A control method and device of an ultrapure water preparation system, a medium and an electronic device

Abstract

The present application relates to the technical field of water treatment control, and discloses a control method, device, medium and electronic equipment of an ultrapure water preparation system, wherein the control method comprises: acquiring real-time operation parameters and membrane model parameters of each pressure container; inverting the equivalent water permeability coefficient decay rate of each membrane element based on a one-dimensional fluid mechanics discrete node method; performing online recursive estimation of the mass transfer coefficient of each membrane element by using an extended Kalman filter; calculating the internal decay heterogeneity index and refined performance matrix of each container, and performing comprehensive performance scoring on each container; solving the optimal water production path configuration of each time period, generating a water production path configuration control instruction; performing water quality safety verification on the optimal water production path configuration; and updating the water production path configuration control instruction based on rolling time domain re-optimization.

Description

Technical Field

[0001] This invention relates to the field of water treatment control technology, specifically to a control method, device, medium, and electronic equipment for an ultrapure water preparation system. Background Technology

[0002] Ultrapure water preparation systems are widely used in industrial settings with stringent water quality requirements, such as semiconductors and pharmaceuticals. These systems typically employ multiple pressure vessels operating in parallel, with multiple reverse osmosis membrane elements installed in series within each vessel. Operating data is collected via pressure and flow sensors at both ends of each vessel, and conductivity meters are installed at the product water outlet of each vessel to monitor the product water quality in real time.

[0003] Existing control methods assess the overall performance degradation of membrane modules on a per-container basis, obtaining average degradation indices at the container level, and then allocating permeate load accordingly. However, multiple membrane elements connected in series within the same pressure vessel exhibit differentiated degradation characteristics due to their different locations: front-end membrane elements are prone to colloidal clogging, while rear-end membrane elements are prone to inorganic scaling due to their high concentrate concentration ratio. When the average degradation rates of two containers are similar, the container-level average index cannot distinguish between uniform, mild degradation of each element and the internal structural differences between severe front-end degradation and intact rear-end elements. This leads to permeate loads being allocated to containers with severe localized degradation, resulting in actual permeate quality and energy efficiency falling short of expectations. Furthermore, existing control methods lack a comprehensive container performance scoring method that incorporates water quality testing data, failing to quantify the performance differences between containers and thus enabling differentiated load allocation with boundary constraints. If permeate is allocated equally based solely on average performance indices, the potential of higher-performing containers is not rationally utilized; if allocation is unrestricted based on performance differences, high-load containers experience accelerated degradation due to factors such as increased concentration polarization and excessive transmembrane pressure differentials, worsening the overall operating condition. Summary of the Invention

[0004] This invention provides a control method, device, medium, and electronic equipment for an ultrapure water preparation system, solving the technical problems in related technologies where ultrapure water preparation systems cannot quantify the performance degradation status of each pressure vessel at the component level and are difficult to achieve differentiated water production path optimization configuration that balances energy efficiency and water quality safety.

[0005] This invention discloses a control method for an ultrapure water preparation system, comprising: Obtain the real-time operating parameters of each pressure vessel, the measured conductivity value of the product water at the product water outlet of each vessel, and the membrane model parameters of the membrane elements connected in series in each vessel; Based on the one-dimensional fluid dynamics discrete node method, the internal flow path of each container is divided into multiple calculation segments corresponding to the number of membrane elements. The local water production flux of each segment is iteratively solved using the pressure at both ends of the container as the boundary condition, and the equivalent permeability coefficient decay rate of each membrane element is obtained. The equivalent permeability coefficient decay rate time series of each membrane element is recursively estimated online using extended Kalman filtering to obtain the mass transfer coefficient estimate, estimation uncertainty, and decay rate of each element. Based on the mass transfer coefficient estimate and decay rate of each element, the internal decay heterogeneity index, refined energy efficiency ratio matrix, and refined removal rate matrix of each container are calculated. Combined with the measured permeate conductivity value, a comprehensive performance score is given to each container. The refined removal rate is the removal rate of the membrane element with the lowest removal rate in the container. Based on the comprehensive performance score, a differentiated load allocation with an upper limit constraint is performed above the equal baseline load. The optimal permeate path configuration for each time period is solved using a dynamic programming algorithm, and a permeate path configuration control command is generated and issued for execution.

[0006] Furthermore, the iterative solution for the local permeate flux of each segment includes: using the container inlet pressure and outlet concentrate pressure as boundary conditions, and the influent flow rate and total permeate flow rate as constraint conditions, iteratively solving for the local transmembrane pressure difference and local permeate flux of each segment based on the effective membrane area and channel geometry parameters of each segment; during the iterative solution process, the flow rate conservation relationship is satisfied between adjacent segments: the concentrate flow rate of the previous segment is equal to the influent flow rate of the next segment, and the effluent concentration of the previous segment is concentrated and used as the influent concentration of the next segment; iterating segment by segment from the first segment to the last segment, so that the calculation results of each segment simultaneously satisfy the boundary conditions and constraint conditions.

[0007] Furthermore, the equivalent permeability coefficient decay rate is the ratio of the difference between the initial permeability coefficient and the current equivalent permeability coefficient to the initial permeability coefficient; wherein the current equivalent permeability coefficient is obtained by dividing the local permeate flux by the local net driving pressure difference of the corresponding segment, and the local net driving pressure difference is the difference between the local transmembrane pressure difference and the local osmotic pressure difference of the corresponding segment.

[0008] Furthermore, the hidden state variables of the extended Kalman filter are the water flux attenuation coefficient and salt permeability increment of each element, and the observed variables are the local product water flux and local product water conductivity estimates of the corresponding element; wherein the local product water conductivity estimates are jointly determined by the salt permeability, local influent concentration and local product water flux of the corresponding segment, and the local influent concentration is obtained recursively from the material balance relationship of the previous segment.

[0009] Furthermore, the internal attenuation heterogeneity index includes the standard deviation of the attenuation rate of each component within the container and the ratio of the maximum attenuation rate to the average attenuation rate within the container; wherein the standard deviation reflects the dispersion of the attenuation rate of each component within the container, and the ratio of the maximum attenuation rate to the average attenuation rate reflects the degree of deviation of the most severe local attenuation from the average level.

[0010] Furthermore, the comprehensive performance score is obtained by weighted summation of the following four indicators after normalization: a score component based on the decay rate of the average mass transfer coefficient at the component level, a score component based on the internal decay heterogeneity index, a score component based on the deviation between the measured product water conductivity value and the conductivity threshold corresponding to the ultrapure water quality standard, and a score component based on the refined energy efficiency ratio; wherein each score component is mapped to the interval of zero to one by the mean normalization method based on the range before participating in the weighted summation.

[0011] Furthermore, the optimization objective of the dynamic programming algorithm is to minimize the sum of the total energy consumption for all time periods within a preset time window and the weighted cumulative loss of the weakest membrane element in each container; wherein the weakest membrane element is the membrane element with the largest decay rate in the container, and the increment of the weighted cumulative loss is obtained by multiplying the decay rate of the weakest membrane element by the load of the corresponding time period; in the transition constraint condition of the dynamic programming algorithm, a decreasing load upper limit constraint is applied to containers whose internal heterogeneity index exceeds a preset threshold.

[0012] Furthermore, in the differentiated load allocation with upper limit constraints, the adjustment range of the water production load of any container shall not exceed the preset maximum offset ratio; the differentiated load allocation scheme shall be updated once with each periodic performance evaluation cycle, and the current allocation scheme shall remain unchanged between two evaluation cycles.

[0013] Furthermore, it also includes: performing water quality safety verification on the optimal permeate path configuration for each time period: cascading and multiplying the refined removal rates corresponding to the container combinations put into operation in each time period to obtain the expected terminal effluent water quality index values, calculating the lower bound of the confidence interval of the water quality index in combination with the estimation uncertainty of each membrane element, and comparing the lower bound of the confidence interval with the ultrapure water quality requirement threshold; if the lower bound of the confidence interval meets the water quality requirement threshold, the verification is passed; if it does not meet the threshold, the safety margin constraint is increased and the permeate path configuration is resolved; one-dimensional hydrodynamic discrete node inversion and element-level extended Kalman filter update are re-executed at a preset period; when the heterogeneity index inside any container suddenly increases or the decay rate of any membrane element deviates from the predicted value beyond the allowable range, rolling time-domain re-optimization is triggered, the refined energy efficiency ratio matrix and refined removal rate matrix are recalculated with the updated element-level mass transfer coefficient estimate, the permeate path configuration for the remaining time periods is resolved, and the corrected permeate path configuration control command is generated and issued for execution.

[0014] This invention discloses a control system for an ultrapure water preparation system, used to execute the control method of the aforementioned ultrapure water preparation system, comprising: The data acquisition module is used to acquire the real-time operating parameters of each pressure vessel, the measured conductivity value of the product water at the product water outlet of each vessel, and the membrane model parameters of the membrane elements connected in series in each vessel. The module for element-level attenuation inversion divides the internal flow path of each container into multiple computational segments corresponding to the number of membrane elements based on a one-dimensional fluid dynamics discrete node method. It iteratively solves for the local permeate flux of each segment using the pressure at both ends of the container as boundary conditions, obtaining the attenuation rate of the equivalent permeability coefficient for each membrane element. The module for recursive estimation of the mass transfer coefficient uses extended Kalman filtering to perform online recursive estimation of the mass transfer coefficient at the element level, obtaining the estimated mass transfer coefficient value, estimation uncertainty, and attenuation rate for each element. The module for comprehensive scoring is used to... The estimated mass transfer coefficient and decay rate of each element are used to calculate the internal decay heterogeneity index, refined energy efficiency ratio matrix, and refined removal rate matrix of each container. Combined with the measured permeate conductivity value, a comprehensive performance score is given for each container. The refined removal rate is the removal rate of the membrane element with the lowest removal rate in the container. The path optimization and instruction generation module is used to perform differentiated load allocation with upper limit constraints on the equal baseline load based on the comprehensive performance score, solve the optimal permeate path configuration for each time period using a dynamic programming algorithm, generate permeate path configuration control instructions, and issue them for execution.

[0015] An electronic device for an ultrapure water preparation system includes a memory and a processor. The memory stores programs or instructions that can run on the processor, and when the programs or instructions are executed by the processor, they implement the steps of the control method described above.

[0016] A readable storage medium storing a program or instructions that, when executed by a processor, implement the steps of the control method described above.

[0017] The present invention has the following beneficial effects.

[0018] By using one-dimensional fluid dynamics discrete node inversion, the differentiated decay state of each membrane element inside the container can be reconstructed using only the sensor data at both ends of the container. This refines the monitoring granularity of membrane performance decay from the container level to the element level, enabling subsequent load distribution strategies to distinguish between two scenarios: uniform internal decay and severe local decay. This solves the problem of the container-level average decay rate masking the internal heterogeneous distribution.

[0019] By using the removal rate of the weakest link instead of the average removal rate of the containers to calculate a refined removal rate matrix, the problem of average values ​​masking local defects and causing the production water task to be incorrectly assigned to containers with severe local degradation, resulting in actual effluent quality lower than expected, is avoided. By incorporating the weighted cumulative loss of the weakest membrane element into the dynamic programming optimization objective, the load allocation strategy proactively avoids containers with an accelerating trend of deterioration, delaying premature replacement of the entire container due to local element failure.

[0020] By incorporating both measured permeate conductivity values ​​and component-level physical model inversion indices into the comprehensive performance score, the scoring results take into account both model estimation and actual measurement dimensions, compensating for scoring biases that may arise from model errors if relying solely on model inversion. Based on the comprehensive performance score, differentiated load allocation with upper limits is implemented. This upper limit on load adjustment prevents high-load vessels from accelerating their degradation due to increased concentration polarization and excessive transmembrane pressure differential, achieving a balance between performance priority and safe operation.

[0021] By using confidence interval water quality verification based on estimation uncertainty, a safety margin is provided for product water quality under conditions of limited accuracy in mass transfer coefficient estimation. Through rolling time-domain reoptimization, the system responds promptly to sudden state changes during the continuous degradation of membrane element performance, ensuring operational stability. Attached Figure Description

[0022] Figure 1 This is a flowchart of the control method for the ultrapure water preparation system provided in the embodiments of the present invention; Figure 2 This is a schematic diagram of the attenuation rate distribution of each membrane element in three pressure vessels provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the inversion results of discrete nodes of each membrane element in container 2 provided in an embodiment of the present invention; Figure 4 This is a schematic diagram comparing the internal attenuation heterogeneity indices of the three containers provided in this embodiment of the invention; Figure 5 This is a schematic diagram comparing the comprehensive performance score components of each container provided in an embodiment of the present invention; Figure 6 This is a schematic diagram showing the estimated mass transfer coefficient and decay rate of each key membrane element in the container provided in the embodiments of the present invention; Figure 7 This is a schematic diagram of a typical high-demand period water production load allocation scheme provided in an embodiment of the present invention; Figure 8 This is a schematic diagram comparing the decay rate of container 2 and element 4 before and after rolling re-optimization provided in an embodiment of the present invention; Figure 9 This is a schematic diagram comparing the measured conductivity of the produced water from three containers with water quality thresholds provided in this embodiment of the invention. Detailed Implementation

[0023] In an ultrapure water production system, multiple pressure vessels operate in parallel, with multiple reverse osmosis membrane elements connected in series within each vessel. The system collects operational data using pressure and flow sensors at both ends of each vessel, and a conductivity meter is installed at the product water outlet of each vessel to monitor the product water quality in real time. Existing control methods assess the overall performance degradation of the membrane modules on a vessel-by-vessel basis, obtaining an average degradation index at the vessel level. However, multiple membrane elements connected in series within the same pressure vessel exhibit differentiated degradation characteristics due to their different locations: front-end membrane elements, directly in contact with high-turbidity feed water, are prone to colloidal clogging; rear-end membrane elements, due to their high concentrate concentration ratio, are prone to inorganic scaling. When the average degradation rates of two vessels are similar, the vessel-level average index cannot distinguish between "uniform, slight degradation of all elements" and "severe degradation at the front end combined with intact rear end," leading to the product water production task being assigned to the vessel with severe localized degradation, resulting in actual product water quality and energy efficiency falling below expectations. Furthermore, existing control methods lack a comprehensive evaluation method for container performance that incorporates water quality testing data. This makes it impossible to quantify the performance differences between containers and to allocate loads with boundary constraints. If permeate water is allocated equally based on average performance indicators, the potential of containers with better performance will not be rationally utilized. If the allocation is tilted without restriction based on performance differences, high-load containers will accelerate their degradation due to practical limitations such as membrane pressure resistance, concentration polarization, and membrane fouling accumulation, which will worsen the overall operating condition.

[0024] To address the aforementioned issues, this embodiment provides a control method for an ultrapure water preparation system. The control method refines the monitoring granularity of membrane performance degradation from the container level to the element level through one-dimensional hydrodynamic discrete node inversion. Based on element-level performance parameters, it calculates a refined energy efficiency ratio matrix and a refined removal rate matrix, and combines this with the container's comprehensive performance score to perform differentiated load allocation with upper limit constraints, generating production water path configuration control commands.

[0025] A control method for an ultrapure water preparation system according to an embodiment of the present invention, such as... Figure 1 As shown, it includes the following steps: Step 1: Obtain the real-time operating parameters and membrane model parameters of each pressure vessel.

[0026] Real-time operating parameters of each pressure vessel in the ultrapure water preparation system are acquired, including time-series data of inlet pressure, outlet permeate pressure, outlet concentrate pressure, feed water flow rate, permeate flow rate, feed water temperature, and feed water conductivity. Real-time permeate conductivity values ​​from conductivity meters at the permeate outlet of each vessel are simultaneously collected as a direct characterization indicator of permeate quality. Membrane model parameters of the series-connected membrane elements within each vessel are obtained, including initial permeability coefficient, effective membrane area, and flow channel geometry parameters.

[0027] It should be noted that conductivity is used as a unified water quality indicator for both the influent and effluent. The conductivity of the influent and the conductivity of the product water together form the basis for calculating the transmembrane desalination rate. Using a unified conductivity index can avoid evaluation biases caused by dimensional conversion errors and differences in index response characteristics when different water quality indicators are used for the influent and effluent.

[0028] Step 2: Invert the equivalent permeability coefficient attenuation rate of each membrane element based on the one-dimensional fluid dynamics discrete node method.

[0029] Based on the flow channel geometry parameters and arrangement order of the membrane elements connected in series in each container, the internal flow path of each container is divided into multiple calculation segments corresponding to the number of membrane elements. The local permeate flux distribution of each membrane element is calculated, and the equivalent permeability coefficient decay rate of each element is obtained.

[0030] Step 2 may specifically include: Step 201: According to the series arrangement of membrane elements in each container, the internal flow path of the container is discretized into multiple calculation segments corresponding to the number of membrane elements, with the inlet and outlet ends of each membrane element as nodes. Each calculation segment corresponds to one membrane element.

[0031] Step 202: Using the container inlet pressure and outlet concentrate pressure as boundary conditions, and the influent flow rate and total product flow rate as constraints, the local transmembrane pressure difference and local product flux of each segment are iteratively solved based on the effective membrane area and channel geometry parameters of each segment to obtain the local product flux distribution of each membrane element. During the iterative solution process, the flow rate conservation relationship is satisfied between adjacent segments: the concentrate flow rate of the previous segment is equal to the influent flow rate of the next segment, and the effluent concentration of the previous segment is concentrated and used as the influent concentration of the next segment. The iteration proceeds segment by segment from the first segment to the last segment, ensuring that the calculation results of each segment simultaneously satisfy the boundary conditions and constraints.

[0032] Step 203: Compare the local permeate flux distribution of each membrane element with the initial permeability coefficient of each element, calculate the equivalent permeability coefficient of each element, and obtain the attenuation rate of the equivalent permeability coefficient of each element.

[0033] It should be noted that the first Equivalent permeability coefficient attenuation rate of membrane element Defined as:

[0034] in, This is the serial number of the membrane element; For the first Initial permeability coefficient of the membrane element; The current equivalent permeability coefficient is obtained by inversion from the local water production flux; both have the same unit. Local water production flux Divide by the local net driving pressure difference of the corresponding segment Calculated, i.e. ,in For the first Local permeate flux of membrane elements For the first The difference between the local transmembrane pressure difference and the local osmotic pressure difference in a segment. The one-dimensional fluid dynamics discrete node inversion method only uses existing sensor data at both ends of the container, and can reconstruct the differentiated decay state of each membrane element without installing additional sensors inside the container.

[0035] Step 3: Use extended Kalman filtering to perform online recursive estimation of the mass transfer coefficient of each membrane element.

[0036] For the time series of the equivalent permeability decay rate of each membrane element in each container, an extended Kalman filter is used to recursively estimate the element-level mass transfer coefficient online, obtaining the current mass transfer coefficient estimate, estimation uncertainty, and decay rate for each membrane element in each container. The input of the extended Kalman filter is the time series of the equivalent permeability decay rate of each element, the hidden state variables are the water flux decay coefficient and salt permeability increment of each element, the observed variables are the estimated local product water flux and local product water conductivity of the corresponding element, and the output is the estimated mass transfer coefficient, estimation uncertainty, and decay rate of each element.

[0037] It should be noted that the estimated local permeate conductivity is calculated based on the concentrate concentration of each segment and the salt permeability of the corresponding element. Specifically, the first... Estimated local water conductivity of the segment Based on the salt permeability of this segment Local influent concentration and local water production flux Jointly determined:

[0038] in, Segment number; It is the linear conversion factor between conductivity and concentration, which is obtained by calibrating the ratio of influent conductivity to influent concentration; For the first The salt permeability of the segment corresponding to the membrane element, and The units are the same, both being units of volumetric flux; For the first Local permeate flux of membrane elements; For the first The local influent concentration of a segment is recursively obtained from the material balance relationship of the previous segment, and the unit is concentration. Since the local permeate conductivity of each component cannot be directly measured inside the container, the estimated local permeate conductivity is... The conductivity values ​​obtained by calculating the material balance relationships of each calculation segment in step 2 are independent of the measured conductivity values ​​of the produced water from the conductivity detector at the outlet of the container. They are used for component-level recursive estimation and container-level comprehensive scoring, respectively.

[0039] Step 4: Calculate the internal attenuation heterogeneity index and refined performance matrix of each container, and give a comprehensive performance score for each container.

[0040] Step 4 may specifically include: Step 401: Based on the estimated mass transfer coefficient and decay rate of each component, calculate the internal decay heterogeneity index of each container. The internal decay heterogeneity index includes the standard deviation of the decay rate of each component within the container, and the ratio of the maximum decay rate to the average decay rate within the container.

[0041] Specifically, for those containing The standard deviation of the attenuation rate of each element in the container of the membrane element. The ratio of maximum attenuation rate to average attenuation rate They are defined as follows:

[0042]

[0043] in, This represents the total number of membrane elements connected in series within the container; This is the serial number of the membrane element, with a value range of [value range missing]. to ; For the first The attenuation rate of the equivalent permeability coefficient of the membrane element; This is the arithmetic mean of the attenuation rates of all components within the container. It reflects the degree of dispersion in the attenuation rate of each component within the container. Both reflect the deviation of the most severe local attenuation from the average level and are dimensionless. It should be noted that when... At this time, none of the components experienced attenuation. Defined as 1.

[0044] Step 402: Substitute the predicted values ​​of the component-level mass transfer coefficients of each container into the material balance equations and energy consumption equations of each calculation segment to calculate the refined energy efficiency ratio matrix and refined removal rate matrix of each container under different water production loads.

[0045] It should be noted that the refined removal rate is taken as the removal rate of the weakest link in the cascade of elements within the container; that is, the removal rate of the membrane element with the lowest removal rate in the container is used as the refined removal rate of that container. Using the removal rate of the weakest link instead of the average removal rate of the container can prevent the average value from masking local defects and causing the water production task to be assigned to containers with severe local degradation.

[0046] Step 403: Perform a comprehensive performance score for each container at a preset period. Before participating in the weighted summation, [the following steps are performed]. , , , and Deviation and refined energy efficiency ratio Each original index was mapped to a mean normalization method based on the range. The interval eliminates the dimensional differences between different indicators. The normalized overall performance score... It is obtained by weighted summation of the following four indicators:

[0047] in, , , , The weighting coefficients for each indicator satisfy the following conditions: ; Attenuation rate based on component-level average mass transfer coefficient The rating components, The smaller The higher; Based on internal decay heterogeneity index and The rating components, and The smaller The higher; Based on the measured conductivity value of the produced water Compared with standard threshold The scoring components of the deviation, among which The measured conductivity value of the product water at the product water outlet of the container is given by a conductivity meter. The conductivity threshold corresponding to the ultrapure water quality standard. The closer to or below but The higher; Based on refined energy efficiency ratio The rating components, The refined energy efficiency ratio calculated in step 402. The higher the The higher the score, the more constrained the differential allocation of permeable water load ratios in subsequent steps.

[0048] It should be noted that the comprehensive performance score combines the inversion indicators from the component-level physical model with the direct measurement indicators from the conductivity meter, forming a multi-dimensional scoring dimension. Among them, the measured conductivity value of the produced water, as a water quality characterization indicator independent of the physical model inversion, can compensate for the errors caused by parameter simplification or boundary condition approximation in the physical model, making the performance evaluation results closer to the actual produced water quality.

[0049] Step 5: Solve for the optimal water production path configuration for each time period and generate water production path configuration control instructions.

[0050] Obtain the time-period water demand forecast sequence within a preset future time window. Based on the time-period water demand forecast sequence, refined energy efficiency ratio matrix, refined removal rate matrix, and comprehensive performance score of each container, use a dynamic programming algorithm to solve for the optimal water production path configuration for each time period. The input of the dynamic programming algorithm is the water demand forecast for each time period, the refined energy efficiency ratio and refined removal rate of each container under different loads, and the output is the water production load allocation scheme for each container in each time period. The optimization objective is to minimize the total energy consumption and the weighted cumulative loss of the weakest membrane element in each container within the preset time window.

[0051] Specifically, optimize the objective function Defined as:

[0052] in, This represents the total number of time periods within the preset time window. This is a time-period index, with a value range of [value range missing]. to ; This represents the total number of pressure vessels operating in parallel. For container index, the value range is to ; For the first Total energy consumption of all containers during the time period; For the first Time period The weighted cumulative loss increment of the weakest membrane element in each container, and Use the same units of measurement; It is a dimensionless weighting coefficient used to adjust the relative weights between the energy consumption term and the loss term. The decay rate of the membrane element with the highest decay rate in the container and the first The loss is calculated by multiplying the load over time periods; the higher the load, the greater the increase in loss. The objective function is optimized accordingly. from to Summation reflects the overall minimization of energy consumption and loss in each period within the preset time window.

[0053] It should be noted that the transition constraints of the dynamic programming algorithm incorporate the decay rate of the membrane element with the maximum decay rate within each container. A decreasing load limit constraint is imposed on containers whose internal heterogeneity index exceeds a preset threshold, where the preset threshold is determined during the system commissioning phase based on membrane element model parameters and operating conditions. The weighted cumulative loss of the weakest membrane element is introduced into the optimization objective to replace the average loss of the containers, enabling the load allocation strategy to proactively avoid containers with an accelerating deterioration trend, thus delaying premature replacement of the entire container due to local component failure.

[0054] In this embodiment, during the load allocation phase, differentiated adjustments with upper limits are made based on the comprehensive performance score of each container, above the equal baseline load. Containers with higher scores have their permeate ratio moderately increased based on the equal baseline load, while containers with lower scores have their permeate load decreased accordingly. The permeate load adjustment of any container does not exceed a preset maximum offset ratio to avoid excessive mapping of performance differences to load allocation, which could lead to accelerated degradation of high-load containers due to factors such as increased concentration polarization and excessive transmembrane pressure difference.

[0055] In one embodiment, the preset maximum offset ratio is ±15 percent of the equal reference load.

[0056] It should be noted that the differentiated load allocation scheme is updated once with each periodic performance evaluation cycle, and the current allocation scheme remains unchanged between two evaluation cycles. The way the differentiated load allocation scheme is updated allows the load allocation strategy to track the long-term evolution trend of membrane vessel performance, while avoiding system operation fluctuations caused by frequent adjustments.

[0057] Generate control commands for the water production path configuration that have passed the constraint verification and send them to the ultrapure water preparation system for execution.

[0058] Step 6: Perform water quality safety verification on the optimal water production path configuration.

[0059] In this embodiment of the application, to ensure the safety of the produced water quality, the following steps are also included in addition to step 5: Water quality safety verification is performed on the optimal permeate path configuration for each time period. The fine removal rates corresponding to the container combinations deployed in each time period are cascaded and multiplied to obtain the expected final effluent water quality index values. The lower bound of the confidence interval for the water quality index is calculated based on the estimated uncertainty of each membrane element. The lower bound of the confidence interval is compared with the ultrapure water quality requirement threshold. If the lower bound of the confidence interval meets the water quality requirement threshold, the verification is passed; if the lower bound of the confidence interval does not meet the water quality requirement threshold, the safety margin constraint is increased and the permeate path configuration is re-solved. The control command for the verified permeate path configuration is issued to the ultrapure water preparation system for execution.

[0060] It should be noted that the calculation of the lower bound of the confidence interval introduces the estimation uncertainty of the extended Kalman filter output in step 3. The estimation uncertainty reflects the reliability of the mass transfer coefficient estimates of each membrane element; elements with higher uncertainty correspond to a wider range of predicted removal rates. Using the lower bound of the confidence interval, rather than the point estimate, for verification can provide a safety margin for the quality of the produced water under conditions of limited estimation accuracy.

[0061] Step 7: Update the water production path configuration control command based on rolling time-domain re-optimization.

[0062] In this embodiment of the application, to address sudden changes in state during the continuous degradation of membrane element performance, the following steps are further included during the execution of the permeate path configuration control command: The one-dimensional hydrodynamic discrete node inversion and element-level extended Kalman filter update are re-executed at a preset period to obtain the latest mass transfer coefficient estimate and decay rate of each membrane element.

[0063] Rolling time-domain reoptimization is triggered when any of the following conditions are met: a sudden increase in any heterogeneity index within the container; or the decay rate of any membrane element deviates from the predicted value beyond the allowable range.

[0064] The refined energy efficiency ratio matrix and refined removal rate matrix are recalculated using the updated component-level mass transfer coefficient estimates. The water production path configuration for the remaining time period is resolved, and the corrected water production path configuration control command is generated and issued for execution.

[0065] This implementation uses one-dimensional hydrodynamic discrete node inversion to reconstruct the differentiated degradation states of each membrane element within the container using only existing sensor data at both ends of the container, refining the monitoring granularity of membrane performance degradation from the container level to the element level. Therefore, subsequent load distribution strategies can distinguish between "uniform, mild degradation of each element" and "severe degradation at the front end combined with intact degradation at the rear end," resolving the problem of the container-level average degradation rate masking the internal heterogeneous distribution.

[0066] By using the removal rate of the weakest link instead of the average removal rate of the containers to calculate a refined removal rate matrix, the removal rate assessment of each container is based on the membrane element with the worst internal performance. This avoids the problem of average values ​​masking local defects, leading to the incorrect allocation of permeate tasks to containers with severe local degradation, resulting in actual effluent quality lower than expected. Furthermore, by incorporating the weighted cumulative loss of the weakest membrane element into the dynamic programming optimization objective, the load allocation strategy proactively avoids containers with accelerating deterioration trends, delaying premature replacement of the entire container due to local element failure and reducing long-term membrane replacement costs.

[0067] By incorporating both measured permeate conductivity values ​​and inversion indices from the component-level physical model into the comprehensive performance score, the score considers both model estimations and actual measurements, compensating for potential scoring biases caused by model errors when relying solely on model inversion. Based on the comprehensive performance score, a differentiated load allocation with upper limits is implemented. This approach rationally utilizes the potential of high-performance membrane vessels while preventing accelerated degradation of high-load vessels due to increased concentration polarization and excessive transmembrane pressure differentials, achieving a balance between performance priority and safe operation. The differentiated allocation scheme is updated synchronously with the periodic performance scoring cycle, ensuring that the load allocation strategy tracks the long-term evolution trend of membrane vessel performance.

[0068] By using confidence intervals based on estimation uncertainty for water quality verification, a safety margin is provided for product water quality even under conditions of limited accuracy in mass transfer coefficient estimation. Through rolling time-domain reoptimization, the system responds promptly to sudden state changes during the continuous degradation of membrane element performance, ensuring the operational stability of the control method for the ultrapure water preparation system.

[0069] The following is an example of an application of the present invention, such as... Figure 2-9 As shown, the implementation process is as follows: A semiconductor manufacturing plant (code name: FAB-C) has three parallel pressure vessels (vessel 1, vessel 2, and vessel 3) in its ultrapure water preparation workshop. Each vessel contains four reverse osmosis membrane elements connected in series. This system provides continuous water supply for the wafer wet cleaning process and has extremely strict requirements on the conductivity of the produced water, with water quality standards reaching specific thresholds. The value is 0.10 μS / cm. The system has been running continuously for approximately 18 months. Due to differences in influent water quality and operating load history, the membrane elements in each container exhibit varying degrees of degradation. The current scoring period is the 7th week of the 3rd quarter of 20XX Year C.

[0070] The system collects real-time operating parameters from sensors at both ends of each container and a conductivity meter at the product water outlet. The total conductivity of the influent is 152 μS / cm, and the influent temperature is 25℃. The real-time operating parameters of each container are shown in the table below.

[0071] Table 1 Real-time operating parameters of each pressure vessel

[0072] Simultaneously read the membrane model parameters of the four membrane elements (element 1 to element 4, arranged sequentially from the influent end to the concentrate end) in each container: initial permeability coefficient. The standard is 7.20 L / (m³). 2 (·h·bar), effective membrane area is 37.2m² 2 / branch, with a hydraulic diameter of 0.86mm.

[0073] Taking container 2 as an example, the one-dimensional fluid dynamics discrete node inversion process is illustrated. Container 2 has an inlet pressure of 12.35 bar, an outlet concentrate pressure of 11.58 bar, and an influent flow rate of 8.15 m³ / s. 3 / h, production water flow rate 5.70m 3 The input of / h serves as the boundary and constraint condition for iterative solution. The internal flow path of container 2 is discretized into four computational segments, and iteration is performed segment by segment from element 1 to element 4, with continuous transfer of concentrate flow rate between adjacent segments. The results of the iterative convergence of local transmembrane pressure difference, osmotic pressure difference, and product flux for each segment are shown in the table below.

[0074] Table 2. Inversion results of discrete nodes for each membrane element in container 2.

[0075] Taking element 1 as an example, the equivalent permeability coefficient L / (m 2 ·h·bar), where For the local water production flux of element 1, Net driving voltage difference for component 1; attenuation rate ,in The initial permeability coefficient is given. Container 2 exhibits a gradient distribution of attenuation rate that gradually increases from the influent end to the concentrate end, with the attenuation rate of the tail-end element 4 reaching as high as 0.64, which is typical of inorganic scaling on the concentrate side. The same method was used to perform inversion on containers 1 and 3, and the element-level attenuation rates of the three containers are summarized in Table 3.

[0076] Table 3 Summary of the equivalent permeability coefficient attenuation rate of each membrane element in the three containers

[0077] The attenuation rate of each component in container 1 is uniformly distributed, while that in container 2 shows a clear gradient increasing trend, and that in container 3 is at an intermediate level. The average attenuation rates of container 1 and container 2 (0.20 and 0.55) differ significantly, but the heterogeneous structure inside container 2 needs to be further quantified in the next step.

[0078] The time series of the equivalent permeability coefficient decay rate of each membrane element in each container is input into an extended Kalman filter. Taking element 4 in container 2 as an example, the historical decay rate series of this element shows a continuously accelerating upward trend, and the water flux decay coefficient and salt permeability increment in the latent state variables are both at a high level. Combining the concentrate concentration and salt permeability of this segment, according to the formula... Calculate the estimated local permeate conductivity, where This is the conductivity conversion factor. For element 4, salt permeability, The concentration of the inlet water for element 4 is... The permeate flux of element 4 is shown in the table below, which corroborates the material balance recursive results during the iteration process. The extended Kalman filter outputs of the three key membrane elements in the container are shown in the table below.

[0079] Table 4. Estimation results of mass transfer coefficients for key membrane elements in each container.

[0080] The estimation uncertainty of component 4 in container 2 (0.21) is significantly higher than that in containers 1 and 3, reflecting that the reliability of its mass transfer coefficient estimation is relatively low. The decay rate of 0.018 / kWh is also much higher than that of other containers, indicating that the component is in the accelerated deterioration stage.

[0081] Step 401 calculates the internal attenuation heterogeneity index. Taking container 2 as an example, The attenuation rates of each component are 0.46, 0.52, 0.57, and 0.64, respectively. Substituting these values ​​into the formula yields the standard deviation of the attenuation rate. ,in For the first Attenuation rate of membrane element; ratio of maximum to mean. ,in This represents the maximum unit decay rate within the container. The heterogeneity indices for the three containers are summarized below.

[0082] Table 5. Attenuation Heterogeneity Indicators within Each Container

[0083] Container 2 and The values ​​are the highest among the three, indicating that the internal attenuation height is not uniform, which is consistent with the inversion result of the gradient attenuation distribution in step 2.

[0084] Step 402 substitutes the predicted mass transfer coefficient at the element level into the material balance and energy consumption equations to calculate the refined energy efficiency ratio and refined removal rate of each container under the current standard load. The refined removal rate is taken as the removal rate of the membrane element with the lowest removal rate in the container (i.e., element 4), rather than the average removal rate of the containers. The salt retention capacity of element 4 in container 2 has decreased significantly, and its refined removal rate is lower than that of containers 1 and 3.

[0085] Step 403 performs a comprehensive performance evaluation of the three containers. Each raw index, after range normalization, is weighted accordingly. , , , The weighted summation and comprehensive scoring formula is as follows: ,in The mean decay rate component, As heterogeneous components, For water quality, The energy efficiency ratio component yields the following comprehensive performance score.

[0086] Table 6. Overall Performance Scores of Each Container

[0087] Container 1 received a comprehensive score of 0.93, Container 3 scored 0.79, and Container 2 scored only 0.17. The measured conductivity of the produced water from Container 2, at 0.118 μS / cm, exceeded the water quality threshold of 0.10 μS / cm. Its score is extremely low, making its overall score far lower than the other two containers.

[0088] Get the next 8 hours (preset time window) The forecast sequence of water demand for the wafer cleaning process (divided into time periods, each lasting 1 hour) shows a total demand ranging from 13.20 to 16.50 m³ / h. 3 Fluctuations between / h. Based on the equal baseline load of the three containers (total demand divided by the number of containers), a differential adjustment with upper limit constraints is made according to the comprehensive performance score, with a maximum deviation of ±15% of the equal baseline load. Container 2... Exceeding the preset heterogeneity threshold (1.10) triggers the decreasing load upper limit constraint, further reducing the upper limit of its water production load. After dynamic programming solution, the typical high demand period (demand 16.50m) 3 The load allocation scheme for ( / h) is as follows.

[0089] Table 7. Water Production Load Allocation Scheme during Typical High Demand Periods

[0090] It should be noted that the total load is 16.33m. 3 / h and demand 16.50m 3 The slight difference in / h is compensated at the execution level by the system control valve. Container 1 obtains the highest allocated load, making full use of its excellent performance; Container 2 is subject to both scoring and heterogeneity constraints, and its actual allocated load is lower than the average benchmark, effectively mitigating the further deterioration of its weakest component (component 4).

[0091] Water quality safety verification was performed on the above allocation scheme. The refined removal rates of containers 1 and 3 under their respective allocated loads were cascaded and multiplied to obtain an estimated terminal effluent conductivity value of 0.068 μS / cm. Combining the estimation uncertainty of the weakest element of each container in Table 4, the lower bound of the confidence interval for water quality indicators (corresponding to the lower bound of the confidence interval for removal rate, i.e., the upper bound of the confidence interval for product water conductivity) was calculated to be 0.091 μS / cm, which is lower than the water quality threshold of 0.10 μS / cm, thus passing the water quality safety verification. Container 2 was not included in the main product water path during this period due to its extremely low comprehensive score and the measured conductivity exceeding the standard. The final product water path configuration control command was issued to the FAB-C ultrapure water preparation workshop for execution.

[0092] The system continuously updates the one-dimensional fluid dynamics discrete node inversion and extended Kalman filter at a preset cycle of 30 minutes. Approximately 2.5 hours into this scoring cycle, the decay rate of element 4 in container 2 suddenly increased from 0.018 / kWh to 0.031 / kWh, deviating from the predicted value beyond the allowable range, triggering a rolling time-domain reoptimization. The refined energy efficiency ratio matrix and refined removal rate matrix were recalculated using the updated element-level mass transfer coefficient estimates, and the permeate path configuration was resolved for the remaining 5.5 time periods. After reoptimization, the load limit of container 2 was further compressed, the allocation ratio of containers 1 and 3 was correspondingly increased, and the corrected permeate path configuration control command was immediately issued and executed.

[0093] Throughout the process, data starts with the raw sensor parameters collected in step 1, is transformed into element-level attenuation rates through discrete node inversion in step 2 (Tables 2 and 3), and then outputs estimated mass transfer coefficients and uncertainties through extended Kalman filtering in step 3 (Table 4). In step 4, heterogeneity indices are calculated and a comprehensive score is generated (Tables 5 and 6) to form a multi-dimensional performance profile for each container. Finally, in step 5, dynamic programming is used to generate constrained differentiated load allocation schemes (Table 7), and the safety is confirmed by confidence interval water quality verification in step 6. Rolling re-optimization in step 7 continuously tracks membrane performance evolution. Data flow between each step strictly adheres to the material and flow conservation relationships, with element-level information throughout, ensuring that load allocation decisions fully consider the heterogeneous structure within the containers.

[0094] The embodiments of the present invention have been described above. However, the embodiments are not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make more equivalent embodiments under the guidance of the present embodiments, and all of them are within the protection scope of the present embodiments.

Claims

1. A control method for an ultrapure water preparation system, characterized in that, include: Obtain the real-time operating parameters of each pressure vessel, the measured conductivity value of the product water at the product water outlet of each vessel, and the membrane model parameters of the membrane elements connected in series in each vessel; Based on the one-dimensional fluid dynamics discrete node method, the internal flow path of each container is divided into multiple calculation segments corresponding to the number of membrane elements. The local water production flux of each segment is iteratively solved using the pressure at both ends of the container as the boundary condition, and the equivalent permeability coefficient decay rate of each membrane element is obtained. The equivalent permeability decay rate time series of each membrane element is recursively estimated online using extended Kalman filtering to obtain the mass transfer coefficient estimate, estimation uncertainty and decay rate of each element. Based on the estimated mass transfer coefficient and decay rate of each element, the internal decay heterogeneity index, refined energy efficiency ratio matrix, and refined removal rate matrix of each container are calculated. Combined with the measured permeate conductivity value, the comprehensive performance score of each container is given; wherein the refined removal rate is the removal rate of the membrane element with the lowest removal rate in the container. Based on the comprehensive performance score, differentiated load allocation with upper limit constraints is carried out on an equal baseline load. The optimal water production path configuration for each time period is solved using a dynamic programming algorithm, and water production path configuration control instructions are generated and issued for execution.

2. The control method for the ultrapure water preparation system according to claim 1, characterized in that, The iterative solution for the local water production flux of each segment includes: Using the container inlet pressure and outlet concentrate pressure as boundary conditions, and the influent flow rate and total product flow rate as constraints, the local transmembrane pressure difference and local product flux of each segment are iteratively solved based on the effective membrane area and flow channel geometry parameters of each segment. During the iterative solution process, the flow rate conservation relationship between adjacent segments is satisfied: the concentrate flow rate of the previous segment is equal to the influent flow rate of the next segment, and the effluent concentration of the previous segment is used as the influent concentration of the next segment after concentration. The iteration proceeds segment by segment from the first segment to the last segment, so that the calculation results of each segment simultaneously satisfy the aforementioned boundary conditions and constraints.

3. The control method for the ultrapure water preparation system according to claim 1, characterized in that, The equivalent permeability coefficient decay rate is the ratio of the difference between the initial permeability coefficient and the current equivalent permeability coefficient to the initial permeability coefficient; wherein the current equivalent permeability coefficient is obtained by dividing the local permeate flux by the local net driving pressure difference of the corresponding segment, and the local net driving pressure difference is the difference between the local transmembrane pressure difference and the local osmotic pressure difference of the corresponding segment.

4. The control method for the ultrapure water preparation system according to claim 1, characterized in that, The hidden state variables of the extended Kalman filter are the water flux attenuation coefficient and salt permeability increment of each element, and the observed variables are the local product water flux and local product water conductivity estimates of the corresponding element; wherein the local product water conductivity estimates are jointly determined by the salt permeability, local influent concentration and local product water flux of the corresponding segment, and the local influent concentration is obtained recursively from the material balance relationship of the previous segment.

5. The control method for the ultrapure water preparation system according to claim 1, characterized in that, The internal attenuation heterogeneity index includes the standard deviation of the attenuation rate of each component in the container and the ratio of the maximum attenuation rate to the average attenuation rate in the container; wherein the standard deviation reflects the dispersion of the attenuation rate of each component in the container, and the ratio of the maximum attenuation rate to the average attenuation rate reflects the deviation of the most severe local attenuation from the average level.

6. The control method for the ultrapure water preparation system according to claim 1, characterized in that, The comprehensive performance score is obtained by weighted summation of the following four indicators after normalization: the score component based on the attenuation rate of the average mass transfer coefficient at the component level, the score component based on the internal attenuation heterogeneity index, the score component based on the deviation between the measured conductivity value of the product water and the conductivity threshold corresponding to the ultrapure water quality standard, and the score component based on the refined energy efficiency ratio. Each score component is mapped to the interval between zero and one using a mean normalization method based on the range before participating in the weighted summation.

7. The control method for the ultrapure water preparation system according to claim 1, characterized in that, The optimization objective of the dynamic programming algorithm is to minimize the sum of the total energy consumption for all time periods within a preset time window and the weighted cumulative loss of the weakest membrane element in each container. The weakest membrane element is the membrane element with the largest decay rate in the container, and the weighted cumulative loss increment is obtained by multiplying the decay rate of the weakest membrane element by the load of the corresponding time period; in the transition constraint condition of the dynamic programming algorithm, a decreasing load upper limit constraint is applied to the container whose internal heterogeneity index exceeds the preset threshold.

8. A control device for an ultrapure water preparation system, applied to electronic equipment, characterized in that, include: The data acquisition module is used to acquire the real-time operating parameters of each pressure vessel, the measured conductivity value of the product water at the product water outlet of each vessel, and the membrane model parameters of the membrane elements connected in series in each vessel. The element-level attenuation inversion module is used to divide the internal flow path of each container into multiple calculation segments corresponding to the number of membrane elements based on the one-dimensional fluid dynamics discrete node method. The local permeate flux of each segment is iteratively solved using the pressure at both ends of the container as the boundary condition to obtain the equivalent permeability coefficient attenuation rate of each membrane element. The mass transfer coefficient recursive estimation module is used to perform online recursive estimation of the mass transfer coefficient at the element level using extended Kalman filtering on the time series of the equivalent permeability coefficient decay rate of each membrane element, and to obtain the estimated mass transfer coefficient, estimation uncertainty and decay rate of each element. The comprehensive scoring module is used to calculate the internal decay heterogeneity index, refined energy efficiency ratio matrix, and refined removal rate matrix of each container based on the estimated mass transfer coefficient and decay rate of each element, and to give a comprehensive performance score to each container in combination with the measured permeate conductivity value; wherein the refined removal rate is the removal rate of the membrane element with the lowest removal rate in the container. The path optimization and instruction generation module is used to perform differentiated load allocation with upper limit constraints on the equal baseline load based on the comprehensive performance score, solve the optimal water production path configuration for each time period using dynamic programming algorithm, generate water production path configuration control instructions and issue them for execution.

9. An electronic device for an ultrapure water preparation system, characterized in that, It includes a memory and a processor, the memory storing a program or instructions that can run on the processor, the program or instructions being executed by the processor to implement the steps of the control method as described in any one of claims 1 to 7.

10. A readable storage medium, characterized in that, The readable storage medium stores a program or instructions that, when executed by a processor, implement the steps of the control method as described in any one of claims 1 to 7.

Images