High-purity water steady-state control method and system based on electrodeionization field intensity dynamic equilibrium

Abstract

The invention discloses a high-purity water steady-state control method and system based on electrodeionization field intensity dynamic equilibrium, and relates to the technical field of water treatment.The method comprises the steps that water quality fluctuation data and electric field signal data are collected from all units of a water treatment system in real time through a sensor array; carrying out preliminary denoising processing on the collected data by adopting fluctuation correlation analysis in a time window, and carrying out segmented extraction on a signal frequency response range to obtain a preliminarily smooth water quality fluctuation sequence and an electric field signal sequence; according to the preliminarily smoothed water quality fluctuation sequence, analyzing spatial transmission path weight and delay time distribution characteristics, constructing a dynamic map of fluctuation transmission, and determining transmission directions and potential unbalance nodes of fluctuation among units; according to the high-purity water steady-state control method and system based on electrodeionization field intensity dynamic equilibrium, the robustness and the overall stability of the system to external disturbance are remarkably improved, and the reliability and the energy efficiency level of the deep desalting process are improved.

Description

Technical Field

[0001] This invention relates to the field of water treatment technology, specifically to a method and system for steady-state control of high-purity water based on dynamic equilibrium of electro-deionization field strength. Background Technology

[0002] In high-purity water preparation systems, especially in scenarios employing electrodeionization (EDI) technology for deep desalination, fluctuations in influent water quality inevitably exhibit randomness and complexity. Since water treatment systems typically consist of multiple series or parallel functional units, water quality disturbances in any unit will propagate sequentially along the water flow direction to subsequent units, posing a stability challenge to the entire system due to dynamic changes. Accurately identifying the propagation path of water quality fluctuations and their cumulative impact on downstream units has become a crucial research topic for the stable operation of high-purity water systems.

[0003] Existing technologies typically focus on monitoring and statically adjusting water quality parameters in individual treatment stages, lacking a systematic analysis of the propagation patterns of water quality fluctuations across multiple units. Under complex operating conditions, water quality changes often exhibit dynamic characteristics across multiple frequencies and scales, making it difficult for local control of a single parameter to accurately reflect the overall system trend. Furthermore, due to the lack of ability to identify the time distribution of fluctuation delays, the system's response to abnormal fluctuations is often lagging, leading to the failure to detect high-risk nodes in a timely manner and causing some subsequent treatment units to experience overload or deviate from their optimal operating state. On the other hand, electro-deionization technology relies on real-time control of electric field strength to maintain ion migration efficiency and desalination effects. Although the electric field signal has the potential for rapid adjustment, in actual operation, the responses of different treatment units to the electric field signal are not consistent. The electric field signal is easily affected by noise, interference, or parameter drift during transmission within the system, resulting in distortion or delay, making it difficult for electric field control to keep pace with the propagation of water quality fluctuations. Especially when influent conditions change abruptly or local units experience anomalies, the uneven response of the electric field signal further exacerbates system instability. Furthermore, existing control strategies lack a collaborative analysis mechanism for water quality fluctuation data and electric field signal data, making it impossible to comprehensively assess the system's dynamic state from multiple dimensions. The lack of in-depth analysis of fluctuation path weights, delay distribution characteristics, and anomaly propagation trends makes it difficult for the system to establish accurate control criteria. Simultaneously, the lack of an iterative update and adaptive optimization mechanism in generating control parameters hinders the formation of a stable and reliable feedback control closed loop. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for steady-state control of high-purity water based on dynamic equilibrium of electro-deionization field strength, thereby solving the problems existing in the prior art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a high-purity water steady-state control method based on dynamic equilibrium of electro-deionization field strength, comprising: real-time acquisition of water quality fluctuation data and electric field signal data from each unit of the water treatment system through a sensor array; preliminary denoising processing of the acquired data using fluctuation correlation analysis within a time window; segmented extraction of the signal frequency response range to obtain a preliminary smoothed water quality fluctuation sequence and electric field signal sequence; analysis of spatial transmission path weights and delay time distribution characteristics based on the preliminary smoothed water quality fluctuation sequence to construct a dynamic spectrum of fluctuation transmission, determining the transmission direction of fluctuations and potential imbalance nodes between units; if the fluctuation transmission direction in the dynamic spectrum shows abnormal clustering and the delay time distribution characteristics exceed a preset threshold, high-risk subsequent units are marked, and the specific distribution of transmission delay is calculated using spatial transmission path weight data to obtain a detailed delay distribution spectrum; matching and fusing the acquired delay distribution spectrum with the electric field signal sequence, and generating a control compensation matrix by combining the signal strength attenuation coefficient and data fusion weight allocation; determining whether the matrix element mapping rules meet preset conditions to obtain a targeted control basis.

[0006] Preferably, the step of acquiring water quality fluctuation data and electric field signal data in real time from each unit of the water treatment system using a sensor array, performing preliminary denoising processing on the acquired data using fluctuation correlation analysis within a time window, and extracting segments based on the signal frequency response range to obtain a preliminary smoothed water quality fluctuation sequence and electric field signal sequence includes: acquiring water quality fluctuation data and electric field signal data from each unit of the water treatment system using a sensor array; performing fluctuation correlation calculation on the water quality fluctuation data using a preset time window; wherein the fluctuation correlation calculation obtains the preliminary denoised data sequence by calculating the correlation coefficient between data points within the time window; and obtaining the signal frequency response range for the preliminary denoised data sequence. The frequency response range of the signal is determined by Fourier transform to determine the frequency distribution. High-frequency and low-frequency components are extracted in segments to determine the segmented electric field signal sequence. Abnormal fluctuation points are obtained from the segmented electric field signal sequence. Abnormal fluctuation points are determined by detecting points in the sequence that deviate from the average value. If the abnormal fluctuation points exceed a preset threshold, a smoothing filter is applied to the signal sequence. The smoothing filter uses a moving average method to obtain a preliminary smoothed water quality fluctuation sequence. Based on the preliminary smoothed water quality fluctuation sequence, the electric field signal sequence is integrated. The integration is achieved by merging sequences with synchronized timestamps. The stability of the integrated sequence is judged. The stability is obtained by calculating the variance value to obtain the final preliminary smoothed sequence.

[0007] Preferably, the step of analyzing the spatial transmission path weights and delay time distribution characteristics based on the initially smoothed water quality fluctuation sequence, constructing a dynamic spectrum of fluctuation transmission, and determining the transmission direction and potential imbalance nodes between units includes: obtaining spatial transmission paths from the initially smoothed water quality fluctuation sequence; obtaining path weight sequences by calculating the Pearson correlation coefficient between sequences; extracting delay time distribution characteristics from the path weight sequences; calculating time offset values ​​using a cross-correlation function, where the input of the cross-correlation function is the path weight sequence pair, and the output is the offset corresponding to the peak value to obtain the delay distribution sequence; constructing a dynamic spectrum of fluctuation transmission based on the delay distribution sequence; representing the connections between units using a weighted directed graph, where the weighted directed graph uses the delay distribution sequence as edge weights to obtain the graph structure; analyzing the transmission direction from the graph structure; determining the directional path from the source unit to the target unit using a depth-first search graph traversal algorithm, where the input of the depth-first search is the graph structure, and the output is a set of paths to obtain the direction sequence; identifying potential imbalance nodes based on the direction sequence by detecting nodes with unbalanced in-degree and out-degree in the graph, where the difference between in-degree and out-degree exceeds a preset threshold to obtain a set of imbalance nodes.

[0008] Preferably, if the wave propagation direction in the dynamic spectrum shows abnormal clustering and the delay time distribution characteristics exceed a preset threshold, then high-risk subsequent units are marked. The specific distribution of the propagation delay is calculated using spatial propagation path weight data to obtain a detailed delay distribution map. This includes obtaining wave propagation direction data from the dynamic spectrum, determining the degree of abnormal clustering, and if the abnormal clustering exceeds a preset threshold, marking high-risk subsequent units to obtain a risk marking sequence; extracting spatial propagation path weight data from the risk marking sequence, and calculating the propagation delay value using a cross-correlation function, where the cross-correlation function uses path weight data as input and calculates the delay value through peak offset to obtain a delay calculation sequence; analyzing the delay time distribution characteristics based on the delay calculation sequence, and constructing a delay distribution model, where the delay distribution model uses the delay calculation sequence as input and determines the distribution parameters through Gaussian fitting to obtain a set of peak distributions; fusing the connection relationships between units using the peak distribution set to generate a detailed delay distribution map, where the fusion uses weighted average to calculate the connection strength, obtains potential wave source nodes in the map, and obtains a source node sequence; evaluating the influence range of subsequent units based on the source node sequence, determining the influence expansion trend, and obtaining an expansion trend map.

[0009] Preferably, the process involves matching and fusing the acquired delay distribution map with the electric field signal sequence, combining the signal intensity attenuation coefficient and data fusion weight allocation to generate a control compensation matrix, determining whether the matrix element mapping rules meet preset conditions, and obtaining targeted control basis. This includes extracting distribution peak data from the delay distribution map, performing preliminary alignment using the electric field signal sequence, calculating the peak offset using a cross-correlation function to obtain an offset calibration sequence; fusing the signal intensity attenuation coefficient with the offset calibration sequence to calculate an attenuation adjustment factor, where the cross-correlation function takes the sequence and coefficient as input, adjusting the fused signal intensity attenuation process through peak position to determine the adjusted signal sequence; allocating data fusion weights according to the adjusted signal sequence to generate a weight vector; if the sum of the weight vectors exceeds a preset threshold, then fusing the sequence and vector by weighted summation to obtain a fused signal matrix; constructing a control compensation matrix using the fused signal matrix, determining whether the matrix elements meet the matrix element mapping rule fusion matching rule to obtain a mapping conformance sequence; evaluating preset condition judgments on the mapping conformance sequence to generate targeted control basis, thus obtaining a control basis set.

[0010] Preferably, the method further includes: if the element values ​​in the control compensation matrix meet the matrix element threshold judgment condition and are greater than zero, then the electric field signal is intensity adjusted according to the signal intensity adjustment amplitude and the signal gain dynamic range; the adjusted optimized signal sequence is obtained through feedback loop correction parameter optimization processing. Specifically, this includes: obtaining element values ​​from the control compensation matrix; judging that the element values ​​meet the matrix element threshold judgment condition and are greater than zero; obtaining a set of element values ​​that meet the condition; for the set of element values ​​that meet the condition, adjusting the electric field signal intensity according to the signal intensity adjustment amplitude and the signal gain dynamic range to determine the adjusted electric field signal; optimizing the adjusted electric field signal through feedback loop correction parameter processing to obtain a corrected parameter set to obtain an optimized signal sequence; fusing the signal phase offset data with the optimized signal sequence to calculate the phase calibration factor, wherein the phase offset data is extracted from the electric field signal sequence, and the phase adjustment sequence is determined through the phase difference adjustment fusion process; allocating phase fusion weights according to the phase adjustment sequence to generate a phase weight vector; if the sum of the phase weight vectors exceeds a preset threshold, then the phase stable signal is obtained by weighted summation of the sequence and vector; judging whether the phase stable signal meets the preset signal stability condition to obtain a targeted control basis.

[0011] Preferably, the method further includes, for the optimized signal sequence, combining the adjusted sequence smoothness and abnormal signal removal logic, simulating the dynamic response of each unit of the water treatment system, controlling the simulation process by the number of sequence optimization iterations, and determining the final synchronization control parameter set. Specifically, this includes obtaining the sequence smoothness index from the optimized signal sequence, removing abnormal parts by a preset abnormal signal removal logic, and generating a preliminary response sequence; for the preliminary response sequence, simulating the dynamic response of each unit of the water treatment system, calculating the response difference through the response mapping relationship between units, and obtaining a response change set; based on the response change set, controlling the iteration loop by the number of sequence optimization iterations, adjusting the deviation values ​​in the response change set, and determining the iterated response sequence; obtaining the iterated response sequence, fusing the initial values ​​of the synchronization control parameters, synchronizing the parameters of the water treatment units through parameter matching rules, and obtaining a synchronization parameter set; evaluating the control stability of the water treatment system through the synchronization parameter set, and generating the final synchronization control parameter set by fusing stability indices.

[0012] Preferably, the method further includes applying the electric field control equipment to each unit of the water treatment system according to the synchronous control parameter set, and monitoring the adjusted water quality fluctuation sequence through real-time feedback. If the deviation value is less than the preset threshold and the stability of the adjusted signal meets the requirements, the system is confirmed to have reached a stable state, and the final operating parameter set is obtained. Specifically, this includes generating a control signal sequence by applying synchronous parameters, integrating electric field equipment control and unit response simulation, and determining the control signal sequence; and obtaining the water quality fluctuation sequence by using real-time feedback monitoring based on the control signal sequence, and judging the water quality fluctuation sequence.

[0013] Preferably, the step of applying the synchronous control parameter set to the electric field control equipment of each unit of the water treatment system, and monitoring the adjusted water quality fluctuation sequence through real-time feedback, confirming that the system has reached a stable state and obtaining the final operating parameter set, also includes comparing the water quality fluctuation sequence with a preset threshold by judging the deviation value and checking the signal stability, and obtaining the deviation comparison result; from the deviation comparison result, integrating the system stability confirmation and the operating parameter acquisition, generating the final operating parameter set, and determining the final operating parameter set.

[0014] A high-purity water steady-state control system based on electro-deionization field strength dynamic equilibrium is used to implement the steps of the high-purity water steady-state control method based on electro-deionization field strength dynamic equilibrium. The system includes an acquisition and preprocessing module, which collects water quality fluctuation data and electric field signal data in real time from each unit of the water treatment system through a sensor array. It performs preliminary noise reduction on the collected data using fluctuation correlation analysis within a time window, and extracts the signal frequency response range in segments to obtain a preliminary smoothed water quality fluctuation sequence and electric field signal sequence. A fluctuation propagation analysis module analyzes the spatial propagation path weights and delay time distribution characteristics based on the preliminary smoothed water quality fluctuation sequence, constructs a dynamic spectrum of fluctuation propagation, and determines the propagation direction and potential imbalance nodes between units. A high-risk identification module marks high-risk subsequent units if the fluctuation propagation direction in the dynamic spectrum shows abnormal clustering and the delay time distribution characteristics exceed a preset threshold. It calculates the specific distribution of propagation delay using spatial propagation path weight data to obtain a detailed delay distribution spectrum. A control and fusion module uses the acquired delay distribution spectrum... The system matches and fuses the spectrum and electric field signal sequences, and generates a control compensation matrix by combining the signal intensity attenuation coefficient and data fusion weight allocation. It then determines whether the matrix element mapping rules meet preset conditions to obtain a basis for targeted control. In the electric field signal adjustment module, if the element values ​​in the control compensation matrix meet the matrix element threshold judgment conditions and are greater than zero, the electric field signal intensity is adjusted based on the signal intensity adjustment amplitude and the signal gain dynamic range. Through feedback loop correction parameter optimization processing, an optimized signal sequence is obtained. In the parameter simulation optimization module, for the optimized signal sequence, the dynamic response of each unit of the water treatment system is simulated by combining the smoothness of the adjusted sequence and the abnormal signal removal logic. The simulation process is controlled by the number of sequence optimization iterations to determine the final set of synchronous control parameters. In the steady-state control module, based on the synchronous control parameter set, the electric field control equipment of each unit of the water treatment system is applied. Through real-time feedback loop monitoring of the adjusted water quality fluctuation sequence, if the deviation value is less than a preset threshold and the stability of the adjusted signal meets the requirements, the system is confirmed to have reached a stable state, and the final operating parameter set is obtained.

[0015] As can be seen from the above technical solution, the present invention has the following beneficial effects: This invention relates to a method and system for steady-state control of high-purity water based on dynamic equilibrium of an electro-deionization field. By introducing a collaborative analysis mechanism between water quality fluctuation sequences and electric field signal sequences into the high-purity water preparation system, and utilizing multi-level data processing steps such as spatial propagation path weight identification, delay distribution map construction, signal fusion compensation matrix generation, and feedback loop optimization, it achieves precise characterization of fluctuation propagation patterns and electric field response characteristics. This effectively solves the technical challenges of existing technologies, such as difficulty in tracking fluctuation propagation, inability to promptly identify abnormal delays, uneven electric field signal response, and lack of dynamic adaptability in control strategies. This invention can adjust the electric field strength in real time according to the dynamic changes of high-risk nodes, ensuring that the electric field effect is consistent with actual water quality fluctuations. Furthermore, by optimizing the signal sequence, it improves simulation accuracy and control stability, ultimately achieving rapid response, precise control, and steady-state operation of the electro-deionization system under complex influent conditions. This significantly improves the system's robustness to external disturbances and overall stability, enhances the consistency of high-purity water quality, extends system lifespan, and improves the reliability and energy efficiency of the deep desalination process. Attached Figure Description

[0016] Figure 1 This is a flowchart of the high-purity water steady-state control method based on dynamic equilibrium of electro-deionization field strength according to the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. 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.

[0018] like Figure 1As shown, this invention provides a technical solution: a high-purity water steady-state control method based on dynamic equilibrium of electro-deionization field strength, comprising: real-time acquisition of water quality fluctuation data and electric field signal data from each unit of a water treatment system using a sensor array; preliminary denoising processing of the acquired data using fluctuation correlation analysis within a time window; segmented extraction based on the signal frequency response range to obtain a preliminary smoothed water quality fluctuation sequence and electric field signal sequence; based on the preliminary smoothed water quality fluctuation sequence, analyzing the spatial transmission path weights and delay time distribution characteristics to construct a dynamic spectrum of fluctuation transmission, determining the transmission direction of fluctuations and potential imbalance nodes between units; if the fluctuation transmission direction in the dynamic spectrum shows abnormal clustering and the delay time distribution characteristics exceed a preset threshold, high-risk subsequent units are marked; the specific distribution of transmission delay is calculated using spatial transmission path weight data to obtain a detailed delay distribution spectrum; and the acquired delay distribution spectrum is matched and fused with the electric field signal sequence. By combining the signal strength attenuation coefficient and data fusion weight allocation, a control compensation matrix is ​​generated. The mapping rules of the matrix elements are then checked to determine if they meet preset conditions, thus obtaining a basis for targeted control. If the element values ​​in the control compensation matrix meet the matrix element threshold judgment conditions and are greater than zero, the electric field signal is adjusted based on the signal strength adjustment amplitude and the dynamic range of the signal gain. Through feedback loop correction parameter optimization, an optimized signal sequence is obtained. For the optimized signal sequence, the dynamic response of each unit in the water treatment system is simulated by combining the smoothness of the adjusted sequence and the abnormal signal removal logic. The simulation process is controlled by the number of sequence optimization iterations to determine the final set of synchronous control parameters. Based on the synchronous control parameter set, the electric field control equipment in each unit of the water treatment system is applied. The adjusted water quality fluctuation sequence is monitored in real-time through feedback loops. If the deviation value is less than a preset threshold and the stability of the adjusted signal meets the requirements, the system is confirmed to have reached a stable state, and the final operating parameter set is obtained.

[0019] This implementation method is based on the coupling relationship between water quality fluctuations and electric field responses in an electro-deionization system. By collecting water quality change parameters and electric field signals from each unit, it achieves dynamic identification of fluctuation sources, fluctuation directions, and imbalance risks within the system. Correlation analysis within a time window is used to reduce transient noise, giving the water quality and electric field sequences stable characteristics for easier subsequent processing. By analyzing the smoothed sequence, the weights and response delay characteristics of spatial propagation paths in the system can be extracted, thereby constructing a dynamic spectrum reflecting the propagation direction of water quality disturbances within the system. When the spectrum shows clustered propagation paths and abnormal delay characteristics, it can be inferred that there is potential imbalance in local units of the system, requiring focused control. The delay distribution spectrum further quantifies the wave propagation time difference between different units and is used for fusion with the electric field signal sequence. During the fusion process, an attenuation coefficient and weight allocation mechanism are used to ensure that the fusion result accurately maps the actual dynamic behavior of each unit. The generation of the compensation matrix is ​​based on mapping rule verification, which can identify electric field control points that require gain adjustment. When the matrix elements meet the threshold conditions, the system gradually optimizes by adjusting the electric field signal strength and combining it with a feedback loop mechanism, so that the signal sequence tends to the optimal state after multiple iterations. The optimized signal sequence, combined with smoothness evaluation and outlier removal strategies, can be used to drive dynamic simulation of the system and obtain a set of synchronous control parameters. Ultimately, these parameters are used to control the equipment, enabling real-time adjustment of the electric field strength, and, under the action of a feedback monitoring mechanism, ensuring that water quality fluctuations converge to a stable range, allowing the entire system to enter steady-state operation.

[0020] This implementation method can identify potential risk units in the early stages of water quality fluctuations through dynamic graphs, enabling early warning and precise location, and significantly improving the system's stable control capability. By fusing the delay distribution graph and the electric field signal, the compensation matrix becomes more targeted, avoiding unnecessary energy consumption caused by overall adjustments in traditional static control. Under the action of signal strength adjustment and feedback optimization mechanisms, the electric field response can be adaptively adjusted in real time according to actual system changes, improving the accuracy and response speed of water quality control. Simultaneously, dynamic simulation and synchronous parameter generation methods improve the reliability of the control strategy, enabling the system to quickly recover to a steady state under different load fluctuations. Phased optimization and iterative strategies reduce the risk of misadjustment, improving system safety and durability. The establishment of the final operating parameter set provides a replicable stable control foundation for long-term operation, enhancing equipment operating efficiency and reducing maintenance frequency.

[0021] In one implementation, firstly, water quality fluctuation data and electric field signal data are continuously collected from each unit of the water treatment system using a sensor array. The water quality and electric field values ​​corresponding to each sampling moment are sequentially stored in a data sequence, with each sampling moment bearing a specific timestamp. A time window length parameter is set, which is a fixed number of sampling points. This number is determined by selecting all events from historical operating data where water quality fluctuates and recovers under normal operating conditions. The time taken for each event from the initial deviation from normal water quality to recovery to stable water quality is calculated. The average duration of all events is then calculated, and this average duration is divided by the sampling time interval to obtain the result. The number of sampling points is defined as the length of the time window. In actual processing, the water quality fluctuation data is segmented in chronological order. Within the first time window, a series of consecutive sampling points are selected as the current time window dataset. Fluctuation-related calculations are performed on the water quality fluctuation data within this time window. Specifically, the average value of all water quality samples within the time window is calculated first, and then the average value is subtracted from each sample value to obtain the deviation value of that sample point. The same processing is then performed on the electric field signal data within the same time window to obtain the deviation value of each electric field sample value from its average value. The water quality deviation value and the electric field deviation value at each moment are multiplied to obtain the product value at that moment. The product values ​​at certain times are summed to obtain a product sum. Simultaneously, the squares of all water quality deviation values ​​within the time window are squared and summed to obtain a water quality deviation sum of squares. The squares of all electric field deviation values ​​within the time window are squared and summed to obtain an electric field deviation sum of squares. This sum of squares is then divided by the number of sampling points within the time window to obtain a value. The square root of this value is used to obtain the water quality standard deviation. Similarly, the square root of the electric field deviation sum of squares is divided by the number of sampling points within the time window to obtain another value. The square root of this value is used to obtain the electric field standard deviation. The aforementioned product sum is then divided sequentially by the water quality standard deviation and the electric field standard deviation to obtain the correlation coefficient value within that time window. A minimum correlation coefficient threshold is preset, and this threshold value is determined using the following method: Under normal operating conditions, collect historical operational data for no less than 30 days. Divide this data into multiple time windows. Calculate the correlation coefficient for each time window using the aforementioned method. Statistically analyze the distribution of all correlation coefficients and arrange them in ascending order. Find the correlation coefficient value that makes the cumulative number of time windows reach 95% of the total number of time windows. Fix this value as the minimum correlation coefficient threshold. In actual operation, when the correlation coefficient value within a certain time window is lower than this minimum correlation coefficient threshold, it is determined that the data within that time window contains noisy data points. Mark the sampling points with the largest deviation within that time window as noise points and remove them from the sequence. The remaining sampling points are used as the data after preliminary denoising.Subsequently, a sliding motion is used to move the time window along the time axis, with each forward step equal to a portion of the sampling points. The process of calculating the correlation coefficient and removing noise is repeated until all data sequences are processed, resulting in a preliminary denoised data sequence covering the entire time period. After obtaining the preliminary denoised water quality fluctuation data and electric field signal data, frequency response analysis is performed on these two types of data sequences. Specifically: first, the sampling time interval is determined, which is the time difference between two adjacent sampling moments. This time difference is fixed as a constant during the system design phase to ensure that the sampling interval remains unchanged throughout the operation. Then, each data sequence is treated as a time sequence sampled at fixed time intervals, and frequency response analysis is performed. The analysis process transforms the time series into a frequency distribution. Specifically, it involves selecting a set of discrete frequency points, starting with the lowest frequency and ensuring the highest frequency does not exceed the maximum analyzable frequency determined by the sampling time interval. For each frequency point, using the sine and cosine functions of that frequency as a benchmark, different sine and cosine weights are calculated based on the product of the time corresponding to each sampling moment and the current frequency. The value of each sampling point is then multiplied by its corresponding sine and cosine weights, and the sum is applied to all sampling points to obtain the amplitude of the sine and cosine components at that frequency. Finally, the total amplitude of the frequency component is calculated based on these two amplitudes. This process is repeated for all frequency points to obtain a frequency distribution covering the entire frequency range. Frequency distribution map; After obtaining the frequency distribution map, by statistically analyzing the total amplitude of all frequency components, the frequency range in which the cumulative amplitude contribution ratio reaches a preset proportion is identified, thereby determining the signal frequency response range. Within this range, a boundary frequency is selected. The portion below this boundary frequency is defined as the low-frequency part, and the portion above this boundary frequency but not exceeding the maximum analysis frequency is defined as the high-frequency part. The method for determining the boundary frequency is as follows: Perform frequency analysis on historical data under normal operating conditions, sort all frequency components in ascending order of frequency, calculate the energy proportion of each frequency component, and then calculate the cumulative energy proportion starting from the lowest frequency. The frequency corresponding to when the cumulative energy proportion reaches 50% is used as the candidate boundary frequency. High and low frequency reconstructed signals are constructed at several frequency points near the candidate boundary frequency, and the mean square error of the reconstructed signals is calculated. The frequency that minimizes the mean square error is selected as the final boundary frequency. Based on the determined boundary frequency, the frequency components corresponding to the electric field signal sequence are divided into low-frequency and high-frequency parts. By retaining all low-frequency components and attenuating or retaining the high-frequency components according to a set ratio, the segmented electric field signal sequence is obtained. Subsequently, abnormal fluctuation points are detected in the segmented electric field signal sequence. Specifically, the average value of each segment of the electric field signal sequence is calculated first, and then the absolute deviation of each sampling point is calculated. The absolute deviation is the absolute value of the difference between the value of the sampling point and the average value.A pre-set deviation threshold is established. This threshold value is determined by statistically analyzing the distribution of absolute deviations of electric field signals under normal operating conditions and the same processing method. Specifically, electric field signals are collected for a period of time under normal operating conditions. The absolute deviation of each sampling point is calculated in the same manner. All absolute deviations are arranged in ascending order. The absolute deviation value that makes the cumulative number of sampling points reach 99% of the total number of sampling points is found and fixed as the deviation threshold. In actual operation, when the absolute deviation of a sampling point exceeds this deviation threshold, the sampling point is identified as an abnormal fluctuation point. Simultaneously, an abnormal point number threshold is set, which represents the maximum allowable number of abnormal fluctuation points within the corresponding time period. The method for determining the number of abnormal fluctuation points is as follows: multiply the total number of sampling points within a time period by a fixed proportion. This proportion is determined statistically under normal operating conditions. For example, the distribution of the number of abnormal fluctuation points within each time period is statistically analyzed in historical normal operating data. The number of abnormal points that makes the cumulative proportion reach 95% is selected as the number corresponding to the proportion. This number is then divided by the total number of sampling points within the time period to obtain the proportion value. This proportion is multiplied by the total number of sampling points in the current time period and rounded up to obtain the threshold number of abnormal points. In actual operation, when the number of abnormal fluctuation points within a certain time period exceeds the threshold number of abnormal points, it is determined that there is a significant abnormal fluctuation in the electric field signal during that time period, and smoothing filtering needs to be performed. The smoothing filtering uses a moving average method, which has... The process is as follows: Set a moving average window length parameter, which is a fixed number of sampling points. This number is determined by dividing the time window length by two and rounding up to obtain the number of sampling points. For the electric field signal sequence to be smoothed, starting from the first sampling point, take several consecutive sampling points to form the current window. Calculate the arithmetic mean of all sampled values ​​within the window. Replace the original value of the sampling point at the center of the window with this average value. Then, move the window forward one sampling point along the time direction and repeat the above calculation until the entire sequence is processed, resulting in the smoothed electric field signal sequence. After the electric field signal smoothing is completed, the corresponding water quality fluctuation sequence is compared using the same sampling time. The smoothed electric field signal sequence and water quality fluctuation data are matched one by one according to time stamp. When both types of data exist at the same time stamp, they are kept as a pair of data. When only a single data exists at a certain time stamp, the missing value is filled in by linear interpolation based on the data of adjacent time stamps, so that each time stamp has both water quality value and electric field value, thus obtaining the integrated sequence. To determine the stability of the integrated sequence, the variance of the integrated water quality fluctuation sequence is calculated. Specifically, the average value of all water quality samples is calculated first, then the deviation of each sample point from the average value is calculated, the squares of each deviation are summed, and finally the sum of squares is divided by the total number of sample points to obtain the variance value.A pre-set variance threshold is established. This threshold value is determined by statistically analyzing the variance distribution of the integrated sequence under normal operating conditions. Specifically, integrated sequence data is collected over a period of time under normal operating conditions. The variance of each sequence segment is calculated using the method described above. All variance values ​​are arranged in ascending order, and the variance value that makes the cumulative sequence count reach 95% of the total sequence count is selected as the variance threshold. In actual operation, when the variance value of the current integrated sequence is less than or equal to this variance threshold, the integrated sequence is considered a stable sequence and output as the final preliminary smoothed sequence for subsequent processing. When the variance value is greater than the variance threshold, the above denoising and smoothing steps can be repeated until the variance value meets the threshold requirement.

[0022] In one implementation, the initially smoothed water quality fluctuation sequence is first divided into units within the water treatment system. Each unit corresponds to a water quality fluctuation sequence on a unified time axis. Each data point in this sequence includes a timestamp of the sampling time and the corresponding water quality value. During the system configuration phase, each unit is assigned a unique number to ensure accurate differentiation of sequences from different units during subsequent processing. When acquiring the spatial transmission path, for any two units with different numbers, segments of the initially smoothed water quality fluctuation sequence from these two units within the same time range are extracted. These two sequence segments have the same length and their sampling timestamps correspond one-to-one. Then, calculate the average water quality value for one unit's sequence. Subtract the average water quality value from the water quality value at each sampling point in the sequence to obtain the water quality deviation value sequence for that unit at each sampling time. Perform the same operation on the sequence of the other unit to obtain the water quality deviation value sequence for that unit. Multiply the water quality deviation values ​​of the two units at the same timestamp and sum the products at all sampling times to obtain a product sum. Simultaneously, square the water quality deviation value of each of the two units point by point and sum them to obtain two deviation sums of squares. Then, divide the two deviation sums of squares by the total number of sampling points and take the square root to obtain the standard deviation values ​​of the two units. Finally, combine the aforementioned... The product sum is divided by the two standard deviations to obtain a value between negative and positive. This value is the Pearson correlation coefficient between the two units within a specified time range, reflecting the degree of linear correlation between the water quality fluctuations of the two units during that time period. The above calculation process is repeated for all unit pairs within the system, and the Pearson correlation coefficients of all unit pairs are recorded in order of unit number, forming a path weight sequence. Each item in the path weight sequence consists of a pair of unit numbers and the Pearson correlation coefficient of that pair of units. To ensure the practical significance of the subsequently constructed spatial transfer path, during the system commissioning phase, under normal and stable operation of the high-purity water system... A preliminary smoothed water quality fluctuation sequence was continuously collected for no less than 30 days. The Pearson correlation coefficients of all unit pairs in each time period were calculated according to the aforementioned method. The absolute values ​​of these correlation coefficients were taken and sorted from largest to smallest. The ratio of the number of correlation coefficients from the first position to a certain position after sorting was calculated to the total number. When the ratio reached or approached 90, the absolute value of the correlation coefficient at the corresponding position was fixed as the path weight threshold. In subsequent operation, only when the absolute value of the Pearson correlation coefficient of a unit pair was greater than or equal to the path weight threshold was it considered that there was an effective spatial transmission path between the unit pairs, and the record of the unit pair was retained in the path weight sequence.When extracting the time delay distribution characteristics, for each pair of units in the path weight sequence that are determined to have an effective spatial transmission path, preliminary smoothed water quality fluctuation sequence fragments of these two units are extracted again within the aforementioned time range. The sequence of one unit is used as the reference sequence, and the sequence of the other unit is used as the sequence to be translated. Based on the design parameters of the water treatment system, the theoretical minimum transmission time and theoretical maximum transmission time along the water flow path between these two units are obtained. The theoretical minimum transmission time is the shortest time that water stays in the pipes and equipment when flowing from the upstream unit to the downstream unit under rated flow conditions. The theoretical maximum transmission time is the longest residence time under the minimum operating flow conditions. Combining the sampling time interval, the theoretical minimum transmission time and theoretical maximum transmission time are divided by the sampling time interval and rounded down and up, respectively, to obtain the corresponding minimum time offset steps and maximum time offset steps. When calculating the cross-correlation, starting from the minimum time offset steps, the sequence to be translated is shifted forward or backward relative to the reference sequence by the corresponding number of sampling points. For the water quality values ​​of the two sequences at the same time within the overlapping time period after translation, multiply and sum them one by one to obtain the cross-correlation value under that time offset step. Then, increase the time offset step by 1 sampling point step size, and repeat the above translation and product summation process for the new time offset until the time offset step is increased to the maximum time offset step. During this process, record each time offset step and its corresponding cross-correlation value. By comparing the cross-correlation values ​​under all time offset steps, find the time offset step with the largest value, and multiply the time offset step by the sampling time interval to obtain a specific time length. This time length is the delay time for the water quality fluctuation between the upstream and downstream units. This delay time is used as an element in the delay distribution sequence. Repeat the above cross-correlation calculation and peak time offset determination process for all unit pairs with effective spatial transmission paths in the path weight sequence to obtain the complete delay distribution sequence. Each item in the delay distribution sequence consists of a pair of unit numbers and the corresponding delay time.When constructing the dynamic graph of wave propagation, each unit in the water treatment system is assigned a node in the graph, with the node number matching the unit number. For each pair of units satisfying the path weight threshold condition in the path weight sequence and delay distribution sequence, the direction of the directed edge is determined based on the sign of the delay time. When the delay time is positive, it indicates that the disturbance occurred first in the unit with the smaller number and then reached the unit with the larger number; the directed edge direction is set from the unit where the disturbance occurred first to the unit where the disturbance occurred later. When the delay time is negative, the directed edge is set in the opposite direction. When recording the edge weight, the corresponding delay time is used as the weight value of the directed edge, and this value is recorded in the graph data. A complete list of directed edges or adjacency tables is established in the structure, where each row records the starting node number, ending node number, and edge weight value of an edge. During system operation, the initially smoothed water quality fluctuation sequence is updated at fixed time intervals, the path weight sequence and delay distribution sequence are recalculated, and the directed edge list is updated synchronously, making the directed graph dynamically changeable in time, thus forming a dynamic spectrum of fluctuation propagation. When analyzing the propagation direction, the unit near the inlet is designated as the source unit in advance according to the water treatment process flow, and the high-purity water effluent unit near the outlet is designated as the target unit. When executing the depth-first search graph traversal algorithm, the source unit is... Starting with the node, initialize all nodes in the graph structure to an unvisited state. Create a path record stack and push the source cell number onto the stack, simultaneously marking the source cell as visited. Then, read the top node of the stack and search for all outgoing edges from that node in the directed graph. For each outgoing edge, sequentially select its endpoint as the next visited node. When a endpoint node is found that has not yet been marked as visited, push its number onto the stack and mark it as visited. Simultaneously, append that node's number to the current path record. Continue this process, using the endpoint node as the new top node, to find outgoing edges and push them onto the stack, until the current top node no longer has outgoing edges pointing to unvisited nodes. Alternatively, if the current top node of the stack is the target cell, when the top node of the stack is the target cell, output all the node numbers from the source cell to the target cell in the stack in the order of the stack and save them as a directional path record. Then pop the top node from the stack, backtrack to the previous node and continue searching for other unvisited outgoing edges. When the top node of the stack no longer has outgoing edges pointing to unvisited nodes, pop the node and backtrack, until all possible paths have been traversed. Finally, a path set consisting of multiple directional path records is obtained. Arrange the node numbers of each path in the path set in order to form a direction sequence. The direction sequence fully reflects all possible transmission direction links of the disturbance between the cells.When identifying potentially imbalanced nodes, the in-degree and out-degree of each node in the dynamic graph are calculated. The in-degree is the number of directed edges ending at that node, and the out-degree is the number of directed edges starting at that node. When calculating the in-degree and out-degree, only directed edges with an absolute weight greater than or equal to the path weight threshold are counted to avoid interference from weakly correlated paths. The difference between the in-degree and out-degree of each node is calculated, and the absolute value of this difference is used as a measure of the node's imbalance. During the system commissioning phase, under the condition of long-term stable operation of the high-purity water system, the dynamic graph is constructed in the above manner, and the in-degree and out-degree of each node in each update cycle are calculated. The absolute difference value is calculated by collecting data from a large number of update cycles, arranging all absolute difference values ​​in ascending order, and calculating the proportion of the total number of differences from the first absolute difference value to a certain position. When this proportion reaches or approaches 95%, the absolute difference value at that position is fixed as the in-degree / out-degree difference threshold. In actual operation, when the absolute difference between the in-degree and out-degree of a node in the current cycle exceeds this threshold, the node is identified as a potential imbalance node and added to the imbalance node set. Nodes in this imbalance node set are given priority in subsequent control compensation matrix generation and electric field adjustment processes.

[0023] In one implementation, based on the previously obtained wave propagation dynamic map, all wave propagation direction data within the current time window are completely read from the dynamic map within each defined time update period. This wave propagation direction data consists of a set of directed connection records. Each connection record contains the starting unit number, the ending unit number, the spatial propagation path weight value of the connection within the current time window, and the corresponding delay time statistics value of the connection within the current time window. The spatial propagation path weight value and the delay time statistics value have been calculated through the aforementioned steps and will not be re-estimated in this step; they are only used as input. To quantitatively determine the degree of abnormal aggregation of dynamic fluctuations among different units, the cumulative output weight is first calculated for each unit. Specifically, for each unit, the weight values ​​of all directed connections in the dynamic graph originating from that unit are added one by one to obtain the total output weight of that unit within the current time window. Then, the total output weight of all units in the entire system is summed again to obtain the total output weight value of the entire system within the time window. Finally, the total output weight of each unit is divided by the total output weight value of the entire system to obtain the output aggregation ratio of that unit within the time window. This ratio is a definite value between 0 and 1, used to represent the degree of concentration of the fluctuation output undertaken by that unit in the entire system within the current time window. To determine the abnormal aggregation threshold, during the system commissioning phase, under the premise that the high-purity water system is under recognized stable operating conditions, operating data for no less than 30 days is continuously selected. The above output aggregation ratio calculation steps are repeated for each time window, and the maximum value is selected from the output aggregation ratios of all units within each time window to form a set of aggregation degree index sequences. Then, this set of aggregation degree indexes is arranged in ascending order, and the ratio of the number of indexes from the first position to a certain position to the total number of indexes is calculated. When this ratio reaches or exceeds 0.95 for the first time, the aggregation degree index value corresponding to that position is taken as the abnormal aggregation threshold. This abnormal aggregation threshold remains fixed during subsequent online operation. In actual operation, for any current time window, the output aggregation ratio of all units is calculated according to the above method and the maximum value is taken. This maximum value is compared with a pre-fixed abnormal aggregation threshold. When the maximum value is greater than the abnormal aggregation threshold, it is determined that there is an abnormal aggregation phenomenon in the direction of fluctuation transmission in the dynamic spectrum. Then, those units whose output aggregation ratio is greater than the abnormal aggregation threshold are selected from all units in the current time window. These units are added to the high-risk subsequent unit candidate set one by one. Each unit is added only once. Then, the unit number is recorded according to the order of addition to form the initial risk label sequence.To simultaneously consider whether the delay time distribution characteristics are abnormal, under the premise that abnormal clustering has been determined, for each unit in the high-risk subsequent unit candidate set, all directed connection records starting from that unit are extracted from the dynamic graph. The delay time statistics of these connections within the current time window are taken. First, the arithmetic mean of these delay time values ​​is calculated and used as the average delay time of the unit's output path. Then, the delay time of each connection is subtracted from the average delay time to obtain the delay deviation value of each connection. Each delay deviation value is then squared and summed. The sum of squares is divided by the number of connections and the square root is calculated to obtain the standard deviation of the delay time of the unit within the current time window. This standard deviation of delay time is used to quantitatively characterize the delay time distribution of the clustered unit on all output paths. To determine the preset threshold for the delay time distribution, dynamic spectrum data for no less than 30 days was continuously collected during the commissioning phase of the high-purity water system's stable operation. For each unit in each time window, the standard deviation of the delay time was calculated using the method described above. The maximum value of the standard deviation of the delay time within each time window was taken to form a set of maximum delay time standard deviation data. This set of data was arranged in ascending order, and the ratio of the cumulative quantity to the total quantity was calculated sequentially starting from the first data point. When this ratio first reached or exceeded 0.95, the maximum delay time standard deviation value corresponding to that position was taken as the preset threshold for the delay time distribution. This threshold was kept unchanged during actual operation. During online operation, when a time window satisfies both the maximum output aggregation ratio exceeding the abnormal aggregation threshold and the standard deviation of the delay time of at least one high-risk subsequent unit candidate exceeding the preset threshold for delay time distribution, it is considered to simultaneously satisfy the conditions of abnormal aggregation in the direction of fluctuation propagation and delay time distribution characteristics exceeding the preset threshold. All units simultaneously satisfying these two conditions are extracted from the candidate set to form the final high-risk subsequent unit set, which is then rearranged in ascending order of unit number to form a new risk labeling sequence. Each element in this risk labeling sequence is a high-risk subsequent unit number, and each number appears only once in the sequence. For each high-risk subsequent unit in this risk labeling sequence, to extract spatial propagation path weight data and calculate the propagation delay value, a continuous time window sequence is selected from the historical records of the dynamic spectrum. The length of this time window sequence is set to a fixed value during the system design phase based on control response requirements, for example, selecting several recent time windows. This fixed number of windows is recorded as the historical window length parameter. This parameter is determined through repeated experiments before the system is put into operation based on the system's design target for fluctuation response time, and once determined, it remains unchanged during normal operation.Then, for each high-risk downstream unit, all directed connection records originating from any upstream unit and ending at that high-risk downstream unit are retrieved from the dynamic graph records corresponding to the historical window. The spatial transmission path weight values ​​of each connection within each historical time window are arranged in chronological order to obtain the path weight time series of that connection. Simultaneously, the path weight time series of all paths entering the connection within the same time period of that high-risk downstream unit are taken and recorded one by one for subsequent cross-correlation calculations. When calculating the path transmission delay value from a certain upstream unit to a high-risk downstream unit, the path weight time series of the upstream unit is used as the reference sequence, and the path weight time series corresponding to the high-risk downstream unit is used as the sequence to be translated. The sampling time interval has been determined as a fixed time value during the system design stage based on the sensor sampling capability and system dynamic characteristics, and will not be changed in this calculation step. Based on the hydraulic design parameters of the high-purity water system, the theoretical minimum and maximum transmission times between the upstream unit and the high-risk downstream unit are calculated first, according to the pipeline length, flow range, and residence volume within the equipment. The theoretical minimum transmission time is divided by the sampling time interval and then rounded down to obtain the minimum translation steps. The theoretical maximum transmission time is divided by the sampling time interval and then rounded up to obtain the maximum translation steps. Both the minimum and maximum translation steps are definite non-negative integers. Then, starting from the minimum translation step, the sequence to be translated is shifted forward or backward by one step relative to the reference sequence on the time axis. During each specific translation, only the temporal overlap between the two sequences is retained. Within this overlap period, for each time point, the path weight value at that time point is extracted from the reference sequence, and the corresponding path weight value at the same time point is extracted from the sequence to be translated. These two values ​​are multiplied, and the product values ​​at all time points within this period are summed one by one to obtain a cross-correlation cumulative value. This cross-correlation cumulative value is bound to the current translation step number. Then, the translation step number is incremented by 1, and the process of translation, truncation of overlap, and point-by-point multiplication and summation is repeated until the translation step number increases to the maximum translation step number. Finally, the cross-correlation cumulative value corresponding to each translation step number from the minimum to the maximum translation step number is obtained. This set of cross-correlation cumulative values ​​is compared, and the translation step number with the largest value is found. This translation step number is multiplied by the sampling time interval to obtain a specific time length, which is the transmission delay value of the spatial transmission path. For each upstream path of each high-risk subsequent unit in the risk-marked sequence, repeat the above cross-correlation calculation and peak shift step determination process to obtain the transmission delay value of all related paths. Arrange these delay values ​​in sequence according to the path identification information and time window order to form a delay calculation sequence. Each element in the delay calculation sequence consists of the starting unit number, the ending unit number, and the transmission delay time value of the corresponding path.To analyze the delay time distribution characteristics and construct a delay distribution model based on the delay calculation sequence, all transmission delay time values ​​in the delay calculation sequence are first treated as a sample. These delay time values ​​are summed one by one and divided by the total number of delay time values ​​to obtain the average delay time. This average value serves as the central parameter of the delay distribution model. Then, the delay deviation is obtained by subtracting each delay time value from the central parameter. All delay deviations are squared and summed. The square root of the sum of squares divided by the total number of delay time values ​​is used to obtain the standard deviation of the delay time. This standard deviation serves as the diffusion parameter of the delay distribution model. After determining the central parameter and diffusion parameter, the time axis is divided into several equally wide intervals with the central parameter as the midpoint. The interval width is set by dividing the diffusion parameter by a fixed positive integer. This positive integer was determined through multiple experiments and error analysis based on the number of delay samples and the distribution width during the initial system design phase and remains fixed after determination, thus ensuring that the width of each interval is a definite value. Then, for each interval, the number of delay time values ​​contained within that interval is counted. This number is divided by the total number of delay time samples to obtain the relative frequency of that interval. The center time position of the interval and the relative frequency are combined into a pair of data points. The center time and relative frequency data points of all intervals are collected. These data points are then fitted to a curve that satisfies a Gaussian distribution using numerical methods. During the fitting process, the center position and diffusion width of the curve are constrained to be equal to the aforementioned center parameter and diffusion parameter, respectively. The curve height is adjusted using the least squares error criterion to minimize the error between the value of the fitted curve at the center time of each interval and the relative frequency of that interval. After fitting, all local maxima are found on the obtained Gaussian curve from the minimum to the maximum value of the time axis. At each local maxima, the corresponding delay time value and the function value of the fitted curve at that point are recorded. The ordered set of delay times and corresponding function values ​​of all local maxima is defined as the peak distribution set, where each peak distribution element consists of peak delay time and peak height.Next, a detailed delay distribution map is generated by fusing the connection relationships between units through the peak distribution set. For each path entry in the delay calculation sequence, the position of the corresponding transmission delay time in the peak distribution set is first found, and the absolute difference between the delay time and each peak delay time is calculated. The peak with the smallest absolute difference is selected, and the peak height corresponding to the peak is used as the time weight factor for the path. Then, within all time windows corresponding to the historical window length parameter, the spatial transmission path weight values ​​of the path in each time window are added one by one. The sum is then divided by the historical window length parameter to obtain the average path weight value of the path in the time dimension. Subsequently, the average path weight value is multiplied by the aforementioned time weight factor to obtain the connection strength value of the path in the detailed delay distribution map. When there are multiple records of the same pair of start and end units in different time windows, the connection strength values ​​of these records are added together and divided by the number of records to obtain the total connection strength value between the pair of units. The above steps are performed sequentially on all paths in the delay calculation sequence. Finally, a connection strength value and a transmission delay time value are determined for each pair of units with a spatial transmission relationship. Each unit is considered a node in the graph, and the transmission relationship between each pair of units is considered a directed edge in the graph. The connection strength value and the transmission delay time value are appended to this directed edge, thus obtaining a detailed delay distribution graph. To obtain potential wave source nodes in this graph, the output strength and input strength are calculated for each unit node in the detailed delay distribution graph. The output strength is the sum of the connection strength values ​​of all directed edges originating from that node, and the input strength is the sum of the connection strength values ​​of all directed edges ending at that node. The output strength is subtracted from the input strength to obtain the strength difference, which is used to measure the source degree of the node in wave transmission. To determine the source identification threshold, during the stable debugging phase of the system, the output intensity, input intensity, and intensity difference are calculated for each node in each time window using the method described above. The node with the largest intensity difference is selected within each time window, and these maximum intensity differences are grouped into a sample. The samples are arranged in ascending order, and the ratio of the number of samples from the first sample to a certain position to the total number of samples is calculated. When this ratio first reaches or exceeds 0.95, the intensity difference corresponding to that position is taken as the source identification threshold. This threshold remains fixed in subsequent operations. During online operation, the intensity difference for the current time window is calculated for each node in the detailed delay distribution map. When the intensity difference of a node is greater than the source identification threshold, that node is identified as a potential source node, and the numbers of these nodes are arranged in descending order of intensity difference, forming a source node sequence.After obtaining the source node sequence, to assess the influence range of the source nodes on subsequent units and determine the trend of influence expansion, starting from each source node in the source node sequence, the influence is expanded layer by layer downstream along all outgoing directed edges on the detailed delay distribution map. In the first layer, all endpoint units connected to the source nodes are directly regarded as subsequent units of the first layer. For each directed edge originating from the source node, the connection strength of the edge is taken as the path influence strength from the source to the subsequent unit of the first layer, and the propagation delay time of the edge is taken as the path delay time. In the second layer, the influence of the first layer is... Each subsequent unit becomes a new starting point, extending downstream along its outgoing directed edge. For each new path, the influence strength of the previous path is multiplied by the current edge connection strength to obtain the new path influence strength. Simultaneously, the delay time of the previous path is added to the current edge propagation delay time to obtain the new path delay time. This process iterates layer by layer until a unit with no outgoing directed edge is reached or a pre-set maximum number of expansion layers is reached. This maximum number of expansion layers is determined through simulation analysis during the system design phase based on the maximum allowable cascade amplification length in the water treatment process and is set as a fixed positive integer. During the expansion process, when the same downstream unit is visited multiple times via different paths, the path with the maximum path influence strength among all paths is recorded for that unit, along with the total delay time corresponding to that path. After repeating the above expansion and path recording process for all source nodes in the source node sequence, the maximum influence intensity between each downstream unit and each source node, as well as the corresponding propagation level and total delay time, are obtained. All unit nodes and their level information, influence intensity information, and source node numbers are associated to form an expansion trend map. In this expansion trend map, by sorting the nodes along the level direction and the magnitude of influence intensity, the path, speed, and strength of the fluctuation expanding layer by layer from the source node to the subsequent units can be intuitively reflected.

[0024] In one implementation, the process first involves processing the detailed delay distribution map and the electric field signal sequence collected at a uniform sampling time interval. Each directed connection record in the detailed delay distribution map includes at least the connection start unit number, the end unit number, the transmission delay time value of the connection, and the peak height value obtained during the delay distribution model fitting process. The transmission delay time value comes from the aforementioned cross-correlation calculation results of water quality fluctuations or path weights, and the peak height value comes from the Gaussian fitting process of the delay time sample, which is the fitted value at the local maximum position on the delay time axis. The electric field signal sequence is a sequence of electric field intensity values ​​recorded by each unit at a fixed sampling time interval on a uniform time axis. This sampling time interval parameter is determined to be a specific time length through multiple experiments based on the electric field response speed and hydraulic residence time of the electro-deionization unit during the system design phase and remains unchanged during operation. The first step is to extract the distribution peak data from the delay distribution map and perform preliminary alignment with the electric field signal sequence. Specifically, for each directed connection record in the map, the corresponding propagation delay time and peak height are read, and these two values ​​are used as the delay peak data for that path. The delay peak data of the same path in multiple historical time windows are arranged in chronological order to obtain the delay peak time sequence for that path. Subsequently, for each path, the corresponding upstream unit electric field signal sequence and downstream unit electric field signal sequence are extracted from the electric field signal database. A certain length of electric field signal segment is extracted within the time range that matches the delay peak time sequence. This length is determined by the coverage time of the delay peak time sequence plus the theoretical maximum propagation delay time. The theoretical maximum propagation delay time has already been determined as a specific time length based on the hydraulic characteristics of the system in the previous steps, and is directly used here without re-estimation.In the specific calculation, the upstream unit electric field signal segment is regarded as the reference sequence, and the downstream unit electric field signal segment is regarded as the sequence to be translated. The minimum and maximum translation steps are determined by dividing the theoretical minimum and maximum propagation delay times by the sampling time interval and rounding down and up, respectively, to obtain two integers. Starting from the minimum translation step number, the sequence to be translated is shifted forward or backward relative to the reference sequence on the time axis. After one translation step, the portion where the two sequences completely overlap in time is extracted. For each sampling moment in the overlapping portion, the electric field intensity value of the reference sequence at that moment is multiplied by the electric field intensity value of the translated sequence at that moment. All products are accumulated over the entire overlapping interval to obtain the cross-correlation accumulation value corresponding to that translation step number. The translation step number is set with a step size of 1. The number of translation steps is increased sequentially until the maximum number of translation steps is reached. The above multiplication and accumulation process is repeated for each translation step to obtain a set of cross-correlation accumulation values ​​corresponding to each of the minimum and maximum translation steps. Then, the item with the largest value is found in this set of values, and the translation step number corresponding to the item is multiplied by the sampling time interval to obtain a specific time offset value. This time offset value is the peak relative offset of the path in the electric field signal layer. Then, the propagation delay time of the path in the delay distribution spectrum is subtracted from the time offset value to obtain the offset error time length of the path. The offset error time lengths of all paths are arranged in order of path number and time window to form an offset calibration sequence. Each item in the offset calibration sequence consists of a path identifier and the corresponding offset error time length. The second step involves fusing the signal strength attenuation coefficient of the offset calibration sequence and calculating the attenuation adjustment factor. The signal strength attenuation coefficient is determined during the system debugging phase using historical electric field signal data under stable operating conditions. Specifically, for each spatial path, multiple time windows are selected. Within each time window, the average amplitude of the absolute value of the electric field signal of the upstream unit and the average amplitude of the absolute value of the electric field signal of the downstream unit are calculated. The average amplitude of the downstream unit is divided by the average amplitude of the upstream unit to obtain the amplitude ratio within that time window. Then, the amplitude ratios of multiple time windows for the same path are averaged to obtain the average amplitude ratio of the path. The reciprocal of this average amplitude ratio is then fixed as the signal strength attenuation coefficient of the path. This coefficient is a real number greater than zero; the larger the value, the more significant the attenuation. In actual operation, for each path in the offset calibration sequence, the signal strength attenuation coefficient and offset error time length corresponding to that path are read. The offset error time length is divided by the sampling time interval and rounded to the nearest integer to obtain the offset calibration step number. The downstream unit electric field signal segment is then shifted again on the time axis according to the offset calibration step number to align the main change position of the downstream electric field signal with the delay peak recorded in the delay distribution spectrum as much as possible. After the shift, each sample value of the downstream electric field signal segment is multiplied by the signal strength attenuation coefficient of that path to obtain the electric field signal segment after considering attenuation.Subsequently, cross-correlation calculations are performed again on the reference sequence and the sequence that has been offset-calibrated and multiplied by the attenuation coefficient. Within the limited translation step range, the aforementioned point-by-point multiplication and accumulation process is repeated to obtain a new set of cross-correlation accumulation values. The maximum value is found among these values, and the maximum cross-correlation accumulation value is divided by the average value of the cross-correlation accumulation values ​​of the path under stable conditions. This average value is pre-calculated and stored during the debugging phase by averaging multiple stable data segments. The division result is the attenuation adjustment factor of the path. This factor is a dimensionless value. A value greater than 1 indicates that the effective response intensity of the path at the current moment is higher than the average level under stable conditions, and a value less than 1 indicates that it is lower than the average level. The attenuation adjustment factors of all paths are stored in path order, and the electric field signal segment after offset calibration and smooth attenuation processing is used as the adjusted signal sequence of the path. The third step involves assigning data fusion weights and generating a weight vector based on the adjusted signal sequences. Specifically, within the current time window, for each water treatment unit, all adjusted signal sequences corresponding to paths ending at that unit are collected. Within the time range of the current time window, the absolute average amplitude of each adjusted signal sequence is calculated. This average amplitude is multiplied by the attenuation adjustment factor of the corresponding path to obtain the basic weight value of that path within the time window. Then, the basic weight values ​​of all paths in that unit are added together to obtain the sum of the basic weights of that unit. Finally, the basic weight value of each path is divided by the sum of the basic weights to obtain the standardized weight of each path at that unit. This standardized weight is a value between zero and one, and the sum of all standardized weights pointing to that unit equals 1. The above calculation process is repeated for all units in the entire system, and all standardized weights are arranged in path order to form a weight vector. To avoid deviations of the weight vector sum from the ideal value due to normalization or rounding during numerical calculations, a large amount of stable operating data is used during the system debugging phase to calculate the weight vector sum for each time window. These sums are sorted in ascending order, and the value at the 95th percentile of the total sum is selected as the preset threshold for the weight vector sum. This threshold is typically slightly greater than 1 and remains unchanged during operation. In actual operation, after calculating the weight vector for the current time window, all elements in the weight vector are added one by one to obtain the current weight sum. The current weight sum is compared with the preset threshold. If the current weight sum is less than or equal to the preset threshold, the weight vector remains unchanged; if the current weight sum is greater than the preset threshold, a scaling factor is calculated by dividing the current weight sum by the preset threshold. Each element in the weight vector is then divided by the scaling factor to obtain a scaled weight vector. The sum of all scaled elements is exactly equal to the preset threshold.Subsequently, the adjusted signal sequence is weighted and fused using the final weight vector. For each unit at each sampling time, the values ​​of all adjusted signals pointing to that unit at that time are multiplied by the corresponding weights and accumulated to obtain the fused electric field signal value of that unit at that time. The fused electric field signals of all units over the entire time window are arranged in unit order and time order to form a fused signal matrix. Each element of this matrix is ​​the fused electric field signal value of a certain unit at a certain sampling time. The fourth step is to construct a control compensation matrix through the fused signal matrix and select the mapping sequence according to the matrix element mapping rules and fusion matching rules. When constructing the control compensation matrix, an electric field signal target benchmark is required. This benchmark is determined in the system design stage through comprehensive simulation and experimentation of high-purity water effluent indicators, influent water quality range, and the performance of the electro-deionization device. It is represented by the target electric field intensity curve of each unit under different operating conditions and is stored in the system parameter library as one of the preset conditions during operation. For each element in the fusion signal matrix within a certain time window, the fusion electric field signal value corresponding to the unit and time is extracted. Then, the target electric field strength value that the unit should reach at that time is read from the target electric field strength curve. The field strength deviation is obtained by subtracting the target electric field strength value from the fusion electric field signal value. The field strength deviation is obtained by dividing the target electric field strength value by the target electric field strength value. The relative deviation is multiplied by an adjustment proportional coefficient. This adjustment proportional coefficient is determined through closed-loop control experiments during the system debugging phase and is a fixed constant. Its magnitude reflects the sensitivity of the control action to the deviation. The above product is the compensation value for the corresponding unit and time in the control compensation matrix. A positive compensation value indicates that the electric field strength needs to be increased at that unit and time, while a negative value indicates that the electric field strength needs to be decreased. After obtaining the control compensation matrix, all elements in the matrix are checked according to the preset matrix element mapping rules and fusion matching rules. The matrix element mapping rules include at least three parts: deviation direction mapping rules, amplitude limit rules, and delay correspondence rules. The deviation direction mapping rules stipulate that when the water quality test results show that the current water purity is lower than the target value, the control compensation value on the corresponding influence path must be positive and negative values ​​are prohibited. When the water purity is higher than the target value and the energy consumption is too high, the corresponding control compensation value must be negative and positive values ​​are prohibited. The amplitude limit rules stipulate that the absolute value of each element in the control compensation matrix must not exceed the maximum allowable adjustment ratio of the unit. This maximum adjustment ratio has been determined as a specific percentage value based on the electrode withstand voltage and water temperature conditions during equipment selection and insulation verification. The delay correspondence rules stipulate that the control compensation value of the unit is allowed to remain non-zero at the corresponding time only when the actual time interval between a unit and the disturbance source and the transmission delay time recorded in the delay distribution map fall within the allowable error range. The upper and lower limits of the allowable error range are determined as two specific time lengths by taking the middle 95% range of the delay time of the path in the delay calculation sequence during the delay statistics stage.The fusion matching rule is used to avoid excessive continuous amplification or attenuation of multiple adjacent units along a spatial path at the same time. Specifically, if the compensation value of an upstream unit is positive and its absolute value exceeds half of its maximum adjustment ratio on the same path, the absolute value of the compensation value of the immediately adjacent downstream unit must not exceed the absolute value of the upstream unit's compensation value. Conversely, if the compensation value of an upstream unit is negative and its absolute value exceeds half of its maximum adjustment ratio, the absolute value of the compensation value of the immediately adjacent downstream unit must not exceed the absolute value of the upstream unit's compensation value. Each element in the matrix is ​​checked against these rules. When an element simultaneously satisfies the deviation direction mapping rule, amplitude limit rule, and delay correspondence rule, and does not violate the fusion matching rule within consecutive units on the same path, that element is considered to conform to all mapping rules. Its corresponding unit number, time index, and compensation value are recorded as a mapping conformance record. All mapping conformance records are arranged in temporal and spatial order to form a mapping conformance sequence. The final step involves evaluating the pre-defined conditions for the mapping conformity sequence and generating targeted control criteria. These pre-defined conditions, determined during the system design phase, include three parts: coverage conditions, priority conditions, and time smoothing conditions. The coverage condition requires that each unit in the set of subsequent units currently marked as high-risk by the delay distribution map has at least one record in the mapping conformity sequence; otherwise, the compensation value of the high-risk unit needs to be increased by reducing the compensation amplitude of less important units and reallocating weights. The priority condition requires that the absolute value of the compensation value of units close to the fluctuation source and with a large connection strength in the detailed delay distribution map should not be less than the absolute value of the compensation value of more distant downstream units on the same path. This priority weight is determined according to the connection strength ranking during the commissioning phase. The time smoothing condition requires that the absolute value of the compensation value change of the same unit within adjacent time windows should not exceed a change threshold. This change threshold is determined by statistically analyzing the changes in electric field regulation amplitude during long-term operation under stable conditions, taking the maximum change within the middle 95% range as the specific value. In actual operation, each record in the mapping sequence is checked sequentially for the above-mentioned preset conditions. When a record meets the coverage condition, priority condition, and time smoothing condition, the record is converted into a targeted control basis. This control basis includes at least the target unit identifier, the action time window, the required adjustment ratio of the electric field intensity, and the adjustment direction. All control basis entries that meet the conditions are sorted according to time order and spatial priority to form a control basis set. This set is directly used as the input of the subsequent electric field control execution module, enabling the system to implement precise compensation based on the fused signal characteristics and delay characteristics of each unit.

[0025] In one implementation, at the beginning of each control cycle, the values ​​of matrix elements are read one by one from the control compensation matrix. Each matrix element uniquely corresponds to the electric field intensity adjustment ratio of a specific unit in the water treatment system within the current control cycle. This ratio is a dimensionless value; a positive value indicates that the electric field intensity needs to be increased, and a negative value indicates that the electric field intensity needs to be decreased. To determine the required matrix element threshold in the "matrix element threshold judgment condition," during the system commissioning phase, the high-purity water system is continuously operated for at least several days under the condition that the changes in influent water quality, flow rate, and load are all within a recognized stable range. During this period, the control compensation matrix generated in each control cycle is recorded. All non-zero valid matrix elements that have passed the aforementioned mapping and matching rules are extracted. The absolute values ​​of these valid elements are grouped into a sample, and this sample is sorted in ascending order. Starting from the first sample, the samples are accumulated sequentially. Dividing this number by the total number of samples yields a cumulative ratio. When the cumulative ratio first reaches or exceeds 95%, the absolute value of the corresponding effective matrix element is taken as the "matrix element threshold," and this value is fixed in the control system parameters, remaining unchanged in subsequent operations. During online operation, in each control cycle, for each matrix element in the control compensation matrix, it is first determined whether the element is greater than zero, and then its absolute value is determined whether it is greater than or equal to the aforementioned matrix element threshold. Only when both conditions of being greater than zero and having an absolute value not less than the matrix element threshold are met simultaneously, the matrix element is considered to represent an enhanced regulation that needs to be actually performed. The unit number, time index, and value of the element are recorded to form a "set of element values ​​that meet the conditions." Each record in the set clearly indicates in which unit and at which time position the electric field signal intensity needs to be increased. Subsequently, the electric field signal strength is adjusted for the set of element values ​​that meet the conditions. The "signal strength adjustment amplitude" parameter is used to limit the maximum allowable change ratio of the electric field strength of each unit in a single control cycle. This parameter is determined during the debugging phase through the following steps: Under safe monitoring conditions, a typical unit is selected, and the influent water quality and flow rate are kept constant. The electric field strength of the unit is gradually increased at a fixed ratio. After each increase, the system is allowed to reach a new steady state. The electrode voltage, operating temperature, and corresponding water quality indicators of the unit are recorded. It is also monitored whether the water temperature approaches the upper limit of the equipment, the electrode voltage approaches the insulation limit, or the water quality shows obvious overshoot. When any of the above situations occurs for the first time, the change ratio of the electric field strength relative to the initial strength is recorded. This ratio is used as the safe upper limit change ratio of this type of unit. The experiment is repeated on multiple units with the same structure, and the minimum value is taken as the "signal strength adjustment amplitude" of this type of unit. The numerical form is then written into the control system parameters.The "signal gain dynamic range" is used to map the dimensionless adjustment ratio in the control compensation matrix to the actual electric field gain factor. It includes two parameters: "minimum gain" and "maximum gain," which are determined during the commissioning phase through simulation and trial operation. First, a series of representative operating conditions are selected, and the elements of the control compensation matrix are set to different positive values ​​within a preset range. Several candidate gain factors are set for each element. After running for a period of time, the convergence speed of water quality deviation and energy consumption are compared. The minimum and maximum gains available under the premise of ensuring stable water quality convergence without excessively increasing energy consumption are recorded. These two determined values ​​are used as the "minimum gain" and "maximum gain," respectively, and are also fixed in the system parameters. During actual operation, for each record in the set of element values ​​that meet the conditions, the positive value of the corresponding matrix element, the signal strength adjustment amplitude of the unit of that type at that moment, and the upper and lower limits of the signal gain dynamic range are read. Then, the maximum value of all matrix elements that meet the conditions within the current control cycle is calculated, and the current element's... Dividing the value by this maximum value yields a "normalized intensity ratio" between zero and one. This ratio reflects the relative importance of the current unit among all units requiring enhancement. The normalized intensity ratio is then multiplied by the signal strength adjustment amplitude of the unit to obtain the target field strength change ratio for the unit in this period. This change ratio is a fixed percentage. The normalized intensity ratio is then multiplied by the difference between the maximum and minimum gain, and added to the minimum gain to obtain the electric field gain coefficient for the unit in this period. Subsequently, all sampling points within the corresponding time window of the unit are found in the electric field signal sequence. The original electric field strength value of each sampling point is multiplied by this electric field gain coefficient. If multiple matrix element records satisfying the conditions exist for the unit in the same time window, the element with the largest value is taken as the calculation basis. After this operation, the "adjusted electric field signal" for each unit in this period is obtained. These adjusted electric field signals are completely consistent with the original electric field signal in terms of time sampling points and indices, only changing in amplitude. To avoid over- or under-adjustment caused by single gain calculations, the adjusted electric field signal undergoes "feedback loop correction parameter optimization processing." At the end of each control cycle, the actual effluent conductivity or target ion concentration parameters at the end of that cycle are read through the water quality sensor array and the difference is calculated with the pre-set target water quality value to obtain the "water quality deviation" for that cycle. Simultaneously, the water quality deviation of the previous cycle is saved in the control system for comparison. During the debugging phase, a "deviation improvement threshold" is pre-set. This threshold is determined by statistically analyzing the typical single-cycle deviation decrease ratio during the stable adjustment process. Specifically, in multiple sets of experiments, the decrease ratio of the water quality deviation in each cycle relative to the previous cycle is recorded, these ratios are sorted from smallest to largest, and the ratio value at the middle position is taken as the deviation improvement threshold. In addition, a "normal fluctuation threshold" is set, which is obtained by statistically analyzing the upper limit of the water quality deviation caused by natural fluctuations in the system when no adjustment is required.In the feedback loop, two adjustable parameters are set for each unit: "field strength gain correction coefficient" and "adjustment amplitude correction coefficient," both initially set to 1. When the percentage decrease in the absolute value of the water quality deviation from the previous cycle is less than the deviation improvement threshold, it indicates insufficient enhancement effect. The field strength gain correction coefficient of all participating units is increased by a fixed proportion, for example, by multiplying the coefficient by a constant factor slightly greater than 1. This constant factor is predetermined during the debugging process based on the system's expected adjustment sensitivity. When the sign of the water quality deviation reverses between two adjacent cycles and the absolute value increases instead of decreasing, it indicates over-adjustment in the previous cycle. The field strength gain correction coefficient of the units participating in this path control is decreased by a fixed proportion, for example, by multiplying it by a constant factor less than 1. This constant factor is also determined during the debugging phase through simulation. Verification confirms that, for the adjustment amplitude correction coefficient, when the water quality deviation is close to the target value but oscillates slightly between positive and negative values ​​in multiple consecutive cycles, and the absolute value of the oscillation amplitude is less than the normal fluctuation threshold, the adjustment amplitude correction coefficient of the unit is multiplied by a fixed factor less than 1, so that the adjustment amplitude of subsequent cycles gradually decreases, thereby suppressing oscillation; after several control cycle iterations, the field strength gain correction coefficient and the adjustment amplitude correction coefficient of each unit will converge to a set of relatively stable values. The control system combines the current regulation compensation matrix elements, the signal gain dynamic range, the signal strength adjustment amplitude, and these two correction coefficients to recalculate the electric field strength at each sampling point of each unit, forming a new electric field signal sequence. This sequence is defined as the "optimized signal sequence", which contains the initial regulation intention and the effect of multiple rounds of feedback correction. Subsequently, an optimized signal sequence was used to finely control the phase of the electric field signal to improve dynamic stability. The extraction process of "signal phase offset data" is as follows: During the debugging phase, the working cycle length of the electro-deionization electric field is first determined. This cycle length can be calculated based on the nominal value of the device drive frequency, or it can be obtained by averaging the time interval between two consecutive positive zero value crossings or fixed threshold crossings on the actual signal. This average time length is used as the "cycle length" parameter and written into the system with a specific time value. During online operation, for each unit, within a certain analysis time window, a signal segment containing several complete cycles is selected from the original electric field signal sequence. The segment is scanned point by point, and the position where the electric field signal first exceeds a certain fixed percentage threshold during the rise of each cycle is recorded, or the time position when it changes from a negative value to a non-negative value is recorded. The time difference between two adjacent characteristic positions of the same type is used as the cycle length for verification. When the cycle length measured multiple times is within the allowable error range of the preset cycle length, the cycle identification is confirmed to be effective.Select a simple unit that serves as the system control benchmark as the "reference unit." Mark the characteristic positions of the same cycle in the signal segments of this unit and any other unit. Calculate the time difference between the characteristic positions of other units and the characteristic positions of the reference unit. Divide this time difference by the cycle length and multiply by 360 to obtain the phase offset angle of this unit relative to the reference unit. Arrange these phase offset angles in an orderly manner according to time windows to form "phase offset data." When calculating the "phase calibration factor" using the optimized signal sequence, compare the main periodic characteristic positions of each unit derived from the optimized signal sequence within the same time window with the original phase offset data. Convert the difference between the two into the remaining phase error angle and record the remaining phase error within the current window for each unit. During the debugging phase, by statistically analyzing the phase offset data of the system under stable operating conditions, calculate the absolute value of the phase offset of each unit in multiple time windows. Sort these absolute values ​​and take the cumulative proportion to reach the desired value. The value at 95 percent of the position is used as the "allowable upper limit of phase deviation", and the specific angle value is written into the parameter. During operation, the absolute value of the remaining phase error of each unit is divided by the allowable upper limit of phase deviation to obtain a dimensionless ratio. When this ratio is not greater than 1, the phase error is considered to be within the acceptable range and no calibration is required. When the ratio is greater than 1, calibration is required. The negative value of this ratio is used as the base value of the phase calibration factor. In order to prevent the phase correction from being too large or too small, the "minimum phase adjustment ratio" and "maximum phase adjustment ratio" are predefined in the debugging stage. These two parameters are determined through multiple experiments to determine the reasonable lower limit and upper limit of the single-cycle phase adjustment without introducing oscillation. The specific method is as follows: different phase adjustment ratios are used in the simulation environment to evaluate the system response time and stability. The minimum ratio that can ensure stable convergence and will not cause obvious slow response is selected as the minimum phase adjustment ratio, and the maximum ratio that will not cause overshoot and oscillation is selected as the maximum phase adjustment ratio.In online calculations, the base value of the phase calibration factor is scaled proportionally and truncated within the range defined by the minimum and maximum phase adjustment ratios to obtain the final phase calibration factor. The phase calibration factors of all units are arranged in order of unit number and time window to form a "phase adjustment sequence". The phase calibration factor is multiplied by the period length and the number of sampling points corresponding to the sampling frequency to obtain the number of sampling points that need to be shifted. The direction of the phase error determines whether the optimized signal is shifted forward or backward on the time axis. After shifting point by point, the phase-corrected electric field signal is obtained. Next, "phase fusion weights" are assigned according to the phase adjustment sequence to generate a "phase weight vector." Specifically, within each time window, the absolute value of the remaining phase error of each unit after phase adjustment is first calculated. The reciprocal of this absolute value is taken to obtain the characteristic that the closer to zero the error, the greater the weight. Then, it is multiplied by the average amplitude of the unit in the optimized signal sequence to obtain the "basic phase weight value" of the unit. Then, the basic phase weight values ​​of all units are added to obtain the "sum of basic phase weights." Finally, the basic phase weight value of each unit is divided by the sum of basic phase weights to obtain a set of standardized phase weights. All phase weights are arranged in unit order to form the initial phase weight vector. During the debugging phase, to prevent the numerical amplification problem caused by an excessively large sum of weights, the phase weight vector of all elements within each time window is calculated using stable operating data. The sum of these values ​​is arranged in ascending order, and the value at which the cumulative percentage reaches 95% is selected as the "preset threshold for the sum of phase weight vectors" and fixed for use during the running phase. During actual operation, the phase weight vectors of the current time window are summed. When the sum is not greater than the preset threshold, the weights remain unchanged. When the sum is greater than the preset threshold, the sum is divided by the preset threshold to obtain a scaling factor. Each phase weight is divided by this scaling factor to obtain the final phase weight vector, and its sum is exactly equal to the preset threshold. Then, for each sampling moment, the phase-adjusted signal values ​​of all units at that moment are multiplied by the corresponding phase weights and added together to obtain the "phase-stable signal" value at that moment. The phase-stable signals of all sampling moments within the entire time window are arranged in chronological order to form a "phase-stable signal sequence". To determine whether a phase-stable signal meets the "signal stability preset conditions," two thresholds are defined during the debugging phase: the "amplitude variance threshold" and the "phase change threshold." The amplitude variance threshold is calculated by statistically analyzing the amplitudes of multiple phase-stable signals during stable system operation, calculating the variance of each signal segment, sorting these variance values, and taking the variance value at the 95th percentile of the cumulative proportion as the amplitude variance threshold. The phase change threshold is calculated by statistically analyzing the distribution of the absolute values ​​of the phase differences between adjacent sampling points during stable operation, and taking the absolute value of the phase difference at the 95th percentile of the cumulative proportion as the phase change threshold.During online judgment, for the phase-stable signal sequence within the current time window, the average amplitude of the signal within the entire window is first calculated. Then, the deviation of each sampling point from the average value is calculated. The squares of the deviations are summed and divided by the total number of sampling points to obtain the amplitude variance. The entire sequence is then scanned to calculate the absolute value of the phase difference between any two adjacent sampling points, and the maximum value is recorded. When the amplitude variance is not greater than the amplitude variance threshold and the maximum phase difference is not greater than the phase change threshold, the phase-stable signal within that time window is considered to meet the preset signal stability conditions. Under these conditions, the optimized signal amplitude adjustment parameters, phase calibration factors, and control compensation matrix elements corresponding to each unit within that time window are recorded as "targeted control basis." All records that meet the conditions form a "control basis set," where each record clearly corresponds to the required electric field strength adjustment ratio and phase correction measures for a specific unit and time window.

[0026] The sequence smoothness index is obtained from the optimized signal sequence. Abnormal parts are removed by a preset abnormal signal removal logic to generate a preliminary response sequence. For the preliminary response sequence, the dynamic response of each unit of the water treatment system is simulated. The response difference is calculated by the response mapping relationship between units to obtain a set of response changes. Based on the set of response changes, the iteration cycle is controlled by the number of sequence optimization iterations to adjust the deviation values ​​in the set of response changes and determine the response sequence after iteration. The response sequence after iteration is obtained, and the initial values ​​of the synchronization control parameters are integrated. The parameters of the water treatment units are synchronized by parameter matching rules to obtain a synchronization parameter set. The control stability of the water treatment system is evaluated by the synchronization parameter set, and the stability index is integrated to generate the final set of synchronization control parameters.

[0027] In this embodiment, firstly, within each control cycle, the time series of electric field signals corresponding to each unit of the water treatment system are read one by one from the optimized signal sequence. For each time series, a "sequence smoothness index" is calculated. This index is used to quantify the fluctuation degree of the electric field signal of that unit within the current cycle. The specific calculation process is as follows: First, the electric field intensity values ​​of all continuous sampling points of that unit within the current cycle are extracted in chronological order. The electric field intensity of two adjacent sampling points is subtracted to obtain a set of adjacent difference values. Then, the absolute value of each difference is taken. All absolute values ​​are added together and divided by the total number of adjacent difference values ​​to obtain a certain value. A fixed average change is used as the smoothness index value for the optimized signal sequence. The smaller the smoothness index value, the smoother the signal. During the system debugging phase, to determine the "smoothness anomaly threshold," optimized signal sequences are continuously collected for several days under stable operation of the high-purity water system. For each sequence, the smoothness index is calculated using the above method. All smoothness indices are sorted in ascending order, and the number of indices is accumulated sequentially starting from the first index and divided by the total number. When the accumulated ratio first reaches or exceeds 0.95, the smoothness index value at this time is recorded and fixed as the "smoothness anomaly." "Threshold"; During online operation, a smoothness index is calculated for the optimized signal sequence of each unit and compared with a smoothness anomaly threshold. When the smoothness index of a sequence exceeds the smoothness anomaly threshold, it is considered that there is abnormal fluctuation in the sequence, and abnormal signal removal logic needs to be executed. The abnormal signal removal logic is as follows: recalculate each adjacent difference for the sequence, use the median of the absolute values ​​of all adjacent differences as the "reference change range", and then set an "abnormal change multiple threshold". This threshold is obtained during the debugging phase by statistically analyzing the distribution of the ratio of the absolute value of adjacent differences to the median under stable operating conditions and taking the cumulative proportion. The ratio at the 0.99 position is used as a fixed value. During operation, when the absolute value of an adjacent difference is greater than the product of the reference change amplitude and the abnormal change multiple threshold, the next sampling point corresponding to the adjacent difference is determined as an abnormal point. The electric field signal values ​​of the two normal points before and after the sampling point are replaced by linear interpolation. If the abnormal point is close to the end of the sequence, the average value of the nearest normal points is used to replace the abnormal point. In this way, the entire sequence is checked and abnormal points are removed point by point to obtain the "preliminary response sequence". After all units have completed the abnormal removal, a set of preliminary response sequence is formed.Subsequently, to simulate the dynamic response of each unit in the water treatment system using the preliminary response sequence set, a "unit-to-unit response mapping relationship" was first established during the system debugging phase. This relationship was identified through actual operating data. The specific steps were as follows: Under known electric field signal input conditions, the outlet water quality indicators of each unit, such as conductivity or ion concentration, were recorded. The temporal offset and amplitude correspondence between the change in the electric field signal of a unit and its own outlet water quality change, as well as the changes in the outlet water quality of several downstream units, were statistically analyzed. The "response coefficient" and "response delay steps" of each unit's electric field signal on its own water quality and downstream water quality were calculated using a regression method. The response coefficient is a non-zero real number, and the response delay steps are integer values ​​representing the delay of several sampling points. After debugging, the response coefficient and response delay steps between each pair of units were fixed in tabular form as the "unit-to-unit response mapping relationship." During online simulation, for each time step, the electric field signal values ​​of all units at that time step and several previous time steps were extracted from the preliminary response sequence and combined with the unit-to-unit response mapping relationship. The system predetermines the response coefficients and response delay steps. For each unit, the simulated outlet water quality is calculated as follows: the change in electric field signal of the unit at the current time step and several previous time steps is aligned according to the response delay step number, multiplied by the corresponding response coefficient, and accumulated. At the same time, the change in electric field signal of the upstream unit at the corresponding delay time step is weighted and accumulated according to its response coefficient and delay step number for that unit. The reference water quality value of the unit under steady-state conditions is added to obtain the "simulated water quality response value" of the unit at that time step. The above calculation is repeated for all time steps in a complete cycle to form the simulated water quality time series of each unit. To quantify the difference between the simulated response and the expected response, for each unit at each time step, the difference between the simulated water quality response value of the unit and the target water quality value is calculated. This difference is recorded as the "response deviation". The response deviations of all time steps are arranged in unit and time order to form a "response change set". Each element in the response change set consists of the unit number, time index, and the response deviation value at that time.Next, based on the response change set, an iterative loop of "sequence optimization iteration count control" is executed to gradually reduce the response deviation. During the system debugging phase, the "maximum number of iterations" and the "deviation convergence threshold" are first determined. The maximum number of iterations is determined by simulating the system computation time and response improvement effect under different iteration rounds, and the minimum number of rounds that can significantly improve the response while ensuring real-time performance is selected as a fixed integer. The deviation convergence threshold is determined by statistically analyzing the distribution of the absolute value of the natural deviation of each unit relative to the target water quality under stable operating conditions, and taking the absolute value of the deviation at the 95th percentile position as a fixed value. During online iteration, the current iteration round is initially defined as 1. In each iteration round, each unit in the response change set is... For each element, the absolute value of its current response deviation is compared with a deviation convergence threshold. If the absolute value is less than or equal to the deviation convergence threshold, the response of the element at that time point is considered to have reached an acceptable range, and the electric field signal of the corresponding preliminary response sequence will not be adjusted in this iteration. If the absolute value is greater than the deviation convergence threshold, the electric field signal intensity needs to be adjusted. The specific adjustment process is as follows: In the preliminary response sequence of the corresponding time step of the element, the current electric field signal value is taken out, and the response deviation is divided by a fixed "response correction coefficient" to obtain the suggested electric field correction increment. The response correction coefficient is determined as a specific constant through experiments during the debugging phase based on the system's sensitivity to electric field changes and the water quality response speed. This represents the change in electric field corresponding to a unit water quality deviation. After calculating the electric field correction increment, this increment is limited to the maximum single-step change ratio allowed for the current unit. This maximum single-step change ratio is obtained by adjusting the aforementioned signal strength adjustment amplitude parameter and a safety factor. For example, half of the signal strength adjustment amplitude is taken as the maximum single-step change ratio to ensure that the single adjustment in the simulation is not too large. Then, the direction of electric field adjustment is determined according to the sign of the response deviation. When the response deviation is positive and indicates that the water quality deviates from the target direction due to insufficient purity, the current electric field signal value is increased by the electric field correction increment. When the response deviation is negative and indicates that the water quality deviates from the target direction due to excessive purity, the electric field signal value is decreased by the corresponding increment. For each time step in the iteration round that exceeds the deviation convergence threshold, the electric field signal is adjusted. After the adjustment is completed, the dynamic response simulation and response change set calculation are re-executed using the updated preliminary response sequence to obtain a new round of response change set. The iteration round number is incremented by 1 and the above iteration process is repeated until either of the following stopping conditions is met: first, the absolute value of the response deviation of all units at all time steps is less than or equal to the deviation convergence threshold; second, the iteration round number reaches the preset maximum number of iterations. At this time, the simulated water quality response sequence of each unit corresponding to the current round is taken as the "iterated response sequence", and the electric field signal in the preliminary response sequence corresponding to the current round is taken as the final optimized simulation input.Subsequently, the iterative response sequence is obtained and fused with the "initial values ​​of the synchronization control parameters." The initial values ​​of the synchronization control parameters are the set of basic control parameters for each unit calculated in the preceding steps based on the control compensation matrix and electric field phase stability. These parameters include the target field strength setpoint, allowable adjustment range, and field strength gain correction coefficient formed in the previous cycle for each unit. These parameters participate in the synchronization process as initial values ​​in the current step. The "parameter matching rule" is used to ensure the spatial and temporal coordination and consistency of the parameters of each unit. Specifically, it includes three items: the first item is the process flow direction consistency rule, i.e., along the water flow direction. The target field strength of the downstream unit must not be lower than the minimum allowable field strength of its directly upstream unit minus a fixed safety margin. This safety margin is set to a specific field strength difference value according to different unit types and water quality requirements during the commissioning phase. The second rule is the sensitive unit priority rule, that is, for key units identified as having a significant impact on effluent water quality in the detailed delay distribution map and response mapping relationship, their field strength adjustment shall prioritize meeting the correction requirements proposed by the response sequence after iteration, while non-key units are allowed to appropriately back off during the synchronization process to maintain overall coordination. The third rule is the time smoothing rule, that is, the parameter change amplitude of the same unit within adjacent time windows shall not be lower than that of the upstream unit. If the time smoothing threshold is exceeded, this threshold is determined as a specific proportion value by statistically analyzing the natural fluctuation range of parameters under stable operating conditions. In the actual synchronization process, for each unit, the simulated water quality response value at each time step in the iterative response sequence is first read and compared with the target water quality value to calculate the latest response deviation. Then, combined with the target field strength setpoint and field strength gain correction coefficient in the initial value of the unit's synchronization control parameters, the target field strength setpoint is fine-tuned according to the direction of the response deviation. The adjustment range is determined by the ratio of the response deviation to the response correction coefficient and the maximum single-step change proportion limited by the time smoothing rule. The target field strength value of the unit in the current synchronization operation is determined. After generating the target field strength value for all units, it is checked one by one whether it conforms to the process flow consistency rule and the sensitive unit priority rule. When it is found that the target field strength value of the downstream unit is lower than the allowable lower limit, the target field strength values ​​at both the upstream and downstream ends are adjusted to the position that meets the safety margin requirements. For sensitive units, the target field strength value is retained first and adjusted on non-critical units to restore consistency. After the above matching rule processing, the target field strength, gain correction coefficient and allowable adjustment range of all units in the current cycle constitute the "synchronization parameter group".Finally, the stability of the water treatment system was evaluated using a set of synchronous parameters, and stability indices were integrated to generate the final set of synchronous control parameters. To this end, a set of "stability indices" was predefined during the commissioning phase, including at least the maximum deviation of the simulated effluent quality, the settling time of the simulated response, the maximum overshoot of the water quality indicators during the simulation, and the time fluctuation amplitude of the electric field parameters of each unit. A threshold was set for each index. The threshold was determined by calculating the corresponding index from a large amount of historically successful stable operation data, sorting these index values ​​from smallest to largest, and taking the value at the 95th percentile of the cumulative percentage as the stability threshold for that index. During online evaluation, the effluent quality and the response of each unit along the entire process link were simulated using the iterative response sequence, from which the corresponding parameters for the current cycle were calculated. Stability indices are compared one by one with their corresponding thresholds. When all indices are not greater than the corresponding thresholds, the current synchronization parameter set is considered to meet the stability requirements in a simulation sense. This synchronization parameter set is then directly output as the "final synchronization control parameter set" and used to guide actual electric field control. When one or more stability indices exceed the threshold, without changing the control results of the aforementioned iteration count, the target field strength values ​​and gain correction coefficients of non-critical units in the synchronization parameter set are appropriately adjusted. For example, the parameter variation amplitude of these units is reduced proportionally according to the proportion exceeding the threshold. A rapid simulation is then performed again to verify the indices. This process is repeated until all indices are compressed within the threshold range or the preset upper limit of the number of fine adjustments is reached. The synchronization parameter set at this point is then solidified as the final synchronization control parameter set.

[0028] By applying synchronous parameters, integrating electric field equipment control and unit response simulation, a control signal sequence is generated and determined. Based on the control signal sequence, real-time feedback monitoring is used to obtain and judge the water quality fluctuation sequence. For the water quality fluctuation sequence, a preset threshold comparison is performed through deviation value judgment and signal stability check to obtain the deviation comparison result. From the deviation comparison result, the system stability confirmation and operating parameter acquisition are integrated to generate and determine the final operating parameter set.

[0029] In this embodiment, at the beginning of a control cycle, the parameters in the synchronous control parameter set are applied to the electric field control devices of each unit in the water treatment system, unit by unit. Each record in the synchronous control parameter set contains at least the unit number, the target electric field strength of the unit, the upper and lower limits of the electric field strength adjustment for the unit, and the electric field gain correction coefficient for the unit. During application, the corresponding record is written into the parameter register area of ​​the electric field control device according to the unit number, so that the electric field control device of each unit calculates the actual output electric field according to the target electric field strength and the gain correction coefficient. Specifically, in each sampling cycle, the electric field control device reads the target electric field strength, multiplies it by the correction coefficient under the current operating condition, and obtains the target output electric field strength for the current cycle. Then, this value is compared with the actual output electric field strength of the previous sampling cycle. If the absolute value of the difference is greater than the maximum adjustment range specified by the synchronous control parameter set for the unit in a single cycle, then the adjustment is limited according to the direction of the maximum adjustment range; otherwise, it is directly adjusted to the target value. The output electric field strength is measured. Through this process, a time-varying electric field output sequence is formed for all units in the entire system throughout the control cycle. The electric field output values ​​of all units at all sampling times are arranged in chronological order to obtain the control signal sequence for this cycle. To ensure that the control signal sequence not only reflects the actual output of the equipment but also the simulation results of the unit response, while generating the control signal sequence, the unit dynamic response mapping relationship established in the previous stage is used to recalculate the simulated water quality response of each unit with the electric field output value in the current cycle as input. The simulated water quality value of the corresponding unit in the synchronous control parameter set is compared with the target water quality value. The deviation between the simulated response and the target water quality value is superimposed on the current electric field output through a pre-set weighting method. For example, based on the response coefficient, part of the deviation information is converted into an electric field correction amount and added to the current output. Thus, the influence of equipment control and unit response simulation are considered at each sampling time. Finally, the control signal sequence is determined and used as the input basis for subsequent feedback monitoring. Subsequently, based on the established control signal sequence, a real-time feedback monitoring process is initiated. At each sampling moment, water quality signals, including but not limited to conductivity, ion concentration, or other indicators characterizing fluctuations in high-purity water quality, are collected by an array of water quality sensors deployed at key locations. These measured water quality values ​​are then combined in chronological order to form a water quality fluctuation sequence. To determine the deviation of the water quality fluctuation sequence, target water quality values ​​and allowable deviation ranges are pre-determined for each monitoring point during the system commissioning phase. The target water quality values ​​are obtained through process design and long-term stable operation data statistics. The allowable deviation range is calculated by recording water quality data over a long period under stable operating conditions, calculating the absolute value of the difference between the measured value and the target value at each sampling moment, sorting these absolute values ​​in ascending order, accumulating the number starting from the first value, and dividing by the total number. The cumulative percentage is calculated when it first reaches or exceeds 0.At 95, the absolute value of the difference at this location is taken as the "deviation threshold" for that monitoring point, and it remains fixed in subsequent operations. During real-time operation, at each sampling time, the difference between the measured value in the water quality fluctuation sequence and the target water quality value at that monitoring point is calculated to obtain the "deviation value." The absolute value of the deviation value is compared with the predetermined deviation threshold. If the absolute value of the deviation value is less than or equal to the deviation threshold, the water quality deviation at that time is considered to be within the acceptable range. If the absolute value of the deviation value is greater than the deviation threshold, the deviation at that time is considered to be exceeding the limit. Simultaneously with the deviation judgment, a "signal stability check" is performed on the water quality fluctuation sequence. The signal stability check uses a statistical method for the degree of sequence variation. Specifically, within a fixed-length time sliding window, the variance of each monitoring sequence and the variance of adjacent sampling points are calculated. The variance index is calculated by taking the average of all water quality samples within the window, then calculating the difference between each sample value and the average, summing the squared differences, and dividing by the number of sampling points within the window. The adjacent difference index is calculated by taking the absolute value of the difference between water quality values ​​of adjacent sampling points within the window, and taking the maximum of these absolute values ​​within the window. During the commissioning phase, a "stability variance threshold" and a "stability adjacent difference threshold" are preset for each monitoring sequence. The method for determining these thresholds is to record water quality sequences for a long time when the system is in a deemed stable operating condition, calculate the variance and adjacent difference index for many sliding windows, sort all variance values ​​from smallest to largest, take the variance value with a cumulative proportion reaching 0.95 as the stability variance threshold, and sort all adjacent difference maximum values ​​from smallest to largest, taking the value with a cumulative proportion reaching 0.The value at position 95 is used as the stability adjacent difference threshold. During online judgment, when the variance within a certain time window is less than or equal to the stability variance threshold and the maximum adjacent difference within the window is less than or equal to the stability adjacent difference threshold, the water quality fluctuation within that time window is considered stable; otherwise, the window is considered unstable. Through deviation judgment and stability check, a "deviation comparison result" is made for each time window. The deviation comparison result includes at least whether the deviation values ​​of all monitoring points within the time window do not exceed the deviation threshold and whether the signal stability indicators of all monitoring points simultaneously meet the preset stability threshold conditions. After obtaining the deviation comparison results for each time window, these results are integrated with the system stability confirmation logic to generate the final operating parameter set. The system stability confirmation logic is explicitly defined in the system design phase: the water treatment system is considered to have reached a stable state only when the absolute value of the deviation at each sampling moment of all key monitoring points does not exceed its corresponding deviation threshold within a consecutive number of control cycles, and the signal stability indicators of all time windows within the same consecutive cycle meet the stability variance threshold and the stability adjacent difference threshold conditions. The specific number of "consecutive control cycles" is determined during the commissioning phase by analyzing the system's recovery time to load disturbances. Specifically, in multiple operating condition disturbance tests, the number of control cycles required from the occurrence of the disturbance to the water quality indicators recovering to the deviation threshold range and maintaining stability is counted, and the maximum value is taken as the number of consecutive cycles required for system stability confirmation, fixed as an integer parameter. In actual operation, when the monitoring module detects that the number of control cycles in a recent period has reached this integer, and the signal stability indicators within this period... When all deviation comparison results indicate that both deviation and stability are satisfied, a stability confirmation event is triggered. Simultaneously, the target electric field strength, gain correction coefficient, and allowable adjustment range of all units in the entire system are read from the current synchronous control parameter set. At the same time, the average actual electric field output, average actual water quality response, and various correction parameters obtained from feedback correction for each unit during the most recent stable period are also read. These currently applied parameters and the measured average values ​​from the stable phase are encapsulated together to form the "final operating parameter set." When generating the final operating parameter set, for each unit, the synchronous target electric field strength value used by that unit, the corresponding electric field gain correction coefficient, the average electric field output value during actual steady-state operation, the corresponding steady-state water quality index, and the associated maximum allowable transient adjustment amplitude are explicitly recorded. These parameters serve as the benchmark parameters for subsequent long-term stable operation. During subsequent control cycles, unless a new large-scale disturbance is detected, the control system prioritizes maintaining the setpoints and ranges in these final operating parameter sets.

[0030] A high-purity water steady-state control system based on dynamic equilibrium of electro-deionization field strength is also provided to implement the steps of the high-purity water steady-state control method based on dynamic equilibrium of electro-deionization field strength. The system includes an acquisition and preprocessing module, which acquires water quality fluctuation data and electric field signal data in real time from each unit of the water treatment system through a sensor array, performs preliminary noise reduction on the acquired data using fluctuation correlation analysis within a time window, and extracts the signal frequency response range in segments to obtain a preliminary smoothed water quality fluctuation sequence and electric field signal sequence; a fluctuation transmission analysis module, which analyzes the spatial transmission path weight and delay time distribution characteristics based on the preliminary smoothed water quality fluctuation sequence, constructs a dynamic spectrum of fluctuation transmission, and determines the transmission direction and potential imbalance nodes of fluctuations between units; a high-risk identification module, which marks high-risk subsequent units if the fluctuation transmission direction in the dynamic spectrum shows abnormal clustering and the delay time distribution characteristics exceed a preset threshold, calculates the specific distribution of transmission delay through spatial transmission path weight data, and obtains a detailed delay distribution spectrum; and a regulation and fusion module, which uses the acquired delay distribution... The spectrum and electric field signal sequence are matched and fused. Combining the signal intensity attenuation coefficient and data fusion weight allocation, a control compensation matrix is ​​generated. It is then determined whether the matrix element mapping rules meet the preset conditions to obtain a targeted control basis. In the electric field signal adjustment module, if the element values ​​in the control compensation matrix meet the matrix element threshold judgment conditions and are greater than zero, the electric field signal intensity is adjusted according to the signal intensity adjustment amplitude and the signal gain dynamic range. Through feedback loop correction parameter optimization processing, the adjusted optimized signal sequence is obtained. In the parameter simulation optimization module, for the optimized signal sequence, the dynamic response of each unit of the water treatment system is simulated by combining the smoothness of the adjusted sequence and the abnormal signal elimination logic. The simulation process is controlled by the number of sequence optimization iterations to determine the final synchronous control parameter set. In the steady-state control module, based on the synchronous control parameter set, the electric field control equipment of each unit of the water treatment system is applied. Through real-time feedback loop monitoring of the adjusted water quality fluctuation sequence, if the deviation value is less than the preset threshold and the stability of the adjusted signal meets the requirements, the system is confirmed to have reached a stable state, and the final operating parameter set is obtained.

[0031] In this embodiment, the dispersed electric field control and water quality monitoring links in traditional electro-deionization high-purity water systems are unified into a closed-loop control framework centered on time-domain sequence analysis and spatial propagation path modeling. The acquisition and preprocessing module processes noisy water quality fluctuation signals and electric field signals coupled at different time scales into a pre-smooth, analyzable sequence. Based on this, the fluctuation propagation analysis module identifies disturbance paths and key nodes in the spatial dimension, the high-risk identification module identifies high-risk units and paths that may cause large-scale fluctuations at a combined spatiotemporal scale, and the regulation and fusion module performs precise matching and weighted fusion between the delay distribution and the electric field signal to obtain a regulation and compensation matrix with time-specific and spatial orientation. The electric field signal adjustment module does not directly execute... Instead of a one-time adjustment, a feedback loop is introduced under matrix drive, allowing the electric field strength adjustment to be dynamically corrected according to the water quality response, forming an optimized signal sequence. The parameter simulation optimization module, without directly affecting the equipment, simulates and tests the expected impact of multiple rounds of signal adjustments on the system, extracting the synchronous control parameter set during the iterative convergence process. Finally, the steady-state control module executes the synchronous parameters on the actual equipment and judges whether the system has truly entered the stable range through real-time feedback. Through this chain structure, a closed-loop control is achieved for the entire process from data acquisition, fluctuation analysis, risk identification, control calculation, signal adjustment to steady-state confirmation, enabling the entire high-purity water electrolysis deionization system to maintain small water quality deviations and high operational stability under fluctuations in operating conditions such as influent, water temperature, and load changes.

[0032] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for steady-state control of high-purity water based on dynamic equilibrium of electro-deionization field strength, characterized in that, include: Water quality fluctuation data and electric field signal data are collected in real time from each unit of the water treatment system using a sensor array. The collected data are initially denoised by using fluctuation correlation analysis within a time window. The signal frequency response range is segmented and extracted to obtain a preliminary smooth water quality fluctuation sequence and electric field signal sequence. Based on the preliminary smoothed water quality fluctuation sequence, the spatial transmission path weight and delay time distribution characteristics are analyzed to construct a dynamic spectrum of fluctuation transmission and determine the transmission direction and potential imbalance nodes of fluctuations between units. If the direction of wave propagation in the dynamic graph shows abnormal clustering and the delay time distribution characteristics exceed the preset threshold, then high-risk subsequent units are marked, and the specific distribution of propagation delay is calculated through spatial propagation path weight data to obtain a detailed delay distribution graph. The acquired delay distribution map is matched and fused with the electric field signal sequence. Combined with the signal intensity attenuation coefficient and data fusion weight allocation, a control compensation matrix is ​​generated. The mapping rules of the matrix elements are judged to meet the preset conditions, so as to obtain the basis for targeted control.

2. The method for steady-state control of high-purity water based on dynamic equilibrium of electro-deionization field strength according to claim 1, characterized in that: The process involves real-time acquisition of water quality fluctuation data and electric field signal data from various units of the water treatment system via a sensor array. Preliminary denoising is performed on the acquired data using fluctuation correlation analysis within a time window. Segmented extraction is then performed based on the signal frequency response range to obtain a preliminary smoothed water quality fluctuation sequence and electric field signal sequence, including: The sensor array acquires water quality fluctuation data and electric field signal data from each unit of the water treatment system. Using a preset time window, the fluctuation correlation calculation is performed on the water quality fluctuation data. The fluctuation correlation calculation obtains the preliminary denoised data sequence by calculating the correlation coefficient between data points within the time window. For the data sequence after preliminary denoising, the signal frequency response range is obtained. The frequency distribution of the signal frequency response range is determined by Fourier transform. The high-frequency part and the low-frequency part are extracted in segments to determine the segmented electric field signal sequence. Abnormal fluctuation points are obtained from the segmented electric field signal sequence. Abnormal fluctuation points are determined by detecting points in the sequence that deviate from the average value. If the abnormal fluctuation points exceed a preset threshold, a smoothing filter is applied to the signal sequence. The smoothing filter uses the moving average method to obtain a preliminary smoothed water quality fluctuation sequence. Based on the initially smoothed water quality fluctuation sequence, the electric field signal sequence is integrated. The integration is achieved by merging sequences through synchronized alignment of timestamps. The stability of the integrated sequence is then determined, and the stability is obtained by calculating the variance value to obtain the final initially smoothed sequence.

3. The method for steady-state control of high-purity water based on dynamic equilibrium of electro-deionization field strength according to claim 1, characterized in that: The process of analyzing the spatial propagation path weights and delay time distribution characteristics based on the initially smoothed water quality fluctuation sequence, constructing a dynamic spectrum of fluctuation propagation, and determining the propagation direction and potential imbalance nodes between units includes: Spatial transmission paths are obtained from a pre-smoothed water quality fluctuation sequence, and path weight sequences are obtained by calculating the Pearson correlation coefficient between sequences. To extract the delay time distribution characteristics of the path weight sequence, a cross-correlation function is used to calculate the time offset value. The cross-correlation function takes the path weight sequence pair as input and outputs the offset corresponding to the peak to obtain the delay distribution sequence. A dynamic graph of wave propagation is constructed based on the delayed distribution sequence. The connections between units are represented by a weighted directed graph, where the delayed distribution sequence is used as the edge weight to obtain the graph structure. The transmission direction is analyzed from the graph structure, and the orientation path from the source unit to the target unit is determined by the depth-first search graph traversal algorithm. The input of the depth-first search is the graph structure, and the output is the path set to obtain the direction sequence. To identify potential imbalanced nodes in directional sequences, the method detects nodes in the graph with unbalanced in-degree and out-degree, where the difference between in-degree and out-degree exceeds a preset threshold to obtain a set of imbalanced nodes.

4. The method for steady-state control of high-purity water based on dynamic equilibrium of electro-deionization field strength according to claim 1, characterized in that: If the wave propagation direction in the dynamic spectrum shows abnormal clustering and the delay time distribution characteristics exceed a preset threshold, then high-risk subsequent units are marked, and the specific distribution of propagation delay is calculated using spatial propagation path weight data to obtain a detailed delay distribution spectrum, including: Data on the direction of fluctuation transmission is obtained from the dynamic spectrum to determine the degree of abnormal clustering. If the abnormal clustering exceeds a preset threshold, high-risk subsequent units are marked to obtain a risk marking sequence. Spatial propagation path weight data is extracted from the risk-marked sequence, and the propagation delay value is calculated using a cross-correlation function. The cross-correlation function takes the path weight data as input and obtains the delay value through the peak offset, thus obtaining the delay calculation sequence. Based on the analysis of the delay time distribution characteristics of the delay calculation sequence, a delay distribution model is constructed. The delay distribution model takes the delay calculation sequence as input, determines the distribution parameters through Gaussian fitting, and obtains the set of peak distributions. By fusing the connection relationships between peak distribution sets, a detailed delay distribution map is generated. The fusion calculates the connection strength through weighted averaging, obtains the potential source nodes of fluctuations in the map, and obtains the source node sequence. The influence range of subsequent units is assessed based on the source node sequence, the trend of influence expansion is determined, and an expansion trend map is obtained.

5. The method for steady-state control of high-purity water based on dynamic equilibrium of electro-deionization field strength according to claim 1, characterized in that: The process involves matching and fusing the acquired delay distribution map with the electric field signal sequence, combining the signal intensity attenuation coefficient and data fusion weight allocation to generate a control compensation matrix, and determining whether the matrix element mapping rules meet preset conditions to obtain targeted control criteria, including: Peak distribution data are extracted from the delay distribution map, and initial alignment is performed using an electric field signal sequence. The peak offset is then calculated using a cross-correlation function to obtain the offset calibration sequence. For the attenuation coefficient of the fused signal intensity of the offset calibration sequence, the attenuation adjustment factor is calculated. The cross-correlation function takes the sequence and coefficient as input, and the attenuation process of the fused signal intensity is adjusted by the peak position to determine the adjusted signal sequence. Data fusion weights are assigned based on the adjusted signal sequence, and a weight vector is generated. If the sum of the weight vectors exceeds a preset threshold, the fused sequence and vector are fused by weighted summation to obtain the fused signal matrix. A control compensation matrix is ​​constructed by fusing the signal matrix. The matrix elements are then judged to meet the matrix element mapping rule and fusion matching rule to obtain the mapping sequence. Based on the judgment that the mapping meets the preset conditions for sequence evaluation, a targeted control basis is generated, resulting in a set of control basis.

6. The method for steady-state control of high-purity water based on dynamic equilibrium of electro-deionization field strength according to claim 1, characterized in that: It also includes adjusting the electric field signal intensity based on the signal intensity adjustment amplitude and the signal gain dynamic range if the element values ​​in the adjustment compensation matrix meet the matrix element threshold judgment condition and are greater than zero. This adjustment is achieved through feedback loop correction parameter optimization processing to obtain the optimized signal sequence, specifically including: Obtain element values ​​from the control compensation matrix, determine if the element values ​​meet the matrix element threshold judgment condition and are greater than zero, and obtain the set of element values ​​that meet the condition; For the set of element values ​​that meet the conditions, the electric field signal is adjusted according to the signal strength adjustment amplitude and the signal gain dynamic range to determine the adjusted electric field signal; The adjusted electric field signal is optimized by feedback loop correction parameters to obtain the corrected parameter set and thus the optimized signal sequence. The phase offset data of the fused signal sequence is optimized, and the phase calibration factor is calculated. The phase offset data is extracted from the electric field signal sequence, and the phase adjustment sequence is determined by adjusting the fusion process through the phase difference. Phase fusion weights are assigned according to the phase adjustment sequence to generate a phase weight vector. If the sum of the phase weight vectors exceeds a preset threshold, the phase-stable signal is obtained by weighted summation of the fusion sequence and vector. It is then determined whether the phase-stable signal meets the preset conditions for signal stability, thus obtaining a basis for targeted regulation.

7. The method for steady-state control of high-purity water based on dynamic equilibrium of electro-deionization field strength according to claim 6, characterized in that, It also includes optimizing the signal sequence, combining the adjusted sequence smoothness and outlier removal logic, simulating the dynamic response of each unit in the water treatment system, controlling the simulation process by the number of sequence optimization iterations, and determining the final set of synchronization control parameters, specifically including: The sequence smoothness index is obtained from the optimized signal sequence, and the abnormal part is removed by the preset abnormal signal removal logic to generate a preliminary response sequence. For the initial response sequence, the dynamic response of each unit of the water treatment system is simulated, and the response differences are calculated through the response mapping relationship between the units to obtain the set of response changes; Based on the set of response changes, the iteration loop is controlled by the number of sequence optimization iterations, the deviation value in the set of response changes is adjusted, and the response sequence after iteration is determined. After obtaining the response sequence after iteration, the initial values ​​of the synchronization control parameters are integrated, and the parameters of the water treatment unit are synchronized through parameter matching rules to obtain the synchronization parameter set. By using a set of synchronous parameters, the control stability of the water treatment system is evaluated, and the stability indicators are combined to generate the final set of synchronous control parameters.

8. The method for steady-state control of high-purity water based on dynamic equilibrium of electro-deionization field strength according to claim 7, characterized in that, It also includes electric field control equipment applied to each unit of the water treatment system based on the synchronous control parameter set. Through real-time feedback and cyclic monitoring of the adjusted water quality fluctuation sequence, if the deviation value is less than a preset threshold and the stability of the adjusted signal meets the requirements, the system is confirmed to have reached a stable state, and the final operating parameter set is obtained, specifically including: By applying synchronous parameters, integrating electric field equipment control and unit response simulation, a control signal sequence is generated and determined. Based on the control signal sequence, real-time feedback monitoring is used to obtain the water quality fluctuation sequence and determine the water quality fluctuation sequence.

9. The method for steady-state control of high-purity water based on dynamic equilibrium of electro-deionization field strength according to claim 8, characterized in that: The electric field control equipment applied to each unit of the water treatment system based on the synchronous control parameter set, through real-time feedback and cyclic monitoring of the adjusted water quality fluctuation sequence, confirms that the system has reached a stable state if the deviation value is less than a preset threshold and the stability of the adjusted signal meets the requirements, and obtains the final operating parameter set, further includes: For water quality fluctuation sequences, deviation comparison results are obtained by judging deviation values ​​and checking signal stability, and comparing preset thresholds. Based on the deviation comparison results, the system stability is confirmed and the operating parameters are obtained, and the final operating parameter set is generated and determined.

10. A high-purity water steady-state control system based on strong dynamic equilibrium of an electro-deionization field, used to implement the steps of the high-purity water steady-state control method based on strong dynamic equilibrium of an electro-deionization field as described in any one of claims 1-9, characterized in that, The system includes: The data acquisition and preprocessing module collects water quality fluctuation data and electric field signal data from each unit of the water treatment system in real time through a sensor array. It performs preliminary noise reduction on the acquired data by using fluctuation correlation analysis within a time window and extracts the signal frequency response range in segments to obtain a preliminary smooth water quality fluctuation sequence and electric field signal sequence. The wave propagation analysis module analyzes the spatial propagation path weights and delay time distribution characteristics based on the initially smoothed water quality wave sequence, constructs a dynamic spectrum of wave propagation, and determines the propagation direction and potential imbalance nodes of the wave between units. The high-risk identification module marks high-risk subsequent units if the direction of wave transmission in the dynamic spectrum shows abnormal clustering and the delay time distribution characteristics exceed a preset threshold. It then calculates the specific distribution of transmission delay through spatial transmission path weight data to obtain a detailed delay distribution spectrum. The control and fusion module matches and fuses the acquired delay distribution map with the electric field signal sequence. It combines the signal intensity attenuation coefficient and data fusion weight allocation to generate a control compensation matrix. It then determines whether the matrix element mapping rules meet the preset conditions to obtain a targeted control basis. The electric field signal adjustment module adjusts the electric field signal intensity based on the signal intensity adjustment amplitude and the dynamic range of the signal gain if the element value in the control compensation matrix meets the threshold judgment condition of the matrix element and is greater than zero. The optimized signal sequence is obtained by feedback loop correction parameter optimization processing. The parameter simulation optimization module, for the optimized signal sequence, combines the adjusted sequence smoothness and abnormal signal removal logic to simulate the dynamic response of each unit of the water treatment system. The simulation process is controlled by the number of sequence optimization iterations to determine the final set of synchronous control parameters. The steady-state control module, based on the synchronous control parameter set, is applied to the electric field control equipment of each unit of the water treatment system. Through real-time feedback and cyclic monitoring of the adjusted water quality fluctuation sequence, if the deviation value is less than the preset threshold and the stability of the adjusted signal meets the requirements, the system is confirmed to have reached a stable state, and the final operating parameter set is obtained.

Images