Digital twin prediction driven circulation cooling water electrochemical intelligent regulation system and method

Abstract

This invention relates to the field of intelligent predictive control technology, and more particularly to a digital twin predictive-driven electrochemical intelligent control system and method for circulating cooling water, comprising a physical operation device and a control device. The control device collects water quality monitoring parameters through a data acquisition module, and a state assessment module uses the entropy weight method to map and calculate a comprehensive index. A trend inference module inputs the comprehensive index into a long short-term memory network model for forward propagation, outputting a predicted water quality index, which is then iteratively optimized by a hyperparameter optimization unit using a sparrow search algorithm to optimize the network connection parameters. An instruction optimization module compares the predicted water quality index with a safety threshold to generate control instructions, driving the physical operation device to perform electric field adjustment, mechanical scaling, and coordinated water replenishment and chemical dosing. A virtual mapping module constructs a three-dimensional digital model to achieve state early warning. This invention solves the technical problem of lag in existing feedback control, achieving precise pre-control of water quality in circulating cooling water systems.

Description

Technical Field

[0001] This invention relates to the field of intelligent predictive control technology, and in particular to a digital twin predictive-driven electrochemical intelligent control system and method for circulating cooling water. Background Technology

[0002] Industrial circulating cooling water systems are critical auxiliary systems in process industries such as power, chemical, and steel, and their stable operation directly affects the safety and energy efficiency of the main production units. The core objective of water quality control in these systems is to prevent scaling and corrosion of heat exchange equipment. Existing water quality control methods generally rely on the addition of chemical reagents, supplemented by timed blowdowns. The control of these chemical reagent addition and timed blowdown methods is primarily based on real-time monitoring of water quality parameters.

[0003] In existing technologies, electrochemical methods have emerged to replace or partially replace chemical dosing. Some improved solutions have introduced automated control, such as automatically adjusting the operating voltage or electrode spacing of the electrochemical device based on real-time monitored water hardness parameters. However, these control modes are essentially still feedback control based on real-time monitoring.

[0004] Real-time monitoring-based control models have inherent limitations. Industrial circulating cooling water systems operate under complex and variable conditions. Water quality parameters are not only time-varying, but also exhibit strong coupling relationships among multiple key parameters such as hardness, alkalinity, pH, and conductivity. Adjusting solely based on the current measured value of a single parameter fails to reflect the dynamic trends of the overall water quality, and cannot predict impending scaling or corrosion. Therefore, control actions always lag behind the actual changes in water quality, constituting a delayed adjustment. Furthermore, single-parameter adjustment strategies neglect the synergistic effects between multiple parameters, making precise control difficult under complex conditions.

[0005] In summary, existing lag control modes based on real-time monitoring are insufficient to effectively address the time-varying and strong coupling of water quality parameters, resulting in unpredictable scaling and corrosion risks in industrial circulating cooling water systems, which in turn affect heat exchange efficiency and reduce the safety of long-term equipment operation. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a digital twin-based predictive-driven electrochemical intelligent control system and method for circulating cooling water. This invention aims to solve the technical problem that existing control modes based on real-time monitoring have response lags and are difficult to avoid scaling and corrosion risks caused by the time-varying and strong coupling of water quality parameters under the operating conditions of large-scale industrial circulating cooling water systems.

[0007] To solve the above-mentioned technical problems, the specific contents of the present invention are as follows: In a first aspect, the digital twin predictive-driven circulating cooling water electrochemical intelligent control system provided by the present invention includes a physical operating device and a control device that establishes a communication connection with the physical operating device. The physical operation device includes a reaction chamber defining the main reaction zone, an inlet and an outlet connecting the main reaction zone, electrochemical cathode reaction groups and electrochemical anode reaction groups interleaved within the main reaction zone, a main shaft motor arranged outside the reaction chamber, a rotary scraper assembly arranged within the main reaction zone and close to the surface of the electrochemical cathode reaction group, a coupling connecting the main shaft motor and the rotary scraper assembly, a scale removal component arranged at the bottom of the reaction chamber and connecting the main reaction zone, a power supply connected to the electrochemical cathode reaction group and the electrochemical anode reaction group, a measuring component, a dosing component, and a water replenishment valve arranged at the inlet and outlet. The control device includes: The data acquisition module is used to collect circulating cooling water quality monitoring parameters and output the circulating cooling water quality monitoring parameters. The status assessment module receives circulating cooling water quality monitoring parameters, performs mapping calculations on the circulating cooling water quality monitoring parameters according to a preset distribution characteristic mapping formula, generates a comprehensive index, and outputs the comprehensive index. The trend projection module receives the comprehensive index, inputs the comprehensive index into a preset neural network model for forward propagation calculation, and outputs the predicted water quality index. The instruction optimization module receives the predicted water quality index, compares the predicted water quality index with the preset safety threshold, generates control instructions, and sends the control instructions to the spindle motor, scale removal components, power supply, chemical dosing components, and water supply valve.

[0008] Furthermore, the digital twin predictive-driven electrochemical intelligent control system for circulating cooling water of the present invention includes a data acquisition module comprising: The data reading submodule is used to receive physical environment measurement values ​​sent by the measurement component, extract the acidity, conductivity, calcium hardness concentration and total alkalinity concentration from the physical environment measurement values, and output the acidity, conductivity, calcium hardness concentration and total alkalinity concentration. The data cleaning submodule receives pH, conductivity, calcium hardness concentration, and total alkalinity concentration, filters out values ​​that exceed the preset range, performs dimensionless mapping processing, generates standardized water quality parameters, and sends the standardized water quality parameters to the status assessment module.

[0009] Furthermore, the digital twin predictive-driven electrochemical intelligent control system for circulating cooling water of the present invention includes a state assessment module comprising: The weight allocation submodule receives standardized water quality parameters, calculates the information entropy corresponding to the standardized water quality parameters according to the entropy weight method, and outputs the influence weight values ​​based on the information entropy. The index normalization submodule receives standardized water quality parameters and their corresponding influence weights, multiplies the standardized water quality parameters by their corresponding influence weights and sums them to calculate the comprehensive index, and then sends the comprehensive index to the trend inference module.

[0010] Furthermore, the trend prediction module of the digital twin predictive-driven circulating cooling water electrochemical intelligent control system of the present invention includes: The feature filtering submodule is used to receive the comprehensive index, extract the comprehensive index according to the preset timestamp, generate time series dimension variables, and output the time series dimension variables. The network inference submodule is used to receive time series dimension variables, input the time series dimension variables into a preset long short-term memory network model for forward propagation calculation, and output the predicted water quality index. The long short-term memory network model includes node connection weights. The predicted water quality index is sent to the instruction optimization module.

[0011] Furthermore, the digital twin predictive-driven electrochemical intelligent control system for circulating cooling water of the present invention further includes a hyperparameter optimization unit connected to the network inference submodule: The hyperparameter optimization unit receives the predicted water quality index and the comprehensive index output by the state assessment module as the actual water quality index. It calculates the error value between the predicted water quality index and the actual water quality index. When the error value is greater than the preset calibration error threshold, it iteratively calculates the network connection parameters according to the preset sparrow search algorithm, inputs the network connection parameters into the network inference submodule, and replaces the node connection weights inside the long short-term memory network model.

[0012] Furthermore, the digital twin predictive-driven electrochemical intelligent control system for circulating cooling water of the present invention includes an instruction optimization module comprising: The status comparison submodule is used to receive the predicted water quality index, calculate the difference between the predicted water quality index and the preset safety threshold, generate the deviation value, and output the deviation value. The power supply regulation submodule is used to receive deviation values. When the deviation value exceeds the preset deviation limit, the deviation value is input into the preset proportional-integral-differential calculation model to calculate the compensation current density. Based on the compensation current density, a power supply adjustment command is generated as a control instruction and sent to the power supply.

[0013] Furthermore, the instruction optimization module of the digital twin predictive-driven circulating cooling water electrochemical intelligent control system of the present invention further includes: The scraping drive submodule is used to receive the compensation current density. When the compensation current density is greater than the preset first calibration threshold, it generates a scraping operation command as a control command and sends the scraping operation command to the spindle motor. The rotating scraper assembly is driven to rotate along the surface of the electrochemical cathode reaction group through the coupling component. The sediment discharge submodule is used to start a preset timer to record the duration of the spindle motor in operation and generate an operation duration value. When the operation duration value is equal to or greater than a preset time threshold, a discharge command is generated as a control command and sent to the scale removal component.

[0014] Furthermore, the instruction optimization module of the digital twin predictive-driven circulating cooling water electrochemical intelligent control system of the present invention further includes: The linkage compensation submodule is used to receive the predicted water quality index and the compensation current density, calculate the slope of the predicted water quality index over time, and when the slope is greater than the preset slope threshold and the compensation current density is equal to the rated maximum value of the power supply, generate a chemical dosing command and a valve opening control command, send the chemical dosing command to the dosing component, and send the valve opening control command to the water supply valve.

[0015] Furthermore, the digital twin predictive-driven electrochemical intelligent control system for circulating cooling water of the present invention further includes the following control device: The virtual mapping module receives circulating cooling water quality monitoring parameters and control commands, establishes a three-dimensional digital model that maps to the physical operating device structure, inputs the circulating cooling water quality monitoring parameters and control commands into the three-dimensional digital model, triggers the update calculation of the state variables inside the three-dimensional digital model, extracts the sediment thickness value output by the three-dimensional digital model, and generates and outputs sediment accumulation warning information when the sediment thickness value is greater than the preset thickness limit value.

[0016] Secondly, the digital twin prediction-driven electrochemical intelligent control method for circulating cooling water provided by the present invention is applied to the digital twin prediction-driven electrochemical intelligent control system for circulating cooling water as described in any one of the claims, comprising: Step 1: Collect circulating cooling water quality monitoring parameters and output the circulating cooling water quality monitoring parameters; Step 2: Receive circulating cooling water quality monitoring parameters, perform mapping calculations on the circulating cooling water quality monitoring parameters according to the preset distribution characteristic mapping formula, generate a comprehensive index, and output the comprehensive index; Step 3: Receive the comprehensive index, input the comprehensive index into the preset neural network model for forward propagation calculation, and output the predicted water quality index; Step 4: Receive the predicted water quality index, compare the predicted water quality index with the preset safety threshold, generate control commands, and send the control commands to the spindle motor, scale removal components, power supply, chemical dosing components, and water supply valve.

[0017] Beneficial effects of this invention: This invention provides a digital twin-based predictive-driven electrochemical intelligent control system and method for circulating cooling water. The data acquisition module collects pH, conductivity, calcium hardness concentration, and total alkalinity concentration, generating standardized water quality parameters. The state assessment module performs mapping calculations on the standardized water quality parameters using the entropy weight method to derive a comprehensive index. This comprehensive index integrates multi-dimensional physicochemical characteristics of water quality, overcoming the objective defect of single-parameter adjustment strategies ignoring the strong coupling effects between multiple parameters. The trend inference module inputs the comprehensive index into a long short-term memory network model for forward propagation to output a predicted water quality index. This predicted water quality index reflects the future dynamic trend of the overall water quality state, solving the response lag problem commonly found in feedback control based on real-time monitoring. The hyperparameter optimization unit within the control device iteratively calculates network connection parameters using a sparrow search algorithm, replacing the original node connection weights with mapping biases within the long short-term memory network model, improving the computational accuracy of the time-series variable network inference calculation. The instruction optimization module compares the predicted water quality index with a preset safety threshold to generate control instructions, and inputs the deviation value converted from the difference into a proportional-integral-differential operation model to calculate the compensation current density. The power supply adjusts the electric field strength within the main reaction zone according to control commands. The main shaft motor drives the rotary scraper assembly to rotate along the surface of the electrochemical cathode reaction group via a coupling to continuously peel off the solid scale layer. The scale removal component performs sediment discharge actions at regular intervals according to control commands. The linkage compensation submodule sends a chemical dosing command to the dosing component when the compensation current density equals the rated maximum value of the power supply, and simultaneously sends a valve opening control command to the water supply valve. This enables the electric field power supply adjustment, mechanical cutting descaling, physical sewage discharge, and chemical intervention to work together to form a multi-dimensional collaborative execution mechanism, effectively mitigating the risks of scaling and corrosion faced by the system under complex and variable operating conditions. The virtual mapping module simultaneously establishes a three-dimensional digital model and extracts the sediment thickness value. When the sediment thickness value exceeds the preset thickness limit value, it outputs sediment accumulation warning information, eliminating the objective hidden danger of water flow channel blockage inside the physical operation device in advance. Attached Figure Description

[0018] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on the drawings without creative effort.

[0019] Figure 1 This is a schematic diagram of the overall three-dimensional structure of the physical operation device provided in an embodiment of the present invention.

[0020] Figure 2 This is a schematic diagram of the internal cross-sectional main view structure of the physical operation device provided in an embodiment of the present invention.

[0021] Figure 3This is a system architecture diagram of the digital twin prediction-driven electrochemical intelligent control system for circulating cooling water according to the present invention.

[0022] Explanation of reference numerals in the attached diagram: 1-Inlet, 2-Main reaction zone, 3-Outlet, 4-Main shaft motor, 5-Coupling assembly, 6-Rotary scraper assembly, 7-Electrochemical cathode reaction group, 8-Electrochemical anode reaction group, 9-Scale removal component. Detailed Implementation

[0023] To make the technical solution of the present invention clearer, the present invention will be clearly and completely described below with reference to specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. The various embodiments of the present invention will be described in detail below with reference to the accompanying drawings. To better understand the purpose of the present invention, the present invention will be described in further detail below.

[0024] Please see Figures 1 to 3 In a first aspect, the digital twin predictive-driven circulating cooling water electrochemical intelligent control system provided by the present invention includes a physical operating device and a control device that establishes a communication connection with the physical operating device. The physical operation device includes a reaction chamber defining the main reaction zone 2, an inlet 1 and an outlet 3 connecting the main reaction zone 2, an electrochemical cathode reaction group 7 and an electrochemical anode reaction group 8 interleaved within the main reaction zone 2, a main shaft motor 4 arranged outside the reaction chamber, a rotary scraper assembly 6 arranged within the main reaction zone 2 and close to the surface of the electrochemical cathode reaction group 7, a coupling component 5 connecting the main shaft motor 4 and the rotary scraper assembly 6, a scale removal component 9 arranged at the bottom of the reaction chamber and connecting the main reaction zone 2, a power supply connected to the electrochemical cathode reaction group 7 and the electrochemical anode reaction group 8, a measuring component, a dosing component, and a water replenishment valve arranged at the inlet 1 and the outlet 3; The control device includes: The data acquisition module is used to collect circulating cooling water quality monitoring parameters and output the circulating cooling water quality monitoring parameters. The status assessment module receives circulating cooling water quality monitoring parameters, performs mapping calculations on these parameters according to a preset distribution feature mapping formula, generates a comprehensive index, and outputs the comprehensive index. This distribution feature mapping formula essentially constitutes a multi-criteria decision fusion model, used to decouple the nonlinear relationship between pH, conductivity, calcium hardness concentration, and total alkalinity concentration, mapping multiple dimensions of physical quantities to a unified evaluation dimension space. In terms of technical implementation logic, the core of the above distribution feature mapping formula lies in achieving objective data dimensionality reduction by quantifying the statistical distribution characteristics of multi-source water quality parameters. Specifically, the objective data here refers to the pH, conductivity, calcium hardness concentration, and total alkalinity concentration values ​​sensed and output in real time by the internal measurement components of the physical operating device. These values ​​constitute the original feature set reflecting the real-time state fluctuations of the circulating cooling water, providing an objective mathematical and statistical basis for the above distribution feature mapping formula.

[0025] In addition to the disclosed entropy weight method calculation logic, the aforementioned distribution feature mapping formula also encompasses multi-dimensional parameter fusion logics based on objective data distribution characteristics, such as the CRITIC weight method, the coefficient of variation method, or the principal component mapping model. These multi-dimensional parameter fusion logics share a common technical effect: by calculating the information entropy, contrast, or correlation coefficients of different water quality indicators, they determine the weight values ​​of each indicator's influence on the overall scaling and corrosion tendency of water quality, thereby providing globally representative feature inputs for subsequent trend extrapolation.

[0026] The trend projection module receives the comprehensive index, inputs the comprehensive index into a preset neural network model for forward propagation calculation, and outputs the predicted water quality index. The instruction optimization module receives the predicted water quality index, compares the predicted water quality index with the preset safety threshold, generates control instructions, and sends the control instructions to the spindle motor 4, scale removal component 9, power supply, chemical dosing component, and water supply valve.

[0027] The physical operating device acts as the execution entity to carry out the water quality improvement and regulation process of the circulating cooling water. Untreated circulating cooling water enters the main reaction zone 2 defined inside the reaction chamber through inlet 1. The circulating cooling water submerges the electrochemical cathode reaction group 7 and the electrochemical anode reaction group 8, which are interleaved and distributed in the main reaction zone 2. The power supply delivers DC power to the electrochemical cathode reaction group 7 and the electrochemical anode reaction group 8, causing an electrolytic reaction to occur in the water in the main reaction zone 2. Under the drive of the electric field, scale-forming ions in the water migrate to the surface of the electrochemical cathode reaction group 7 and are reduced and crystallized on the surface of the electrochemical cathode reaction group 7 to form a solid scale layer. The spindle motor 4 is arranged outside the reaction chamber and is connected to the rotary scraper assembly 6, which is arranged in the main reaction zone 2 and close to the surface of the electrochemical cathode reaction group 7, through the coupling component 5. After receiving the operation command, the spindle motor 4 outputs rotational torque, and the coupling component 5 transmits the rotational torque to the rotary scraper assembly 6, which drives the rotary scraper assembly 6 to perform tangential mechanical movement along the surface of the electrochemical cathode reaction group 7, peeling off the attached solid scale layer along the tangential direction. The solid scale layer after being stripped settles under gravity to the scale removal component 9 located at the bottom of the reaction chamber, and is discharged outside the reaction chamber through the scale removal component 9. The circulating cooling water after descaling flows out through the outlet 3. Measuring components located outside the reaction chamber at the inlet 1 and outlet 3 sense the water quality status in real time. The dosing component and water replenishment valve perform chemical agent replenishment and system water replacement according to the received instructions.

[0028] The control device's data reading submodule receives physical environment measurement values ​​sent by the measurement components. These physical environment measurement values ​​are specifically represented as a continuous digital electrical signal sequence generated in real-time by sensor probes at a preset sampling frequency (e.g., once per second) and after analog-to-digital conversion, reflecting the changes in ion distribution and conductivity within the circulating cooling water. The data reading submodule extracts data streams of pH, conductivity, calcium hardness concentration, and total alkalinity concentration from the physical environment measurement values, constructing a multi-dimensional water quality feature matrix. The data cleaning submodule receives the pH, conductivity, calcium hardness concentration, and total alkalinity concentration data streams and uses a smoothing filtering algorithm to identify and remove abnormal peak values ​​that exceed the preset numerical range due to electromagnetic interference at the measurement site and sensor thermal noise. The preset digital range is set according to the physical extreme value range under normal operating conditions of the circulating cooling water system. Values ​​exceeding the range are invalidated, and missing time node data are supplemented by a linear interpolation algorithm. Subsequently, the range transformation method is used to eliminate the differences in different physical dimensions of pH, conductivity, calcium hardness concentration, and total alkalinity concentration, mapping their respective values ​​to a preset dimensionless range (such as the [0,1] range), ultimately forming standardized water quality parameters. A two-way synchronization mechanism based on a fixed time step is established between the above control device and the physical operation device. The query cycle of the data acquisition module is matched with the update frequency of the physical environment measurement values ​​(e.g., between 500 milliseconds and 1000 milliseconds), thereby enabling the predicted water quality index output by the digital twin model to map the electrochemical dynamic evolution process inside the physical main reaction zone in real time.

[0029] The weight allocation submodule within the state assessment module receives standardized water quality parameters and performs dimensionality reduction and weighting of multi-dimensional water quality characteristics using the entropy weighting method. The weight allocation submodule extracts the historical time series of the standardized water quality parameters and calculates the information entropy corresponding to each parameter. Information entropy reflects the degree of dispersion and volatility of pH, conductivity, calcium hardness concentration, and total alkalinity concentration within the monitoring period. The greater the dispersion of a feature parameter, the more drastic the chemical state change within the system, and the richer the information provided. The weight allocation submodule calculates and outputs the influence weight values ​​for each parameter based on the inverse mapping relationship of the information entropy. The index normalization submodule receives the standardized water quality parameters and influence weight values, uses a linear weighting operator to perform matrix multiplication of the standardized water quality parameters and their corresponding influence weight values, and performs a summation operation, transforming the multi-dimensional water quality variables into a single-dimensional scalar, calculating a comprehensive index reflecting the overall scaling and corrosion tendency of the current circulating cooling water.

[0030] The feature filtering submodule within the trend inference module receives the comprehensive index and performs time-series alignment and sliding window extraction according to a preset timestamp, converting the discrete index sequence into a time-series dimension variable with a time-shifting relationship. The network inference submodule receives the time-series dimension variable and inputs it into a preset long short-term memory (LSM) network model. The forgetting gate within the LSM model determines the proportion of historical comprehensive index information to be discarded based on the hidden state at the previous moment, the input gate determines the update amount of the current time-series dimension variable, and the updated cell state is transformed by the output gate for forward propagation, outputting a predicted water quality index that maps the trajectory of water quality changes in the future. The LSM network model includes node connection weights, which determine the scaling ratio of information transmission between neurons in the hidden layer of the network. The hyperparameter optimization unit receives the predicted water quality index and the comprehensive index synchronously output by the state evaluation module, using the comprehensive index as the true water quality index as a verification benchmark. The hyperparameter optimization unit calculates the mean square error (MSE) between the predicted and true water quality indices. When the MSE exceeds a preset calibration error threshold, a preset sparrow search algorithm is triggered for optimization. The hyperparameter optimization unit maps node connection weights to the spatial position vectors of search individuals within the sparrow search algorithm. Based on the mean squared error and the rate of change of the predicted water quality index over time, a fitness evaluation function is constructed. The rate of change of fluctuation is introduced as a penalty term to prevent iterative optimization from getting trapped in local optima. This drives the search individuals to update their positions and perform multiple rounds of iterative calculations to obtain the globally optimal network connection parameters. The hyperparameter optimization unit inputs the network connection parameters into the network inference submodule, directly replacing the original node connection weights with mapping biases within the long short-term memory network model.

[0031] The state comparison submodule within the instruction optimization module receives the predicted water quality index, performs arithmetic subtraction to calculate the absolute difference between the predicted water quality index and the preset safety threshold, and generates a deviation value representing the deviation from the normal water quality benchmark. The power supply regulation submodule receives the deviation value; when the deviation value exceeds the preset deviation limit, it inputs the deviation value into a preset proportional-integral-differential (PID) arithmetic model. The PID model calculates the current proportional amplification term, the historical cumulative integral term, and the future trend differential term of the deviation value, sums these three values, and outputs a compensation current density to suppress further water quality deterioration. The power supply regulation submodule generates a power supply adjustment command as a control instruction based on the numerical encoding of the compensation current density, sends the power supply adjustment command to the power supply, and dynamically changes the electric field intensity distribution between the electrochemical cathode reaction group 7 and the electrochemical anode reaction group 8.

[0032] In industrial circulating cooling water systems operating at full load during high temperatures in summer, the massive evaporation of water leads to a rapid concentration of inorganic salts. The scale removal drive submodule receives a significantly increased compensation current density. When the compensation current density exceeds a preset first calibration threshold, it indicates that the enhanced electric field strength significantly accelerates the rate of crystal precipitation and scaling on the surface of the electrochemical cathode reaction assembly 7. The scale removal drive submodule generates a scale removal operation command as a control instruction and sends it to the spindle motor 4. The speed control parameters included in the scale removal operation command exhibit a positive linear correlation with the compensation current density, and the upper limit of the speed control parameters is set below the rated speed of the spindle motor 4 to ensure that the mechanical cutting action always operates within a physically safe range.

[0033] The main spindle motor 4 operates at the speed set by the scraping operation command, driving the rotary scraper assembly 6 to continuously rotate and cut along the surface of the electrochemical cathode reaction group 7 via the coupling component 5. The sediment discharge submodule simultaneously starts a preset timer to record the duration of continuous operation of the main spindle motor 4, generating an operation duration value. When the operation duration value is equal to or greater than a preset time threshold, it indicates that a large amount of physically stripped solid sludge has accumulated at the bottom of the reaction chamber. The sediment discharge submodule generates a discharge command as a control command, sending it to the scale discharge component 9 to open the bottom discharge solenoid valve and discharge the high-concentration sludge-laden water.

[0034] Under extreme operating conditions, the rate of water quality deterioration may exceed the maximum design load limit of the physical-electrochemical treatment. The linkage compensation submodule continuously receives the predicted water quality index and compensation current density, and uses a differential algorithm to calculate the slope of the predicted water quality index over time. When the slope is greater than a preset slope threshold and the compensation current density is equal to the rated maximum value of the power supply, the linkage compensation submodule determines that the power supply adjustment and physical scaling methods cannot suppress scaling. To prevent algorithm oscillation caused by the infinite expansion of system parameters, the selection range of the above slope threshold should be calibrated with reference to the historical maximum rate of change of water quality deterioration in industrial circulating cooling water systems under full load conditions, ensuring the timeliness and accuracy of the linkage compensation mechanism triggering.

[0035] The linkage compensation submodule generates chemical dosing instructions and valve opening control instructions. It sends chemical dosing instructions to the dosing component to introduce scale inhibitors and dispersants for emergency chemical intervention, and sends valve opening control instructions to the water supply valve to inject fresh desalinated water to dilute the salt content of the system body.

[0036] The virtual mapping module of the control device receives circulating cooling water quality monitoring parameters and control commands covering power supply adjustment and physical scaling. It reads a pre-constructed three-dimensional digital model that has a topological mapping relationship with the geometric dimensions and spatial position of the physical operating device. The virtual mapping module inputs the circulating cooling water quality monitoring parameters and control commands as boundary constraints and dynamic excitation signals into the three-dimensional digital model. The compensation current density included in the control commands is specifically mapped to the second type of boundary conditions on the corresponding electrode surface in the three-dimensional digital model, that is, specifying the normal current flux values ​​of the walls where the electrochemical cathode reaction group and the electrochemical anode reaction group are located; the circulating cooling water quality monitoring parameters are mapped to the first type of boundary conditions at inlet 1, that is, specifying the mass concentration of scale-forming ions of each component at inlet 1. The state variables inside the three-dimensional digital model are triggered to perform discrete time step update calculations. Based on the input boundary conditions, the three-dimensional digital model simulates the evolution of the electric field distribution and the microscopic process of scale particle deposition inside the reaction chamber, and extracts the simulated precipitate thickness values ​​output from the grid surface of the three-dimensional digital model. When the thickness of the sediment exceeds the preset thickness limit, the virtual mapping module determines that there is a risk of flow channel blockage inside the physical equipment, generates sediment accumulation warning information, and outputs the sediment accumulation warning information to the central control terminal interface.

[0037] The state assessment module inside the control device receives standardized water quality parameters from the data cleaning submodule and performs mapping calculations using the entropy weight method. The state assessment module extracts multiple sets of standardized water quality parameters from the data acquisition period to construct an evaluation matrix. It is assumed that the data acquisition period includes... Monitoring data at each time point, including pH, conductivity, calcium hardness concentration, and total alkalinity concentration. The evaluation matrix is ​​constructed using standardized water quality parameters across several dimensions. Elements in the matrix Indicates the first The first time node Standardized water quality parameters in each dimension. The state assessment module calculates the first... The information entropy corresponding to each dimension of standardized water quality parameters is calculated using the following formula:

[0038] In the formula, For the first The first time node The proportion of standardized water quality parameters in each dimension at all time points. For the first The first time node Standardized water quality parameters in several dimensions The total number of time points. For the first The information entropy corresponding to each dimension of standardized water quality parameters. It is a logarithmic function with the natural constant as the base. The state assessment module calculates the influence weight value based on the information entropy. The formula for calculating the influence weight value is:

[0039] In the formula, For the first The influence weight values ​​corresponding to each dimension of standardized water quality parameters. For the first The information entropy corresponding to each dimension of standardized water quality parameters. This represents the total number of dimensions for standardized water quality parameters.

[0040] The status assessment module multiplies and sums the standardized water quality parameters with their corresponding influence weights to generate a comprehensive index. The formula for calculating the comprehensive index is as follows:

[0041] In the formula, For the first The comprehensive index corresponding to each time point. For the first The influence weight values ​​corresponding to each dimension of standardized water quality parameters. For the first The first time node Standardized water quality parameters in several dimensions This represents the total number of dimensions for standardized water quality parameters. The trend projection module extracts time-series dimensional variables from the comprehensive index according to timestamps, and inputs these time-series dimensional variables into a preset Long Short-Term Memory (LSTM) network model for forward propagation. The LTM network model includes a forget gate, an input gate, a cell state update gate, and an output gate. The formula for its internal computational logic is as follows:

[0042]

[0043]

[0044]

[0045]

[0046]

[0047] In the formula, This is the output vector of the forget gate at the current time. The output vector of the input gate at the current time. Let be the candidate cell state vector at the current moment. This is the updated cell state vector at the current moment. This is the cell state vector from the previous time step. This is the output vector of the output gate at the current moment. This is the predicted water quality index for the current moment, output by the Long Short-Term Memory network model. Let be the hidden state vector from the previous time step. The current time-series dimension variable is input into the network model. It is the Sigmoid activation function. The hyperbolic tangent activation function is used. This is an element-wise multiplication operation within a matrix. , , , These are all node connection weights within the Long Short-Term Memory (LSTM) network model. , , , All are bias vectors.

[0048] The network topology of the Long Short-Term Memory (LSTM) network model is configured as follows: the number of features in the input layer is the same as that in the time series dimension; the hidden layer consists of two cascaded LSM computation units, with 64 computation nodes in the first layer and 32 in the second layer; and the output layer is a linear fully connected layer with a single computation node. The time series dimension variable serves as the input data for the LSM network model. The physical relationship between the input and output of the model lies in the fact that equipment scaling and pipeline corrosion within the physical operating device are electrochemical kinetic evolution processes that accumulate nonlinearly over time. The time series dimension variable carries the historical evolution trajectory of multiple water quality parameters coupled and superimposed. The output data corresponds to the predicted water quality index, numerically mapping the water quality deterioration trend of the physical operating device within a set future time span, establishing a nonlinear mathematical time mapping relationship from multidimensional historical water quality monitoring parameters to a single future predicted indicator.

[0049] The hyperparameter optimization unit receives the predicted water quality index and the comprehensive index output by the state assessment module as the true water quality index, and calculates the error between the two using the mean square error formula:

[0050] In the formula, To determine the error between the predicted water quality index and the actual water quality index, To compare the number of samples, For the first The true water quality index corresponding to each sample For the first The predicted water quality index corresponding to each sample.

[0051] The hyperparameter optimization unit employs a sparrow search algorithm for iterative optimization of node connection weights. Specifically, the hyperparameter optimization unit performs the following iterative calculation steps: The node connection weight matrix and bias vector within the Long Short-Term Memory (LSTM) network model are recombined into a one-dimensional numerical vector, defined as the spatial location vector of the search algorithm. The total number of individuals searched by the algorithm is set to 50, and the maximum number of iterations is 100. To avoid the optimization process getting stuck in local optima that only focus on the mean squared error, the fitness evaluation function adds a penalty term for predicted water quality index fluctuations to the mean squared error formula. Specifically, the calculation logic involves extracting the sum of the absolute values ​​of the differences in predicted water quality index values ​​between adjacent time nodes as the penalty term. This penalty term is then added to the mean squared error value to calculate the final fitness value of each spatial location vector, which is then sorted in ascending order. In a single optimization iteration, spatial position vectors with fitness values ​​in the top 20% are extracted as global search individuals, and their position components are updated according to a multidimensional normal distribution rule. The remaining 80% of spatial position vectors are extracted as local follower individuals, and gradient approximation is performed towards the optimal vector position with the smallest current fitness value. Simultaneously, 10% of spatial position vectors are randomly selected from all search individuals as random perturbation individuals, and Gaussian white noise is introduced into the current coordinate position of the random perturbation individuals to enhance the computational ability to escape local minima. The iteration operation terminates when the number of iterations reaches the set upper limit of 100, or when the mean square error value decreases below the calibration error threshold for five consecutive iterations. The hyperparameter optimization unit extracts the optimal one-dimensional numerical vector with the smallest fitness value, reconstructs it in reverse according to the network hierarchy dimension, and transforms it into the node connection weight matrix and bias vector inside the Long Short-Term Memory network model.

[0052] The state comparison submodule within the instruction optimization module compares the predicted water quality index with the preset safety threshold to generate a deviation value. The calculation formula is as follows:

[0053] In the formula, The deviation value generated at the current moment. The predicted water quality index is input at the current moment. The preset safety threshold is used. The power supply regulation submodule within the instruction optimization module inputs the deviation value into the preset proportional-integral-differential (PI-DE) calculation model to calculate the compensation current density. The calculation formula is:

[0054] In the formula, To calculate the compensation current density, This is the proportionality coefficient. The integral coefficient is... These are the differential coefficients. The deviation value generated at the current moment. This is the cumulative integral of the deviation value over the time interval. In the actual operation of the digital control unit, the above cumulative integral is specifically converted into the sum of all deviation values ​​within the discrete sampling period. The rate of change of the deviation value over time is specifically converted into the difference between the deviation values ​​of two adjacent sampling periods divided by the sampling time step in the actual operation of the digital control unit.

[0055] In a specific numerical operation scenario, the physical operation device collects circulating cooling water quality monitoring parameters and generates a total of parameters including pH, conductivity, calcium hardness concentration, and total alkalinity concentration. Standardized water quality parameters across multiple dimensions. Assuming extraction... The data at each time point is used for status evaluation, where the data at the first time point is... The standardized water quality parameters at each time point, after normalization, are as follows: , , , The state assessment module calculates the information entropy of standardized water quality parameters for each dimension, and obtains the results. , , , The state assessment module calculates the influence weights for each dimension based on information entropy. , , , The state assessment module multiplies the standardized water quality parameters by their corresponding influence weights and sums them to obtain the result. The comprehensive index corresponding to each time point is: The trend projection module generates a time-series variable from the composite index across multiple consecutive time points and inputs it into the Long Short-Term Memory (LSTM) network model. The LTM network model performs forward propagation according to its internally defined node connection weights, outputting a predicted water quality index. The hyperparameter optimization unit extracts the corresponding real water quality index as follows: The calculated error value is It did not exceed the calibration error threshold. No node connection weights need to be updated. The instruction optimization module receives the predicted water quality index. Compare with preset security thresholds The state comparison submodule calculates the deviation value as follows: The power supply regulation submodule determines the deviation value. Exceeding the preset deviation limit The deviation value is input into the proportional-integral-differential (PID) calculation model. The proportional coefficient is set to... The integral coefficient is The differential coefficients are Combining historical cumulative errors and rates of change, the proportional-integral-differential operational model calculates the compensation current density as follows: Amperes per square meter. The instruction optimization module calculates based on the compensation current density. Each ampere per square meter generates a control command, which is sent to the power supply to adjust the current output between the electrochemical cathode reaction group and the electrochemical anode reaction group.

[0056] The control device sets preset digital ranges in the data cleaning submodule. These preset ranges include pH limits between 6.0 and 9.5, and conductivity limits between 500 microsiemens per centimeter and 3000 microsiemens per centimeter. The instruction optimization module internally sets preset safety thresholds. The instruction optimization module sets the preset safety threshold to a predicted water quality index of 0.60. The deviation value calculated by the state comparison submodule is associated with a preset deviation limit. The state comparison submodule sets the preset deviation limit to 0.05. The hyperparameter optimization unit determines the error and associates it with a preset calibration error threshold. The hyperparameter optimization unit sets the preset calibration error threshold to a mean square error of 0.01. The scale removal drive submodule activates the rotary scraper assembly and associates it with a preset first calibration threshold. The scale removal drive submodule sets the preset first calibration threshold to a compensated current density of 40 amperes per square meter. The sediment discharge submodule executes the discharge action and associates it with a preset time threshold. The sediment discharge submodule sets the preset time threshold to an operating duration of 45 minutes. The linkage compensation submodule activates chemical reagent replenishment associated with a preset slope threshold. The linkage compensation submodule sets the preset slope threshold to a predicted water quality index change rate of 0.02 per minute. The virtual mapping module triggers an early warning associated with a preset thickness limit value. The virtual mapping module sets the preset thickness limit value to a sediment thickness of 20 mm. All of the above specific values ​​are pre-programmed into the control device's storage unit to support the logical judgment and calculation of various modules within the system.

[0057] Secondly, the digital twin prediction-driven electrochemical intelligent control method for circulating cooling water provided by the present invention is applied to the digital twin prediction-driven electrochemical intelligent control system for circulating cooling water as described in any one of the claims, comprising: Step 1: Collect circulating cooling water quality monitoring parameters and output the circulating cooling water quality monitoring parameters; Step 2: Receive circulating cooling water quality monitoring parameters, perform mapping calculations on the circulating cooling water quality monitoring parameters according to the preset distribution characteristic mapping formula, generate a comprehensive index, and output the comprehensive index; Step 3: Receive the comprehensive index, input the comprehensive index into the preset neural network model for forward propagation calculation, and output the predicted water quality index; Step 4: Receive the predicted water quality index, compare the predicted water quality index with the preset safety threshold, generate control commands, and send the control commands to the spindle motor 4, scale removal component 9, power supply, chemical dosing component, and water supply valve.

[0058] During the operation of the industrial circulating cooling water system, measuring components deployed at inlet 1 and outlet 3 continuously monitor the physicochemical characteristics of the fluid flowing through the pipeline network. The measuring components output physical environment measurements including pH, conductivity, calcium hardness concentration, and total alkalinity concentration. pH characterizes the acid-base balance of the water body, conductivity reflects the total dissolved solids content, and calcium hardness and total alkalinity concentrations reveal the enrichment degree of scale-forming ions. The data reading submodule extracts the physical environment measurements and constructs a multi-dimensional water quality characteristic data stream. The data cleaning submodule receives the multi-dimensional water quality characteristic data stream, uses a smoothing filtering algorithm to identify and remove abnormal jump values ​​exceeding a preset numerical range, and generates continuous and stable standardized water quality parameters. The standardized water quality parameters eliminate high-frequency noise caused by electromagnetic interference at the measurement site, presenting the physical background state of the current time-series changes in water quality parameters.

[0059] The state assessment module receives standardized water quality parameters and performs multi-dimensional feature reduction and fusion calculations according to a preset distribution feature mapping formula. The weight allocation submodule uses the entropy weight method to quantify the historical fluctuation range of each indicator within the standardized water quality parameters. This submodule extracts the time series of the standardized water quality parameters and calculates the information entropy corresponding to pH, conductivity, calcium hardness concentration, and total alkalinity concentration. Information entropy reflects the dispersion of data distribution within a single monitoring dimension; the greater the dispersion, the more drastic the chemical state changes within the water body, and the richer the information provided. The weight allocation submodule outputs the influence weight values ​​corresponding to each water quality feature based on the reciprocal mapping relationship of the information entropy. The index normalization submodule receives the standardized water quality parameters and influence weight values, uses a linear weighting operator to multiply the standardized water quality parameters and their corresponding influence weight values, and sums them. The multi-dimensional water quality variables are reduced in dimensionality through linear weighting calculations and transformed into single-dimensional scalar values, generating a comprehensive index reflecting the overall scaling and corrosion tendency of the current water body.

[0060] The trend projection module receives the comprehensive index, and the feature selection submodule performs a sliding window extraction operation on the comprehensive index according to a preset timestamp, generating a time-series dimension variable with a time-shifting relationship. In high-temperature, full-load operation scenarios during summer, massive evaporation of water leads to a rapid concentration of inorganic salts, resulting in a highly nonlinear scaling trend. The network projection submodule receives the time-series dimension variable and inputs it into a preset long short-term memory (LSTM) network model for forward propagation. The gating unit within the LSM model determines the proportion of historical comprehensive index information to discard based on the previous hidden state. The input gate determines the update amount of the current time-series dimension variable. The updated cell state is transformed by the output gate and forward propagated, outputting a predicted water quality index that maps the future water quality change trajectory. The hyperparameter optimization unit receives the predicted water quality index, extracts the comprehensive index output by the state evaluation module as the true water quality index as a verification benchmark, and calculates the mean square error (MSE) values ​​of both the predicted and true water quality indices. When the MSE value exceeds a preset calibration error threshold, the hyperparameter optimization unit iteratively calculates and finds the globally optimal network connection parameters using a preset sparrow search algorithm. The hyperparameter optimization unit inputs network connection parameters into the network inference submodule, directly replacing the node connection weights with mapping biases within the long short-term memory network model, and reconstructing the scaling ratio of information transmission between neurons.

[0061] The instruction optimization module receives the predicted water quality index, and the state comparison submodule calculates the absolute difference between the predicted water quality index and the preset safety threshold, generating a deviation value representing the deviation from the normal water quality benchmark. The power supply regulation submodule receives the deviation value, and when the deviation value exceeds the preset deviation limit, it inputs the deviation value into the preset proportional-integral-differential (PID) calculation model. The PID model calculates the current proportional amplification term, the historical cumulative integral term, and the future trend differential term of the deviation value, respectively, and sums the three values ​​to output a compensation current density used to suppress water quality deterioration. The power supply regulation submodule generates a power supply adjustment command as a control instruction based on the compensation current density, and sends the power supply adjustment command to the power supply to dynamically change the electric field intensity distribution between the electrochemical cathode reaction group 7 and the electrochemical anode reaction group 8. The scale removal drive submodule receives the compensation current density, and when the compensation current density is greater than the preset first calibration threshold, it indicates that the enhanced electric field intensity leads to a significant acceleration in the crystal precipitation and scaling rate on the surface of the electrochemical cathode reaction group 7. The scale removal drive submodule generates a scale removal operation command as a control instruction and sends it to the spindle motor 4. The main spindle motor 4 operates at a set speed, driving the rotary scraper assembly 6 to continuously rotate and cut along the surface of the electrochemical cathode reaction group 7 via the coupling component 5. The sediment discharge submodule starts a preset timer to record the duration of the main spindle motor 4's operation, generating an operating duration value. When the operating duration value is equal to or greater than a preset time threshold, the sediment discharge submodule generates a discharge command as a control instruction, sending a discharge command to the scale discharge component 9 to open the bottom valve and discharge the high-concentration sludge and water that has been physically stripped.

[0062] The trend projection module, including the Long Short-Term Memory (LSTM) network model, underwent an offline initial training phase before deployment. The control device collected historical circulating cooling water quality monitoring parameters, which were then converted by the state evaluation module to generate 10,000 sets of historical comprehensive indices as a training sample dataset. When constructing the training sample dataset, a sliding window mechanism was used to divide the historical comprehensive indices into supervised learning sample pairs with temporal correlation. Specifically, a continuous comprehensive index with a fixed preceding time length was extracted as the input feature sequence, and the comprehensive index of the next time node immediately following the above input feature sequence was extracted as the true label value. This constructed a network topology including an input layer, two hidden layers, and an output layer. The number of neurons in the input layer matched the number of features in the time-series dimension variable. The two hidden layers were configured with 64 and 32 neurons respectively. The output layer was configured with one neuron to output the predicted water quality index. The computing device used the time backpropagation algorithm to calculate the gradient vector of the loss function. The computing device used an adaptive moment estimation optimizer to update the node connection weights within the LSM network model based on the gradient vector. The network model training process continued until the training set loss function converged and the mean squared error of the validation set was less than 0.02. The virtual mapping module establishes a 3D digital model using computational fluid dynamics (CFD) software. It imports real geometric drawings of the physical operating device into the CFD software to generate a 3D geometric boundary. The computing device then meshes this boundary, generating a fluid domain mesh with 500,000 computational nodes. Multiphysics coupling equations are then defined based on this mesh. These equations include the Navier-Stokes equations describing fluid flow, the Laplace equation describing electric field distribution, and the Nernst-Planck equation describing ion migration reactions. The virtual mapping module uses circulating cooling water quality monitoring parameters and control commands as dynamic boundary conditions, substituting them into the multiphysics coupling equations for discretization. During this discretization process, the computing device sets the convergence criterion to ensure that the residuals of the continuity equation, momentum equation, electric field equation, and migration equations for each component all decrease to a certain value. The calculation results are considered to have reached convergence when the fluctuation of the ion concentration at the outlet with the time step is less than 0.1%. The virtual mapping module uses the discrete solution results to update the state variables inside the three-dimensional digital model in real time, enabling the three-dimensional digital model to complete the physical mapping of the microscopic state of the physical operating device.

[0063] Embodiment 1 of this invention: In chemical process industries, which operate continuously year-round, the circulating cooling water system faces frequent fluctuations in water quality parameters. Measurement components deployed at inlet 1 and outlet 3 collect real-time measurements of the water's physical environment, which are represented as a continuous electrical signal matrix. The data reading submodule extracts pH, conductivity, calcium hardness concentration, and total alkalinity concentration from the physical environment measurements. During actual operation, electromagnetic interference at the measurement site often causes conductivity to abruptly spike to abnormally high values ​​outside the normal physical range. The data cleaning submodule, based on a preset numerical range, sets a reasonable upper limit for conductivity at 3000 microsiemens per centimeter, removing peak values ​​exceeding 3000 microsiemens per centimeter to generate continuous and stable standardized water quality parameters. The status assessment module receives the standardized water quality parameters, and the weight allocation submodule extracts the historical fluctuation variance of the standardized water quality parameters using the entropy weighting method, calculating the corresponding information entropy. Information entropy values ​​map the dispersion and weighting of different parameters within a specific period. The index normalization submodule multiplies and sums the standardized water quality parameters with their corresponding influence weights, outputting a single-dimensional comprehensive index. The trend projection module receives the comprehensive index, and the feature selection submodule extracts it according to 15-minute time stamps, generating a time-series dimension variable with physical shift relationships. The network projection submodule inputs the time-series dimension variable into a pre-trained long short-term memory network model for forward propagation, projecting the predicted water quality index for the next two hours. The instruction optimization module compares the predicted water quality index with the set safety threshold, and the state comparison submodule performs subtraction to generate a deviation value. The power supply regulation submodule inputs the deviation value into a proportional-integral-differential (PID) calculation model, outputs a compensation current density of 20 amperes per square meter, sends a power supply adjustment command to the power supply, and dynamically adjusts the output voltage levels at both ends of the electrochemical cathode reaction group 7 and the electrochemical anode reaction group 8.

[0064] Embodiment 2 of the present invention: In the high-load operation scenario of the steel smelting industry in summer, the water temperature in the main reaction zone 2 rises sharply, leading to a significant increase in the activity of scale-forming ions. The power supply continuously applies DC power to the electrochemical cathode reaction group 7 and the electrochemical anode reaction group 8, accelerating the migration and crystallization of scale-forming ions on the surface of the electrochemical cathode reaction group 7 in the main reaction zone 2. As the scaling rate accelerates, the compensation current density calculated by the power supply regulation submodule continues to rise. The scale-scraping drive submodule in the instruction optimization module monitors the compensation current density in real time. When the compensation current density exceeds the preset first calibration threshold of 40 amperes per square meter, the scale-scraping drive submodule generates a scale-scraping operation command. The main spindle motor 4, located outside the reaction chamber, starts after receiving the scale-scraping operation command, outputs rotational torque, and transmits mechanical kinetic energy to the rotary scraper assembly 6 located in the main reaction zone 2 through the coupling component 5. The rotary scraper assembly 6 performs continuous cutting motion close to the surface of the electrochemical cathode reaction group 7, peeling off the attached solid calcium carbonate scale layer along the tangential direction. During the stripping process, the sediment discharge submodule simultaneously starts its internal timer to record the duration of continuous operation of the spindle motor 4. When the cumulative operating time reaches the threshold of 45 minutes, the sediment discharge submodule determines that sufficient solid suspended matter has accumulated at the bottom of the reaction chamber and generates a discharge command, sending it to the scale removal component 9 located at the bottom of the reaction chamber. The scale removal component 9 opens the discharge solenoid valve, discharging the high-concentration mud-water mixture that has been physically stripped out of the system. The virtual mapping module within the control device simultaneously receives the control commands and circulating cooling water quality monitoring parameters, performing multiphysics field reconstruction calculations within the three-dimensional digital model. When the sediment thickness extracted from the surface of the three-dimensional digital model mesh nodes exceeds the 20 mm thickness limit, the system directly outputs sediment accumulation warning information to the central control terminal interface.

[0065] Embodiment 3 of this invention: A thermal power plant faces a severe operating condition where the source water quality has deteriorated significantly, compounded by the system's long-term lack of demineralized water replacement. The hyperparameter optimization unit within the control device continuously verifies the computational accuracy of the network inference submodule. The hyperparameter optimization unit receives the predicted water quality index and the comprehensive index synchronously output by the state assessment module, comparing the current comprehensive index with the actual water quality index. When the calculated mean square error exceeds the set calibration error threshold of 0.05, the system determines that the long short-term memory network model has experienced fitting deviation. The hyperparameter optimization unit triggers a sparrow search algorithm, transforming the node connection weights within the long short-term memory network model into search individual position vectors in a multi-dimensional solution space. After 50 rounds of iterative computation, the globally optimal network connection parameters are obtained. The hyperparameter optimization unit inputs the new network connection parameters into the network inference submodule, replacing the original node connection weights and reconstructing the scaling ratio of information transmission between neurons. Under extreme conditions, the linkage compensation submodule receives the predicted water quality index and calculates the physical slope over the time series. When the slope of the predicted water quality index change over time continuously exceeds the set slope threshold, and the compensation current density output by the power supply regulation submodule equals the maximum rated output limit of the power supply (60 amperes per square meter), the system determines that the purely physical-electrochemical treatment method has reached the physical limit of the installed capacity. The linkage compensation submodule generates chemical dosing commands and valve opening control commands, sending signals to the dosing component to force the injection of scale inhibitor and dispersant, and sending signals to the water supply valve to increase the flow area to introduce fresh external cooling water, forcibly diluting the high salt concentration inside the system.

Claims

1. A digital twin prediction driven cycle cooling water electrochemical intelligent regulation system, characterized in that, This includes physical operating devices and control devices that establish communication connections with the physical operating devices; The physical operation device includes a reaction chamber defining the main reaction zone (2), an inlet (1) and an outlet (3) connecting the main reaction zone (2), an electrochemical cathode reaction group (7) and an electrochemical anode reaction group (8) interleaved in the main reaction zone (2), a main shaft motor (4) arranged outside the reaction chamber, a rotary scraper assembly (6) arranged in the main reaction zone (2) and close to the surface of the electrochemical cathode reaction group (7), a coupling component (5) connecting the main shaft motor (4) and the rotary scraper assembly (6), a scale removal component (9) arranged at the bottom of the reaction chamber and connected to the main reaction zone (2), a power supply connected to the electrochemical cathode reaction group (7) and the electrochemical anode reaction group (8), a measuring component, a dosing component, and a water replenishment valve arranged at the inlet (1) and the outlet (3); The control device includes: The data acquisition module is used to collect circulating cooling water quality monitoring parameters and output the circulating cooling water quality monitoring parameters. The status assessment module receives circulating cooling water quality monitoring parameters, performs mapping calculations on the circulating cooling water quality monitoring parameters according to a preset distribution characteristic mapping formula, generates a comprehensive index, and outputs the comprehensive index. The trend projection module receives the comprehensive index, inputs the comprehensive index into a preset neural network model for forward propagation calculation, and outputs the predicted water quality index. The instruction optimization module is used to receive the predicted water quality index, compare the predicted water quality index with the preset safety threshold, generate control instructions, and send control instructions to the spindle motor (4), scale removal component (9), power supply, chemical dosing component and water supply valve.

2. The digital twin prediction-driven electrochemical intelligent regulation system for circulating cooling water according to claim 1, wherein, The data acquisition module includes: The data reading submodule is used to receive physical environment measurement values ​​sent by the measurement component, extract the acidity, conductivity, calcium hardness concentration and total alkalinity concentration from the physical environment measurement values, and output the acidity, conductivity, calcium hardness concentration and total alkalinity concentration. The data cleaning submodule receives pH, conductivity, calcium hardness concentration, and total alkalinity concentration, filters out values ​​that exceed the preset range, performs dimensionless mapping processing, generates standardized water quality parameters, and sends the standardized water quality parameters to the status assessment module.

3. The digital twin prediction-driven electrochemical intelligent regulation system for circulating cooling water according to claim 2, characterized in that, The status assessment module includes: The weight allocation submodule receives standardized water quality parameters, calculates the information entropy corresponding to the standardized water quality parameters according to the entropy weight method, and outputs the influence weight values ​​based on the information entropy. The index normalization submodule receives standardized water quality parameters and their corresponding influence weights, multiplies the standardized water quality parameters by their corresponding influence weights and sums them to calculate the comprehensive index, and then sends the comprehensive index to the trend inference module.

4. The digital twin prediction-driven electrochemical intelligent regulation system for circulating cooling water according to claim 3, characterized in that, The trend projection module includes: The feature filtering submodule is used to receive the comprehensive index, extract the comprehensive index according to the preset timestamp, generate time series dimension variables, and output the time series dimension variables. The network inference submodule is used to receive time series dimension variables, input the time series dimension variables into a preset long short-term memory network model for forward propagation calculation, and output the predicted water quality index. The long short-term memory network model includes node connection weights. The predicted water quality index is sent to the instruction optimization module.

5. The digital twin prediction-driven electrochemical intelligent regulation system for circulating cooling water according to claim 4, wherein, The control unit also includes a hyperparameter optimization unit that connects to the network inference submodule: The hyperparameter optimization unit receives the predicted water quality index and the comprehensive index output by the state assessment module as the actual water quality index. It calculates the error value between the predicted water quality index and the actual water quality index. When the error value is greater than the preset calibration error threshold, it iteratively calculates the network connection parameters according to the preset sparrow search algorithm, inputs the network connection parameters into the network inference submodule, and replaces the node connection weights inside the long short-term memory network model.

6. The digital twin prediction-driven electrochemical intelligent regulation system for circulating cooling water according to claim 5, wherein, The instruction optimization module includes: The status comparison submodule is used to receive the predicted water quality index, calculate the difference between the predicted water quality index and the preset safety threshold, generate the deviation value, and output the deviation value. The power supply regulation submodule is used to receive deviation values. When the deviation value exceeds the preset deviation limit, the deviation value is input into the preset proportional-integral-differential calculation model to calculate the compensation current density. Based on the compensation current density, a power supply adjustment command is generated as a control instruction and sent to the power supply.

7. The digital twin prediction-driven electrochemical intelligent regulation system for circulating cooling water according to claim 6, wherein, The instruction optimization module also includes: The scraping drive submodule is used to receive the compensation current density. When the compensation current density is greater than the preset first calibration threshold, it generates a scraping operation command as a control command and sends the scraping operation command to the spindle motor (4). Through the coupling component (5), it drives the rotary scraper assembly (6) to rotate along the surface of the electrochemical cathode reaction group (7). The sediment discharge submodule is used to start a preset timer to record the duration of the spindle motor (4) in operation and generate an operation duration value. When the operation duration value is equal to or greater than the preset time threshold, a discharge command is generated as a control command and sent to the scale removal component (9).

8. The digital twin prediction-driven electrochemical intelligent regulation system for circulating cooling water according to claim 7, wherein, The instruction optimization module also includes: The linkage compensation submodule is used to receive the predicted water quality index and the compensation current density, calculate the slope of the predicted water quality index over time, and when the slope is greater than the preset slope threshold and the compensation current density is equal to the rated maximum value of the power supply, generate a chemical dosing command and a valve opening control command, send the chemical dosing command to the dosing component, and send the valve opening control command to the water supply valve.

9. The digital twin prediction-driven electrochemical intelligent regulation system for circulating cooling water according to claim 8, wherein, The control device also includes: The virtual mapping module receives circulating cooling water quality monitoring parameters and control commands, establishes a three-dimensional digital model that maps to the physical operating device structure, inputs the circulating cooling water quality monitoring parameters and control commands into the three-dimensional digital model, triggers the update calculation of the state variables inside the three-dimensional digital model, extracts the sediment thickness value output by the three-dimensional digital model, and generates and outputs sediment accumulation warning information when the sediment thickness value is greater than the preset thickness limit value.

10. The digital twin prediction driven circulating cooling water electrochemical intelligent regulation method is applied to the digital twin prediction driven circulating cooling water electrochemical intelligent regulation system as claimed in any one of claims 1-9, characterized in that, include: Step 1: Collect circulating cooling water quality monitoring parameters and output the circulating cooling water quality monitoring parameters; Step 2: Receive circulating cooling water quality monitoring parameters, perform mapping calculations on the circulating cooling water quality monitoring parameters according to the preset distribution characteristic mapping formula, generate a comprehensive index, and output the comprehensive index; Step 3: Receive the comprehensive index, input the comprehensive index into the preset neural network model for forward propagation calculation, and output the predicted water quality index; Step 4: Receive the predicted water quality index, compare the predicted water quality index with the preset safety threshold, generate control instructions, and send the control instructions to the spindle motor (4), scale removal component (9), power supply, chemical dosing component and water supply valve.

Images