Leachate concentrate softening control method based on double closed-loop predictive control

Abstract

The embodiment of the present invention discloses a leachate concentrate softening control method based on double closed-loop predictive control, which relates to the field of environmental engineering automatic control technology. The method includes: obtaining multi-dimensional data collected in real time; constructing a dynamic mass balance equation for magnesium removal / calcium removal reaction, and based on the model predictive control algorithm, rolling optimization of NaOH / Na2CO3 dosage to achieve precise control of pH value and hardness removal rate; constructing a membrane fouling state space equation, combining a dynamic programming algorithm to optimize membrane type switching and cleaning strategy, and achieve Pareto optimal control of flux maintenance and cost minimization. The present invention solves the industry problems of "low control accuracy, high operating cost, and poor system robustness" in the softening treatment of high-hardness wastewater, and has significant technological advancement and engineering application value.

Description

Technical Field

[0001] The invention relates to the technical field of environmental engineering automatic control, in particular to a leachate concentrate softening control method based on double closed-loop predictive control. Background Art

[0002] In the softening treatment of leachate concentrate, traditional control methods have the following technical bottlenecks:

[0003] 1. Chemical softening process: relying on fixed pH value setting and empirical dosing, without establishing Mg 2+ / Ca 2+ The real-time coupling relationship between the dynamic change of concentration and the dosage of the reagent results in a reagent waste rate of 20%-30%, and the hardness removal rate drops by more than 15% when the pH fluctuation exceeds ±0.5.

[0004] 2. Ultrafiltration membrane control link: The cleaning strategy based on the pressure difference threshold lags behind the membrane fouling process, the membrane flux attenuation rate reaches 5% / day, and the switching of multiple membrane types relies on manual experience, and an intelligent decision-making mechanism with operating costs as the target has not been formed.

[0005] 3. System synergy: There is a lack of linkage control between the chemical reaction section and the membrane separation section, and membrane blockage problems often occur due to the mismatch between the solid particle size distribution after the reaction and the membrane pore size, affecting the continuous operation time of the system.

[0006] In the existing technology, invention patent application CN119087802A discloses a wastewater treatment control strategy based on fuzzy neural networks. The core technology is to use the PSO algorithm to optimize the BP neural network, build a wastewater treatment prediction model, and achieve high-precision prediction of effluent COD and suspended solids (SS) through data normalization, model training and performance evaluation, and improve prediction accuracy through intelligent algorithms.

[0007] The invention patent application has the following shortcomings: It does not address the specific chemical reactions in the softening process of leachate concentrate (such as Mg 2+ / Ca 2+ Sedimentation dynamics) modeling has strong universality but weak industry specificity.

[0008] Invention patent application CN118963154A discloses a wastewater treatment control optimization method and system based on monitoring data. The core technology is to establish a process benchmark and compensate for abnormal nodes through real-time monitoring of wastewater treatment process node data, realize dynamic feedback control, and propose an "abnormal node detection-downstream compensation" mechanism.

[0009] This invention patent application suffers from the following shortcomings: insufficient model depth, reliance on empirical benchmarks, and the absence of established chemical reaction kinetics or membrane fouling mathematical models, making it difficult to address drastic fluctuations in water quality. Furthermore, the patent application fails to correlate the coupling between the chemical softening and membrane separation stages (e.g., the effect of reaction product particle size on membrane pore blockage), making it impossible to achieve cross-stage coordinated optimization. Summary of the Invention

[0010] In view of this, an embodiment of the present invention provides a leachate concentrate softening control method based on double closed-loop predictive control to solve the industry problems of "low control accuracy - high operating cost - poor system robustness" in the softening treatment of high-hardness wastewater.

[0011] A leachate concentrate softening control method based on double closed-loop predictive control, comprising:

[0012] Step S101: Acquire multi-dimensional data collected in real time;

[0013] Step S102: constructing a dynamic mass balance equation for the magnesium / calcium removal reaction, and optimizing the NaOH / Na2CO3 dosage based on a model predictive control algorithm to achieve precise control of the pH value and hardness removal rate;

[0014] Step S103: Construct a membrane fouling state space equation, and optimize the membrane type switching and cleaning strategy by combining a dynamic programming algorithm to achieve Pareto optimal control for flux maintenance and cost minimization.

[0015] Furthermore, in step S102, the dynamic mass balance equation of the magnesium removal / calcium removal reaction is constructed as follows:

[0016] Magnesium removal reaction:

[0017] ;

[0018] in, : the rate of change of magnesium ion concentration in the reaction cell with time, k1: reaction rate constant, [OH - ]: hydroxide ion concentration in the reaction tank, V: reaction tank volume, Q: water flow rate, : Mass flow rate of magnesium ions brought in by the influent, : Mass flow rate of magnesium ions carried out by the outlet water;

[0019] Decalcification reaction:

[0020] ;

[0021] Where, k2: reaction rate constant, : Carbonate ion concentration in the reaction tank. The meanings of other items are the same as those of magnesium removal reaction.

[0022] Furthermore, the step S102 includes:

[0023] Step S1021: Based on real-time water quality data, the dynamic mass balance equation of the magnesium / calcium removal reaction is used to optimize the NaOH / Na2CO3 dosage in real time, control the pH value and calcium and magnesium hardness removal rate, and dynamically correct the reaction kinetic parameters to correct water quality fluctuations;

[0024] Step S1022: In each control cycle, a quadratic programming solution is performed to generate an optimal drug addition sequence within a future preset time period.

[0025] Furthermore, in step S1021, the formula for dynamically correcting the reaction kinetic parameters is:

[0026] ;

[0027] in, : reaction rate constant at time t, : the reaction rate constant at the previous moment, : adaptive learning rate, : The measured water hardness at the previous moment, : The model predicts the water hardness at the previous moment, y target : Target water hardness;

[0028] And / or, in step S1022, the formula for solving the quadratic programming is:

[0029] ;

[0030] in, : The hardness of the water produced at time t+k predicted at time t, : target water hardness at time t+k, Q: hardness error weight matrix, : The dosage of the reagent at time t+k calculated at time t, R: control input weight matrix, N: control time domain.

[0031] Furthermore, in step S103, the membrane fouling state space equation constructed is:

[0032] ;

[0033] in, : Membrane pore blockage rate, : Membrane pore blockage rate The rate of change over time, : blockage rate constant, : Cleaning recovery constant, u clean : cleaning control variable, h: thickness of filter cake layer on membrane surface, : The rate of change of the thickness of the filter cake layer on the membrane surface h with time, : filter cake growth coefficient, : backwash stripping coefficient, J: membrane flux, C solid : Solid concentration of mixed solution.

[0034] Furthermore, the step S103 includes:

[0035] Step S1031: estimating the membrane pore blockage rate and the thickness of the filter cake layer on the membrane surface through real-time data, and updating the membrane pollution level;

[0036] Step S1032: establishing a cost-efficiency matrix and using the Q-learning algorithm to select the optimal membrane type;

[0037] Step S1033: When the membrane pore blockage rate is greater than or equal to the first preset threshold or the filter cake layer thickness on the membrane surface is greater than or equal to the second preset threshold, chemical cleaning is triggered and the cleaning solution formula is automatically switched according to the membrane type.

[0038] Furthermore, in step S1032, the membrane type switching condition is:

[0039] ;

[0040] Among them, m: membrane type variable, A: BUFF membrane, B: ceramic membrane, C: hollow fiber membrane, : membrane module cost, : Energy consumption cost, : Cleaning cost.

[0041] Furthermore, the method further comprises:

[0042] Step S104: A dynamic programming algorithm is used to solve a multi-objective optimization problem including hardness removal rate, membrane flux, and operating cost, to achieve coordinated control of the chemical section and the membrane section.

[0043] Furthermore, in step S104, the objective function including hardness removal rate, membrane flux maintenance, and reagent cost is constructed as follows:

[0044] , ;

[0045] Where, F: objective function, : Hardness removal rate weight coefficient, : Real-time total hardness of produced water, y target : Target total hardness of produced water, : membrane flux weight coefficient, : Cumulative value of inner membrane flux in period T, : Pharmaceutical cost weight coefficient, c i: Unit price of the i-th drug, m i : dosage of the i-th agent, n: number of agent types;

[0046] And / or, in step S104, when the turbidity of the membrane pool is greater than a third preset threshold, the sedimentation time of the reaction pool is automatically extended;

[0047] And / or, in step S104, when the membrane flux is lower than 80% of the design value for a certain period of time, the dynamic mass balance equation of the magnesium / calcium removal reaction is reversely corrected and the Na2CO3 dosage is increased.

[0048] Furthermore, the method further comprises:

[0049] Step S105: Dynamically adjust control parameters based on the particle swarm optimization algorithm to adapt to water quality fluctuations and equipment aging.

[0050] The present invention has the following beneficial effects:

[0051] Through the deep integration of theoretical modeling, intelligent algorithms and engineering practice, this invention solves the industry problems of "low control accuracy - high operating costs - poor system robustness" in the softening treatment of high-hardness wastewater, and has significant technological advancement and engineering application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0053] Figure 1 Schematic diagram of the process of the leachate concentrate softening control method based on double closed-loop predictive control of the present invention;

[0054] Figure 2 This is a diagram of the dual closed-loop predictive control architecture of the present invention, which shows the coupling relationship between the inner-loop chemical reaction control and the outer-loop membrane system decision-making;

[0055] Figure 3 This is a flow chart of the multi-objective optimization algorithm of the present invention, which presents the complete logic from data acquisition to control output. DETAILED DESCRIPTION

[0056] The embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0057] It should be understood that the embodiments described are only a portion of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without creative work are within the scope of protection of the present invention.

[0058] The embodiment of the present invention provides a leachate concentrate softening control method based on double closed-loop predictive control, such as Figure 1-3 Shown, including:

[0059] Step S101 (real-time acquisition of multi-dimensional data): acquiring multi-dimensional data acquired in real time;

[0060] This step corresponds to Figure 2 The "data collection layer" and Figure 3 In the "data layer", during specific implementation, multi-dimensional data is at least partially collected through core sensors. The core sensors may include a pH meter and an online magnesium ion detector set in the magnesium removal reaction tank, an online calcium ion detector and a conductivity meter set in the calcium removal reaction tank, and a membrane pressure difference sensor and a water production turbidity meter set in the membrane system. The collection frequency can be flexibly set as needed, for example, 100ms / time.

[0061] Step S102 (inner loop model predictive control): Construct a dynamic mass balance equation for the magnesium / calcium removal reaction, and based on the model predictive control algorithm, perform rolling optimization of the NaOH / Na2CO3 dosage to achieve precise control of the pH value and hardness removal rate;

[0062] This step corresponds to Figure 2 Inner loop (chemical reaction control): through Mg 2+ / Ca 2+ The concentration dynamic balance equation is combined with the MPC (Model Predictive Control) algorithm to continuously optimize the dosage of the reagent, control the pH value at 11.0-11.5, and ensure that the calcium and magnesium removal rate is ≥90%;

[0063] Input parameters: real-time ion concentration, pH value, inlet flow rate;

[0064] Output control: NaOH / Na2CO3 metering pump frequency, agitator speed.

[0065] As an optional embodiment, in step S102, the dynamic mass balance equation of the magnesium removal / calcium removal reaction is constructed as follows:

[0066] Magnesium removal reaction (Mg 2+ Remove the main reaction):

[0067] ;

[0068] in, is the magnesium ion (Mg 2+ ) concentration change rate over time (unit: );

[0069] is the reaction rate constant, describing Mg 2+ With OH - The rate at which the reaction produces Mg(OH)2 precipitate;

[0070] is the hydroxide ion concentration in the reaction tank (unit: mol / L), which is controlled by the amount of NaOH added;

[0071] V=0.54m 3 is the volume of the reaction tank, Q=0.5m³ / h is the water inlet flow rate;

[0072] is the mass flow rate of magnesium ions brought into the influent (unit: mol / s);

[0073] is the mass flow rate of magnesium ions carried out by the outlet water (unit: mol / s).

[0074] Calcium removal reaction (Ca 2+ Remove the main reaction):

[0075] ;

[0076] in, is the reaction rate constant;

[0077] is the carbonate ion concentration in the reaction tank, which is determined by the addition of NaOH and HCO3 - It is determined jointly by hydrolysis and direct addition of Na2CO3;

[0078] The meanings of other items are the same as those of magnesium removal reaction.

[0079] This step corresponds to Figure 2 Chemical Reaction Kinetics Model.

[0080] As another optional embodiment, step S102 includes:

[0081] Step S1021 (prediction model update): Based on the real-time water quality data (Mg 2+ / Ca 2+concentration), through the chemical reaction kinetics model, i.e. the dynamic mass balance equation of the magnesium / calcium removal reaction, the NaOH / Na2CO3 dosage is optimized in real time to control the pH value and the calcium and magnesium hardness removal rate, thereby solving the hysteresis problem of the traditional control method and dynamically correcting the reaction kinetic parameters (i.e. the reaction rate constants k1 / k2 mentioned above) to correct water quality fluctuations (such as a ±10% change in the total hardness of the raw water);

[0082] This step corresponds to Figure 3 Chemical Reaction Model Updates.

[0083] Preferably, in step S1021, the formula for dynamically correcting the reaction kinetic parameters is:

[0084] ;

[0085] in, is the reaction rate constant at time t (i=1 is magnesium removal reaction, i=2 is calcium removal reaction);

[0086] is the reaction rate constant at the previous moment;

[0087] It is the adaptive learning rate, which controls the parameter correction amplitude;

[0088] The actual water hardness measured at the previous moment (such as calcium / magnesium hardness);

[0089] The model predicts the water hardness at the previous moment;

[0090] y target The target water hardness is (e.g. calcium hardness ≤ 300 mg / L).

[0091] Step S1022 (rolling optimization): In each control cycle (such as 1 minute), a quadratic programming solution is performed to generate an optimal drug addition sequence within a future preset time period (which can be flexibly set according to needs, 30 minutes in this embodiment).

[0092] This step corresponds to Figure 3 "Rolling Optimization: Quadratic Programming Solver" in [1].

[0093] Preferably, in step S1022, the formula for solving the quadratic programming is:

[0094] ;

[0095] in, The hardness of the produced water at time t+k predicted at time t;

[0096] The target water hardness at time t+k (e.g. total hardness ≤ 1800 mg / L);

[0097] Q is the hardness error weight matrix, which prioritizes ensuring that water quality meets standards;

[0098] The dosage of reagent at time t+k calculated at time t (NaOH / Na2CO3);

[0099] R is the control input weight matrix to avoid drastic fluctuations in the dosage of the reagent.

[0100] N is the control time domain, which refers to the number of time steps of future control actions that need to be determined at the current moment. For example, N=60, which means that each optimization generates 60 future control cycles (each cycle is 1 minute, for a total of 60 minutes).

[0101] The constraints can be flexibly set according to actual production conditions (such as reactor size and cost) and production needs. In the embodiment of the present invention, the constraints are NaOH dosage ≤ 15 L / h, Na2CO3 dosage ≤ 12 L / h, and pH value maintained at 11.0-11.5.

[0102] The above steps S1021-S1022 correspond to Figure 3 Inner-loop (MPC) optimization: Based on real-time water quality data, chemical reaction kinetic parameters (such as k1 / k2) are updated. Quadratic programming is used to solve the optimal chemical dosing sequence for achieving a hardness removal rate of ≥90% within the next 30 minutes, outputting control signals to the dosing pump. In specific implementation, a model predictive control algorithm is used to continuously optimize the NaOH / Na2CO3 dosage, targeting a hardness removal rate of ≥90% within the next 30 minutes. This optimization is done over a control horizon of 60 seconds and a prediction horizon of 120 seconds, addressing the lag issues inherent in traditional control.

[0103] Step S103 (intelligent decision-making of the outer ring membrane system): Construct a membrane fouling state space equation, and combine it with a dynamic programming algorithm to optimize the membrane type switching and cleaning strategy to achieve Pareto optimal control for flux maintenance and cost minimization.

[0104] This step corresponds to Figure 2 The outer loop in (membrane system control).

[0105] As an optional embodiment, in step S103, the membrane fouling state space equation constructed is as follows:

[0106] First, define the membrane fouling state variables:

[0107] Membrane pore blockage rate ( , To clean the membrane, is completely blocked);

[0108] Thickness of filter cake layer on membrane surface h (unit: ).

[0109] Then, establish the state transfer equation:

[0110] ;

[0111] in, Membrane pore blockage rate The rate of change over time, Indicates cleaning film, Indicates complete blockage;

[0112] is the clogging rate constant, describing the effect of membrane flux on clogging;

[0113] is the cleaning recovery constant, which describes the effect of cleaning on clogging relief;

[0114] u clean : Cleaning control variable (0-1), 0 means no cleaning, 1 means maximum cleaning intensity;

[0115] is the rate of change of the thickness of the filter cake layer on the membrane surface with time (unit: );

[0116] is the filter cake growth coefficient, which indicates the filter cake growth rate under unit flux and solid concentration;

[0117] is the backwash stripping coefficient, which indicates the stripping efficiency of the filter cake layer by backwashing;

[0118] J is the membrane flux (unit: L / (m²·h)), which indicates the water production per unit membrane area per unit time;

[0119] C solid is the solid concentration of the mixed liquid, that is, the concentration of solid particles such as Mg(OH)2 and CaCO3 generated by the reaction;

[0120] u clean It is a cleaning control variable with a value range of 0-1.

[0121] As another optional embodiment, step S103 includes:

[0122] Step S1031 (membrane fouling status estimation): estimate the membrane pore blockage rate and the thickness of the filter cake layer on the membrane surface through real-time data, and update the membrane fouling level (mild / moderate / severe);

[0123] This step corresponds to Figure 3"Membrane fouling state estimation" in.

[0124] Step S1032 (membrane type switching strategy): establish a cost-efficiency matrix and use the Q-learning algorithm to select the optimal membrane type;

[0125] This step corresponds to Figure 3 The cost-efficiency matrix established can be shown in Table 1:

[0126] Table 1: Membrane Type Cost-Efficiency Matrix

[0127]

[0128] In Table 1:

[0129] Design flux: BUFF membrane flux is significantly higher than the other two membranes;

[0130] Membrane life: Based on the corrosion resistance of the membrane material, the ePTFE BUFF membrane has the longest life;

[0131] Cleaning cycle: In the embodiment, "50% increase in membrane cleaning cycle" corresponds to a BUFF membrane cleaning cycle of 4.5 hours (conventionally 3 hours);

[0132] Comprehensive cost-efficiency ratio: a dimensionless indicator that combines reagent cost, energy consumption and flux. The lower the value, the higher the cost-effectiveness. The calculation formula is: Comprehensive cost-efficiency ratio = (cleaning cost + energy consumption cost) / (design flux × membrane life).

[0133] Preferably, in step S1032, the membrane type switching condition is:

[0134] ;

[0135] Among them, m: membrane type variable, A: BUFF membrane, B: ceramic membrane, C: hollow fiber membrane, : membrane module cost, : Energy consumption cost, : Cleaning cost.

[0136] The specific meanings of each item in the formula are shown in Table 2:

[0137] Table 2: Meaning of each item in the formula

[0138]

[0139] Step S1033 (cleaning cycle optimization): When the membrane pore blockage rate When the thickness h of the filter cake layer on the membrane surface is greater than or equal to the first preset threshold or the thickness h of the filter cake layer on the membrane surface is greater than or equal to the second preset threshold, chemical cleaning is triggered and the cleaning liquid formula is automatically switched according to the membrane type.

[0140] This step corresponds to Figure 3 In the "Cleaning Cycle Optimization" section, for example, when θ ≥ 0.6 or h ≥ 50 μm, chemical cleaning is triggered, and the cleaning solution formula is automatically switched based on the membrane type (e.g., 0.1% NaOH for ceramic membranes and 0.05% citric acid for BUFF membranes).

[0141] The above steps S1031-S1033 correspond to Figure 3 Outer loop optimization / dynamic programming in: Estimating membrane fouling status using real-time data ( ), select the membrane type with the lowest cost under the current pollution state from Table 1 through the Q-learning algorithm (such as BUFF membrane is preferred for high hardness conditions), and at the same time according to Threshold (such as ) to trigger the cleaning process. In specific implementation, the membrane pore blockage rate can be The state equation of the filter cake layer thickness h is used to estimate the membrane fouling state in real time; the Q-learning algorithm is used to select the optimal membrane type from the cost-efficiency matrix, and the dynamic programming is combined to determine the cleaning cycle (such as Chemical cleaning is triggered when the device is powered on); the input parameters are membrane pressure difference, produced water turbidity, and membrane flux; the output control is the membrane module switching instruction, backwash pump start and stop, and cleaning fluid formula selection.

[0142] Through the above steps S1031-S1033, the system can automatically balance the treatment effect and operating costs under complex working conditions, significantly improving the intelligent level of leachate concentrate softening treatment. Compared with the existing technology, it has the following advantages:

[0143] Membrane type switching strategy: For the first time, membrane material characteristics (flux, lifespan, cleaning cost) are quantified into a computable cost matrix, combined with Q-learning to achieve intelligent decision-making, solving the inefficiency of traditional manual switching (efficiency increased by 40%, membrane life extended by 25%).

[0144] Data support for the embodiment:

[0145] Compared with traditional control, the method of the present invention reduces reagent consumption by 20% (1.2kg / ton vs 1.5kg / ton), extends the membrane cleaning cycle by 50% (4.5h vs 3h), and reduces the comprehensive cost per ton of water by 15%-20%.

[0146] As an optional embodiment, the method further includes:

[0147] Step S104 (double closed-loop coordinated control): A dynamic programming algorithm is used to solve a multi-objective optimization problem including hardness removal rate, membrane flux, and operating cost, to achieve coordinated control of the chemical section and the membrane section.

[0148] In this step, a linkage mechanism between the reaction section and the membrane section is established. The inner loop provides the basic conditions for the outer loop, and the outer loop feedback affects the inner loop. This step corresponds to Figure 2 The collaborative control module in .

[0149] Preferably, in step S104, the objective function including hardness removal rate, membrane flux maintenance, and reagent cost is constructed as follows:

[0150] , ;

[0151] Among them, objective function F: comprehensive optimization goal, which needs to be minimized;

[0152] : Hardness removal rate weight coefficient (dimensionless), giving priority to ensuring the quality of produced water;

[0153] : Real-time total hardness of produced water (unit: mg / L);

[0154] y target : Target total hardness of produced water (design value ≤ 1800 mg / L);

[0155] The squared norm term represents the deviation between the actual hardness and the target value, ensuring that the produced water quality meets the standard;

[0156] : membrane flux weight coefficient (dimensionless), equilibrium flux stability;

[0157] : Cumulative membrane flux value within period T (unit: L / m²), reflecting the operating efficiency of the membrane system;

[0158] : Pharmaceutical cost weight coefficient (dimensionless), controlling operating costs;

[0159] c i : Unit price of the i-th reagent (unit: yuan / L), such as NaOH, Na2CO3;

[0160] m i : dosage of the i-th agent (unit: L);

[0161] n: The number of reagent types, that is, the number of reagent types that need to be optimized (such as NaOH and Na2CO3, in this case n=2).

[0162] The constraints can be flexibly set according to actual production conditions (such as reactor size and cost) and production requirements. In the embodiment of the present invention, the constraints are:

[0163] Total hardness of produced water ≤1800mg / L, turbidity ≤2NTU;

[0164] Membrane flux J ≥ 0.7 J max (J max Design flux for each membrane type);

[0165] Cleaning cycle T clean ≥2h (avoid excessive cleaning).

[0166] In this way, the dynamic programming algorithm is used to solve the optimal control sequence and realize the coordinated optimization of the chemical section and the membrane section.

[0167] Preferably, in step S104, when the turbidity of the membrane pool is greater than the third preset threshold, the sedimentation time of the reaction pool is automatically extended; when the membrane flux is lower than 80% of the design value for a certain period of time, the dynamic mass balance equation of the magnesium removal / calcium removal reaction is reversely corrected and the Na2CO3 dosage is increased.

[0168] In specific implementation, when the turbidity of the membrane pool is greater than 1.5NTU, the sedimentation time of the reaction pool will be automatically extended by 10 minutes, and the electrostatic adsorption strength of the crystallization catcher will be increased (by adjusting the voltage of the PTFE coating on the inner wall of the reactor); when the membrane flux is lower than 80% of the design value for 30 consecutive minutes, the chemical reaction model will be reversely corrected and the Na2CO3 dosage will be increased by 5%-10% to promote the formation of large-particle crystals and reduce membrane pore blockage.

[0169] The above step S104 corresponds to Figure 3 Collaborative control and self-correction in the process: Linking the reaction pool and membrane system, such as adjusting the dosage of reagents to promote large particle crystallization when the membrane flux decreases; optimizing the Q / R weight matrix of MPC and the membrane fouling model through the PSO algorithm every 72 hours coefficient to adapt to water quality fluctuations (e.g. automatic parameter correction when total hardness changes by ±10%).

[0170] As another optional embodiment, the method further includes:

[0171] Step S105 (adaptive parameter optimization): dynamically adjust control parameters based on the particle swarm optimization algorithm to adapt to water quality fluctuations and equipment aging.

[0172] This step corresponds to Figure 3 Adaptive parameter optimization: With cost per ton of water as the core objective, the PSO algorithm dynamically adjusts control parameters to adapt to water quality fluctuations (for example, when the total hardness changes by ±10%, the parameter adjustment time is less than 15 minutes), achieving the unity of "water quality compliance" and "economic operation."

[0173] The role of the PSO algorithm in the optimization mechanism is:

[0174] Optimization variables: weight matrix Q / R of inner loop MPC (adjusting the priority of hardness removal rate and reagent dosage); outer loop membrane fouling model parameters (clogging rate and cleaning recovery constant);

[0175] Iterative goal: By adjusting the above parameters, f is minimized while meeting the water quality constraints (total hardness ≤ 1800 mg / L, turbidity ≤ 2 NTU).

[0176] In specific implementation, the control parameters (such as the Q / R weight of MPC, the membrane fouling model) can be dynamically adjusted using the particle swarm optimization (PSO) algorithm based on historical operation data (≥72h). coefficient), the objective function is to minimize the comprehensive cost per ton of water, and the optimization formula is:

[0177] , ;

[0178] Among them, f is the comprehensive cost of water treatment per ton, unit: yuan / ton;

[0179] C chemical is the total cost of chemicals, including the consumption of NaOH and Na2CO3, unit: yuan;

[0180] C membrane is the membrane module loss cost, including membrane replacement and maintenance costs, unit: yuan;

[0181] C energy is the total energy consumption cost of the system, including the power consumption of water pumps, mixers and other equipment, unit: yuan;

[0182] Q treated It is the total water treatment volume, unit: tons.

[0183] Implementation case: Pilot plant softening system for leachate concentrate from a waste incineration power plant in Zhejiang

[0184] 1. Control parameter initialization

[0185] Reaction tank control target: pH = 11.5 ± 0.05, calcium hardness removal rate ≥ 91%, magnesium hardness removal rate ≥ 75%;

[0186] Membrane system parameters: BUFF membrane design flux 50L / h·m², ceramic membrane 25L / h·m², hollow fiber membrane 20L / h·m²;

[0187] Optimization algorithm parameters: MPC prediction time domain 120 minutes, control time domain 60 minutes, PSO particle number 30, and iteration number 50.

[0188] 2. Operational performance comparison

[0189] The comparison between the method of the present invention and the traditional control is shown in Table 3:

[0190] Table 3: Comparison of operating results

[0191]

[0192] In summary, the present invention relates to a dual closed-loop predictive control method for the leachate nanofiltration / reverse osmosis concentrate softening process. By constructing a chemical reaction kinetic model and a membrane fouling state-space equation, the coordinated optimization control of the calcium and magnesium hardness removal rate and the membrane system operating efficiency is achieved. It is suitable for pilot and industrial-grade treatment scenarios of high-salt and high-hardness wastewater.

[0193] Inner loop: Based on the chemical reaction kinetics model, the model predictive control (MPC) of NaOH / Na2CO3 dosage is implemented to solve the problems of dynamic hysteresis of pH value and fluctuation of hardness removal rate;

[0194] Outer loop: Constructs the membrane fouling state space equation and combines it with the dynamic programming algorithm to optimize membrane type switching and cleaning strategies to achieve Pareto optimal control for flux maintenance and cost minimization.

[0195] Dual closed-loop collaborative control mechanism: The inner and outer loops are linked through real-time data sharing (such as solid particle size distribution after reaction and membrane flux attenuation rate), dynamically adjusting the reagent addition and membrane cleaning strategies.

[0196] The present invention also has the following innovative features:

[0197] 1. Innovation in modeling dimensions: For the first time, a dual-scale model of the leachate concentrate softening process was established, analyzing the Mg(OH)2 / CaCO3 crystallization kinetics at the micro level and constructing the membrane fouling state-space equation at the macro level, breaking through the limitations of traditional control that relies on empirical formulas.

[0198] 2. Innovation in control structure: A dual-closed-loop predictive control architecture for the chemical reaction section and the membrane separation section is proposed. Through the cross-scale collaboration of MPC and dynamic programming, Pareto optimal control of hardness removal rate and membrane operation efficiency is achieved, solving the problem of multi-objective conflicts.

[0199] 3. Innovation in decision-making mechanism: The Q-learning algorithm is introduced to build a multi-membrane type intelligent decision-making system, which dynamically selects the optimal membrane type based on real-time pollution status and cost model. Compared with manual switching, the efficiency is improved by 40% and the membrane life is extended by 25%.

[0200] 4. Innovation in adaptive capabilities: A parameter self-correction mechanism based on particle swarm optimization is designed to automatically adapt to water quality fluctuations (for example, when the total hardness changes by ±10%, the control parameter adjustment time is less than 15 minutes), significantly improving the robustness of the system.

[0201] 5. Dual closed-loop collaborative control mechanism: The inner and outer loops are linked through real-time data sharing (such as solid particle size distribution after reaction and membrane flux attenuation rate), dynamically adjusting the reagent addition and membrane cleaning strategies.

[0202] Through the deep integration of theoretical modeling, intelligent algorithms and engineering practice, this invention solves the industry problems of "low control accuracy - high operating costs - poor system robustness" in the softening treatment of high-hardness wastewater, and has significant technological advancement and engineering application value.

[0203] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A leachate concentrate softening control method based on double closed-loop predictive control, characterized in that: include: Step S101: Acquire multi-dimensional data collected in real time; Step S102: constructing a dynamic mass balance equation for the magnesium / calcium removal reaction, and optimizing the NaOH / Na2CO3 dosage based on a model predictive control algorithm to achieve precise control of the pH value and hardness removal rate; Step S103: Constructing a membrane fouling state space equation, and optimizing membrane type switching and cleaning strategies in combination with a dynamic programming algorithm to achieve Pareto optimal control for flux maintenance and cost minimization; Wherein, the step S103 includes: Step S1031: estimating the membrane pore blockage rate and the thickness of the filter cake layer on the membrane surface through real-time data, and updating the membrane pollution level; Step S1032: establishing a cost-efficiency matrix and using the Q-learning algorithm to select the optimal membrane type; Step S1033: When the membrane pore blockage rate is greater than or equal to the first preset threshold or the filter cake layer thickness on the membrane surface is greater than or equal to the second preset threshold, chemical cleaning is triggered and the cleaning solution formula is automatically switched according to the membrane type.

2. The leachate concentrate softening control method based on double closed-loop predictive control according to claim 1, characterized in that: In step S102, the dynamic mass balance equation of the magnesium / calcium removal reaction is constructed as follows: Magnesium removal reaction: ； in, : the rate of change of magnesium ion concentration in the reaction cell with time, k1: reaction rate constant, [OH - ]: hydroxide ion concentration in the reaction tank, V: reaction tank volume, Q: water flow rate, : Mass flow rate of magnesium ions brought in by the influent, : Mass flow rate of magnesium ions carried out by the outlet water; Decalcification reaction: ； Where, k2: reaction rate constant, : Carbonate ion concentration in the reaction tank. The meanings of other items are the same as those of magnesium removal reaction.

3. The leachate concentrate softening control method based on double closed-loop predictive control according to claim 1, characterized in that: The step S102 includes: Step S1021: Based on real-time water quality data, the dynamic mass balance equation of the magnesium / calcium removal reaction is used to optimize the NaOH / Na2CO3 dosage in real time, control the pH value and calcium and magnesium hardness removal rate, and dynamically correct the reaction kinetic parameters to correct water quality fluctuations; Step S1022: In each control cycle, a quadratic programming solution is performed to generate an optimal drug addition sequence within a future preset time period.

4. The leachate concentrate softening control method based on double closed-loop predictive control according to claim 3, characterized in that: In step S1021, the formula for dynamically correcting the reaction kinetic parameters is: ； in, : reaction rate constant at time t, : the reaction rate constant at the previous moment, : adaptive learning rate, : The measured water hardness at the previous moment, : The model predicts the water hardness at the previous moment, : Target water hardness; And / or, in step S1022, the formula for solving the quadratic programming is: ； in, : The hardness of the water produced at time t+k predicted at time t, : target water hardness at time t+k, Q: hardness error weight matrix, : The dosage of the reagent at time t+k calculated at time t, R: control input weight matrix, N: control time domain.

5. The leachate concentrate softening control method based on double closed-loop predictive control according to claim 1, characterized in that: In step S103, the membrane fouling state space equation is constructed as follows: ； in, : Membrane pore blockage rate, : Membrane pore blockage rate The rate of change over time, : blockage rate constant, : Cleaning recovery constant, u clean : cleaning control variable, h: thickness of filter cake layer on membrane surface, : The rate of change of the thickness of the filter cake layer on the membrane surface h with time, : filter cake growth coefficient, : backwash stripping coefficient, J: membrane flux, C solid : Solid concentration of mixed solution.

6. The leachate concentrate softening control method based on double closed-loop predictive control according to claim 1, characterized in that: In step S1032, the membrane type switching condition is: ； Among them, m: membrane type variable, A: BUFF membrane, B: ceramic membrane, C: hollow fiber membrane, : membrane module cost, : Energy consumption cost, : Cleaning cost.

7. The leachate concentrate softening control method based on double closed-loop predictive control according to any one of claims 1 to 6, characterized in that: The method further comprises: Step S104: A dynamic programming algorithm is used to solve a multi-objective optimization problem including hardness removal rate, membrane flux, and operating cost, to achieve coordinated control of the chemical section and the membrane section.

8. The leachate concentrate softening control method based on double closed-loop predictive control according to claim 7, characterized in that: In step S104, the objective function including hardness removal rate, membrane flux maintenance, and reagent cost is constructed as follows: ， ； Where, F: objective function, : Hardness removal rate weight coefficient, : Real-time total hardness of produced water, y target : Target total hardness of produced water, : membrane flux weight coefficient, : Cumulative value of inner membrane flux in period T, : Pharmaceutical cost weight coefficient, c i : Unit price of the i-th drug, m i : dosage of the i-th agent, n: number of agent types; And / or, in step S104, when the turbidity of the membrane pool is greater than a third preset threshold, the sedimentation time of the reaction pool is automatically extended; And / or, in step S104, when the membrane flux is lower than 80% of the design value for a certain period of time, the dynamic mass balance equation of the magnesium / calcium removal reaction is reversely corrected and the Na2CO3 dosage is increased.

9. The leachate concentrate softening control method based on double closed-loop predictive control according to claim 7, characterized in that: The method further comprises: Step S105: Dynamically adjust control parameters based on the particle swarm optimization algorithm to adapt to water quality fluctuations and equipment aging.

Images