Operating parameter optimization method and system for reverse osmosis membrane water production line

Abstract

The invention discloses an operation parameter optimization method and system for a reverse osmosis membrane water production line, and relates to the technical field of water treatment.The method comprises the steps that saline water characteristics of saline water are obtained, pretreatment prediction is conducted, and pretreated saline water characteristics are obtained; the method comprises the following steps: preprocessing saline water characteristics, obtaining desalination rate demand information, performing reverse osmosis operation parameter optimization in combination with the preprocessing saline water characteristics to obtain optimal reverse osmosis operation parameters, and performing reverse osmosis membrane blockage prediction, scaling prediction, desalination rate prediction and reverse osmosis fitness calculation of the reverse osmosis operation parameters according to the preprocessing saline water characteristics and the desalination rate demand information. A cluster optimization rule is set, and optimization is carried out; and carrying out production control on the reverse osmosis membrane water production line by adopting the optimal reverse osmosis operation parameters. The problems that in the prior art, an operation strategy cannot be adjusted in real time, so that a system operates under the non-optimal working condition for a long time, the overall energy efficiency is low, the operation control energy consumption is large, the maintenance cost is high, and the production efficiency is poor are solved.

Description

Technical Field

[0001] This invention relates to the field of water treatment technology, and more specifically to a method and system for optimizing operating parameters of a reverse osmosis membrane water production line. Background Technology

[0002] Existing reverse osmosis membrane water production line operation control methods typically set operating parameters based on fixed operating experience or simple indicators such as water flow rate and energy consumption.

[0003] Because the quality of the feed water and the risk of contamination to the reverse osmosis membrane were not considered, impurities and salts in the brine caused blockage or scaling, affecting the operation of the reverse osmosis membrane and the water purification effect. At the same time, when faced with fluctuations in water quality, the operating strategy could not be adjusted in real time, causing the system to operate under suboptimal conditions for a long time, resulting in low overall energy efficiency, high energy consumption for operation and control, high maintenance costs, and poor production efficiency. Summary of the Invention

[0004] This application provides a method and system for optimizing operating parameters of a reverse osmosis membrane water production line, which addresses the problems in the prior art where the operating strategy cannot be adjusted in real time, resulting in the system operating under suboptimal conditions for a long time, leading to low overall energy efficiency, high energy consumption for operation and control, high maintenance costs, and poor production efficiency.

[0005] In view of the above problems, this application provides a method and system for optimizing the operating parameters of a reverse osmosis membrane water production line.

[0006] In a first aspect, this application provides a method for optimizing operating parameters of a reverse osmosis membrane water production line, the method comprising: The characteristics of the brine are obtained, and preprocessing and prediction are performed to obtain the characteristics of the preprocessed brine. Obtain the desalination rate requirement information, combine it with the characteristics of the pretreated brine, optimize the reverse osmosis operating parameters, and obtain the optimal reverse osmosis operating parameters. Specifically, based on the characteristics of the pretreated brine and the desalination rate requirement information, predict the reverse osmosis membrane blockage, scaling, and desalination rate of the reverse osmosis operating parameters, calculate the reverse osmosis fitness, and set optimization cluster rules for optimization. The optimal reverse osmosis operating parameters are used to control the production of the reverse osmosis membrane water production line.

[0007] Secondly, the present invention provides an operating parameter optimization system for a reverse osmosis membrane water production line, comprising: The preprocessing prediction module is used to obtain the saline characteristics of the saline, perform preprocessing prediction, and obtain the preprocessed saline characteristics. The operating parameter optimization module is used to obtain desalination rate requirement information, combine it with the characteristics of the pretreated brine, optimize the reverse osmosis operating parameters, and obtain the optimal reverse osmosis operating parameters. Specifically, based on the characteristics of the pretreated brine and the desalination rate requirement information, the reverse osmosis operating parameters are predicted for reverse osmosis membrane blockage, scaling, and desalination rate. The reverse osmosis fitness is calculated, and optimization cluster rules are set for optimization. The production control module is used to control the production of the reverse osmosis membrane water production line using the optimal reverse osmosis operating parameters.

[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages: This application first obtains the salt concentration and impurity characteristics of the brine and performs pretreatment prediction to monitor the actual water quality characteristics entering the reverse osmosis membrane in real time. This overcomes the prediction bias caused by the pretreatment process and provides a reliable water quality data foundation for subsequent precise optimization. Secondly, by combining desalination rate requirements with the characteristics of the pretreated brine, reverse osmosis operating parameters such as membrane fouling, scaling, and desalination rate are predicted, providing ample optimization space for subsequent optimization. Simultaneously, a dual-objective dynamic optimization is performed, comprehensively considering the real-time and long-term operational risks of parameter combinations. Combining reverse osmosis membrane scaling prediction results with optimization cluster rules, a global search is conducted in the complex parameter space to output the optimal reverse osmosis operating parameters that effectively delay membrane fouling, improve system stability, and enhance economy while meeting desalination rate product requirements. Finally, the optimal reverse osmosis operating parameters are distributed for automatic production line control. Based on real-time water quality and production needs, production is dynamically adjusted to the optimal operating state, reducing reliance on operator experience and ensuring a highly efficient and low-consumption production optimization mode, thereby improving the automation level of the water production line. Attached Figure Description

[0009] Figure 1 This is a flowchart illustrating the method for optimizing operating parameters of a reverse osmosis membrane water production line, as described in this application. Figure 2 This is a schematic diagram of the operating parameter optimization system for a reverse osmosis membrane water production line used in this application.

[0010] In the attached diagram, the components represented by each number are as follows: Preprocessing prediction module 11, running parameter optimization module 12, production control module 13. Detailed Implementation

[0011] This application provides a method for optimizing operating parameters for reverse osmosis membrane water production lines, specifically addressing the problems in existing technologies where the operating strategy cannot be adjusted in real time, resulting in the system operating under suboptimal conditions for extended periods, leading to low overall energy efficiency, high energy consumption for operation and control, high maintenance costs, and poor production performance.

[0012] The present invention will now be described in detail with reference to the accompanying drawings.

[0013] Example 1, as Figure 1 As shown, this application provides a method for optimizing operating parameters of a reverse osmosis membrane water production line, characterized in that the method includes: S10: Obtain the saline characteristics of the saline, perform preprocessing prediction, and obtain the preprocessed saline characteristics; In this embodiment, the brine is the raw water to be treated, which has a high salt content and impurities. The brine characteristics are a set of key parameters describing the quality of the raw water, which is the basis for determining the difficulty and method of subsequent treatment. Pretreatment is a process of coarse filtration and adsorption of the raw water before it enters the reverse osmosis membrane. Pretreatment prediction is a process of discharging the raw water entering the reverse osmosis membrane and outputting the pretreated brine characteristics based on the raw water quality characteristics after passing through the pretreatment process unit.

[0014] Specifically, before reverse osmosis purification, the brine to be treated undergoes pretreatment such as coarse filtration and adsorption, followed by pretreatment prediction to predict the impurity content and salt content characteristics of the brine after pretreatment.

[0015] Step S10 in the method provided in this application embodiment includes: The characteristics of the brine are obtained, including salt concentration and impurity characteristics. The saline features are input into the preprocessing predictor, and the output is the preprocessed saline features.

[0016] In this embodiment, the brine characteristics are first obtained, including salt concentration and impurity characteristics. The brine characteristics are descriptive indicators of the brine in the reverse osmosis system; the salt concentration is the total amount of dissolved inorganic salts in the water, which directly determines the osmotic pressure that the reverse osmosis process needs to overcome and the potential quality of the produced water; the impurity characteristics are the properties of all substances in the water, excluding soluble salts, that may cause physical blockage, chemical contamination, or biological fouling of the reverse osmosis membrane.

[0017] Specifically, firstly, before pretreatment prediction, the characteristics of the brine are obtained, including salt concentration and impurity content. These two indicators are then used to quantify the salt and impurity content in the brine, serving as the basis for subsequent reverse osmosis optimization.

[0018] For example, assuming a salt concentration of 8500 mg / L is obtained, impurity characteristics may include: calcium ion concentration: 280 mg / L.

[0019] Secondly, the saline features are input into the preprocessing predictor, and the output is the preprocessed saline features. Specifically, the obtained saline feature vector is passed as a calling parameter to the preprocessing predictor model that has been deployed on the server or edge computing device. The preprocessing prediction model performs preprocessing prediction, and finally, the model outputs the preprocessed saline features.

[0020] In step S10 of the method provided in this application embodiment, the preprocessing predictor training step includes: Collect the feature set of sample saline solution, and collect the feature set of pretreated saline solution after different sample saline solution feature preprocessing. Based on machine learning, a preprocessing predictor is constructed. The preprocessed predictor is trained and tested using a sample saline feature set and a sample preprocessed saline feature set. The network parameters are iteratively optimized until the test converges, thus completing the training and testing process.

[0021] In this embodiment, a sample saline feature set is first collected, and a sample pre-processed saline feature set is collected after different sample saline features are pre-processed. The sample saline feature set is a dataset composed of historical or experimental data points; the sample pre-processed saline feature set is a dataset corresponding to the records in the sample saline feature set.

[0022] Specifically, at the inlet of the pretreatment system, the characteristics of the sample brine are collected; at the outlet of the pretreatment system and before the reverse osmosis high-pressure pump, the characteristics of the corresponding sample pretreated brine are collected. The two sets of data are strictly aligned according to the timestamp to ensure that they come from the same batch of water samples. Similarly, multiple samplings are performed in this way to finally obtain paired datasets.

[0023] Secondly, a preprocessing predictor is constructed based on machine learning. Specifically, a neural network is used to build the preprocessing predictor. This involves constructing the neural network framework, defining the dimensions of the input layer, the number of hidden layers and their neurons, the dimensions of the output layer, and the structural layers between the input and output layers. At this point, the model has a structure, but the weights and bias parameters of the internal connections are randomly initialized, and it does not yet possess predictive capabilities.

[0024] Next, using both sample saline feature sets and preprocessed sample saline feature sets, the preprocessed predictor is subjected to supervised training and testing. The network parameters are iteratively optimized until the test converges, completing the training and testing process. Supervised training involves calculating the predicted output through forward propagation of the input data, then using a loss function to calculate the difference between the predicted output and the true label, and reducing this difference. Iterative optimization of the network parameters uses an optimization algorithm to calculate the gradient based on the loss value, and then backpropagates along the gradient in the opposite direction, sequentially adjusting the weights and bias parameters of each layer of the network.

[0025] Specifically, the sample data is first randomly divided into training, validation, and test sets. Then, the saline water features from the training set are input into the constructed preprocessing prediction model to obtain the predicted preprocessed water quality. The loss between the predicted value and the training set label is calculated using the preprocessing prediction model, and the network parameters are updated through backpropagation and an optimizer. After a certain number of training rounds, the performance of the current model is evaluated using test set data. This process is repeated until the loss value on the test set no longer decreases or fluctuates very little over multiple consecutive evaluation periods, at which point the test is considered converged. Training is then stopped, the current network parameters are saved, and the preprocessed predictor is obtained.

[0026] A backpropagation (BP) neural network can be used to construct a preprocessing predictor, predicting the direction of output parameter adjustment. The BP neural network model is a feedforward neural network trained through backpropagation of error and is commonly used to predict continuous values.

[0027] For example, the steps to construct a preprocessing predictor based on a BP neural network are as follows: First, a set of sample saline features was collected, and a set of preprocessed sample saline features was also collected after different sample saline features were preprocessed. The sample data was then divided into training, validation, and test sets in a 7:2:1 ratio.

[0028] Secondly, the preprocessing predictor mainly consists of an input layer, a hidden layer, and an output layer. The input layer receives the sample saline feature set and the sample preprocessed saline feature set; the hidden layer performs nonlinear transformation through an activation function; and the output layer outputs the prediction result through an activation function.

[0029] Finally, using the sample saline feature set and the sample preprocessed saline feature set as input, an initial learning rate and weights are set and weights are assigned. The mean squared error function is used to calculate the error between the predicted and actual results. Weights are adjusted and calculated, and this process is repeated iteratively until the error is minimized. The input data is processed through weighted summation and activation functions via forward propagation, and then passed layer by layer to the output layer. The gradient of the loss function is calculated via backpropagation, and the parameters are updated. Performance is evaluated using a validation set after each training epoch to avoid overfitting. When the MSE loss decreases by less than 1e in five consecutive training epochs, the algorithm is considered successful.-5 When the MSE loss on the validation set stabilizes below 0.01, the model is considered converged, and the preprocessed predictor is obtained.

[0030] In this embodiment, the specific composition of brine characteristics is obtained, and then a machine learning-based preprocessing predictor is used to establish the ability to predict the water quality from raw water to the reverse osmosis membrane feed water, thereby achieving precise optimization input. The brine characteristics are concretized into salt concentration and impurity characteristics, providing standardized input dimensions for the evaluation of all subsequent models and ensuring the objectivity and comprehensiveness of the optimization basis.

[0031] S20: Obtain desalination rate requirement information, combine it with the characteristics of the pretreated brine, optimize the reverse osmosis operating parameters, and obtain the optimal reverse osmosis operating parameters. Specifically, based on the characteristics of the pretreated brine and the desalination rate requirement information, predict the reverse osmosis membrane blockage, scaling, and desalination rate of the reverse osmosis operating parameters, calculate the reverse osmosis fitness, and set optimization cluster rules for optimization. In this embodiment, the desalination rate requirement information is the requirement for product water quality in the production target, usually expressed as desalination rate; the reverse osmosis operating parameters are the directly controllable variables for operating the RO membrane unit; the reverse osmosis membrane fouling prediction is a process of predicting the membrane fouling rate or risk based on feed water characteristics and operating parameters; the optimization cluster rule is a mechanism that classifies and groups a large number of candidate operating parameter solutions according to their performance, and guides parameter solutions in different clusters to collaboratively explore and compete towards a better direction.

[0032] Specifically, the desalination rate requirement information for this operation is first obtained from external or internal databases. Then, this desalination rate requirement information, along with the characteristics of the pretreated brine, is used as input to the optimization problem, generating numerous candidate combinations of reverse osmosis operating parameters. Next, a prediction model is used to predict reverse osmosis membrane blockage, scaling, and desalination rate for each parameter set under the current feed water characteristics. Based on the predicted blockage rate, scaling rate, and desalination rate, the reverse osmosis fitness is calculated. Finally, based on the reverse osmosis fitness, the long-term operational risk is assessed, optimization cluster rules are set, and the reverse osmosis operating parameters are obtained through an optimization algorithm.

[0033] Step S20 in the method provided in this application embodiment includes: Obtain desalination rate requirements; Obtain the reverse osmosis operating parameter space; Multiple first reverse osmosis operating parameters are generated within the reverse osmosis operating parameter space, wherein each first reverse osmosis operating parameter includes a pressure parameter; Based on the characteristics of the pretreated brine, pretreatment blockage prediction and permeation scaling prediction are performed by combining multiple first reverse osmosis operating parameters to obtain multiple first blockage rates and multiple first scaling rates, as well as multiple first energy cost information and multiple first desalination rates. Based on multiple first blockage rates, multiple first scaling rates, multiple first energy cost information, and multiple first desalination rates, multiple first reverse osmosis fitness rates are calculated. Based on multiple first blockage rates and multiple first fouling rates, optimization clusters are divided to obtain multiple blockage operation parameter clusters and multiple fouling operation parameter clusters. Iterative optimization is continued until convergence, resulting in multiple converged blockage operation parameter clusters and multiple converged fouling operation parameter clusters. The optimization balance is then analyzed. Based on the optimization balance, the optimized operating parameter set is obtained through screening, and the optimal reverse osmosis operating parameters are then selected.

[0034] In this embodiment, the desalination rate requirement information is first obtained. Specifically, the desalination rate requirement information for this water production task is actively read or received from the upstream production management system or the formula input or preset by the operator's human-machine interface. The obtained desalination rate requirement information is used as the fundamental guide and evaluation standard for all subsequent optimization calculations to ensure that the optimized operating parameters can meet the requirements of the produced water quality.

[0035] For example, in a coastal industrial park, the required desalination rate of a reverse osmosis membrane water production line is 97%.

[0036] Furthermore, the reverse osmosis operating parameter space is obtained. This space is the set of all legal values ​​for adjustable operating parameters, with each dimension representing a controllable variable. Specifically, predefined operating limits are retrieved from the equipment database or engineering configuration files. This results in a multi-dimensional parameter space, including the pressure range for safe operation. The optimization algorithm will then search within this space to ensure the feasibility of all recommended parameter combinations.

[0037] For example, taking a coastal industrial park as an example, the operating parameter space of a certain production line can be read from the equipment database as follows: operating pressure range: 1.2MPa~2.0MPa.

[0038] Furthermore, multiple first reverse osmosis operating parameters are generated within the reverse osmosis operating parameter space, where each first reverse osmosis operating parameter includes a pressure parameter. Additionally, the first reverse osmosis operating parameters are a series of candidate parameter combinations generated in the first round of the optimization iteration process, each combination being a point within the parameter space.

[0039] Specifically, multiple first reverse osmosis operating parameters can be generated in the reverse osmosis operating parameter space using sampling methods such as random sampling. Each first reverse osmosis operating parameter includes a pressure parameter during operation. This pressure parameter is the operating pressure or feed water pressure on the feed water side of the reverse osmosis membrane module, provided by a high-pressure feed water pump. It is mainly used to overcome the driving force of the following two main resistances: the inherent resistance and fouling resistance of the membrane, and the osmotic pressure of the water. The osmotic pressure of the water is the physical pressure determined by the salt concentration difference between the feed water and the product water. The higher the salt concentration, the greater the osmotic pressure, and the higher the required operating pressure must be to drive water molecules to pass through the semi-permeable membrane in the reverse direction. Therefore, the higher the pressure parameter, the more power is consumed, and the more prone it is to clogging or scaling, but the higher the output and the higher the desalination rate. The inherent resistance and fouling resistance of the membrane include the inherent resistance encountered by water when passing through the membrane micropores, as well as the additional flow resistance generated by fouling on the membrane surface. Since pressure is the core factor driving the reverse osmosis process and directly affecting the product water output, desalination rate, and energy consumption, the pressure parameter can reflect the operating effect of the reverse osmosis operation.

[0040] Furthermore, based on the characteristics of the pretreated brine, pretreatment blockage prediction and permeate scaling prediction are performed by combining multiple first reverse osmosis operating parameters, resulting in multiple first blockage rates and multiple first scaling rates, as well as multiple first energy cost information and multiple first desalination rates. Specifically, the pretreatment blockage prediction predicts the risk or rate of physical blockage of membrane channels caused by impurities under given feed water quality and operating parameters; the permeate scaling prediction predicts the risk or rate of soluble salts such as calcium carbonate crystallizing and precipitating on the membrane surface to form a scale layer under the same conditions due to concentration exceeding the solubility product; the first energy cost information is the energy cost consumed to drive equipment such as the high-pressure pump under each first reverse osmosis operating parameter; and the first desalination rate is the expected desalination performance achievable under each first reverse osmosis operating parameter.

[0041] Specifically, the first reverse osmosis operating parameters and the known characteristics of the pretreated brine are input into the reverse osmosis membrane fault predictor. The model predicts the blockage rate and scaling rate after reverse osmosis purification, as well as the desalination rate, in parallel. The power consumption at different pressures is obtained as energy cost information. Finally, this process is repeated for all initial parameters to obtain an array of data containing the first blockage rate, the first scaling rate, the first energy cost information, and the first desalination rate.

[0042] Furthermore, multiple first reverse osmosis fitness values ​​are calculated based on multiple first blockage rates, multiple first scaling rates, multiple first energy cost information, and multiple first desalination rates. The first reverse osmosis fitness value quantifies the merits of a particular combination of operating parameters. Specifically, it is calculated based on the first blockage rate, first scaling rate, first energy cost information, and first desalination rate to obtain the corresponding first reverse osmosis fitness value. A higher value indicates better overall performance of the parameter combination.

[0043] Furthermore, based on multiple first-level blocking rates and multiple first-level scaling rates, optimization clusters are partitioned to obtain multiple blocking operation parameter clusters and multiple scaling operation parameter clusters. Iterative optimization continues until convergence, resulting in multiple convergent blocking operation parameter clusters and multiple convergent scaling operation parameter clusters. The optimization equilibrium is then analyzed. Specifically, optimization cluster partitioning is based on clustering, grouping individuals according to a key feature to form different exploratory clusters. Iterative optimization involves, after initial population evaluation and cluster partitioning, the algorithm performs fitness-based selection and parameter adjustment within each cluster to generate a new generation of parameter populations. This process is repeated, with re-evaluation and re-partitioning of clusters. Convergence occurs when, after multiple iterations, the population's fitness no longer significantly improves, or parameter changes stabilize, reaching an optimal or near-optimal state. Optimization equilibrium is obtained by analyzing the performance similarity between different convergent clusters, reflecting the existence of a unique optimal solution region.

[0044] Specifically, the system first performs clustering based on the initial blocking rate of the array parameters to obtain corresponding blocking operation parameter clusters; simultaneously, it performs another round of clustering based on the initial fouling rate, similarly obtaining fouling operation parameter clusters. For each cluster, the parameters of other individuals are updated, guided by the optimal individual within each cluster, generating a new generation of parameter populations. This new generation of populations undergoes iterative optimization and evaluation again until a preset number of iterations is reached or a convergence condition is met. Ultimately, several convergent clusters that achieve optimal performance in different directions of blocking and fouling are obtained. By analyzing the similarity of the overall performance of the optimal convergent clusters, the optimization equilibrium can be calculated.

[0045] Finally, based on the optimization balance, an optimized set of operating parameters is obtained through screening, and the optimal reverse osmosis operating parameters are selected. The screening process involves targeted selection of the final convergence clusters based on the optimization balance to determine the ultimately recommended optimal parameters; the optimized set of operating parameters is the set selected from all convergence clusters to construct the final decision candidate pool; and the optimal reverse osmosis operating parameters are the parameter combination ultimately selected by the system and recommended to the production line control system for execution.

[0046] Specifically, firstly, the cluster with the highest average fitness is selected from all convergent clusters as the optimized operating parameter set. Then, using the optimization equilibrium as a proportional coefficient, the top-ranked parameter individuals are selected from the clusters according to their fitness ranking to form the optimized operating parameter set. Finally, the parameter combination with the highest absolute fitness value is directly selected from the optimized operating parameter set as the optimal reverse osmosis operating parameters.

[0047] In step S20 of the method provided in this application embodiment, based on the characteristics of the pretreated brine, pretreatment blockage prediction and permeation scaling prediction are performed by combining multiple first reverse osmosis operating parameters to obtain multiple first blockage rates and multiple first scaling rates, including: A reverse osmosis membrane failure predictor is obtained, which includes a blockage prediction branch and a scaling prediction branch. The inputs during the training process of the blockage prediction branch are the sample pretreated brine characteristics and the sample reverse osmosis operating parameters, and the label is the sample blockage rate of the reverse osmosis membrane blocked by impurities. The inputs during the training process of the scaling prediction branch are the sample pretreated brine characteristics and the sample reverse osmosis operating parameters, and the label is the sample scaling rate of salt scaling on the reverse osmosis membrane. The pretreated brine characteristics are combined with multiple first reverse osmosis operating parameters and input into the reverse osmosis membrane fault predictor to obtain multiple first blockage rates and multiple first scaling rates.

[0048] In this embodiment, a reverse osmosis membrane fault predictor is first obtained. This predictor includes a blockage prediction branch and a scaling prediction branch. The inputs to the blockage prediction branch during training are the characteristics of the sample pretreated brine and the sample reverse osmosis operating parameters, labeled as the sample blockage rate of the reverse osmosis membrane due to impurities. Similarly, the inputs to the scaling prediction branch are the characteristics of the sample pretreated brine and the sample reverse osmosis operating parameters, labeled as the sample scaling rate of salts forming on the reverse osmosis membrane. The blockage prediction branch is responsible for predicting physical blockage, learning the patterns of impurities in the feed water depositing on the reverse osmosis membrane surface and within the flow channels, leading to blockage. The scaling prediction branch is responsible for predicting chemical scaling, learning the patterns of dissolved inorganic salt ions precipitating into a crystalline scale layer on the membrane surface due to excessively high concentrations.

[0049] A backpropagation (BP) neural network can be used to construct a reverse osmosis membrane fault predictor to predict the direction of output parameter adjustments. The BP neural network model is a feedforward neural network trained through error backpropagation and is commonly used to predict continuous values.

[0050] For example, the steps to construct a reverse osmosis membrane failure predictor based on a BP neural network are as follows: First, the inputs to the blockage prediction branch training process are the sample pretreated brine characteristics and sample reverse osmosis operating parameters, with the label being the sample blockage rate of the reverse osmosis membrane due to impurities. Similarly, the inputs to the scaling prediction branch training process are the sample pretreated brine characteristics and sample reverse osmosis operating parameters, with the label being the sample scaling rate of salt on the reverse osmosis membrane. The sample data are divided into training, validation, and test sets in a 7:2:1 ratio.

[0051] Secondly, the reverse osmosis membrane fault predictor mainly consists of an input layer, a hidden layer, and an output layer. The input layer receives the input sample pretreated brine characteristics and sample reverse osmosis operating parameters for the blockage prediction branch, as well as the input sample pretreated brine characteristics and sample reverse osmosis operating parameters for the scaling prediction branch. The hidden layer performs nonlinear transformation through an activation function and is responsible for extracting high-level nonlinear comprehensive features from the original input. The input and hidden layers of the two prediction branches have the same structure and share both soft and hard input data. However, the output layer feeds the output of the hidden layer into two parallel and structurally identical blockage prediction branches and scaling prediction branches. The output of each branch is a neuron as the output node, so the blockage prediction branch and the scaling prediction branch can output the predicted blockage rate and the predicted scaling rate, respectively.

[0052] Finally, using the pretreated brine characteristics and reverse osmosis operating parameters of the input samples as input, the blockage rate of the reverse osmosis membrane due to impurities and the scaling rate of the reverse osmosis membrane due to salt fouling are used as labels for the blockage prediction branch and the scaling prediction branch, respectively. Initial learning rates and weights are set and weights are assigned. Mini-batch gradient descent is used for iterative optimization, sequentially passing through the input layer and both branches, ultimately calculating the predicted blockage rate and predicted scaling rate of the samples. Subsequently, the loss between the predicted values ​​and the true labels is calculated. The mean squared error loss function is applied to both branches, and the gradient of the total loss with respect to each trainable parameter in the network is accurately calculated using the backpropagation algorithm. An adaptive learning rate optimization algorithm (Adam optimizer) is used to update all weights and bias parameters based on the calculated gradient, reducing error. This process is repeated, with performance evaluated on a validation set after each training epoch to avoid overfitting. Training is stopped early if the loss no longer decreases within 20 consecutive training epochs to prevent overfitting, indicating model convergence and the preprocessed predictor is obtained.

[0053] Secondly, the pretreated brine characteristics are combined with multiple first reverse osmosis operating parameters and input into the reverse osmosis membrane fault predictor, resulting in multiple first blockage rates and multiple first scaling rates. Specifically, the previously generated N first reverse osmosis operating parameters and the current pretreated brine characteristics are used as inputs to the reverse osmosis membrane fault predictor. The blockage prediction branch and scaling prediction branch of the reverse osmosis membrane fault predictor work in parallel, performing calculations separately, and finally outputting the first blockage rate and first scaling rate that may occur for the first reverse osmosis operating parameters within a certain operating period. Subsequently, this process is repeated for all first reverse osmosis operating parameters, and the N first reverse osmosis operating parameters are evaluated. Finally, N predicted blockage rate values ​​and N predicted scaling rate values ​​are obtained.

[0054] For example, the first reverse osmosis operating parameter A, pressure = 1.67 MPa, is input into the reverse osmosis membrane failure predictor to obtain a first blockage rate = 0.07 and a first fouling rate = 0.12.

[0055] In step S20 of the method provided in this application embodiment, multiple first energy cost information and multiple first desalination rates are obtained, including: Based on the energy consumption and pressure correspondence of the reverse osmosis membrane water production line, multiple first energy cost information of multiple first reverse osmosis operating parameters is obtained; The pretreated brine features and multiple first reverse osmosis operating parameters are input into the desalination predictor, and multiple first desalination rates are output. The desalination predictor is built based on machine learning, and the input features for training are the sample pretreated brine features and sample reverse osmosis operating parameters, and the label is the sample desalination rate.

[0056] In this embodiment, firstly, based on the energy consumption-pressure correspondence of the reverse osmosis membrane water production line, multiple first energy cost information for multiple first reverse osmosis operating parameters is obtained. The energy consumption-pressure correspondence is a defined function or mapping relationship between the energy consumption of core energy-consuming equipment such as the high-pressure feed water pump and key operating parameters in the reverse osmosis system.

[0057] Specifically, based on the pressure parameters in the energy consumption-pressure correspondence and the system design water production, the first energy cost information of the high-pressure pump is calculated, and all the first reverse osmosis operating parameters are executed sequentially to finally obtain the first energy cost information corresponding to the first reverse osmosis operating parameters.

[0058] For example, taking a coastal industrial park as an example, if the production line is designed to produce 100 tons of water per hour, the energy consumption-pressure relationship is: unit water production energy cost = a × pressure + b, where a is 0.2 and fixed loss b is 0.05. Then, for the first reverse osmosis operating parameter A, with a pressure parameter of 1.67 MPa, its first energy cost information is 0.2 × 1.67 + 0.05 ≈ 0.38 yuan / ton. Calculations are performed on all the first reverse osmosis operating parameters to obtain several corresponding first energy cost information.

[0059] Secondly, the pretreated brine characteristics and multiple first reverse osmosis operating parameters are input into the desalination predictor, and multiple first desalination rates are output. The desalination predictor is built based on machine learning, and the input features for training are the sample pretreated brine characteristics and sample reverse osmosis operating parameters, and the label is the sample desalination rate.

[0060] Specifically, a pre-trained desalination predictor is invoked. For each combination of the first reverse osmosis operating parameters and the current pretreated brine characteristics, the parameter combination is input into the desalination predictor, which ultimately outputs a predicted first desalination rate. This process is repeated for all input data consisting of the first reverse osmosis operating parameters and pretreated brine characteristics, resulting in several first desalination rates. Additionally, a desalination predictor is constructed based on machine learning, using sample pretreated brine characteristics and sample reverse osmosis operating parameters as input data, and the sample desalination rate as the label.

[0061] A backpropagation (BP) neural network can be used to construct a desalination predictor to predict the direction of output parameter adjustment. The BP neural network model is a feedforward neural network that is trained through backpropagation of errors and is commonly used to predict continuous values.

[0062] For example, the steps to construct a desalination predictor based on a BP neural network are as follows: First, the pretreated saline characteristics and reverse osmosis operating parameters of the samples were used as input data for the model, and the sample desalination rate was used as the label. The samples were divided into training set, validation set and test set in a ratio of 7:2:1.

[0063] Secondly, the desalination predictor mainly consists of an input layer, a hidden layer, and an output layer. The input layer is used to receive the characteristics of the sample pretreated saline and the sample reverse osmosis operating parameters; the hidden layer performs nonlinear transformation through an activation function; and the output layer outputs the prediction results through an activation function.

[0064] Finally, using the pretreated saline characteristics and reverse osmosis operating parameters as inputs, and the sample desalination rate as the label, an initial learning rate and weights are set and weights are assigned. The mean squared error function is used to calculate the error between the predicted and actual results. Weight adjustments are made and calculations are repeated iteratively until the error is minimized. The gradient of the loss function is generated through forward propagation and calculated through backpropagation to update the parameters. Performance is evaluated using a validation set after each training epoch to avoid overfitting. When the MSE loss decreases by less than 1e in five consecutive training epochs, the target is considered achieved. -5 When the MSE loss on the validation set stabilizes below 0.01, the model is considered converged, and the desalination predictor is obtained.

[0065] Taking a coastal industrial park as an example, if the first reverse osmosis operating parameter A has a pressure of 1.67 MPa, the first desalination rate obtained after inputting it into the desalination predictor is 97.8%.

[0066] In step S20 of the method provided in this application embodiment, multiple first reverse osmosis fitness values ​​are calculated based on multiple first blockage rates, multiple first scaling rates, multiple first energy cost information, and multiple first desalination rates, including: Multiple first fusion failure rates are calculated based on multiple first blockage rates and multiple first fouling rates; Based on multiple first desalination rates and desalination rate requirements, multiple first desalination coefficients are calculated. Based on multiple primary energy cost information and benchmark energy cost information, multiple primary energy coefficients are calculated and obtained; Multiple first reverse osmosis fitness values ​​are calculated based on multiple first fusion failure rates, multiple first desalination coefficients, and multiple first energy coefficients. Among them, the fusion failure rate and energy coefficient are negatively correlated with the reverse osmosis fitness value, while the desalination coefficient is positively correlated with the reverse osmosis fitness value.

[0067] In this embodiment, multiple first fusion failure rates are first calculated based on multiple first blockage rates and multiple first fouling rates. The first fusion failure rate is a comprehensive measure of blockage risk and fouling risk, reflecting the total contamination load or health threat level faced by the membrane under a specific operating scheme. Specifically, the first fusion failure rate is the average of the multiple first blockage rates and the multiple first fouling rates, and is calculated as: First fusion failure rate = (First blockage rate + First fouling rate) / (Number of first blockage rates and first fouling rates).

[0068] For example, taking a coastal industrial park as an example, if the first reverse osmosis operating parameter A is pressure = 1.67MPa, the first blockage rate is 0.2, 0.3, 0.25, 0.15, 0.4, the first scaling rate is 0.3, 0.35, 0.25, 0.1, 0.2, and the first fusion failure rate is (0.2+0.3+0.25+0.15+0.4+0.3+0.35+0.25+0.1+0.2) / 10=0.225.

[0069] Furthermore, based on multiple first desalination rates and desalination rate demand information, multiple first desalination coefficients are calculated. The desalination coefficient is a scalar coefficient that measures the degree to which the predicted desalination rate meets and exceeds the demand target. Specifically, the first desalination coefficient is the ratio of the first desalination rate to the desalination rate demand information: First Desalination Coefficient = First Desalination Rate / Desalination Rate Demand Information. If the first desalination coefficient is 1, it indicates that the target is just met; if the coefficient is <1, it indicates that the target is not met; if the coefficient is >1, it indicates that the target is exceeded.

[0070] For example, taking a coastal industrial park as an example, the first desalination rate is 97.8%, the required desalination rate is 97%, and the first desalination coefficient = 97.8% / 97%≈1.008.

[0071] Furthermore, multiple first energy cost information and benchmark energy cost information are used to calculate multiple first energy coefficients. The first energy coefficient measures the energy efficiency of candidate solutions. Specifically, the first energy coefficient is the ratio of the first energy cost information to the benchmark energy cost information. If the ratio is greater than or equal to 1, it indicates that the energy consumption is equal to or lower than the benchmark, and the coefficient is excellent; if the ratio is less than 1, it indicates that the energy consumption is higher than the benchmark, and the coefficient decreases. The benchmark energy cost information is a standard value for judging the level of energy consumption, which can be obtained based on historical average advanced levels or current energy consumption benchmarks for influent salinity and water production.

[0072] For example, taking a coastal industrial park as an example, if the benchmark energy cost information is 0.7 yuan / ton and the first energy cost information is 0.38 yuan / ton, then the first energy coefficient is 0.38 / 0.7≈0.54.

[0073] Furthermore, based on multiple first fusion failure rates, multiple first desalination coefficients, and multiple first energy coefficients, multiple first reverse osmosis fitness values ​​are calculated. The fusion failure rate and energy coefficient are negatively correlated with the reverse osmosis fitness value, while the desalination coefficient is positively correlated. The first reverse osmosis fitness value is an evaluation of the overall merits of each candidate operating parameter scheme. Specifically, the first reverse osmosis fitness value is calculated as desalination coefficient - fusion failure rate - energy coefficient. Therefore, the first reverse osmosis fitness value = first desalination coefficient - first fusion failure rate - first energy coefficient. For the fusion failure rate and energy coefficient, the larger the index value, the smaller the calculated fitness value should be; for the desalination coefficient, the larger the index value and the larger the desalination coefficient, the greater the fitness value should be.

[0074] For example, taking a coastal industrial park as an example, the first reverse osmosis fitness is obtained as 1.008-0.54-0.225=0.243.

[0075] In step S20 of the method provided in this application embodiment, optimization clusters are divided according to multiple first blockage rates and multiple first fouling rates to obtain multiple blockage operation parameter clusters and multiple fouling operation parameter clusters. Iterative optimization continues until convergence is obtained to obtain multiple converged blockage operation parameter clusters and multiple converged fouling operation parameter clusters. The optimization balance is analyzed to obtain the optimization balance, including: For multiple first blocking rates and multiple first fouling rates, clustering is performed to obtain multiple first blocking rate clusters and multiple first fouling rate clusters. Each of the multiple first blocking rate clusters includes the smallest first blocking rate and multiple other following first blocking rates, and each of the multiple first fouling rate clusters includes the smallest first fouling rate and multiple other following first fouling rates. Based on multiple first blockage rate clusters and multiple first scaling rate clusters, multiple first reverse osmosis operating parameters are divided to obtain multiple blockage operating parameter clusters and multiple scaling operating parameter clusters. The operating parameters corresponding to the first blocking rate and the first scaling rate in each operating parameter cluster are used as the adjustment direction. The operating parameters corresponding to the first blocking rate and the first scaling rate are adjusted to update the multiple blocking operating parameter clusters and the multiple scaling operating parameter clusters. The blocking rate and scaling rate are predicted and the adjustment direction is updated. Continue iterative optimization until the convergence iteration count is reached, and obtain the final multiple convergence blocking operation parameter clusters and multiple convergence fouling operation parameter clusters; Calculate the average reverse osmosis fitness of multiple clusters of convergent blocking operating parameters and multiple clusters of convergent scaling operating parameters, and calculate the similarity to obtain the optimized balance.

[0076] In this embodiment, firstly, multiple first blocking rates and multiple first fouling rates are clustered to obtain multiple first blocking rate clusters and multiple first fouling rate clusters. Each first blocking rate cluster includes a minimum first blocking rate and multiple following first blocking rates, and each first fouling rate cluster includes a minimum first fouling rate and multiple following first fouling rates. Clustering involves grouping data objects according to the similarity of their values, ensuring that objects within the same cluster are similar to each other, while objects in different groups are dissimilar. The first blocking rate / first fouling rate cluster is the grouping result obtained after clustering the blocking rate values ​​and fouling rate values ​​respectively. The minimum first blocking rate / mineralization rate is the member with the smallest value in each blocking rate or fouling rate cluster; the following first blocking rate / following first fouling rate is the value in each cluster that is next in value to the minimum first blocking rate / mineralization rate.

[0077] Specifically, similar blocking rates and similar fouling rates are clustered together to obtain multiple first blocking rate clusters and multiple first fouling rate clusters. Since lower blocking and fouling rates are better, the minimum blocking and fouling rates within each cluster are used as the adjustment direction, and the other blocking and fouling rates are adjusted accordingly. Subsequently, within each cluster, the minimum blocking and fouling rates are identified and labeled as the first blocking rate / first fouling rate of that cluster; the remaining blocking and fouling rates within the cluster are designated as the following first blocking and following first fouling rates. Similarly, the same clustering operation is performed on all fouling rate values ​​to obtain multiple first fouling rate clusters and multiple first fouling rate clusters.

[0078] Furthermore, based on multiple first blockage rate clusters and multiple first scaling rate clusters, multiple first reverse osmosis operating parameters are divided to obtain multiple blockage operating parameter clusters and multiple scaling operating parameter clusters. Specifically, a blockage operating parameter cluster is the set of first reverse osmosis operating parameters corresponding to all blockage rate values ​​within the same first blockage rate cluster; a scaling operating parameter cluster is the set of first reverse osmosis operating parameters corresponding to all scaling rate values ​​within the same first scaling rate cluster.

[0079] Specifically, based on the cluster number to which each blockage rate value belongs, the first reverse osmosis operating parameter that generated that blockage rate is identified and assigned to its corresponding blockage operating parameter cluster, forming multiple blockage operating parameter clusters. Similarly, based on the clustering results of scaling rates, multiple scaling operating parameter clusters are also formed. All the first reverse osmosis operating parameters are divided into two groups according to their blockage characteristics and scaling characteristics.

[0080] Furthermore, taking the operating parameters corresponding to the first blocking rate and the first fouling rate within each operating parameter cluster as the adjustment direction, the operating parameters corresponding to multiple following first blocking rates and multiple following first fouling rates are adjusted, updating multiple blocking operating parameter clusters and multiple fouling operating parameter clusters, and predicting the blocking rate and fouling rate to update the adjustment direction. Here, the adjustment direction serves as a guide for searching and modifying within the parameter space; adjustment involves applying optimization algorithms to modify the parameter values ​​of the following individuals, causing their positions in the parameter space to move closer to or near the head individual; and updating means that after adjustment, the old parameter clusters are replaced by a newly generated batch of parameter combinations.

[0081] Specifically, based on pressure adjustment step size or other adjustment step size, the operating parameters corresponding to the first head blockage rate and the first head scaling rate within each operating parameter cluster are adjusted in the direction of adjustment. Multiple operating parameters corresponding to the first following blockage rate and multiple operating parameters corresponding to the first following scaling rate are adjusted. A certain degree of movement is made with the adjustment step size, while random perturbation is added to maintain exploratory nature. The adjusted blockage operating parameter cluster and the adjusted scaling operating parameter cluster are obtained. Then, combined with the current pretreated brine characteristics, a fault predictor is invoked to predict and obtain new blockage rates and scaling rates. Based on the new blockage rates and new scaling rates, new first head blockage rates and first head scaling rates are determined again within the adjusted blockage operating parameter cluster and the adjusted scaling operating parameter cluster, and these are used as the adjustment direction for the next iteration.

[0082] Furthermore, iterative optimization continues until the convergence iteration count is reached, resulting in multiple convergence blocking / scaling operating parameter clusters and multiple convergence scaling operating parameter clusters. The convergence blocking / scaling operating parameter clusters are the final parameter clusters obtained upon reaching the termination condition. These clusters have undergone multiple evolutions, and their performance tends to stabilize in their respective optimization directions, representing the optimal or near-optimal parameter set in their corresponding dimensions.

[0083] Specifically, a set number of convergence iterations is set, and the process is repeated in a loop. Once the required number of iterations is reached, the cluster is adjusted, evaluated, and updated. In each round, all clusters evolve independently and in parallel. The final results are obtained as multiple convergence blocking parameter clusters and multiple convergence fouling parameter clusters.

[0084] Finally, the mean reverse osmosis fitness of multiple convergent blocking operating parameter clusters and multiple convergent scaling operating parameter clusters is calculated, and the similarity is calculated to obtain the optimization balance. Here, the mean reverse osmosis fitness is the average of the reverse osmosis fitness of all individual parameters within a convergent cluster; the similarity of the mean reverse osmosis fitness is the degree of similarity in the reverse osmosis fitness distribution characteristics between different convergent clusters.

[0085] Specifically, firstly, the mean reverse osmosis fitness of the convergence blockage operating parameter cluster and the convergence scaling operating parameter cluster are calculated, and the mean reflects the overall level of the cluster. Then, the similarity between the mean reverse osmosis fitness of the convergence blockage operating parameter cluster and the convergence scaling operating parameter cluster is calculated, and this is used as the optimization balance degree. Optimization balance degree = 1 - (|mean reverse osmosis fitness of convergence blockage operating parameter cluster - mean reverse osmosis fitness of convergence scaling operating parameter cluster| / mean of both). A higher optimization balance degree indicates a relatively balanced approach to preventing blockage and scaling, allowing for greater tolerance of fluctuations in subsequent control.

[0086] For example, suppose the mean reverse osmosis fitness values ​​of a certain convergence blockage operating parameter cluster A and its corresponding convergence scaling operating parameter cluster B are 1.4 and 1.45 respectively, and the similarity = 1 - (|1.4 - 1.45| / 1.425) ≈ 0.965, then the optimization balance is ≈ 0.965. Similarly, the similarity is calculated for the mean reverse osmosis fitness values ​​of all convergence blockage operating parameter clusters and their corresponding convergence scaling operating parameter clusters to obtain the optimization balance.

[0087] In step S20 of the method provided in this application embodiment, an optimized operating parameter set is obtained by screening based on the optimization balance, and the optimal reverse osmosis operating parameters are obtained by further screening, including: Obtain the operating parameter cluster with the largest average reverse osmosis fitness as the optimized operating parameter set; Using the optimization balance as the selection ratio, the reverse osmosis operating parameters with the largest selection ratio of reverse osmosis fitness within the optimized operating parameter set are selected to construct reverse osmosis operating parameter constraints. The reverse osmosis operating parameters with the highest reverse osmosis adaptability are extracted as the optimal reverse osmosis operating parameters.

[0088] In this embodiment, the operating parameter cluster with the largest average reverse osmosis fitness is first obtained as the optimized operating parameter set. The optimized operating parameter set is the set of all reverse osmosis operating parameters contained in the convergent cluster with the highest similarity. Specifically, after calculating the average reverse osmosis fitness, the operating parameter cluster corresponding to the reverse osmosis fitness with the largest value is taken as the optimized operating parameter set, and all operating parameters in the cluster are taken as optimized operating parameters.

[0089] Secondly, using the optimization balance as the selection ratio, the reverse osmosis operating parameters with the highest selection ratio within the optimized operating parameter set are chosen to construct reverse osmosis operating parameter constraints. Here, the selection ratio is a proportional coefficient used for further filtering from the optimized operating parameter set. Specifically, using the optimization balance as the selection ratio, the optimized operating parameters in the optimized operating parameter set are arranged in descending order. Then, according to the selection ratio, a corresponding number of optimized operating parameters are selected from the arranged optimized operating parameters, and their corresponding reverse osmosis operating parameters are used to construct reverse osmosis operating parameter constraints.

[0090] For example, if the optimization balance is 0.9 and the optimized operating parameter set includes 10 reverse osmosis operating parameters, then the top 9 reverse osmosis operating parameters with the highest reverse osmosis fitness are selected, and reverse osmosis operating parameter constraints are constructed based on the maximum and minimum values ​​among the 9 reverse osmosis operating parameters.

[0091] Finally, the reverse osmosis operating parameter with the highest fitness is extracted as the optimal reverse osmosis operating parameter. Specifically, after obtaining the reverse osmosis operating parameter constraints, each parameter in the reverse osmosis operating parameter constraint set is traversed, its corresponding first reverse osmosis fitness value is read, the magnitude of all first reverse osmosis fitness values ​​is compared, and the reverse osmosis operating parameter with the highest first reverse osmosis fitness value is taken as the optimal reverse osmosis operating parameter.

[0092] For example, the reverse osmosis operating parameters in the reverse osmosis operating parameter constraints are compared numerically. Among the nine reverse osmosis operating parameters, the maximum reverse osmosis fitness is 0.7. The reverse osmosis operating parameter corresponding to 0.7 is taken as the optimal reverse osmosis operating parameter.

[0093] In this embodiment, firstly, by defining a parameter space, generating an initial population, and combining the characteristics of pretreated brine, a fault predictor and a desalination predictor are used to evaluate the reverse osmosis operating parameters, quantifying the complex operational optimization. Secondly, fitness calculation provides a clear data foundation for subsequent steps. Then, through optimized cluster partitioning and iterative evolution, bidirectional optimization is performed from the directions of reducing blockage and scaling, and a global search is conducted to effectively avoid getting trapped in local optima and ensure sufficient exploration of the solution space. Finally, the similarity of the average reverse osmosis fitness values ​​of the convergent blockage operating parameter cluster and the convergent scaling operating parameter cluster is used as the optimization equilibrium degree. Reverse osmosis operating parameters are dynamically selected based on the optimization equilibrium degree. Finally, by comparing the reverse osmosis fitness of the operating parameters, the reverse osmosis operating parameter with the highest fitness is selected as the optimal reverse osmosis operating parameter, achieving an optimal balance between energy consumption and cost while meeting rigid permeable water quality requirements.

[0094] S30: Uses optimal reverse osmosis operating parameters to control the production of the reverse osmosis membrane water production line.

[0095] In this embodiment of the application, production control is a process in which the execution equipment on the production line is adjusted in real time through a programmable logic controller and a distributed control system in an industrial automation system, so that the key parameters of the production process are stabilized near the target value.

[0096] Specifically, the optimal reverse osmosis operating parameters are first transmitted to the production line's control system via an industrial communication network, serving as the new control setpoint. The control system then drives real-time production control, enabling the entire reverse osmosis membrane water production line to operate in the optimized mode.

[0097] In this embodiment, the optimal parameter settings output by the optimization algorithm are directly sent to the production line control system via an industrial automation interface. This drives actuators such as high-pressure pumps and regulating valves to make precise adjustments, ensuring that the reverse osmosis system operates at the theoretically calculated optimal operating point in real time. This automation of production control ensures that the entire production line can operate continuously and stably at the highest overall efficiency, thereby significantly improving production efficiency and reducing operating costs.

[0098] The embodiments of this application, through the specific implementation methods described above, achieve the following technical effects: In this embodiment, the specific composition of brine characteristics is first obtained. Then, a machine learning-based preprocessing predictor is used to establish the ability to predict the water quality from raw water to the reverse osmosis membrane feed water, thereby achieving precise optimization input. The brine characteristics are concretized into salt concentration and impurity characteristics, providing standardized input dimensions for the evaluation of all subsequent models and ensuring the objectivity and comprehensiveness of the optimization basis.

[0099] Secondly, by defining a parameter space, generating an initial population, and combining the characteristics of pretreated brine, a fault predictor and a desalination predictor are used to evaluate the reverse osmosis operating parameters, quantifying the complex operational optimization. Next, fitness calculation provides a clear data foundation for subsequent steps. Furthermore, through optimized cluster partitioning and iterative evolution, bidirectional optimization is performed from the directions of reducing blockage and scaling, and a global search is conducted to effectively avoid getting trapped in local optima and ensure a thorough exploration of the solution space. Finally, the similarity of the mean reverse osmosis fitness values ​​of the convergent blockage operating parameter cluster and the convergent scaling operating parameter cluster is used as the optimization equilibrium degree. Based on the optimization equilibrium degree, reverse osmosis operating parameters are dynamically selected. Finally, by comparing the reverse osmosis fitness of the operating parameters, the reverse osmosis operating parameter with the highest fitness is selected as the optimal reverse osmosis operating parameter, achieving an optimal balance between energy consumption and cost while meeting rigid permeable water quality requirements.

[0100] Ultimately, the optimal parameter settings output by the optimization algorithm are directly sent to the production line control system via the industrial automation interface. This drives actuators such as high-pressure pumps and regulating valves to make precise adjustments, ensuring that the reverse osmosis system operates at the theoretically calculated optimal operating point in real time. This automation of production control ensures that the entire production line can operate continuously and stably at the highest overall efficiency, thereby significantly improving production efficiency and reducing operating costs.

[0101] Example 2, as Figure 2 As shown, based on the same inventive concept as the method for optimizing operating parameters of a reverse osmosis membrane water production line provided in Embodiment 1, this embodiment of the invention also provides an operating parameter optimization system for a reverse osmosis membrane water production line, including: Preprocessing prediction module 11 is used to obtain the saline characteristics of the saline, perform preprocessing prediction, and obtain the preprocessed saline characteristics. The operating parameter optimization module 12 is used to obtain desalination rate requirement information, combine it with the characteristics of pretreated brine, optimize the reverse osmosis operating parameters, and obtain the optimal reverse osmosis operating parameters. Specifically, based on the characteristics of pretreated brine and desalination rate requirement information, reverse osmosis membrane blockage prediction, scaling prediction, and desalination rate prediction are performed for the reverse osmosis operating parameters. The reverse osmosis fitness is calculated, and optimization cluster rules are set for optimization. Production control module 13 is used to control the production of the reverse osmosis membrane water production line using optimal reverse osmosis operating parameters.

[0102] In one embodiment, the preprocessing prediction module 11 is used for: The characteristics of the brine are obtained, including salt concentration and impurity characteristics. The saline features are input into the preprocessing predictor, and the output is the preprocessed saline features.

[0103] The training steps for the preprocessor predictor include: Collect the feature set of sample saline solution, and collect the feature set of pretreated saline solution after different sample saline solution feature preprocessing. Based on machine learning, a preprocessing predictor is constructed. The preprocessed predictor is trained and tested using a sample saline feature set and a sample preprocessed saline feature set. The network parameters are iteratively optimized until the test converges, thus completing the training and testing process.

[0104] In one embodiment, the runtime parameter optimization module 12 is used for: Obtain desalination rate requirements; Obtain the reverse osmosis operating parameter space; Multiple first reverse osmosis operating parameters are generated within the reverse osmosis operating parameter space, wherein each first reverse osmosis operating parameter includes a pressure parameter; Based on the characteristics of the pretreated brine, pretreatment blockage prediction and permeation scaling prediction are performed by combining multiple first reverse osmosis operating parameters to obtain multiple first blockage rates and multiple first scaling rates, as well as multiple first energy cost information and multiple first desalination rates. Based on multiple first blockage rates, multiple first scaling rates, multiple first energy cost information, and multiple first desalination rates, multiple first reverse osmosis fitness rates are calculated. Based on multiple first blockage rates and multiple first fouling rates, optimization clusters are divided to obtain multiple blockage operation parameter clusters and multiple fouling operation parameter clusters. Iterative optimization is continued until convergence, resulting in multiple converged blockage operation parameter clusters and multiple converged fouling operation parameter clusters. The optimization balance is then analyzed. Based on the optimization balance, the optimized operating parameter set is obtained through screening, and the optimal reverse osmosis operating parameters are then selected.

[0105] Specifically, based on the characteristics of the pretreated brine, pretreatment blockage prediction and permeation scaling prediction are performed by combining multiple first reverse osmosis operating parameters, resulting in multiple first blockage rates and multiple first scaling rates, including: A reverse osmosis membrane failure predictor is obtained, which includes a blockage prediction branch and a scaling prediction branch. The inputs during the training process of the blockage prediction branch are the sample pretreated brine characteristics and the sample reverse osmosis operating parameters, and the label is the sample blockage rate of the reverse osmosis membrane blocked by impurities. The inputs during the training process of the scaling prediction branch are the sample pretreated brine characteristics and the sample reverse osmosis operating parameters, and the label is the sample scaling rate of salt scaling on the reverse osmosis membrane. The pretreated brine characteristics are combined with multiple first reverse osmosis operating parameters and input into the reverse osmosis membrane fault predictor to obtain multiple first blockage rates and multiple first scaling rates.

[0106] This includes obtaining multiple primary energy cost information and multiple primary desalination rates, including: Based on the energy consumption and pressure correspondence of the reverse osmosis membrane water production line, multiple first energy cost information of multiple first reverse osmosis operating parameters is obtained; The pretreated brine features and multiple first reverse osmosis operating parameters are input into the desalination predictor, and multiple first desalination rates are output. The desalination predictor is built based on machine learning, and the input features for training are the sample pretreated brine features and sample reverse osmosis operating parameters, and the label is the sample desalination rate.

[0107] Among them, based on multiple first blockage rates, multiple first scaling rates, multiple first energy cost information, and multiple first desalination rates, multiple first reverse osmosis fitness values ​​are calculated, including: Multiple first fusion failure rates are calculated based on multiple first blockage rates and multiple first fouling rates; Based on multiple first desalination rates and desalination rate requirements, multiple first desalination coefficients are calculated. Based on multiple primary energy cost information and benchmark energy cost information, multiple primary energy coefficients are calculated and obtained; Multiple first reverse osmosis fitness values ​​are calculated based on multiple first fusion failure rates, multiple first desalination coefficients, and multiple first energy coefficients. Among them, the fusion failure rate and energy coefficient are negatively correlated with the reverse osmosis fitness value, while the desalination coefficient is positively correlated with the reverse osmosis fitness value.

[0108] Specifically, based on multiple first blockage rates and multiple first fouling rates, optimization clusters are divided to obtain multiple blockage operation parameter clusters and multiple fouling operation parameter clusters. Iterative optimization continues until convergence, resulting in multiple convergent blockage operation parameter clusters and multiple convergent fouling operation parameter clusters. The optimization balance is then analyzed, including: For multiple first blocking rates and multiple first fouling rates, clustering is performed to obtain multiple first blocking rate clusters and multiple first fouling rate clusters. Each of the multiple first blocking rate clusters includes the smallest first blocking rate and multiple other following first blocking rates, and each of the multiple first fouling rate clusters includes the smallest first fouling rate and multiple other following first fouling rates. Based on multiple first blockage rate clusters and multiple first scaling rate clusters, multiple first reverse osmosis operating parameters are divided to obtain multiple blockage operating parameter clusters and multiple scaling operating parameter clusters. The operating parameters corresponding to the first blocking rate and the first scaling rate in each operating parameter cluster are used as the adjustment direction. The operating parameters corresponding to the first blocking rate and the first scaling rate are adjusted to update the multiple blocking operating parameter clusters and the multiple scaling operating parameter clusters. The blocking rate and scaling rate are predicted and the adjustment direction is updated. Continue iterative optimization until the convergence iteration count is reached, and obtain the final multiple convergence blocking operation parameter clusters and multiple convergence fouling operation parameter clusters; Calculate the average reverse osmosis fitness of multiple clusters of convergent blocking operating parameters and multiple clusters of convergent scaling operating parameters, and calculate the similarity to obtain the optimized balance.

[0109] Specifically, based on the optimization balance, an optimized set of operating parameters is obtained through screening, and the optimal reverse osmosis operating parameters are then selected, including: Obtain the operating parameter cluster with the largest average reverse osmosis fitness as the optimized operating parameter set; Using the optimization balance as the selection ratio, the reverse osmosis operating parameters with the largest selection ratio of reverse osmosis fitness within the optimized operating parameter set are selected to construct reverse osmosis operating parameter constraints. The reverse osmosis operating parameters with the highest reverse osmosis adaptability are extracted as the optimal reverse osmosis operating parameters.

[0110] The embodiments of this application, through the specific implementation methods described above, achieve the following technical effects: In this embodiment, the specific composition of brine characteristics is first obtained through the pretreatment prediction module 11. Then, based on a machine learning-based pretreatment predictor, the ability to predict the water quality from raw water to the reverse osmosis membrane feed water is established, thereby achieving precise optimization input. The brine characteristics are concretized into salt concentration and impurity characteristics, providing standardized input dimensions for the evaluation of all subsequent models and ensuring the objectivity and comprehensiveness of the optimization basis.

[0111] Secondly, through the operating parameter optimization module 12, a parameter space is defined, an initial population is generated, and pretreated brine characteristics are combined. Fault predictors and desalination predictors are used to evaluate the reverse osmosis operating parameters, quantifying the complex operational optimization. Secondly, fitness calculation provides a clear data foundation for subsequent steps. Furthermore, through optimized cluster partitioning and iterative evolution, bidirectional optimization is performed from the directions of reducing blockage and scaling, and a global search is conducted to effectively avoid getting trapped in local optima and ensure sufficient exploration of the solution space. Finally, the similarity of the average reverse osmosis fitness values ​​of the convergent blockage operating parameter cluster and the convergent scaling operating parameter cluster is used as the optimization equilibrium degree. Reverse osmosis operating parameters are dynamically selected based on the optimization equilibrium degree. Finally, by comparing the reverse osmosis fitness of the operating parameters, the reverse osmosis operating parameter with the highest fitness is selected as the optimal reverse osmosis operating parameter, achieving an optimal balance between energy consumption and cost while meeting rigid permeable water quality requirements.

[0112] Finally, through the production control module 13, the optimal parameter settings output by the optimization algorithm are directly sent to the production line control system via the industrial automation interface. This drives actuators such as high-pressure pumps and regulating valves to make precise adjustments, ensuring that the reverse osmosis system operates at the theoretically calculated optimal operating point in real time. This automation of production control ensures that the entire production line can operate continuously and stably at the highest overall efficiency, thereby significantly improving production efficiency and reducing operating costs.

Claims

1. A method for optimizing operating parameters of a reverse osmosis membrane water production line, characterized in that, The method includes: The characteristics of the brine are obtained, and preprocessing and prediction are performed to obtain the characteristics of the preprocessed brine. Obtain the desalination rate requirement information, combine it with the characteristics of the pretreated brine, optimize the reverse osmosis operating parameters, and obtain the optimal reverse osmosis operating parameters. Specifically, based on the characteristics of the pretreated brine and the desalination rate requirement information, predict the reverse osmosis membrane blockage, scaling, and desalination rate of the reverse osmosis operating parameters, calculate the reverse osmosis fitness, and set optimization cluster rules for optimization. The optimal reverse osmosis operating parameters are used to control the production of the reverse osmosis membrane water production line.

2. The method for optimizing operating parameters of a reverse osmosis membrane water production line according to claim 1, characterized in that, Obtain the saline characteristics, perform preprocessing prediction, and obtain the preprocessed saline characteristics, including: The characteristics of the brine are obtained, including salt concentration and impurity characteristics. The brine features are input into the preprocessing predictor, and the output is the preprocessed brine features.

3. The method for optimizing operating parameters of a reverse osmosis membrane water production line according to claim 2, characterized in that, The training steps for the preprocessor predictor include: Collect the feature set of sample saline solution, and collect the feature set of pretreated saline solution after different sample saline solution feature preprocessing. Based on machine learning, a preprocessing predictor is constructed. The preprocessed predictor is trained and tested using the sample saline feature set and the sample preprocessed saline feature set, and the network parameters are iteratively optimized until the test converges, thus completing the training and testing.

4. The method for optimizing operating parameters of a reverse osmosis membrane water production line according to claim 1, characterized in that, Obtain the required desalination rate information, and combine it with the characteristics of the pretreated brine to optimize the reverse osmosis operating parameters, thereby obtaining the optimal reverse osmosis operating parameters, including: Obtain desalination rate requirements; Obtain the reverse osmosis operating parameter space; Multiple first reverse osmosis operating parameters are generated within the reverse osmosis operating parameter space, wherein each first reverse osmosis operating parameter includes a pressure parameter; Based on the characteristics of the pretreated brine, pretreatment blockage prediction and permeation scaling prediction are performed by combining the multiple first reverse osmosis operating parameters respectively, to obtain multiple first blockage rates and multiple first scaling rates, and to obtain multiple first energy cost information and multiple first desalination rates. Based on multiple first blockage rates, multiple first scaling rates, multiple first energy cost information, and multiple first desalination rates, multiple first reverse osmosis fitness rates are calculated. Based on multiple first blockage rates and multiple first fouling rates, optimization clusters are divided to obtain multiple blockage operation parameter clusters and multiple fouling operation parameter clusters. Iterative optimization is continued until convergence, resulting in multiple converged blockage operation parameter clusters and multiple converged fouling operation parameter clusters. The optimization balance is then analyzed. Based on the optimized balance, an optimized set of operating parameters is obtained through screening, and the optimal reverse osmosis operating parameters are then selected.

5. The method for optimizing operating parameters of a reverse osmosis membrane water production line according to claim 4, characterized in that, Based on the characteristics of the pretreated brine, pretreatment blockage prediction and permeation scaling prediction are performed in conjunction with the multiple first reverse osmosis operating parameters to obtain multiple first blockage rates and multiple first scaling rates, including: A reverse osmosis membrane failure predictor is obtained, wherein the reverse osmosis membrane failure predictor includes a blockage prediction branch and a scaling prediction branch. The inputs during the training process of the blockage prediction branch are the characteristics of the sample pretreated brine and the sample reverse osmosis operating parameters, and the label is the sample blockage rate of the reverse osmosis membrane blocked by impurities. The inputs during the training process of the scaling prediction branch are the characteristics of the sample pretreated brine and the sample reverse osmosis operating parameters, and the label is the sample scaling rate of salt scaling on the reverse osmosis membrane. The characteristics of the pretreated brine are combined with the multiple first reverse osmosis operating parameters and input into the reverse osmosis membrane fault predictor to obtain multiple first blockage rates and multiple first scaling rates.

6. The method for optimizing operating parameters of a reverse osmosis membrane water production line according to claim 4, characterized in that, Obtain multiple primary energy cost information and multiple primary desalination rates, including: Based on the energy consumption and pressure correspondence of the reverse osmosis membrane water production line, multiple first energy cost information of multiple first reverse osmosis operating parameters is obtained; The pretreated brine characteristics and multiple first reverse osmosis operating parameters are input into the desalination predictor, and multiple first desalination rates are output. The desalination predictor is built based on machine learning, and the input features for training are the sample pretreated brine characteristics and sample reverse osmosis operating parameters, and the label is the sample desalination rate.

7. The method for optimizing operating parameters of a reverse osmosis membrane water production line according to claim 4, characterized in that, Based on multiple first blockage rates, multiple first scaling rates, multiple first energy cost information, and multiple first desalination rates, multiple first reverse osmosis fitness values ​​are calculated, including: Multiple first fusion failure rates are calculated based on multiple first blockage rates and multiple first fouling rates; Based on multiple first desalination rates and desalination rate requirements, multiple first desalination coefficients are calculated. Multiple primary energy cost information and benchmark energy cost information are used to calculate multiple primary energy coefficients; Multiple first reverse osmosis fitness values ​​are calculated based on multiple first fusion failure rates, multiple first desalination coefficients, and multiple first energy coefficients. Among them, the fusion failure rate and energy coefficient are negatively correlated with the reverse osmosis fitness value, while the desalination coefficient is positively correlated with the reverse osmosis fitness value.

8. The method for optimizing operating parameters of a reverse osmosis membrane water production line according to claim 4, characterized in that, Based on multiple first-level blocking rates and multiple first-level scaling rates, optimization clusters are divided to obtain multiple blocking operation parameter clusters and multiple scaling operation parameter clusters. Iterative optimization continues until convergence, resulting in multiple convergent blocking operation parameter clusters and multiple convergent scaling operation parameter clusters. The optimization balance is then analyzed, including: For multiple first blocking rates and multiple first fouling rates, clustering is performed to obtain multiple first blocking rate clusters and multiple first fouling rate clusters. Each of the multiple first blocking rate clusters includes the smallest first blocking rate and multiple other following first blocking rates, and each of the multiple first fouling rate clusters includes the smallest first fouling rate and multiple other following first fouling rates. According to the multiple first blockage rate clusters and multiple first scaling rate clusters, the multiple first reverse osmosis operating parameters are divided to obtain multiple blockage operating parameter clusters and multiple scaling operating parameter clusters. The operating parameters corresponding to the first blocking rate and the first scaling rate in each operating parameter cluster are used as the adjustment direction. The operating parameters corresponding to the first blocking rate and the first scaling rate are adjusted to update the multiple blocking operating parameter clusters and the multiple scaling operating parameter clusters. The blocking rate and scaling rate are predicted and the adjustment direction is updated. Continue iterative optimization until the convergence iteration count is reached, and obtain the final multiple convergence blocking operation parameter clusters and multiple convergence fouling operation parameter clusters; Calculate the average reverse osmosis fitness of multiple clusters of convergent blocking operating parameters and multiple clusters of convergent scaling operating parameters, and calculate the similarity to obtain the optimized balance.

9. The method for optimizing operating parameters of a reverse osmosis membrane water production line according to claim 4, characterized in that, Based on the optimized balance, an optimized set of operating parameters is obtained through screening, and the optimal reverse osmosis operating parameters are further selected, including: Obtain the operating parameter cluster with the largest average reverse osmosis fitness as the optimized operating parameter set; Using the optimization balance as the selection ratio, the reverse osmosis operating parameters with the largest reverse osmosis fitness within the optimized operating parameter set are selected to construct reverse osmosis operating parameter constraints. The reverse osmosis operating parameters with the highest reverse osmosis adaptability are extracted as the optimal reverse osmosis operating parameters.

10. An operating parameter optimization system for a reverse osmosis membrane water production line, characterized in that, The system is used to implement the method for optimizing operating parameters of a reverse osmosis membrane water production line according to any one of claims 1-9, the system comprising: The preprocessing prediction module is used to obtain the saline characteristics of the saline, perform preprocessing prediction, and obtain the preprocessed saline characteristics. The operating parameter optimization module is used to obtain desalination rate requirement information, combine it with the characteristics of the pretreated brine, optimize the reverse osmosis operating parameters, and obtain the optimal reverse osmosis operating parameters. Specifically, based on the characteristics of the pretreated brine and the desalination rate requirement information, the reverse osmosis operating parameters are predicted for reverse osmosis membrane blockage, scaling, and desalination rate. The reverse osmosis fitness is calculated, and optimization cluster rules are set for optimization. The production control module is used to control the production of the reverse osmosis membrane water production line using the optimal reverse osmosis operating parameters.

Images