimal operating state can be maintained without human intervention.

[0096] Step 7: System linkage operation and security protection.

[0097] 7.1 Full-process time-series linkage By integrating steps 3-6, an automated operation process is formed, which includes raw water input → electrolytic desalination → alkaline scaling → sedimentation prediction → stirring and sludge discharge → intelligent backwashing → dynamic reuse → neutralization optimization → global adjustment. Each link is seamlessly connected through communication between the edge computing gateway and the PLC controller, requiring no manual operation.

[0098] 7.2 Security protection mechanism.

[0099] The original PLC controller's alarm protection function is retained. When faults such as electrode short circuit, abnormal flow, excessively high / low liquid level, or excessive filter differential pressure occur, the PLC controller immediately issues an audible and visual alarm and cuts off the power supply to the electrolysis module. At the same time, the fault information is uploaded to the gateway, which records the fault data for subsequent analysis.

[0100] By adopting the above treatment, the system's operational stability is improved to over 99%, downtime due to failure is reduced by 60%, and fully unattended intelligent operation is achieved.

[0101] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0102] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. An electrochemical desalination system based on alkaline solution recycling and automatic sewage discharge filtration, characterized in that, include: An electrolysis module includes an electrolytic cell with a double-membrane, three-chamber structure. The electrolytic cell comprises, in sequence, an anode plate (320), an anode chamber (300), an anion exchange membrane (310), a raw water chamber (100), a cation exchange membrane (210), a cathode chamber (200), and a cathode plate (220). Water discharged from the industrial circulation equipment is buffered in a circulating water tank and then transported to the raw water chamber (100). After electrolysis by the anode plate (320) and the cathode plate (220), and screening by the anion exchange membrane (310) and the cation exchange membrane (210), Mg... 2+ and Ca 2+ Ions accumulate in the cathode chamber (200) and form an alkaline solution, Cl - Ions migrate to the anode chamber (300) and are removed by oxidation; The alkali treatment module includes an alkali collection tank, an automatic sewage discharge device, an automatic backwash filter, and a sedimentation tank connected in sequence. The alkali collection tank is connected to the bottom of the cathode chamber (200) and is used to collect the alkali and precipitate discharged from the cathode chamber (200). The automatic sewage discharge device is connected to the sewage outlet of the alkali collection tank and is used for sewage discharge control. The automatic backwash filter is connected to the output end of the automatic sewage discharge device and is used to filter and backwash the precipitate in the alkali. The filtered alkali is discharged into the sedimentation tank, and the supernatant of the alkali collection tank and the sedimentation tank is transported back to the cathode chamber (210). An acid reflux module, connected to the anode chamber, is used to discharge the acid in the anode chamber into the circulating water tank; The automatic control module is equipped with an LSTM time-series prediction model, an XGBoost regression model, and a reinforcement learning model. The LSTM time-series prediction model is used for adaptive sewage discharge and stirring control, the XGBoost regression model is used for intelligent backwashing control, and the reinforcement learning model is used for synergistic optimization of alkali reuse and acid reflux and multi-objective global optimization control.

2. The electrochemical desalination system based on alkali solution recycling and automatic sewage discharge filtration according to claim 1, characterized in that, The anode plate (320) is a Ti / IrO2-Ta2O5 coated electrode, and the cathode plate (220) is made of 316L stainless steel.

3. The electrochemical desalination system based on alkali solution recycling and automatic sewage discharge filtration according to claim 1, characterized in that, The bottom of the cathode chamber (200) is provided with an inverted conical discharge port, the inlet of the alkali collection tank is connected to the discharge port, and the bottom of the alkali collection tank is provided with a stirring unit, which is used to stir the precipitate in the alkali collection tank to a suspended state.

4. The electrochemical desalination system based on alkali solution recycling and automatic sewage discharge filtration according to claim 1, characterized in that, The raw water chamber is equipped with a turbidity sensor and a raw water hardness monitor. The alkaline solution collection tank is equipped with a level sensor and a turbidity sensor. The sedimentation tank is equipped with a level sensor. The inlet and outlet ends of the automatic backwash filter are respectively equipped with pressure sensors. The pipeline that transports the supernatant back to the cathode chamber (210) is equipped with an effluent hardness monitor. The raw water hardness monitor, the level sensor, the turbidity sensor, the pressure sensor, and the effluent hardness monitor are all electrically connected to the automatic control module.

5. The electrochemical desalination system based on alkali solution recycling and automatic sewage discharge filtration according to claim 1, characterized in that, The acid reflux module includes an acidic water reflux pipeline and a flow regulating valve.

6. The electrochemical desalination system based on alkali recycling and automatic sewage discharge filtration according to claim 1, characterized in that, The LSTM time-series prediction model sequentially includes an input layer, an embedded layer, an LSTM layer, a fully connected layer, and an output layer. The ion concentration of the raw water, the liquid level and turbidity of the alkaline solution collection tank, the pH value of the cathode chamber, and the electrolysis running time are input into the LSTM time-series prediction model. The model outputs the predicted precipitation accumulation rate and the time to reach the discharge threshold. When the time to reach the discharge threshold is within 12 hours, the stirring speed is controlled according to the precipitation accumulation rate. After stirring, the automatic discharge device is turned on to transport the alkaline solution containing suspended precipitates to the automatic backwash filter.

7. The electrochemical desalination system based on alkali solution recycling and automatic sewage discharge filtration according to claim 1, characterized in that, The pressure difference between the inlet and outlet of the automatic backwash filter, the suspended solids concentration of the raw water, and the discharge flow rate of the automatic discharge device are input into the XGBoost regression model, which outputs the predicted trend of pressure difference change within the next 1 hour. If the pressure difference will reach 0.1 MPa within the next 1 hour, the backwashing procedure is started in advance. If the filter element contamination degree calculated based on the pressure difference reaches 30% or more after the discharge is completed, the backwashing process is also started.

8. The electrochemical desalination system based on alkali recycling and automatic sewage discharge filtration according to claim 1, characterized in that, The state space of the reinforcement learning model includes the ion concentration of the raw water, the pH value of the cathode chamber (200), the turbidity and liquid level of the alkali collection tank, the pH value of the circulating water tank, and the flow rate of the raw water chamber. The action space includes the alkali reflux flow rate, the acid reflux ratio, the current density, and the flow velocity of the raw water chamber.

Images