A heavy metal contaminated water remediation system and method combining adsorption and extraction

Abstract

The application discloses a kind of heavy metal pollution water body remediation system and method of adsorption-extraction combination, belong to water pollution control field.The system is equipped with sensing unit, and data is collected in real time by heavy metal concentration sensor;Model unit is based on the collected data and adsorbent saturation degree prediction model, and key information such as adsorbent real-time saturation is output;Decision unit calculates dynamic threshold according to these information, accurately judges the regeneration opportunity and optimizes regeneration parameter;Execution unit executes regeneration operation according to instruction.Feedback unit feeds back data after regeneration to model unit, realizes system dynamic optimization.In method, based on the cooperation of each unit of system, realize the closed-loop operation of data acquisition, processing, decision and feedback.The application effectively improves the heavy metal pollution water body remediation efficiency, accurately controls adsorbent regeneration, reduces operating energy consumption, can better cope with complex and changeable pollution working condition, and guarantees effluent water quality.

Description

Technical Field

[0001] This invention relates to the field of water pollution control, and in particular to a combined adsorption-extraction system and method for the remediation of heavy metal polluted water bodies. Background Technology

[0002] Against the backdrop of rapid industrial development, the problem of heavy metal pollution in water bodies is becoming increasingly severe. From an industry perspective, many industrial sectors, such as chemical, mining, and electroplating, generate large amounts of wastewater containing heavy metals during their production processes. With increasing environmental awareness and increasingly stringent regulations, effective treatment of heavy metal-polluted water has become an urgent industry need. In terms of technology, adsorption methods, due to their advantages of high efficiency, economy, and ease of operation, have been widely used in the treatment of heavy metal wastewater. Common adsorption materials include activated carbon, bentonite, natural minerals, waste resources, and novel functional materials. These adsorbents can be specifically selected based on different water quality conditions and pollutant types. Moreover, with continuous technological advancements, the application scope of adsorption technology is constantly expanding, gradually extending from traditional industrial wastewater treatment to new areas such as domestic sewage and urban water body management.

[0003] However, existing technologies for treating heavy metal-contaminated water still have many problems. On the one hand, some adsorbents lack selectivity, making it difficult to accurately adsorb specific heavy metal ions when faced with polluted water of complex composition. This results in some heavy metals remaining in the treated water that are difficult to remove, failing to achieve the desired purification effect. On the other hand, in the adsorbent regeneration stage, many traditional methods suffer from high energy consumption and complex processes, significantly increasing the cost of adsorbent reuse and limiting the widespread application of adsorption technology in large-scale water treatment. In addition, some treatment technologies lack sufficient adaptability when dealing with water bodies with large fluctuations in water quality and quantity, resulting in unstable treatment effects and difficulty in consistently ensuring that the effluent water quality meets standards. Summary of the Invention

[0004] The purpose of this invention is to overcome one or more deficiencies of the prior art and provide an adsorption-extraction combined system and method for the remediation of heavy metal polluted water.

[0005] The objective of this invention is achieved through the following technical solution:

[0006] A method for remediating heavy metal-contaminated water bodies using a combined adsorption-extraction approach includes the following steps:

[0007] S1. Data Acquisition: Heavy metal concentration sensors, flow meters, and differential pressure sensors are deployed at the inlet and outlet of the adsorption unit to collect data on influent heavy metal concentration, effluent heavy metal concentration, water flow velocity, and differential pressure of the adsorption unit in real time, and to simultaneously acquire historical adsorption capacity and cumulative adsorbent usage time data.

[0008] S2. Data Processing: Based on historical adsorption capacity, real-time concentration, flow rate, pressure difference, and cumulative usage time, a machine learning model for predicting adsorbent saturation is constructed. The model is used to calculate the real-time saturation and remaining adsorption capacity of the adsorbent.

[0009] S3. Decision generation: Calculate the dynamic threshold based on the real-time saturation. When the saturation reaches the dynamic threshold or the remaining adsorption capacity is lower than the safety margin, trigger the regeneration command. Combine the saturation, heavy metal type and historical regeneration data to generate the regeneration parameter combination with the lowest energy consumption.

[0010] S4. Execution and Feedback: Based on the regeneration instructions and parameter combinations, control the regeneration unit to perform the regeneration operation, collect the regenerated data and feed it back to the prediction model, and dynamically adjust the model parameters.

[0011] Furthermore, step S2 includes the following sub-steps:

[0012] S2.1 Collect historical data to form a dataset, which includes historical adsorption amount, concentration at the corresponding time, flow rate, pressure difference, cumulative usage time and actual saturation;

[0013] S2.2 Divide the dataset into a training set and a test set, train the prediction model, and calculate the model prediction bias rate using the test set. The formula is: Bias rate = |Spredicted - Sactual| / Sactual × 100%; where Spredicted is the model prediction saturation and Sactual is the actual saturation.

[0014] S2.3 If the deviation rate exceeds the preset value, adjust the model parameters and retrain until the deviation rate meets the requirements.

[0015] Furthermore, step S3 includes the following sub-steps:

[0016] S3.1 Calculates the dynamic threshold using the following formula:

[0017] Dynamic threshold = Sbase × [1 + k1 × (Cin - Cflat) / Cflat + k2 × tcumulative / ttotal];

[0018] Wherein, Sbase is the basic threshold, k1 is the concentration influence coefficient, Cin is the real-time influent concentration, Clevel is the historical average influent concentration, k2 is the duration influence coefficient, tcumulative is the cumulative usage time, and ttotal is the total designed lifespan of the adsorbent.

[0019] S3.2 Compare the real-time saturation with the dynamic threshold, and generate a regeneration signal when the triggering condition is met;

[0020] S3.3 Based on the regeneration signal, regeneration parameters are generated with the goal of minimizing energy consumption. These parameters include current density and regeneration time. Further, the data preprocessing in step S1 includes outlier removal, and the outlier determination formula is:

[0021] |x-μ|>3σ;

[0022] Where x represents real-time collected data, μ represents the mean of this type of data, and σ represents the standard deviation.

[0023] Furthermore, in step S2, the update cycle of the prediction model is as follows: after collecting n sets of new data, the new data is added to the dataset and the model is retrained, where n is a preset positive integer.

[0024] Furthermore, in step S3, when multiple heavy metals coexist, the regeneration parameters are weighted according to the adsorption equilibrium constant of each heavy metal, as shown in the formula: w i =K i / ΣK j Among them, w i Let K be the parameter weight for the i-th heavy metal. i Its adsorption equilibrium constant, ΣK j This is the sum of the adsorption equilibrium constants of all coexisting heavy metals.

[0025] Furthermore, step S5 is included: real-time monitoring of energy consumption during the regeneration process. If the ratio of actual energy consumption to predicted energy consumption exceeds 1.2, the process returns to step S3 to re-optimize the regeneration parameters.

[0026] Furthermore, during regeneration in step S4, the process is first run at 70% of the optimized parameters, and after a preset time, the adsorbent activity is detected. Then, the regeneration intensity is adjusted according to the degree of activity recovery.

[0027] A combined adsorption-extraction system for the remediation of heavy metal-contaminated water is provided, comprising:

[0028] The sensing unit is used to collect real-time data on influent concentration, effluent concentration, flow rate, and differential pressure.

[0029] The model unit, connected to the sensing unit, runs the adsorbent saturation prediction model and outputs the real-time saturation and remaining adsorption capacity.

[0030] The decision-making unit, connected to the model unit, calculates dynamic thresholds, determines regeneration timing, and generates regeneration parameters.

[0031] The execution unit, connected to the decision-making unit, performs adsorbent regeneration operations based on the regeneration parameters.

[0032] Furthermore, it also includes a feedback unit, which is connected to both the execution unit and the model unit, and is used to feed back the regenerated data to the model unit to update the prediction model.

[0033] The beneficial effects of this invention are:

[0034] (1) Through intelligent sensing and dynamic modeling technology, the precise monitoring and simulation of water pollution characteristics and adsorbent status can be achieved.

[0035] (2) Through dynamic threshold calculation and regeneration parameter optimization technology, the effect of flexibly deciding the timing of regeneration and precisely controlling the regeneration process based on the pollution conditions can be achieved;

[0036] (3) By using feedback unit data feedback and model iteration technology, we can achieve the effect of continuously optimizing system operation, improving repair efficiency and adapting to complex working conditions. Attached Figure Description

[0037] Figure 1 A flowchart of the steps of a combined adsorption-extraction method for remediating heavy metal contaminated water bodies provided by the present invention;

[0038] Figure 2 This is a functional architecture diagram of the adsorption-extraction combined heavy metal polluted water remediation system provided in Embodiment 2 of the present invention;

[0039] Figure 3 This is a schematic diagram of an adsorption-extraction combined heavy metal pollution water remediation system provided in Embodiment 3 of the present invention. Detailed Implementation

[0040] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. 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.

[0041] Example 1

[0042] See Figure 1 This paper provides a method for remediating heavy metal-polluted water bodies using a combined adsorption-extraction approach, comprising the following steps:

[0043] S1. Data Acquisition: Heavy metal concentration sensors, flow meters, and differential pressure sensors are deployed at the inlet and outlet of the adsorption unit to collect data on influent heavy metal concentration, effluent heavy metal concentration, water flow velocity, and differential pressure of the adsorption unit in real time, and to simultaneously acquire historical adsorption capacity and cumulative adsorbent usage time data.

[0044] S2. Data Processing: Based on historical adsorption capacity, real-time concentration, flow rate, pressure difference, and cumulative usage time, a machine learning model for predicting adsorbent saturation is constructed. The model is used to calculate the real-time saturation and remaining adsorption capacity of the adsorbent.

[0045] S3. Decision generation: Calculate the dynamic threshold based on the real-time saturation. When the saturation reaches the dynamic threshold or the remaining adsorption capacity is lower than the safety margin, trigger the regeneration command. Combine the saturation, heavy metal type and historical regeneration data to generate the regeneration parameter combination with the lowest energy consumption.

[0046] S4. Execution and Feedback: Based on the regeneration instructions and parameter combinations, control the regeneration unit to perform the regeneration operation, collect the regenerated data and feed it back to the prediction model, and dynamically adjust the model parameters.

[0047] Step S2 includes the following sub-steps:

[0048] S2.1 Collect historical data to form a dataset, which includes historical adsorption amount, concentration at the corresponding time, flow rate, pressure difference, cumulative usage time and actual saturation;

[0049] S2.2 Divide the dataset into a training set and a test set, train the prediction model, and calculate the model prediction bias rate using the test set. The formula is: Bias rate = |Spredicted - Sactual| / Sactual × 100%; where Spredicted is the model prediction saturation and Sactual is the actual saturation.

[0050] S2.3 If the deviation rate exceeds the preset value, adjust the model parameters and retrain until the deviation rate meets the requirements.

[0051] Step S3 includes the following sub-steps:

[0052] S3.1 Calculates the dynamic threshold using the following formula:

[0053] Dynamic threshold = Sbase × [1 + k1 × (Cin - Cflat) / Cflat + k2 × tcumulative / ttotal];

[0054] Wherein, Sbase is the basic threshold, k1 is the concentration influence coefficient, Cin is the real-time influent concentration, Clevel is the historical average influent concentration, k2 is the duration influence coefficient, tcumulative is the cumulative usage time, and ttotal is the total designed lifespan of the adsorbent.

[0055] S3.2 Compare the real-time saturation with the dynamic threshold, and generate a regeneration signal when the triggering condition is met;

[0056] S3.3 generates regeneration parameters based on the regeneration signal with the goal of minimizing energy consumption. The parameters include current density and regeneration time.

[0057] Step S1 data preprocessing includes outlier removal, and the outlier determination formula is as follows:

[0058] |x-μ|>3σ;

[0059] Where x represents real-time collected data, μ represents the mean of this type of data, and σ represents the standard deviation.

[0060] In step S2, the update cycle of the prediction model is as follows: after collecting n sets of new data, the new data is added to the dataset and the model is retrained, where n is a preset positive integer.

[0061] In step S3, when multiple heavy metals coexist, the regeneration parameter weights are allocated according to the adsorption equilibrium constants of each heavy metal, using the formula: wi=Ki / ΣKj; where wi is the parameter weight of the i-th heavy metal, Ki is its adsorption equilibrium constant, and ΣKj is the sum of the adsorption equilibrium constants of all coexisting heavy metals.

[0062] It also includes step S5: real-time monitoring of energy consumption during the regeneration process. If the ratio of actual energy consumption to predicted energy consumption exceeds 1.2, return to step S3 to re-optimize the regeneration parameters.

[0063] When performing regeneration in step S4, the process is first run at 70% of the optimized parameters, and after a preset time, the adsorbent activity is detected. Then, the regeneration intensity is adjusted according to the degree of activity recovery.

[0064] The specific implementation steps of this method are as follows:

[0065] S1. Data Acquisition:

[0066] Data acquisition is fundamental to the entire remediation approach. By deploying multi-dimensional sensors and data acquisition mechanisms, real-time data on water pollution characteristics and system operational status are obtained, providing reliable input for subsequent model calculations and decision generation. This step includes three sub-steps: sensor deployment, data acquisition, and data preprocessing, as detailed below:

[0067] S1.1 Sensor Deployment: Multiple sensors are deployed in the inlet and outlet pipes of the adsorption unit and at key locations inside the adsorption unit to construct a real-time sensing network, specifically including:

[0068] Heavy metal concentration sensors: Online heavy metal concentration sensors (such as online spectral probes) are deployed at both the inlet and outlet to collect real-time data on the influent heavy metal concentration (Cin) and effluent heavy metal concentration. These sensors can quickly identify the types and concentrations of heavy metals in the water using spectral analysis technology, eliminating the need for offline laboratory testing and enabling real-time data output.

[0069] Flow meter: A flow meter is deployed on the inlet pipe to continuously monitor the flow velocity (v) of the water entering the adsorption unit. The flow velocity data is a key parameter for calculating the heavy metal load per unit time and directly affects the saturation rate of the adsorbent.

[0070] Differential pressure sensor: A differential pressure sensor is deployed between the inlet and outlet of the adsorption unit to monitor the pressure difference (ΔP) between the two ends of the adsorption unit. As the amount of heavy metals adsorbed by the adsorbent increases, the pores inside the adsorption unit are gradually occupied, the water flow resistance increases, and the pressure difference changes accordingly. Therefore, the pressure difference data can indirectly reflect the degree of clogging and adsorption state of the adsorbent.

[0071] All of the aforementioned sensors are connected to the subsequent data processing module via data transmission lines, ensuring that the collected data can be transmitted to the system central hub in real time, providing a foundation for subsequent processing.

[0072] S1.2 Data Acquisition: After the sensor deployment is complete, start the data acquisition program to synchronously acquire the following types of data:

[0073] Real-time dynamic data includes influent heavy metal concentration (Cin), effluent heavy metal concentration, water flow velocity (v), and adsorption unit pressure difference (ΔP). The acquisition frequency is set according to the water pollution fluctuation characteristics to ensure that instantaneous changes in water quality can be captured.

[0074] Historical cumulative data includes the historical adsorption capacity of the adsorption unit (i.e., the total amount of heavy metals adsorbed by the adsorbent over a period of time) and the cumulative usage time of the adsorbent (tcumulative, the total operating time from when the adsorbent was put into use to the present moment). This historical data is retrieved through the system storage module and used as training samples to build machine learning models.

[0075] During data collection, it is necessary to ensure that the timestamps of various types of data are aligned to avoid data misalignment caused by collection delays, which could affect the accuracy of subsequent model calculations.

[0076] S1.3 Data Preprocessing: To ensure the quality of the data input to the subsequent model, the collected real-time and historical data need to be preprocessed, specifically including data cleaning and outlier removal. The outlier removal process is as follows:

[0077] S1.3.1 Outlier Removal: Due to factors such as sensor malfunction and water disturbance, outliers may exist in the real-time collected data (such as concentration values ​​that change instantaneously or flow velocity values ​​that exceed reasonable ranges). These outliers will interfere with the calculation accuracy of subsequent models, so they need to be identified and removed through a preset algorithm.

[0078] Outlier determination uses the formula: |x-μ|>3σ, where:

[0079] x represents a specific data point collected in real time, such as the influent concentration or flow rate at a specific moment.

[0080] μ is the mean of this type of data over the same historical period or the most recent stable period;

[0081] σ is the standard deviation of this type of data over the same historical period or the most recent stable period.

[0082] When a data point x satisfies |x-μ|>3σ, the data point is determined to be an outlier. The system automatically removes it and fills it with the previous valid value or moving average of the data point x to ensure the continuity of the data sequence.

[0083] S1.3.2 Data Standardization: To eliminate dimensional differences between different types of data (e.g., concentration unit is mg / L, flow rate unit is m). 3 / h (pressure difference unit is kPa), the preprocessed data needs to be standardized. The standardization process can use the min-max standardization method to map the data to the [0,1] interval, with the formula: x'=(xx min ) / (x max -x min ), where x is the original data, x min x is the minimum value of this type of data. max Let x' be the maximum value of this type of data, and x' be the standardized data. Standardized data is more suitable as input to machine learning models, improving the convergence speed and computational accuracy of the models.

[0084] S2. Data Processing:

[0085] Data processing is the process of transforming the raw data obtained in step S1 into key parameters that can be used for decision-making (such as real-time adsorbent saturation and remaining adsorption capacity) by building a machine learning model. This step includes four sub-steps: dataset construction, model training, model optimization, and model updating, as detailed below:

[0086] S2.1 Dataset Construction: To train the adsorbent saturation prediction model, a dataset containing historical data and feature parameters needs to be constructed first. The specific process is as follows:

[0087] S2.1.1 Historical Data Collection: The collected historical data includes, but is not limited to:

[0088] Historical adsorption capacity: the total amount of heavy metals actually adsorbed by the adsorbent during a certain period of time in the past;

[0089] Concentration at the corresponding time: including the heavy metal concentration at the inlet and outlet during that time period;

[0090] Flow velocity at corresponding time: the flow velocity of water entering the adsorption unit during this time period;

[0091] Pressure difference at corresponding time: The pressure difference between the two ends of the adsorption unit during this time period;

[0092] Cumulative usage time: The total operating time of the adsorbent up to the end of this period;

[0093] Actual saturation: The actual degree of saturation of the adsorbent at the end of the time period, obtained through offline detection or indirect calculation, that is, the ratio of the amount of heavy metals adsorbed to the maximum adsorption capacity.

[0094] The above data needs to be organized in time series to ensure that each data sample contains a complete "input-output" correspondence (i.e., the correspondence between "historical adsorption capacity, concentration, flow rate, pressure difference, and cumulative usage time" and "actual saturation").

[0095] S2.1.2 Dataset Partitioning: The organized historical dataset is divided into training, validation, and test sets according to a preset ratio. The training set is used for model parameter learning, the validation set is used to adjust model hyperparameters (such as learning rate and number of iterations), and the test set is used to evaluate the model's generalization ability. The partitioning ratio can be flexibly set according to the dataset size. For example, when the dataset has a large sample size, a 7:1:2 ratio can be used (70% training set, 10% validation set, and 20% test set).

[0096] S2.2 Model Training and Bias Rate Calculation: Based on the partitioned dataset, a machine learning model for predicting adsorbent saturation is constructed and trained. The specific process is as follows:

[0097] S2.2.1 Model Selection and Construction: Select machine learning algorithms suitable for time series forecasting (such as random forest, gradient boosting tree, long short-term memory network, etc.) to construct the prediction model. The input features of the model are historical adsorption amount, real-time concentration (including influent and effluent concentration), flow rate, pressure difference, and cumulative usage time; the output of the model is the real-time saturation of the adsorbent and the remaining adsorption capacity (remaining adsorption capacity = maximum adsorption capacity - adsorbed amount, which can be indirectly calculated through real-time saturation).

[0098] During model building, a reasonable network structure needs to be set according to the data characteristics (e.g., for neural network models, the number of hidden layers and the number of neurons per layer need to be determined) to ensure that the model can fully capture the nonlinear relationship between input features and output parameters.

[0099] S2.2.2 Model Training: Input the training set data into the initial model and adjust the model parameters (such as weights and biases) through the backpropagation algorithm to make the predicted saturation of the model output as close as possible to the actual saturation of the training set. During training, the mean squared error (MSE) between the predicted and actual values ​​is used as the loss function, and the loss function is minimized by the gradient descent method until the model's loss on the training set converges to a preset threshold.

[0100] S2.2.3 Bias Rate Calculation and Model Evaluation: After model training is completed, the prediction accuracy of the model is evaluated using test set data. This is specifically achieved by calculating the model prediction bias rate, using the following formula:

[0101] Deviation rate = |Spredicted - Sactual| / Sactual × 100%;

[0102] Where Spre is the predicted saturation output by the model, and Sactual is the actual saturation in the test set.

[0103] After calculating the bias rate for all samples in the test set, the average value is taken as the overall bias rate of the model. If the average bias rate exceeds a preset threshold (e.g., 5%), the model training step needs to be returned to, the model's hyperparameters adjusted (e.g., increasing the number of training iterations, optimizing the network structure) or supplemented with additional dataset samples, and the model retrained until the bias rate meets the preset requirements.

[0104] S2.3 Model Optimization: Model optimization is a crucial step in ensuring that prediction accuracy continues to meet requirements, and specifically includes the following operations:

[0105] When the model's bias rate on the test set does not meet the requirements, first check the completeness and representativeness of the dataset. If there is uneven data distribution (e.g., samples are missing in a certain concentration range), historical data for the corresponding range needs to be added, the dataset needs to be re-divided, and the model needs to be trained.

[0106] If the dataset has no obvious defects, adjust the algorithm parameters of the model. For example, for a random forest model, the number of decision trees can be increased or the depth of the trees can be adjusted; for a neural network model, parameters such as the learning rate and batch size can be adjusted, and the bias rate of the model can be reduced through multiple iterations of optimization.

[0107] During optimization, model performance needs to be monitored in real time using the validation set to avoid overfitting (i.e., the model performs well on the training set, but the bias rate spikes on the validation set). If overfitting occurs, regularization (such as L1 regularization or L2 regularization) or early stopping strategies (stop training when the loss on the validation set no longer decreases) can be used to correct it.

[0108] S2.4 Model Update: As water pollution load and adsorbent performance change over time (e.g., fluctuations in influent concentration and adsorbent aging), the prediction model needs to be updated regularly to adapt to these changes. The specific update cycle is set as follows: after collecting n sets of new data, the new data is added to the dataset and the model is retrained (n is a preset positive integer that can be adjusted according to the frequency of water quality fluctuations).

[0109] The new data includes the latest data collected in real time during step S1 (such as newly added historical adsorption capacity, concentration, flow rate, pressure difference, and cumulative usage time) and the corresponding actual saturation (obtained through subsequent regeneration feedback or offline detection). During the update process, a sliding window mechanism is used to retain the latest dataset samples and remove outdated or potentially unrepresentative data (such as data that has exceeded the adsorbent performance decay period), ensuring that the model is always trained based on the actual state of the current water body and system.

[0110] S3. Decision Generation:

[0111] Decision generation is the process of determining the regeneration timing and parameters based on the real-time saturation and remaining adsorption capacity output from step S2, through dynamic threshold calculation and parameter optimization. This step includes four sub-steps: dynamic threshold calculation, regeneration signal generation, regeneration parameter optimization, and energy consumption monitoring, as detailed below:

[0112] S3.1 Dynamic Threshold Calculation: The dynamic threshold is a core indicator for determining whether the adsorbent needs regeneration. Its value is adaptively adjusted based on factors such as fluctuations in influent concentration and adsorbent usage time. The calculation formula is as follows:

[0113] Dynamic threshold = Sbase × [1 + k1 × (Cin - Cflat) / Cflat + k2 × tcumulative / ttotal];

[0114] The meanings and functions of each parameter are as follows:

[0115] S-base: The basic threshold is the critical value for safe saturation of the adsorbent under ideal conditions (such as fresh adsorbent and stable influent concentration), which is determined by the material properties and design standards of the adsorbent.

[0116] k1: Concentration influence coefficient, a positive number, used to quantify the impact of influent concentration fluctuations on the threshold. When the real-time influent concentration Cinfluent is higher than the historical average influent concentration Caverage, (Cinfluent - Caverage) / Caverage is positive, the dynamic threshold increases, allowing the adsorbent to regenerate at a higher saturation level (because when the influent concentration is high, the adsorbent needs to accommodate more heavy metals to avoid exceeding the effluent standard); when Cinfluent is lower than Caverage, the dynamic threshold decreases, triggering regeneration earlier (because when the influent concentration is low, the adsorbent saturates slowly, and early regeneration can reduce energy consumption);

[0117] Cin: Real-time influent concentration, collected in real time by the heavy metal concentration sensor in step S1;

[0118] C-level: Historical average influent concentration, which is the average influent concentration over a certain period (e.g., 30 days) in the past, and can be calculated from historical data collected in step S1;

[0119] k2: Duration-based influence coefficient, a positive number used to quantify the impact of cumulative adsorbent usage time on the threshold. As the cumulative usage time tcumulative increases (i.e., the adsorbent gradually ages), the tcumulative / ttotal ratio increases, the dynamic threshold decreases, prompting the adsorbent to regenerate earlier (because the adsorption performance of the aged adsorbent decreases, requiring more frequent regeneration to maintain its activity);

[0120] t_cumulative: Cumulative usage time, i.e., the total operating time of the adsorbent from the time it is put into use to the current moment;

[0121] t_total: Total design life of the adsorbent, which is the expected total operating time of the adsorbent under normal maintenance, and is determined by the material properties and the operating environment.

[0122] S3.2 Regeneration Signal Generation: Regeneration signal generation determines whether regeneration is triggered by comparing real-time saturation with the dynamic threshold, remaining adsorption capacity with the safety margin. Specifically, this includes:

[0123] The real-time saturation and remaining adsorption capacity of the adsorbent output from step S2 are obtained in real time.

[0124] Compare the real-time saturation with the dynamic threshold calculated in step S3.1: if the real-time saturation is greater than or equal to the dynamic threshold, then the regeneration trigger condition is met.

[0125] At the same time, the remaining adsorption capacity is compared with the preset safety margin: the safety margin is the minimum remaining adsorption capacity to ensure that the effluent meets the standards (that is, when the remaining adsorption capacity is lower than this value, the adsorbent may not be able to effectively adsorb heavy metals, resulting in the effluent exceeding the standards). If the remaining adsorption capacity is less than or equal to the safety margin, the regeneration triggering condition is met.

[0126] A regeneration signal can be generated if either of the above two conditions is met, and the regeneration signal will be used as input for subsequent parameter optimization; if neither condition is met, the process returns to step S2 and the state of the adsorbent is continuously monitored.

[0127] S3.3 Regeneration Parameter Optimization: Based on the regeneration signal, a combination of regeneration parameters (including current density, regeneration time, etc.) is generated with the goal of minimizing energy consumption. The specific process is as follows:

[0128] S3.3.1 Parameter optimization under single heavy metal pollution: When only one heavy metal exists in the water body, the optimization of regeneration parameters needs to take into account the real-time saturation of the adsorbent and historical regeneration efficiency data, and achieve this by establishing a mapping relationship between energy consumption and parameters.

[0129] Historical regeneration efficiency data includes the correspondence between "current density, regeneration time" and "desorption amount, energy consumption" in past regeneration processes: the desorption amount is the amount of heavy metals released by the adsorbent after regeneration, and the energy consumption is the total energy consumed in the regeneration process (positively correlated with current density and time).

[0130] Using real-time saturation as a constraint (e.g., the higher the saturation, the greater the required desorption amount, requiring a higher current density or a longer duration), and with the minimum energy consumption as the objective function, the system searches for the optimal combination of current density and regeneration time by traversing historical data or using intelligent algorithms (such as genetic algorithms or particle swarm optimization) to ensure that energy consumption is minimized while meeting desorption requirements (i.e., the adsorbent saturation drops to a preset safe value after regeneration).

[0131] S3.3.2 Parameter Optimization for Multiple Heavy Metals Coexisting: When multiple heavy metals exist in the water, the regeneration parameters need to be weighted according to the adsorption equilibrium constant of each heavy metal to ensure that each heavy metal can be effectively desorbed. The specific formula used is: w i =K i / ΣK j ;

[0132] Where wi is the parameter weight of the i-th heavy metal, Ki is the adsorption equilibrium constant of the i-th heavy metal (reflecting the affinity of the adsorbent for the heavy metal; the larger the Ki, the stronger the affinity and the more difficult it is to desorb), and ΣKj is the sum of the adsorption equilibrium constants of all coexisting heavy metals.

[0133] During parameter optimization, the regeneration parameters need to be adjusted according to the weight wi: for heavy metals with a larger wi (i.e., heavy metals that are difficult to desorb), a higher current density or a longer regeneration time ratio is allocated; for heavy metals with a smaller wi (i.e., heavy metals that are easy to desorb), a lower parameter ratio is allocated, ultimately forming a parameter combination that takes into account the desorption needs of multiple heavy metals and has the lowest energy consumption.

[0134] S3.4 Energy Consumption Monitoring During Regeneration: To ensure the optimization effect of regeneration parameters, actual energy consumption needs to be monitored in real time during the regeneration process and compared with the predicted energy consumption (i.e., the energy consumption estimated during the parameter optimization stage). Specifically, this includes:

[0135] The voltage and current data of the regeneration unit are collected in real time. The instantaneous energy consumption during the regeneration process is calculated using the formula "instantaneous energy consumption = voltage × current". The actual total energy consumption is obtained by accumulating the instantaneous energy consumption.

[0136] Calculate the ratio of actual energy consumption to predicted energy consumption: If the ratio is ≤1.2, it means that the regeneration parameter optimization is effective and the regeneration operation continues; if the ratio is >1.2, it means that the actual energy consumption exceeds expectations, which may be due to unreasonable parameter settings (such as excessive current density) or system abnormalities (such as electrode wear), and it is necessary to return to step S3.3 to re-optimize the regeneration parameters.

[0137] The monitoring data will be used as part of the historical regeneration efficiency data for subsequent parameter optimization model training, thereby improving optimization accuracy.

[0138] S4. Implementation and Feedback:

[0139] Execution and feedback involves performing the regeneration operation based on the regeneration parameters generated in step S3, and feeding back the data from the regeneration process to the model, thus achieving dynamic system adjustment. This step includes two sub-steps: regeneration execution and data feedback, as detailed below:

[0140] S4.1 Regeneration Execution: Based on the optimized regeneration parameters from step S3, the regeneration unit is controlled to perform adsorbent regeneration. A segmented execution strategy is adopted to ensure a balance between regeneration efficiency and energy consumption. The specific process is as follows:

[0141] S4.1.1 First stage regeneration: Regeneration is started at 70% of the optimized parameters: For example, if the optimized current density is J and the regeneration time is t, then the first stage uses a current density of 0.7J and continues for a short preset period (such as t1, where t1 is 20%-30% of t).

[0142] The purpose of this stage is to initially desorb easily desorbed heavy metals from the adsorbent surface through low-intensity regeneration, while avoiding energy waste caused by excessively high initial parameters.

[0143] S4.1.2 Adsorbent activity detection: After the first stage, the activity recovery rate of the adsorbent is detected: The activity recovery rate is the ratio of the adsorption capacity of the regenerated adsorbent to the initial capacity of the fresh adsorbent, which can be indirectly calculated by collecting the initial adsorption rate of the regenerated adsorbent.

[0144] If the activity recovery rate is greater than or equal to the preset target value (e.g., 80%), it means that the first stage of regeneration has basically met the requirements and the second stage can begin.

[0145] If the activity recovery rate is less than the preset target value, it indicates that the regeneration intensity needs to be increased, and the process should proceed to the second stage.

[0146] S4.1.3 Second-stage regeneration: Adjust the regeneration intensity of the second stage based on the activity recovery rate of the first stage.

[0147] If the activity recovery rate is greater than or equal to the preset target value, the second stage uses 50%-60% of the optimized parameters at the intensity for the remaining duration (t-t1) to gently desorb the remaining small amount of heavy metals and reduce energy consumption.

[0148] If the activity recovery rate is less than the preset target value, the second stage uses 100%-120% intensity of the optimized parameters and continues for the remaining time to ensure that the adsorbent activity is fully recovered.

[0149] After the second stage, the regeneration operation is completed, and the adsorbent is put back into the adsorption unit for use.

[0150] S4.2 Data Feedback: After the regeneration operation is completed, the various data collected during the regeneration process are fed back to the prediction model in step S2 to achieve dynamic adjustment of the model. Specifically, this includes:

[0151] S4.2.1 Feedback Data Collection: The collected feedback data includes:

[0152] Data on the state of the regenerated adsorbent: such as the actual saturation after regeneration (calculated by detecting the amount of desorption) and the activity recovery rate;

[0153] Operational data of the regeneration process: such as actual current density, actual regeneration time, and total energy consumption;

[0154] Data on the regenerated water body: such as the concentration of heavy metals in the waste liquid discharged from the regeneration unit (reflecting the desorption effect).

[0155] S4.2.2 Model Adjustment: Add the feedback data to the dataset from step S2.1 as new training samples for model updates (see step S2.4). By introducing actual data after regeneration, the model's prediction bias regarding the "adsorption-regeneration" cycle can be corrected, for example:

[0156] If the actual saturation after regeneration deviates significantly from the remaining adsorption capacity predicted by the model, the model can adjust the weight parameters for saturation prediction by learning feedback data.

[0157] If there is a discrepancy between historical regeneration efficiency data and actual energy consumption, the model can optimize the mapping relationship between energy consumption and regeneration parameters to improve the accuracy of subsequent parameter optimization.

[0158] The feedback process forms a closed loop of "adsorption-regeneration-feedback-optimization", ensuring that the system can continuously adapt to changes in water pollution status and adsorbent performance, and maintain long-term stable operation.

[0159] Step S5: Full-cycle system monitoring:

[0160] To further ensure the stability and economy of the remediation method, this method also includes step S5: full-cycle system monitoring, which involves long-term monitoring of the operating status of the adsorption unit, regeneration unit, and data processing module, specifically including:

[0161] Adsorption unit monitoring: Regularly check the calibration status of the sensors (such as the detection accuracy of the concentration sensor) and the physical state of the adsorbent (such as whether there is damage or loss) to ensure the accuracy of data acquisition and the effectiveness of the adsorption process.

[0162] Regeneration unit monitoring: Regularly check electrode performance (such as whether corrosion or scaling occurs) and the stability of the power supply to avoid deviations in regeneration parameters due to hardware failures;

[0163] Data module monitoring: Regularly check the integrity of the data transmission link (e.g., whether there is data loss) and the real-time performance of model calculations (e.g., whether there is delay) to ensure timely decision generation.

[0164] Anomalies detected during monitoring (such as sensor drift or electrode performance degradation) will be input as adjustment items into the parameter optimization process of step S3 (e.g., when the actual current density decreases due to electrode corrosion, the parameter setting value needs to be appropriately increased), or trigger system maintenance operations to ensure that the entire life cycle operating cost of the repair system is in the optimal state.

[0165] This embodiment achieves accurate judgment of adsorbent regeneration timing and dynamic optimization of regeneration parameters through real-time data acquisition, machine learning prediction, dynamic decision optimization, and closed-loop feedback adjustment. This effectively avoids resource waste and the risk of effluent exceeding standards, and significantly reduces system operating costs.

[0166] Example 2

[0167] See Figure 2 This paper presents a combined adsorption-extraction system for the remediation of heavy metal-contaminated water, comprising a core functional layer and an auxiliary feedback layer. The core functional layer consists of a sensing unit, a model unit, a decision-making unit, and an execution unit. The sensing unit connects to a heavy metal concentration sensor, a flow meter, and a differential pressure sensor to collect and transmit real-time data on heavy metal concentration, flow rate, and differential pressure in the water. The model unit incorporates an adsorbent saturation prediction model, receives data from the sensing unit, and outputs real-time saturation and remaining adsorption capacity. The decision-making unit includes modules for dynamic threshold calculation, regeneration timing determination, and parameter optimization. Based on the model unit's results, it outputs regeneration commands and optimal regeneration parameters (such as current density and duration). The execution unit connects to the regeneration tank, power supply, and valve control components to perform the regeneration operation. The feedback unit in the auxiliary feedback layer collects data on adsorbent performance and energy consumption after regeneration and feeds it back to the model unit for model updates and optimization, forming a complete closed loop of "collection-computation-decision-execution-feedback." This achieves efficient remediation of heavy metal-contaminated water and intelligent regeneration management of the adsorbent, ensuring stable and low-energy system operation.

[0168] Example 3

[0169] See Figure 3This invention provides a combined adsorption-extraction system for the remediation of heavy metal-contaminated water. The contaminated water flows into the adsorption unit through the inlet, first passing through a data processing module integrating heavy metal concentration sensors, flow meters, and differential pressure sensors. This module collects water quality (heavy metal concentration) and operational (flow rate, differential pressure) data, which is then transmitted to the data processing and decision-making module via a data transmission link. Next, the water flows into the reaction zone filled with pre-adsorbent, where the adsorbent's properties intercept heavy metal ions for initial purification. When the adsorbent becomes saturated (determined by the data processing and decision-making module), the valve switches, and the adsorbent enters the regeneration unit. The regeneration unit relies on a power supply to provide electrochemical conditions and a temperature control device to maintain the reaction temperature, allowing the adsorbent to desorb heavy metals and regain its activity. The desorbed water is then discharged from the outlet. The data processing and decision-making module serves as the system's "control center." The data acquisition and preprocessing area receives and cleans raw data, the dynamic adsorption model calculation area simulates the adsorbent state based on the data and outputs key indicators, and the intelligent regeneration decision-making area combines the model results to determine the regeneration timing and parameters, generates instructions to drive the operation of valves, power supplies, and temperature devices, and optimizes decisions based on feedback data from the regeneration unit. This forms a closed loop of "data acquisition - model calculation - strategy output - execution feedback - optimization adjustment," enabling the recycling of adsorbents and efficient water remediation, and constructing a complete system for pollution control and resource recycling.

[0170] The above description is merely a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the forms disclosed herein and should not be construed as excluding other embodiments. It can be used in various other combinations, modifications, and environments, and can be altered within the scope of the concept described herein through the above teachings or related technologies or knowledge. Modifications and variations made by those skilled in the art that do not depart from the spirit and scope of the present invention should be within the protection scope of the appended claims.

Claims

1. A method for remediating heavy metal-contaminated water bodies using a combined adsorption-extraction approach, characterized in that, Includes the following steps: S1. Data Acquisition: Heavy metal concentration sensors, flow meters, and differential pressure sensors are deployed at the inlet and outlet of the adsorption unit to collect data on influent heavy metal concentration, effluent heavy metal concentration, water flow velocity, and differential pressure of the adsorption unit in real time, and to simultaneously acquire historical adsorption capacity and cumulative adsorbent usage time data. S2. Data Processing: Based on historical adsorption capacity, real-time concentration, flow rate, pressure difference, and cumulative usage time, a machine learning model for predicting adsorbent saturation is constructed. The model calculates the real-time saturation and remaining adsorption capacity of the adsorbent. Specifically, model selection and construction: A machine learning algorithm suitable for time series forecasting is selected to construct the prediction model. The input features of the model are historical adsorption capacity, real-time concentration, flow rate, pressure difference, and cumulative usage time. The output of the model is the real-time saturation and remaining adsorption capacity of the adsorbent. S3. Decision Generation: Calculate the dynamic threshold based on real-time saturation. When the saturation reaches the dynamic threshold or the remaining adsorption capacity is lower than the safety margin, a regeneration command is triggered. Combine saturation, heavy metal type, and historical regeneration data to generate the lowest energy consumption regeneration parameter combination. Step S3 includes the following sub-steps: S3.1 Calculates the dynamic threshold using the following formula: ； Wherein, Sbase is the basic threshold, k1 is the concentration influence coefficient, Cin is the real-time influent concentration, Clevel is the historical average influent concentration, k2 is the duration influence coefficient, tcumulative is the cumulative usage time, and ttotal is the total designed lifespan of the adsorbent. S3.2 Compare the real-time saturation with the dynamic threshold, and generate a regeneration signal when the triggering condition is met; S3.3 generates regeneration parameters based on the regeneration signal with the goal of minimizing energy consumption. The parameters include current density and regeneration time. S4. Execution and Feedback: Based on the regeneration instructions and parameter combinations, control the regeneration unit to perform the regeneration operation, collect the regenerated data and feed it back to the prediction model, and dynamically adjust the model parameters.

2. The method according to claim 1, characterized in that, Step S2 includes the following sub-steps: S2.1 Collect historical data to form a dataset, which includes historical adsorption amount, concentration at the corresponding time, flow rate, pressure difference, cumulative usage time and actual saturation; S2.2 Divide the dataset into a training set and a test set, train the prediction model, and calculate the model prediction bias rate using the test set. The formula is: Bias rate = |Spredicted - Sactual| / Sactual × 100%; where Spredicted is the model prediction saturation and Sactual is the actual saturation. S2.3 If the deviation rate exceeds the preset value, adjust the model parameters and retrain until the deviation rate meets the requirements.

3. The method according to claim 1, characterized in that, Step S1 data preprocessing includes outlier removal, and the outlier determination formula is as follows: |x-μ|>3σ; Where x represents real-time collected data, μ represents the mean of this type of data, and σ represents the standard deviation.

4. The method according to claim 1, characterized in that, In step S2, the update cycle of the prediction model is as follows: after collecting n sets of new data, the new data is added to the dataset and the model is retrained, where n is a preset positive integer.

5. The method according to claim 1, characterized in that, In step S3, when multiple heavy metals coexist, the regeneration parameters are weighted according to the adsorption equilibrium constant of each heavy metal, using the formula: w i =K i / ΣK j ; Among them, w i Let K be the parameter weight for the i-th heavy metal. i Its adsorption equilibrium constant, ΣK j This is the sum of the adsorption equilibrium constants of all coexisting heavy metals.

6. The method according to claim 1, characterized in that, It also includes step S5: real-time monitoring of energy consumption during the regeneration process. If the ratio of actual energy consumption to predicted energy consumption exceeds 1.2, return to step S3 to re-optimize the regeneration parameters.

7. The method according to claim 1, characterized in that, When performing regeneration in step S4, the process is first run at 70% of the optimized parameters, and after a preset time, the adsorbent activity is detected. Then, the regeneration intensity is adjusted according to the degree of activity recovery.

8. A combined adsorption-extraction system for the remediation of heavy metal-contaminated water, characterized in that, The system for performing the method according to any one of claims 1 to 7 includes: The sensing unit is used to collect real-time data on influent concentration, effluent concentration, flow rate, and differential pressure. The model unit, connected to the sensing unit, runs the adsorbent saturation prediction model and outputs the real-time saturation and remaining adsorption capacity. The decision-making unit, connected to the model unit, calculates dynamic thresholds, determines regeneration timing, and generates regeneration parameters. The execution unit, connected to the decision-making unit, performs adsorbent regeneration operations based on the regeneration parameters.

9. The system according to claim 8, characterized in that, It also includes a feedback unit, which is connected to the execution unit and the model unit respectively, and is used to feed back the regenerated data to the model unit to update the prediction model.

Images