An oxidation flocculation treatment system for mancozeb production sulfur-containing wastewater

Abstract

The present application relates to the field of industrial wastewater treatment, in particular to a kind of oxidation flocculation treatment system of mancozeb production sulfur-containing wastewater;Contain parameter acquisition, consumption evaluation, boundary prediction, strategy choice and dynamic execution module;System combines fluid multidimensional state parameter and carries out feature extraction, utilizes mapping model and quantifies the buffer capacity consumption degree of sulfur-containing wastewater;Its core is to predict the speed of approaching system failure boundary based on the change of consumption degree, and generate fluid regulation strategy combined with danger threshold, dynamically control the addition of oxidant and / or flocculant to reconstruct the fluid state;The present application discards the traditional passive dosing relying on single instantaneous index, expands the control object to implicit system buffer toughness, identifies potential risks such as colloidal sulfur enrichment in advance, overcomes the lag defect of conventional sensor, realizes forward-looking early warning and effectively avoids sludge floating and harmful gas emission.

Description

Technical Field

[0001] This invention relates to the field of industrial wastewater treatment, specifically to an oxidation flocculation treatment system for sulfur-containing wastewater from manganese zinc production. Background Technology

[0002] In the production of mancozeb pesticides, under the condition of supporting environmental water treatment, the workshop will periodically or intermittently generate highly fluctuating sulfur-containing wastewater during batch switching and equipment cleaning. This type of wastewater has a complex composition and often alternately flows into upstream cleaning liquid, mother liquor switching liquid and intermittent discharge liquid, resulting in the water body containing a mixture of thiocarbamate decomposition products, free sulfides, heavy metal ions and trace amounts of surface-active impurities. The water quality parameters inside the fluid exhibit strong nonlinear fluctuation characteristics.

[0003] To treat this type of sulfur-containing wastewater, existing oxidation-flocculation solutions generally adopt a passive dosing architecture based on a single instantaneous water quality indicator. This involves online monitoring using oxidation-reduction potential or pH sensors. When the absolute value of the indicator deviates from a set range, the control system fine-tunes the dosage of oxidant or flocculant according to a fixed theoretical stoichiometric ratio. While this approach has a certain continuous treatment capacity under steady-state influent conditions, conventional sensors are only sensitive to changes in the total amount of pollutants and are severely lagging in responding to changes in the interfacial state of the fluid and the colloidal protective effect caused by trace amounts of surface-active impurities. This results in the control system being unable to identify the implicit colloidal sulfur enrichment process within the system. When faced with high-fluctuation loads or impurity impacts, the traditional dosing method, which fine-tunes according to theoretical values, often loses its corrective ability due to insufficient action. This easily leads to the system crossing the failure boundary before the effluent indicators show obvious exceedances, resulting in serious accidents such as floc disintegration, sludge floating, water blackening and deterioration, and the large-scale release of toxic hydrogen sulfide gas.

[0004] Therefore, how to extend the control object from explicit hysteresis effluent indicators to implicit system buffer resilience, and realize forward-looking early warning and adaptive dynamic reconfiguration control of the failure risk of low-redundancy oxidation flocculation system, has become an urgent technical problem to be solved. Summary of the Invention

[0005] To solve the above-mentioned technical problems, the present invention provides an oxidation flocculation treatment system for sulfur-containing wastewater from manganese zinc production. Specifically, the technical solution of the present invention includes:

[0006] A reaction tank, a sensor group and a dosing mechanism disposed within the reaction tank, and a control terminal communicatively connected to the sensor group and the dosing mechanism, the control terminal comprising:

[0007] The parameter acquisition module is used to monitor the sulfur-containing wastewater in the reaction tank in real time through the sensor group and acquire a set of fluid state parameters.

[0008] The loss assessment module is used to extract features from the set of fluid state parameters and calculate the buffer capacity loss of the sulfur-containing wastewater based on a preset fluid state mapping model that takes fluid state parameters as input and loss degree as output.

[0009] The boundary prediction module is used to predict the failure approach time of the sulfur-containing wastewater to reach the preset system failure boundary based on the time series change of the buffer capacity depletion. The system failure boundary has a preset limit threshold.

[0010] The strategy selection module is used to compare the buffer capacity depletion with a preset danger threshold and, in conjunction with the failure approach time, generate a corresponding fluid control strategy.

[0011] The dynamic execution module is used to control the dosing mechanism to add oxidants, flocculants and / or acid-base regulators to the sulfur-containing wastewater according to the fluid control strategy, and to perform corresponding fluid state reconstruction operations.

[0012] Furthermore, the parameter acquisition module is specifically used for:

[0013] Obtain the redox potential sequence of the sulfur-containing wastewater;

[0014] Obtain the stirring current sequence of the sulfur-containing wastewater;

[0015] Obtain the fluid viscosity sequence of the sulfur-containing wastewater;

[0016] The redox potential sequence, the stirring current sequence, and the fluid viscosity sequence are simultaneously processed and filtered to generate the fluid state parameter set.

[0017] Furthermore, the wear assessment module is specifically used for:

[0018] Calculate the rate of change of potential based on the redox potential sequence;

[0019] The fluctuation features of the stirring current sequence are extracted, and the variance features and high-frequency fluctuation amplitudes of the stirring current sequence are obtained as current fluctuation features.

[0020] The potential change rate, the current fluctuation characteristics, and the fluid viscosity sequence are combined into a multidimensional feature vector;

[0021] The multidimensional feature vector is input into the pre-trained fluid state mapping model, and the buffer capacity depletion of the sulfur-containing wastewater is output. The fluid state mapping model is a neural network model trained based on historical multidimensional feature vectors and corresponding depletion labels.

[0022] Furthermore, the boundary prediction module is specifically used for:

[0023] Record the buffer capacity wear-out in chronological order to construct a wear-out time series.

[0024] Regression analysis was performed on the time series of wear and tear to obtain the slope of wear and tear growth;

[0025] Obtain the limit threshold corresponding to the system failure boundary;

[0026] The failure approach time is calculated by dividing the difference between the limit threshold and the current buffer capacity wear-out rate by the wear-out rate growth slope.

[0027] Furthermore, the strategy selection module is specifically used for:

[0028] The buffer capacity depletion is compared with the preset danger threshold;

[0029] If the buffer capacity depletion is less than the danger threshold, a normal steady-state maintenance strategy is generated.

[0030] If the buffer capacity depletion is greater than or equal to the danger threshold, a deep destabilization and reconstruction strategy is generated in conjunction with the failure approach time.

[0031] Furthermore, when executing the conventional steady-state maintenance strategy, the dynamic execution module is specifically used for:

[0032] Obtain the latest potential value of the redox potential sequence in the fluid state parameter set, and calculate the theoretical reagent requirement for the sulfur-containing wastewater by combining it with the preset stoichiometric ratio.

[0033] Obtain the preset steady-state maintenance coefficient;

[0034] The conventional dosage is calculated by multiplying the theoretical drug requirement by the steady-state maintenance coefficient.

[0035] The dosing mechanism is controlled to add the corresponding reagent to the sulfur-containing wastewater according to the conventional dosing amount in order to maintain the metastable equilibrium of the sulfur-containing wastewater.

[0036] Furthermore, when executing the deep destabilization and reconstruction strategy, the dynamic execution module is specifically used for:

[0037] Obtain the current pH data of the sulfur-containing wastewater;

[0038] Obtain a preset target pH level, and calculate the acid-base shock dosage by multiplying the difference between the target pH level and the current pH level data by a dynamic adjustment coefficient that is negatively correlated with the failure approach time.

[0039] The sacrificial agent dosage is calculated by multiplying the inversely proportional normalized mapping value of the failure approach time with the preset agent dosage benchmark.

[0040] The dosing mechanism is controlled to inject an acid-base regulator corresponding to the acid-base shock dosage into the sulfur-containing wastewater, and to add a sacrificial agent corresponding to the sacrificial agent dosage, so as to break the colloidal stability of the sulfur-containing wastewater.

[0041] Furthermore, the reaction tank includes multi-stage reaction tanks, and the sulfur-containing wastewater is distributed in the multi-stage reaction tanks. When performing the corresponding fluid state reconstruction operation, the dynamic execution module is also used for:

[0042] Obtain the cascade sequence and pipeline topology information of the multi-stage reaction tanks;

[0043] Based on the deep destabilization and reconstruction strategy and the volume ratio of each level of reaction tank in the pipeline topology information, the reagent dosing distribution ratio of each level of reaction tank in the multi-level reaction tank is calculated.

[0044] The failure approach time is multiplied by a preset reflux mapping coefficient to calculate the fluid reflux ratio between the multi-stage reaction tanks;

[0045] The multi-stage reaction tank is controlled in a coordinated manner based on the reagent addition distribution ratio and the fluid reflux ratio.

[0046] Furthermore, the control terminal also includes a model update module, which is specifically used for:

[0047] After executing the fluid control strategy, the post-execution fluid state parameters of the sulfur-containing wastewater are collected again.

[0048] The difference between the fluid state parameters after execution and the preset target state parameters is calculated and used as a state recovery index.

[0049] Based on the state recovery index, the internal weight parameters of the fluid state mapping model are adaptively updated using the backpropagation algorithm.

[0050] Furthermore, the buffer capacity depletion is used to characterize the enrichment index of colloidal sulfur in the sulfur-containing wastewater;

[0051] The system failure boundary is used to characterize the critical state in which colloidal sulfur is released in large quantities and accompanied by the release of harmful gases in the sulfur-containing wastewater.

[0052] The sacrificial agents include strong oxidants used for forced oxidation of surface-active impurities.

[0053] The present invention has the following beneficial effects:

[0054] 1. This system collects multidimensional fluid state parameters and extracts features, uses a fluid state mapping model to quantify the buffer capacity depletion, and combines time series to predict the failure approach time to the system failure boundary. Compared with the traditional passive dosing that relies on a single instantaneous index, this invention extends the control object to the implicit buffer capacity, which can identify potential risks such as colloidal sulfur enrichment caused by surface-active impurities in advance, overcome the shortcomings of conventional sensor response lag, realize the early warning before failure, and effectively avoid the large-scale precipitation of colloidal sulfur and the release of harmful gases.

[0055] 2. This system constructs a conventional steady-state maintenance strategy and a deep instability reconstruction strategy by comparing the buffer capacity depletion with a preset danger threshold and combining it with the failure approach time. When the depletion is greater than or equal to the danger threshold, the system actively generates a deep instability reconstruction strategy by combining the failure approach time, dynamically calculates and adds acid-base regulators and sacrificial agents. This mechanism breaks the limitation of traditional methods that lose the ability to correct deviations under high fluctuating loads due to fine-tuning the dosage according to theoretical values. It can break the unfavorable colloidal stability state in advance and implement fluid state reconstruction operation when the system is close to the failure boundary, ensuring continuous and stable operation under extreme water quality shocks. Attached Figure Description

[0056] The following drawings, illustrating embodiments of this application, are incorporated herein by reference and are used to understand this application. The drawings illustrate embodiments of this application and their descriptions, serving to explain the principles of this application. In the drawings,

[0057] Figure 1 This is a structural diagram of the system. Detailed Implementation

[0058] In the following description, numerous specific details are set forth to provide a more thorough understanding of this application. However, it will be apparent to those skilled in the art that embodiments of this application may be practiced without one or more of these details. In other instances, certain technical features well-known in the art have not been described to avoid confusion with embodiments of this application.

[0059] Example 1:

[0060] Please see Figure 1 An oxidation flocculation treatment system for sulfur-containing wastewater from manganese zinc production includes:

[0061] The reaction tank, the sensor group and the dosing mechanism installed in the reaction tank, and the control terminal that is communicatively connected to the sensor group and the dosing mechanism. The control terminal includes: a parameter acquisition module, which is used to monitor the sulfur-containing wastewater in the reaction tank in real time through the sensor group and acquire a set of fluid state parameters.

[0062] The loss assessment module is used to extract features from the set of fluid state parameters and calculate the buffer capacity loss of sulfur-containing wastewater based on a preset fluid state mapping model that takes fluid state parameters as input and loss degree as output.

[0063] The boundary prediction module is used to predict the failure approach time of sulfur-containing wastewater to reach the preset system failure boundary based on the time series change of buffer capacity depletion. The system failure boundary has a preset limit threshold.

[0064] The strategy selection module is used to compare the buffer capacity depletion with the preset danger threshold and, in conjunction with the failure approach time, generate the corresponding fluid control strategy.

[0065] The dynamic execution module is used to control the dosing mechanism to add oxidants, flocculants and / or acid-base regulators to sulfur-containing wastewater according to the fluid control strategy, and to perform the corresponding fluid state reconstruction operation.

[0066] This embodiment provides an oxidation-flocculation control mechanism for the continuous treatment of sulfur-containing wastewater from mancozeb production; specifically, the main application scenario is set as a continuous operation scenario in which a large pesticide plant's environmental protection workshop treats highly fluctuating sulfur-containing wastewater during batch switching.

[0067] The system operates in a low-redundancy state with the dosage of reagents close to the theoretical lower limit for a long time. The upstream cleaning liquid, mother liquor switching liquid and intermittent discharge liquid will enter the reaction tank alternately, causing the decomposition products of thiocarbamates, free sulfides, heavy metal ions and trace amounts of surface active impurities in the wastewater to exhibit strong nonlinear fluctuations.

[0068] To avoid relying solely on a single instantaneous water quality value for passive chemical dosing, this embodiment expands the control target from whether the current pollutant exceeds the standard to whether the current system's buffer capacity is being exhausted and the failure approach time to the failure boundary, thereby enabling the system to have the ability to destabilize and reconstruct in advance.

[0069] Specifically, the reaction tank can be a single tank or a multi-stage reaction tank arranged in series; the sensor group is installed in the reaction tank or on the circulation pipeline connected to the reaction tank, and includes at least a redox potential sensor, a stirring load acquisition unit and an online viscosity detection unit.

[0070] In some implementations, acid-base sensors, liquid level sensors, and temperature sensors can be added as auxiliary inputs; the control terminal can be deployed in a workshop programmable logic controller, industrial computer, or edge controller, and the software is divided into a parameter acquisition module, a loss assessment module, a boundary prediction module, a strategy selection module, and a dynamic execution module. Each module can be logically independent or integrated into different functional segments in the same control program.

[0071] The parameter acquisition module collects a set of fluid state parameters in the reaction tank at a preset sampling period. This set of parameters is not limited to concentration values, but emphasizes process quantities that can be obtained in real time and continuously and can reflect the stability of the system. The loss assessment module does not directly output the concentration of a single chemical component, but outputs the buffer capacity loss degree, which is used to characterize the remaining ability of the system to withstand sudden disturbances.

[0072] This wear level can be denoted as... And can be classified into The interval; where, The closer the value is to 1, the closer the system is to the critical boundary of colloidal sulfur enrichment, flocculation collapse, and release of harmful gases; the boundary prediction module further relies on multiple time points... The time series is constructed to predict the failure approach time to the system failure boundary;

[0073] The strategy selection module uses the danger threshold and failure approach time as the core two variables to determine whether the system should maintain its original metastable state or actively adopt a reconstruction strategy with stronger regulation intensity but greater safety. The dynamic execution module sends control instructions to the dosing mechanism to control the dosing rhythm, dosing location and dosing amount of oxidant, flocculant and auxiliary regulators when necessary.

[0074] To facilitate the explanation of the overall chain of the present invention, an exemplary data description is given below; assuming five consecutive sampling times... to The collected wear rates were 0.42, 0.47, 0.53, 0.61, and 0.70, respectively. Assuming the system failure boundary threshold is set to 1.00 and the danger threshold to 0.65, then... At this point, although the water outlet may not have exceeded the standard significantly, the loss rate has already exceeded the danger threshold, indicating that the risk of colloidal sulfur enrichment hidden in the system has already materialized.

[0075] In actual engineering implementation, the logic for setting the danger threshold and the limit threshold is as follows: calibration is carried out by combining on-site step disturbance experiments with historical water quality safety margin statistics; the specific calibration process is as follows: a known concentration of surface active impurities is artificially injected into the reaction tank to simulate the limit working conditions, and water quality indicators and harmful gas concentrations are monitored simultaneously.

[0076] The system buffer capacity depletion degree when the hydrogen sulfide emission concentration just reaches the lower limit of the plant safety warning is calibrated as the limit threshold; the depletion degree corresponding to the inflection point when the heavy metal index in the effluent shows a non-linear upward trend and the average particle size of colloidal flocs begins to show a statistically significant decrease is calibrated as the danger threshold.

[0077] If the boundary prediction module gives a higher boundary failure approach time, it means that even if the current effluent indicators are maintained, floc disintegration, sludge floating, or hydrogen sulfide release is very likely to occur in a short period of time. At this time, the strategy selection module no longer follows the conventional control with the minimum chemical consumption, but switches to deep instability reconstruction control to prioritize maintaining the system safety boundary. The following boundary situations will occur in actual operation.

[0078] Firstly, if a certain type of sensor in the sensor group loses connection for a short period of time, the parameter acquisition module can use the estimated value within the most recent effective window to fill in the data and assign a lower confidence level to the evaluation result; when the continuous loss of connection exceeds the preset time, the system will be forced to enter the conservative dosing mode.

[0079] Secondly, if the output of the wear and tear assessment module is... If a sudden jump occurs and is inconsistent with the changes in other auxiliary parameters, it can be determined as a measurement anomaly or local eddy current interference. The control terminal needs to perform an anomaly confirmation cycle. During the confirmation period, the drug is administered in a steady state according to the intermediate safety strategy, without immediately performing a large-scale destabilization operation.

[0080] Third, if the failure approach time calculated by the boundary prediction module is negative or close to zero, it indicates that the system's wear and tear is decreasing or basically stabilizing. Therefore, even if the current... When approaching the danger threshold, high-intensity reconstruction can be temporarily suspended, and only the monitoring frequency and the frequency of local drug application correction can be increased;

[0081] For example, within 30 minutes after the manganese zinc production line switched from product A to product B, the upstream cleaning waste liquid carried a trace amount of surface-active impurities into the reaction tank; at this time, the absolute value of the conventional redox potential did not change drastically, but the stirring current fluctuated more and the online viscosity slowly increased. After the wear assessment, the control terminal determined that the buffer capacity wear rate rapidly increased from 0.51 to 0.72.

[0082] Although the total manganese and total zinc in the effluent are still within the permissible range, the system has already initiated the fluid state reconstruction operation to redistribute the dosage ratio of oxidant and flocculant, and is prepared to perform acid-base shock and sacrificial oxidation if necessary, thereby avoiding serious deterioration of water quality and sudden release of sulfides.

[0083] The purpose of this step is to expand the control target from explicit effluent indicators to implicit system resilience indicators, so that the system can identify and cut off the deterioration path before failure occurs, thereby achieving feedforward risk control for low-redundancy oxidation flocculation systems.

[0084] Example 2:

[0085] The parameter acquisition module is specifically used to: acquire the redox potential sequence of sulfur-containing wastewater; acquire the stirring current sequence of sulfur-containing wastewater; acquire the fluid viscosity sequence of sulfur-containing wastewater; and perform synchronous processing and filtering on the redox potential sequence, stirring current sequence, and fluid viscosity sequence to generate a set of fluid state parameters.

[0086] The loss assessment module is specifically used for: calculating the rate of change of potential based on the redox potential sequence; extracting the fluctuation characteristics of the stirring current sequence and obtaining the variance characteristics and high-frequency fluctuation amplitude of the stirring current sequence as current fluctuation characteristics; and combining the rate of change of potential, current fluctuation characteristics and fluid viscosity sequence into a multi-dimensional feature vector.

[0087] The multidimensional feature vector is input into the pre-trained fluid state mapping model, and the output is the buffer capacity depletion degree of sulfur-containing wastewater. The fluid state mapping model is a neural network model trained based on historical multidimensional feature vectors and corresponding depletion degree labels.

[0088] The boundary prediction module is specifically used for: recording the buffer capacity wear-out in chronological order to construct a wear-out time series; performing regression analysis on the wear-out time series to obtain the wear-out growth slope; obtaining the limit threshold corresponding to the system failure boundary; and calculating the failure approach time with time dimensions based on the difference between the limit threshold and the buffer capacity wear-out at the current moment, divided by the wear-out growth slope, which is an equivalent parameter representing the remaining time to reach the failure boundary.

[0089] This embodiment provides a linkage mechanism for parameter construction, loss assessment and boundary prediction to address the risk of hidden colloidal instability. Specifically, in the main scenario, if the decision to add a drug is still based solely on the redox potential or absolute value of pH at a certain moment, it is easy to make misjudgments in the scenario of surface-active impurity impact.

[0090] The reason is that conventional sensors are more sensitive to changes in total amount, but are slow to respond to changes in interface state and colloidal protection effect. Therefore, although the previous solution can initially establish the concept of loss, if it lacks multi-dimensional process characteristics and time trend inference suitable for this scenario, it is difficult to identify the latent stage before the large-scale floating and deterioration of sludge in time. To solve this problem, this embodiment introduces a combination of limited parameter synchronization, feature extraction and boundary prediction as additional features.

[0091] Specifically, the parameter acquisition module acquires the redox potential sequence, stirring current sequence, and fluid viscosity sequence. The sampling frequencies for these three sequences can differ; for example, the redox potential can be acquired every 2 seconds, the stirring current every 0.2 seconds, and the viscosity every 5 seconds. For easier subsequent unified analysis, the sampling frequency can be based on a control cycle. Synchronize it;

[0092] Synchronization processing can employ timestamp alignment and window resampling, for example, using a 10-second analysis window to map sampled values ​​of different frequencies within the window to a unified time. Filtering can use moving average, median filtering, or low-pass filtering to suppress high-frequency noise introduced by transient bubbles, mechanical jitter, and local eddy currents. The rate of change of potential can be approximated using the first-to-last difference, and its calculation formula is as follows:

[0093]

[0094] in, To calculate the rate of change of potential, To analyze the redox potential at the beginning of the window, To analyze the redox potential at the end of the analysis window, To analyze the time span of the window; in a simplified example, suppose that within a 10-second window, the redox potential is resampled to obtain the sequence. The rate of change of potential can then be approximated by the difference between the first and last parts:

[0095]

[0096] in, The calculated potential change rate is given; this value is negative and has a large absolute value, indicating that the reducing component is rapidly consuming the oxidizing capacity; the stirring current samples within the same window are set to {8.0, 8.4, 7.9, 8.7, 8.1} A, and the total number of stirring current samples within the analysis window is given. In this example Then its variance characteristics can be expressed by the following formula:

[0097]

[0098] in, The variance characteristics of the extracted stirring current sequence; For the first in the window One stirring current sampling value, For all within this analysis window The average value of a sample stirring current; specifically, the cutoff frequency is preset to a certain value. A high-pass filter is used to filter the stirring current sequence, separating the high-frequency current sequence, and the root mean square value of the high-frequency current sequence within an analysis window is calculated as the high-frequency fluctuation amplitude. ;

[0099] After separating the high-frequency components of the current sequence, the amplitude of the high-frequency fluctuations can be obtained. As an indirect characterization indicator of abnormal collisions, adhesion, and local aggregation of micro-flocs within the fluid; further, the online viscosity sequence resampling value is set as... It can extract the current viscosity characterization value. and viscosity increment relative to the previous analysis window Thus, the current moment can be formed. The corresponding multidimensional feature vector:

[0100]

[0101] This multidimensional feature vector Input to a pre-trained fluid state mapping model, output buffer capacity depletion The fluid state mapping model employs a feedforward neural network with a single hidden layer; specifically, the model includes an input layer, a hidden layer, and an output layer that receive 5-dimensional input.

[0102] The hidden layer contains a preset number of neurons. Therefore, the weight matrix from the input layer to the hidden layer The dimension is Hidden layer bias vector The dimension is Similarly, the weight matrix from the hidden layer to the output layer... The dimension is Output layer bias It is a scalar;

[0103] Let the model be denoted as ,in Represents the set of weight parameters inside the model Then the multidimensional feature vector Input the model and output the buffer capacity consumption. The specific forward propagation calculation logic is as follows:

[0104]

[0105]

[0106] in, This is the weight matrix from the input layer to the hidden layer. This is the hidden layer bias vector; This is the weight matrix from the hidden layer to the output layer. For output layer bias; For the hidden layer activation function, Activation function for output layer to ensure depletion The output is normalized and limited to Within the range; This is the hidden state vector output by the hidden layer;

[0107] During the training phase, historical samples consist of multi-dimensional feature vectors and corresponding wear degree labels. The wear degree labels are weighted and normalized based on offline colloidal sulfur concentration measurements and sedimentation supernatant turbidity. Specifically, they correspond to a specific historical moment. True wear and tear label The quantification formula is:

[0108]

[0109] in, This refers to the offline colloidal sulfur concentration measured at that specific moment. and These are the critical colloidal sulfur concentrations for system collapse based on historical statistics and the normal baseline value; This corresponds to the turbidity of the supernatant from the sedimentation of the water sample. and These are the maximum permissible value and the normal reference value for turbidity of the supernatant, respectively. and The weighting coefficients are determined based on experiments and satisfy the following conditions: ;

[0110] Thus, test data with diverse physical forms are transformed into... The precise label values ​​of the intervals are obtained, thus completing the supervised training of the fluid state mapping model; in the online inference phase, the model only needs to read the features of the current window to provide real-time results. And without having to wait for offline test results;

[0111] The boundary prediction module obtains continuous... Then, a time series of attrition is constructed in chronological order; for example, within five adjacent control cycles. Regression analysis was performed on the sequence to extract the slope of the wear and tear growth. In simplified scenarios, the slope of linear regression or the moving average of the differences can be used to represent the value. If the calculated value is... The system failure boundary limit threshold is denoted as Then the boundary approximation result at the current moment can be obtained by the following formula:

[0112]

[0113] when At that time, we can obtain:

[0114]

[0115] The control terminal can be based on the execution layer. The urgency control level is calculated to be consistent with the direction of risk. The specific conversion calculation logic uses an inverse proportional normalized mapping function, as shown in the following formula:

[0116]

[0117] in, The preset timescale penalty constant, whose dimension is set to the reciprocal of time, is used to adjust the sensitivity of the control quantity and ensure the dimensionless nature of the formula denominator; when the failure approaches time Tend to At that time, the degree of urgency control trending toward the upper limit ;when At its maximum, Tend to This completes the data direction conversion; It is used only as an intermediate control variable for subsequent dosing intensity mapping, and does not replace the value calculated from the difference in the limit threshold and the growth slope. itself;

[0118] If the growth slope obtained from regression analysis This indicates that the wear and tear has not continued to increase or is recovering. In this case, the formula result should not be used directly to trigger a high-level alarm. The boundary prediction module can... Denote it as infinity or as a failure-free approximation; if If the absolute value is too small, causing the divisor to approach zero, then a preset minimum slope threshold is used. Replacement is used to avoid numerical instability; if the viscosity sensor fails within a certain analysis window, the viscosity dimension can be temporarily deleted, and a dimensionality-reduced backup model can be invoked. Calculate the approximate loss rate; if the redox potential and current are abnormal at the same time, the system no longer trusts the model output and directly enters the conservative safety mode; for cases where the samples exceed the training distribution, such as when a large amount of high-salt cleaning solution suddenly enters and causes all features to be far beyond the historical range, the feature vector can be truncated and normalized first, and this time it is recorded as an abnormal domain sample to prevent the model output from having physically meaningless abnormal values ​​or abnormal jumps.

[0119] For example, in the main scenario, after a certain upstream reactor cleaning is completed, a small amount of rinsing liquid containing surface-active residue enters the primary reaction tank; the absolute value of the redox potential only slowly decreases from 220mV to 205mV, seemingly still within the controllable range, but the variance of the stirring current rapidly increases from 0.03 to 0.26, and the online viscosity increases from 1.7mPa·s to 2.3mPa·s;

[0120] Based on this, the fluid state mapping model assesses the buffer capacity depletion as 0.76, rather than solely judging it as normal to low based on the redox potential; the boundary prediction module further provides... Approximately 3 minutes, and then converted into corresponding high-urgency control quantities at the control interface layer. Therefore, the control terminal issues high-priority intervention commands in advance, before traditional water quality indicators exceed the standards.

[0121] The purpose of this step is to transform the colloidal sulfur enrichment and system buffer decay, which are difficult to detect directly online, into the loss degree and its boundary approximation results that can be inferred in real time from multidimensional process signals, thereby achieving the quantification and forward-looking early warning of hidden risks.

[0122] Example 3: The strategy selection module is specifically used to: compare the buffer capacity depletion with a preset danger threshold; if the buffer capacity depletion is less than the danger threshold, a normal steady-state maintenance strategy is generated; if the buffer capacity depletion is greater than or equal to the danger threshold, a deep destabilization and reconstruction strategy is generated based on the failure approach time.

[0123] When executing a conventional steady-state maintenance strategy, the dynamic execution module is specifically used to: obtain the latest potential value of the redox potential sequence in the fluid state parameter set, calculate the theoretical reagent requirement for sulfur-containing wastewater in combination with the preset stoichiometry, and obtain the preset steady-state maintenance coefficient;

[0124] The theoretical reagent requirement is multiplied by the steady-state maintenance coefficient to calculate the conventional dosage; the dosing mechanism is controlled to add the corresponding reagent to the sulfur-containing wastewater according to the conventional dosage in order to maintain the metastable equilibrium of the sulfur-containing wastewater.

[0125] When executing the deep destabilization and reconstruction strategy, the dynamic execution module is specifically used to: obtain the current pH data of sulfur-containing wastewater; obtain the preset target pH; and calculate the acid-base shock dosage by multiplying the difference between the target pH and the current pH data by a dynamic adjustment coefficient that is negatively correlated with the failure approach time.

[0126] The inversely proportional normalized mapping value of the failure approach time is multiplied by the preset reagent dosage benchmark to calculate the sacrificial reagent dosage; the dosing mechanism is controlled to inject acid-base regulators corresponding to the acid-base shock dosage into the sulfur-containing wastewater, and sacrificial reagents corresponding to the sacrificial reagent dosage are added to break the colloidal stability of the sulfur-containing wastewater.

[0127] This embodiment provides a dual-mode control mechanism driven by risk assessment. After constructing the wear rate and boundary failure approach time, if all operating conditions are still uniformly mapped to fine-tuning the dosage according to the theoretical value, although a certain degree of drug saving can be achieved, when the metastable state approaches the failure boundary, the traditional fine-tuning method often loses its correction ability due to insufficient action range.

[0128] In other words, while the higher-level scheme can identify risks, it may not be able to execute different types of control instructions based on the risk level. To this end, this embodiment introduces a combination of limited strategy selection and differentiated execution features to clearly distinguish control actions into conventional steady-state maintenance strategies and deep instability reconstruction strategies.

[0129] In detail, the strategy selection module will consider the current buffer capacity consumption. With preset danger threshold Compare; if This indicates that although the system experiences fluctuations, it still retains sufficient buffering capacity, generating a regular steady-state maintenance strategy; if This indicates that the system has entered a high-risk zone. At this point, it is no longer sufficient to rely solely on the rate of attrition; the output must also be considered. Classify;

[0130] To ensure consistency of terminology, in this embodiment... Always represents the remaining approximation time; while the control strength actually used by the execution layer is determined by... Urgency control quantity obtained by monotonic conversion Characterization, therefore The smaller the value, the more dangerous it is. The larger the value, the more dangerous it is; a feasible logic is: when Exceeding the threshold but Still greater than the preset value At that time, the level of instability was broken during execution; when Less than or equal to At that time, a high-level deep stabilization process should be performed; this can avoid unnecessary chemical consumption and sludge load caused by a full-volume shock just past the threshold.

[0131] Under the conventional steady-state maintenance strategy, the dynamic execution module calculates the theoretical reagent requirement based on the latest redox potential value and a preset stoichiometry; the theoretical oxidant requirement... Specifically, it is calculated using the following conversion formula:

[0132]

[0133] in, The preset system oxidation equivalent conversion coefficient, To maintain the preset target redox potential of the metastable state, This is the latest potential value. The instantaneous flow rate of wastewater currently flowing into the reaction tank, with dimensions of For ease of explanation, let the theoretical oxidant requirement corresponding to the current redox potential be... The steady-state maintenance coefficient is The standard dosage can then be expressed as:

[0134]

[0135] in, This is the standard dosage. Typically greater than 1 but close to 1, for example, taking values ​​between 1.05 and 1.15, to reflect the characteristics of low-redundancy operation; if the estimated theoretical demand at a certain moment is 100L / h, take... The standard dosage is 108 L / h; the flocculant can also be adjusted in a similar way based on empirical stoichiometry or upstream load estimation; the core of this strategy is not absolute minimum chemical consumption, but maintaining the system in a controllable metastable state under the premise of safety.

[0136] Under the deep destabilization and reconstruction strategy, the system's objective shifts from maintaining the status quo to actively disrupting unfavorable colloidal structures and reconstructing a settleable state; to this end, the dynamic execution module obtains the current pH level. Then read the target pH value. Considering that stronger acid-base disturbances are needed closer to the failure boundary, let the dynamic adjustment coefficient be... To achieve this coefficient and urgency control quantity The positive correlation is calculated using the following linear interpolation logic:

[0137]

[0138] in, The lower limit of the pre-set basic acid-base adjustment coefficient, This represents the upper limit of the allowable acid-base adjustment coefficient for the equipment, and and All units are set to flow rate units (e.g., To ensure dimensional balance on both sides of the equation; based on the obtained... For clarity, the final acid-base shock dosage can be expressed by the following formula:

[0139]

[0140] in, To determine the final acid-base shock dosage, The preset target pH level, The current pH level This is the absolute value operator; for example, if the current pH is 8.6 and the target pH is set to 6.2, the difference is 2.4; if the system has very little remaining safety time, then the corresponding... A larger value is selected to obtain a higher acid-base shock dosage, so as to rapidly change the surface charge environment of the colloidal substance;

[0141] Meanwhile, the deep destabilization and reconstruction strategy also includes the calculation of the sacrificial agent dosage; assuming a preset agent dosage baseline is... The dosage of the sacrificial agent can then be expressed as:

[0142]

[0143] in, The calculated amount of sacrificial agent is used to treat surface-active impurities or other protective colloids in the forced oxidation system. Essentially, this increases the short-term agent dosage and some sludge volume in exchange for a rapid reconstruction window after the colloidal stability is disrupted. The dynamic execution module controls the dosing mechanism to inject the corresponding acid-base regulator and sacrificial agent to help the system break free from its original unstable trajectory.

[0144] To avoid misapplying deep instability to normal operating conditions, this embodiment provides a simplified judgment example; let a danger threshold be set. At a certain moment At that time, the system only generated a conventional steady-state maintenance strategy and fine-tuned the oxidant flow rate from 95 L / h to 102 L / h; during the subsequent 15 minutes It rose to 0.74, and the boundary prediction gave... After only 4 minutes, the system switched to a deep destabilization and reconstruction strategy;

[0145] In this implementation, the control terminal first converts the remaining 4 minutes of safety time into a high-level urgency control quantity. Based on this, the pH level was lowered from the current 8.4 to around 6.5, and a certain amount of strong oxidant was added. Then, flocculant was added downstream to restore settling properties.

[0146] If the current Slightly above the danger threshold, but the boundary forecast shows that the growth slope has clearly fallen back, indicating that the previous round of intervention is taking effect. At this time, a delayed confirmation mechanism can be adopted, that is, the steady-state maintenance coefficient can be increased slightly first without immediately implementing acid-base shock. If the acid-base sensor fails, the acid-base shock part of the deep stability reconstruction strategy will be suspended, but the sacrificial agent part can still be added within the conservative upper limit.

[0147] If the sacrificial agent storage tank level is insufficient, the control terminal should automatically downgrade to an alternative solution of acid-base shock only + increased oxidant redundancy, and simultaneously issue a replenishment alarm; if the calculated... or If the instantaneous dosage limit of the equipment is exceeded, the dynamic execution module should adopt segmented or pulsed dosage to prevent local over-dosing from causing new side reactions.

[0148] For example, in the main scenario, the system runs at 108% of the theoretical demand for an extended period. After a batch switch, the attrition rate increases from 0.60 to 0.68, without any obvious signs of pool overflow. The system first switches its strategy from maintaining a normal steady state to a medium-level deep instability. After 20 seconds, the attrition rate further increases to 0.77. The time was shortened to 3 minutes, indicating that trace amounts of surface-active impurities significantly enhanced the colloidal protective effect.

[0149] The control terminal immediately followed the higher... The corresponding control intensity is increased to increase acid and alkali shock, and the sacrificial agent line is activated; the black suspended fine particles in the primary tank are rapidly reduced, the downstream flocs re-aggregate and form, and the sedimentation interface becomes clear again;

[0150] The purpose of this mechanism is to elevate control behavior from simply and precisely following theoretical values ​​to a control philosophy that switches according to risk levels, thereby achieving a dynamic balance between normal steady-state operation and the survival baseline under extreme conditions.

[0151] Example 4:

[0152] The reaction tank includes multiple reaction tanks, in which sulfur-containing wastewater is distributed. When the dynamic execution module performs the corresponding fluid state reconstruction operation, it is also used to: obtain the cascade sequence of the multiple reaction tanks and the pipeline topology information; and calculate the reagent dosing distribution ratio of each reaction tank in the multiple reaction tanks based on the deep destabilization reconstruction strategy and the volume ratio of each reaction tank in the pipeline topology information.

[0153] The failure approach time is multiplied by a preset reflux mapping coefficient to calculate the fluid reflux ratio between the multi-stage reaction tanks; the multi-stage reaction tanks are then coordinated and controlled based on the reagent dosing distribution ratio and the fluid reflux ratio.

[0154] The control terminal also includes a model update module, which is specifically used for: after executing the fluid control strategy, re-collecting the post-execution fluid state parameters of the sulfur-containing wastewater; calculating the difference between the post-execution fluid state parameters and the preset target state parameters as a state recovery index; and adaptively updating the internal weight parameters of the fluid state mapping model based on the state recovery index using the backpropagation algorithm.

[0155] Buffer capacity depletion is used to characterize the enrichment index of colloidal sulfur in sulfur-containing wastewater; system failure boundary is used to characterize the critical state in which colloidal sulfur in sulfur-containing wastewater precipitates in large quantities and is accompanied by the release of harmful gases; sacrificial agents include strong oxidants used for forced oxidation of surface-active impurities.

[0156] This embodiment provides a collaborative reconstruction and adaptive learning mechanism for multi-stage reaction links. Specifically, the system already has the ability to dosing in a dual-modal manner based on risk level. However, if the wastewater treatment target is expanded from a single tank to a multi-stage reaction tank, and the concentrated shock dosing is only done in a certain tank, new bottlenecks may occur: local over-dosing in the upstream tank, insufficient reaction in the downstream tank, shearing and disintegration of flocs between stages, and even migration and amplification of hazardous components between stages.

[0157] The composition of sulfur-containing wastewater is batch-dependent, and historical models will gradually deviate from reality under new formulations, new cleaning regimes, or new impurities. Therefore, this embodiment introduces a combination of additional features including limited multi-level collaborative control, online model updates, and key physicochemical meanings.

[0158] Specifically, in a multi-stage reaction pool scenario, the dynamic execution module obtains the cascading order and network topology information of each stage of the reaction pool; assuming the system consists of three stages of reaction pools, their effective volumes are respectively , , The volume ratio is 4:3:2. When the system enters a deep instability reconstruction strategy, all reagents cannot be simply added to the primary pool at once. Instead, the reagent distribution ratio should be calculated based on the volume ratio, the functional positioning of each level, and the topological coupling relationship. In a simplified implementation, the basic allocation coefficient can be obtained first based on the volume ratio, and then a risk weight correction term can be added. For example, the total sacrificial reagent dosage is... Then the first The dosage obtained in the primary reactor can be expressed as:

[0159]

[0160] In the formula, The weights are related to the function of the stage pool; the pre-stage pool emphasizes the rapid destruction of protective colloids and can be given a higher weight, while the post-stage pool emphasizes floc growth and stable sedimentation and can appropriately reduce the proportion of sacrificial oxidant and increase the proportion of flocculant compensation. For the first The effective volume of the stage reaction tank; For the first The effective volume of the stage reaction tank; For the first The functional weights corresponding to the stage reactors; the summation formula It is only used to traverse all available reaction pools, i.e. Values ​​range from 1 to ;

[0161] Regarding reflux control, to enhance the buffering and equalization capabilities between different pool levels, the dynamic execution module will output... Urgency control quantity obtained through interface conversion Multiplying this by a preset reflux mapping coefficient yields the interstage fluid reflux ratio; here... Consistent with the above, the remaining approximation time is used uniformly; This is a derived control variable that aligns with the direction of risk; therefore, the closer the system gets to the failure boundary... The higher the value, the higher the reflux ratio, without the reverse control effect of a longer remaining safety time leading to a larger reflux; let the reflux mapping coefficient be... The reflux ratio can then be simplified as follows:

[0162]

[0163] in, The calculated fluid recirculation ratio between the multi-stage reaction tanks is used to prevent the calculated recirculation ratio from exceeding the actual pumping capacity or causing overflow of the upstream reaction tank under extremely urgent operating conditions. Before issuing control commands, the dynamic execution module adjusts the fluid recirculation ratio. Apply hard safety constraints:

[0164]

[0165] in, The final actual return rate. This is the safest backflow ratio limit calculated based on the maximum hydraulic load of the equipment. The function is designed to minimize the value. When the risk increases, the reflux ratio is increased, allowing the mixture that has not yet become completely unstable in the downstream stage to be returned to the upstream stage. This can both dilute the excessively high levels of colloidal sulfur in the local area and prolong the residence time of high-risk components in the oxidation zone. Based on this, the control terminal coordinates the reflux pump speed, valve opening, and the frequency of each stage of the reagent pump.

[0166] To illustrate the necessity of multi-stage synergy, a set of exemplary descriptions is provided. Assume the total depth of the stabilizing agent dosage in the three-stage tank is 90L. If the dosage is added entirely to the first-stage tank without distinguishing between tanks, the local redox environment in the first-stage tank will change drastically, potentially forming a large number of fine particles. The downstream second-stage tank will then experience high shear and high ionic strength impacts, which is detrimental to floc growth. However, if the dosage is allocated according to the corrected ratio of 50%:30%:20%, the surface active protective layer in the first-stage tank will be destroyed, the second-stage tank will continue to supplement oxidation and begin to form bridging, and the third-stage tank will maintain floc integrity with lower intensity, resulting in a more stable overall sedimentation effect.

[0167] The model update module addresses the issue of model aging under varying operating conditions. Specifically, after each execution of a fluid control strategy, the system re-collects the fluid state parameters and compares them with preset target state parameters to calculate a state recovery index. This index can be denoted as... For example, a weighted sum of squares of multi-parameter deviations can be used:

[0168]

[0169] in, The total number of dimensions of the state parameters participating in the state recovery evaluation, and the number of dimensions after execution. The state parameters are denoted as The corresponding target value is denoted as The weighting coefficient is denoted as ;like A smaller value indicates that the control of this wheel is effective; if A larger value indicates a discrepancy between the model's predicted loss and the actual recovery effect. The model update module constructs a loss term based on this state recovery index and adaptively updates the internal weight parameters of the fluid state mapping model using the backpropagation algorithm. The core expression is as follows:

[0170]

[0171] Specifically, the loss function Constructed using the mean squared error form, its calculation formula is as follows:

[0172]

[0173] This structure ensures that the gradient of the internal weight parameters can be calculated subsequently. It is continuous and smoothly differentiable; among which, For learning rate, The loss function is constructed based on the recovery error; and These represent the internal weight parameters of the fluid state mapping model before and after the update, respectively; This represents the gradient calculation of the loss function with respect to the internal weight parameters; the model will gradually absorb new working condition samples, continuously improving its adaptability to specific workshops, specific reagent systems, and specific impurity backgrounds; in the above formula... As the parameter dimension number participating in the state recovery evaluation, it is used to distinguish the redox potential change rate, current fluctuation, viscosity or other target state parameters;

[0174] If a stage of a multi-stage reaction tank is shut down for maintenance, the dynamic execution module should recalculate the volume ratio and topology path based on the currently available tanks to avoid using the original allocation formula. If the reflux pump fails, the collaborative control should be downgraded to a non-reflux mode with multi-point dosing and the ratio of the pre-stage reagents should be appropriately increased.

[0175] If the model update module detects that the loss function increases instead of decreasing after several consecutive updates, it determines that the current online sample may contain abnormal contamination or label distortion, and online updates need to be paused and rolled back to the previous stable model version. If the state recovery index shows a contradiction with the field observation, such as the model believing that the recovery is good but the settlement interface continues to deteriorate, the field process constraints should be given priority, the model self-learning function should be temporarily frozen and the level of manual verification should be increased.

[0176] For example, in the main scenario, the workshop uses three-stage reaction tanks in series to treat manganese zinc wastewater; after a certain impact of surface-active impurities, the system determines that a deep destabilization and reconstruction needs to be performed; the control terminal distributes the strong oxidizing sacrificial agent in a manner that is high in the first stage and low in the last stage according to the volume of the three-stage tank and the process function, while increasing the reflux ratio from the second stage to the first stage, so that some of the intermediate flocs that have begun to form can be returned to the previous stage for re-oxidation;

[0177] The enhanced reflux here is based on high The system triggers the process, rather than simply amplifying the remaining time linearly. After 15 minutes of processing, the system re-collects data on the rate of change of redox potential, current fluctuations, and viscosity. It finds that the actual recovery effect is better than the model prediction. Based on this, the model update module fine-tunes the neural network weights so that similar impurity impacts can be identified earlier in the future.

[0178] This application has been described through the above embodiments; however, it should be understood that the above embodiments are for illustrative purposes only and are not intended to limit this application to the described embodiments. Those skilled in the art will understand that many more variations and modifications can be made based on the teachings of this application, and all such variations and modifications fall within the scope of protection claimed in this application.

Claims

1. An oxidation-flocculation treatment system for sulfur-containing wastewater from mancozeb production, characterized in that, The system includes a reaction tank, a sensor array and a dosing mechanism disposed within the reaction tank, and a control terminal communicatively connected to the sensor array and the dosing mechanism. The control terminal includes: The parameter acquisition module is used to monitor the sulfur-containing wastewater in the reaction tank in real time through the sensor group and acquire a set of fluid state parameters. The loss assessment module is used to extract features from the set of fluid state parameters and calculate the buffer capacity loss of the sulfur-containing wastewater based on a preset fluid state mapping model that takes fluid state parameters as input and loss degree as output. The boundary prediction module is used to predict the failure approach time of the sulfur-containing wastewater to reach the preset system failure boundary based on the time series change of the buffer capacity depletion. The system failure boundary has a preset limit threshold. The strategy selection module is used to compare the buffer capacity depletion with a preset danger threshold and, in conjunction with the failure approach time, generate a corresponding fluid control strategy. The dynamic execution module is used to control the dosing mechanism to add oxidants, flocculants and / or acid-base regulators to the sulfur-containing wastewater according to the fluid control strategy, and to perform corresponding fluid state reconstruction operations. The parameter acquisition module is specifically used for: Obtain the redox potential sequence of the sulfur-containing wastewater; Obtain the stirring current sequence of the sulfur-containing wastewater; Obtain the fluid viscosity sequence of the sulfur-containing wastewater; The redox potential sequence, the stirring current sequence, and the fluid viscosity sequence are simultaneously processed and filtered to generate the fluid state parameter set. The wear assessment module is specifically used for: Calculate the rate of change of potential based on the redox potential sequence; The fluctuation features of the stirring current sequence are extracted, and the variance features and high-frequency fluctuation amplitudes of the stirring current sequence are obtained as current fluctuation features. The potential change rate, the current fluctuation characteristics, and the fluid viscosity sequence are combined into a multidimensional feature vector; The multidimensional feature vector is input into the pre-trained fluid state mapping model, and the buffer capacity depletion of the sulfur-containing wastewater is output. The fluid state mapping model is a neural network model trained based on historical multidimensional feature vectors and corresponding depletion labels. The boundary prediction module is specifically used for: Record the buffer capacity wear-out in chronological order to construct a wear-out time series. Regression analysis was performed on the time series of wear and tear to obtain the slope of wear and tear growth; Obtain the limit threshold corresponding to the system failure boundary; The failure approach time is calculated by dividing the difference between the limit threshold and the buffer capacity wear-out at the current moment by the wear-out growth slope. The strategy selection module is specifically used for: The buffer capacity depletion is compared with the preset danger threshold; If the buffer capacity depletion is less than the danger threshold, a normal steady-state maintenance strategy is generated. If the buffer capacity depletion is greater than or equal to the danger threshold, a deep destabilization and reconstruction strategy is generated in conjunction with the failure approach time. When executing the conventional steady-state maintenance strategy, the dynamic execution module is specifically used for: Obtain the latest potential value of the redox potential sequence in the fluid state parameter set, and calculate the theoretical reagent requirement for the sulfur-containing wastewater by combining it with the preset stoichiometric ratio. Obtain the preset steady-state maintenance coefficient; The conventional dosage is calculated by multiplying the theoretical drug requirement by the steady-state maintenance coefficient. The dosing mechanism is controlled to add the corresponding reagent to the sulfur-containing wastewater according to the conventional dosing amount in order to maintain the metastable equilibrium of the sulfur-containing wastewater; When executing the deep destabilization and reconstruction strategy, the dynamic execution module is specifically used for: Obtain the current pH data of the sulfur-containing wastewater; Obtain a preset target pH value, and calculate the acid-base shock dosage by multiplying the difference between the target pH value and the current pH value data by a dynamic adjustment coefficient that is negatively correlated with the failure approach time. The sacrificial agent dosage is calculated by multiplying the inversely proportional normalized mapping value of the failure approach time with the preset agent dosage benchmark. The dosing mechanism is controlled to inject an acid-base regulator corresponding to the acid-base shock dosage into the sulfur-containing wastewater, and to add a sacrificial agent corresponding to the sacrificial agent dosage, so as to break the colloidal stability of the sulfur-containing wastewater. The buffer capacity depletion is used to characterize the enrichment index of colloidal sulfur in the sulfur-containing wastewater; The system failure boundary is used to characterize the critical state in which colloidal sulfur is released in large quantities and accompanied by the release of harmful gases in the sulfur-containing wastewater. The sacrificial agents include strong oxidants used for forced oxidation of surface-active impurities.

2. The oxidation flocculation treatment system for sulfur-containing wastewater from mancozeb production according to claim 1, characterized in that, The reaction tank includes multiple reaction tanks, and the sulfur-containing wastewater is distributed in the multiple reaction tanks. When performing the corresponding fluid state reconstruction operation, the dynamic execution module is also used for: Obtain the cascade sequence and pipeline topology information of the multi-stage reaction tanks; Based on the deep destabilization and reconstruction strategy and the volume ratio of each level of reaction tank in the pipeline topology information, the reagent dosing distribution ratio of each level of reaction tank in the multi-level reaction tank is calculated. The failure approach time is multiplied by a preset reflux mapping coefficient to calculate the fluid reflux ratio between the multi-stage reaction tanks; The multi-stage reaction tank is controlled in a coordinated manner based on the reagent addition distribution ratio and the fluid reflux ratio.

3. The oxidation flocculation treatment system for sulfur-containing wastewater from manganese zinc production according to claim 2, characterized in that, The control terminal also includes a model update module, which is specifically used for: After executing the fluid control strategy, the post-execution fluid state parameters of the sulfur-containing wastewater are collected again. The difference between the fluid state parameters after execution and the preset target state parameters is calculated and used as a state recovery index. Based on the state recovery index, the internal weight parameters of the fluid state mapping model are adaptively updated using the backpropagation algorithm.

Images