A sewage treatment intelligent dosing control system and method for metering pump constraints

Abstract

This invention relates to the field of automatic control technology for wastewater treatment, and discloses an intelligent dosing control system and method for wastewater treatment oriented towards metering pump constraints. The system includes a multi-source data acquisition module, a data preprocessing module, a reagent demand calculation module, a signal conversion module, a PLC control module, and a reagent dosing module. The method includes data acquisition and preprocessing, target reagent dosing calculation, drive signal conversion, execution constraint compensation, PLC control execution, and feedback correction. This invention can reduce the deviation between the estimated reagent demand and the actual output of the actuator, improve the stability and feasibility of dosing control under complex operating conditions, and is suitable for precise reagent dosing and closed-loop control under complex, multi-variable, and time-varying operating conditions.

Description

Technical Field

[0001] This invention relates to the field of automatic control technology for wastewater treatment, specifically to a wastewater treatment dosing control system and method that combines online monitoring, fuzzy neural network prediction, and metering pump execution constraint compensation in a coordinated coupling. Background Technology

[0002] Water is an essential resource for human survival and social development, and wastewater treatment capacity is directly related to residents' health, the ecological environment, and socio-economic development. With the continuous advancement of industrialization and urbanization, wastewater quality fluctuates more and treatment requirements become more stringent. Traditional dosing methods that rely on manual experience or simple automatic control are no longer sufficient to meet the requirements of stable compliance with standards and energy conservation.

[0003] In wastewater treatment, processes such as coagulation, flocculation, sedimentation, phosphorus removal, and disinfection are all closely related to the addition of chemicals. Insufficient chemical dosage leads to incomplete pollutant removal and affects effluent quality; excessive dosage results in waste, increased operating costs, and even secondary impacts. Therefore, achieving stable and precise chemical dosing has always been a key issue in the automatic control of wastewater treatment.

[0004] In existing technologies, common methods for chemical dosing include manual experience, beaker experiments, and traditional automatic control. Manual experience is highly subjective and lacks stability; beaker experiments have long testing cycles, making them unsuitable for continuous operation; currently, water plant dosing mostly uses automatic control, which involves using instruments to detect water quality indicators and a PID controller to compare the input detected values ​​with the setpoints, correcting deviations and automatically controlling the dosage. While this achieves a certain degree of automation, wastewater treatment processes are characterized by multivariability, nonlinearity, time-varying nature, randomness, and large time delays, making it difficult to adapt to complex operating conditions by relying solely on fixed parameter control.

[0005] Existing technologies include dosing schemes based on artificial neural networks, LSTM, multi-parameter feedback, or real-time monitoring. These schemes primarily focus on predicting the dosage or calculating control parameters. While these schemes can improve the crudeness of manual experience-based control to some extent, a type of easily overlooked source of deviation still exists in engineering implementation: the estimated dosage does not directly equate to the stable output of the metering pump. Distortions on the execution side arise from factors such as differences in drug concentration, pump start-up dead zone, stroke hysteresis, calibration curve nonlinearity, and switching between different metering pump control modes.

[0006] Especially under conditions of low load, shock load, or frequent fluctuations in water quality, even if the estimation results are relatively accurate, if the execution side does not consider the minimum stable pump start-up threshold, duty cycle change rate limit, and hysteresis compensation, the actual output of the metering pump may still be inconsistent with the estimated target, resulting in fluctuations in effluent indicators or waste of chemicals.

[0007] Therefore, it is necessary to propose a wastewater treatment chemical co-dosing control scheme that combines chemical demand estimation with constraint compensation of the actuator, so as to reduce the deviation between the estimation end and the execution end and improve the feasibility and operational stability of the chemical dosing system. Summary of the Invention

[0008] To address the problem of significant discrepancies between estimation and execution results in existing technologies, this invention provides an intelligent control system and method for wastewater dosing constrained by metering pumps.

[0009] To achieve the above objectives, the present invention provides, on the one hand, a wastewater treatment dosing control system oriented towards metering pump execution constraints, including a multi-source data acquisition module, a data preprocessing module, a reagent demand calculation module, a signal conversion module, a PLC control module, and a reagent dosing module; on the other hand, it provides a wastewater treatment dosing control method based on the system, including data acquisition and preprocessing, target reagent dosage calculation, drive signal conversion, execution constraint compensation, PLC control execution, and feedback correction.

[0010] The specific steps of the control method are as follows:

[0011] S1: Data Acquisition: The data acquisition module includes a pH meter, a temperature meter, a flow meter, a turbidity meter, a DO meter, a COD meter, a total phosphorus and total nitrogen meter, a colorimeter, a level gauge, and a pressure gauge. It is used to collect water quality parameters of raw water and effluent, influent and effluent flow rates, return flow rates, liquid level in the chemical tank, and outlet pressure of the metering pump. After detection and storage, the data is transmitted to the data preprocessing module for preprocessing.

[0012] S2: Data processing of the data collected in step S1: The collected data is cleaned and outlier filtered, including missing value imputation, outlier removal, time alignment, and sample reliability evaluation, to improve the comparability between parameters of different dimensions and the model training effect; the Z-score method is used to normalize the monitoring data. In the formula This represents the data after normalization. This represents the input sample data. Represents the sample mean. This represents the sample standard deviation.

[0013] S3: Input the data processed in step S2 into the reagent demand calculation module to calculate and output the target reagent dosage. The dosage prediction module includes an input layer, a fuzzification layer, a fuzzy rule layer, a normalization layer, and an output layer. The input layer receives water quality parameters and reagent-related data. The fuzzification layer uses a membership function to fuzzify the input variables. The fuzzy rule layer is used for rule matching. The normalization layer is used to calculate the applicability of the rules. The output layer outputs the target reagent dosage.

[0014] S3-1: Input layer. Input values ​​are water quality parameters and reagent dosage data during wastewater treatment, represented as follows: Where n represents the number of input parameters, and n is an integer greater than or equal to 1. These represent the water quality parameters of the influent and effluent during the wastewater treatment process, namely pH value, temperature, color, turbidity, DO, COD, total phosphorus, total nitrogen, influent and effluent flow rates, return flow rate, and type and dosage of added chemicals. The input-output relationship of this node is as follows: .

[0015] S3-2: Fuzzification Layer. Each node in this layer corresponds to a linguistic variable value. Its function is to convert the input deterministic quantity into a fuzzy vector. The input variable is transformed into a corresponding membership degree through the membership function defined on its fuzzy subset. A Gaussian function is selected as the membership function. The membership degree of each input component corresponding to different fuzzy linguistic values ​​is as follows: In the formula: , Let be the width and center of the membership function of the j-th fuzzy set of the i-th input variable.

[0016] S3-3: Fuzzy Rule Layer. Each node in this layer represents a fuzzy rule. It connects with the fuzzification layer to match fuzzy rules, enabling fuzzy computation between nodes. The output of each node is the product of all its signals, i.e.: In the formula It is the fuzzy segmentation number of the i-th input.

[0017] S3-4: Normalization layer, implementing normalized calculations, the formula is:

[0018] S3-5: Output layer, converts the fuzzy output into a sharp output, and outputs PID parameters: in, For connection weights, For the applicability of the rule, r is 1 to... Integers between [a certain range].

[0019] S3-6: The drug demand calculation module uses the RBF learning algorithm to adjust the network connection weights, membership function center values, and width. Its characteristic is that the RBF neural network learning algorithm adjusts the connection weights through the learning process. The central value of the membership function and width The formula for calculating the error cost function is: In the formula, , For the ideal output and actual output of each iteration, This indicates the control error.

[0020] The algorithm for learning network weights is as follows:

[0021] The learning methods for the center vector and width vector of the fuzzy layer are as follows:

[0022] If momentum factor is considered, the weights of the output layer are: In the formula, For the iterative steps of the network; For learning rate, ; To learn the momentum factor, .

[0023] The training samples for the reagent demand calculation module include historical operation data of wastewater treatment, expert experience data, and online data that is continuously updated during system operation.

[0024] S4: The target drug dosage input signal output in step S3 is converted into a PWM control signal based on the calibration relationship between the metering pump flow rate and the PWM duty cycle.

[0025] S4-1: Calculation of pesticide dosage: In the formula, M is the mass concentration of the reagent added, mg / h; C is the target concentration, mg / L; and Q is the influent flow rate, L / h.

[0026] S4-2: Conversion between mass flow rate and volumetric flow rate for reagent concentration: In the formula, The target reagent volumetric flow rate, The density of the drug solution, This represents the effective mass fraction of the drug solution.

[0027] S4-3: Establish the calibration relationship between the drug pump flow rate and the PWM: When dealing with nonlinear relationships, piecewise linear regression, lookup table methods, and other curve fitting techniques are used. For linear approximations, it can be expressed as:

[0028] Duty cycle limit is: In the formula, D is the PWM duty cycle (0~100%), and k and b are calibration parameters.

[0029] S4-4: Based on the controller's PWM register range, this can be converted to the actual PWM output value: In the formula, This is the full-scale value of the register.

[0030] S4-5: Based on the feedback signal, perform real-time correction of the PWM: In the formula, Calculated from the target dosage. Corrections are made based on feedback from water quality parameters such as flow meter readings, pH, and conductivity.

[0031] S5: The PLC control module generates the corresponding operating frequency, stroke position or motor speed according to the PWM signal to control the metering pump. Specifically, the PLC control module converts the PWM signal into a digital quantity and calculates the corresponding output parameters according to the preset control algorithm to control the metering pump to add the corresponding amount of medicine.

[0032] S5-1: Action frequency control. The PLC will output a variable frequency pulse signal to drive the electromagnet or stepper motor of the metering pump to change the number of strokes per unit time, thereby adjusting the flow rate.

[0033] S5-2: Stroke position control. The PLC controls the servo motor or linear actuator through analog output or digital pulse to precisely adjust the stroke length of the metering pump and achieve flow regulation.

[0034] S5-3: Motor speed control. The PLC uses a frequency converter or speed control driver to change the rotational speed of the metering pump motor, thereby changing the reciprocating speed of the pump.

[0035] S6: The chemical dosing module controls the metering pump to add an appropriate amount of chemical to the corresponding water treatment unit.

[0036] After adopting the above technical solution, the present invention has the following beneficial effects compared with the prior art: First, by collecting multiple water quality parameters in the sewage treatment process online and performing missing value imputation, outlier removal, time alignment, and sample credibility evaluation on the original data, the present invention can improve the stability and usability of the input data. Second, by using a fuzzy neural network to model the multivariable, nonlinear, and time-varying conditions in the sewage treatment process, the present invention can combine historical data, expert experience, and online operation data to predict the target dosage of chemicals, thereby improving the accuracy of dosing control. Third, by considering the estimation of chemical demand in conjunction with the constraint compensation of the actuator, the present invention can reduce the deviation between the estimation end and the execution end. Fourth, by constructing a complete control link of "water quality monitoring - data processing - dosage prediction - execution control - effect feedback" through a signal conversion module, a PLC control module, and a metering pump, the present invention can realize the automated control of chemical dosage. Fifth, by using sample credibility discrimination and safe dosing mode switching, the present invention can maintain the safe operation of the system when the sensor, model, or actuator malfunctions, which is beneficial to improving the dosing stability under complex conditions and reducing chemical waste. Attached Figure Description

[0037] Figure 1 This is an overall framework diagram of an intelligent chemical dosing control system and method for wastewater treatment oriented towards metering pump constraints, as per the present invention.

[0038] Figure 2 This is a schematic diagram of the fuzzy neural network structure described in this invention.

[0039] Figure 3 This is a flowchart of the fuzzy neural network training process described in this invention. Detailed Implementation

[0040] The preferred technical solutions of the present invention will be described in detail below with reference to the accompanying drawings. These embodiments are for illustrative purposes only and are not intended to limit the scope of protection of the present invention. Equivalent substitutions or conventional modifications made by those skilled in the art without departing from the concept of the present invention should fall within the scope of protection of the present invention.

[0041] Example 1: See Figure 1 A smart dosing control system for wastewater treatment constrained by metering pumps

[0042] A multi-source data acquisition module is installed at one or more locations in the inlet and outlet sections of the water treatment unit to collect influent and effluent water quality parameters during the wastewater treatment process, such as pH, temperature, flow rate, turbidity, color, total phosphorus, total nitrogen, and COD. The data acquisition module may include online sensors, signal conditioning circuits, acquisition interfaces, and data buffer units. The acquired data is transmitted to the data preprocessing module via wired or wireless means.

[0043] The data preprocessing module performs noise reduction, outlier handling, and standardization on the collected data.

[0044] The drug demand calculation module is trained using historical data and human experience to predict the optimal target drug dosage under current operating conditions.

[0045] The signal conversion module converts the target drug dosage into a PWM signal and transmits it to the PLC controller. The logic analysis module is also used to perform closed-loop correction of the PWM control signal based on the feedback signal after drug dosage, and to perform real-time correction of the PWM based on the feedback signal.

[0046] The PLC control module controls the metering pump in the drug dosing module according to the PWM signal to complete the precise control of drug dosing. The metering pump control includes, but is not limited to, controlling the operating frequency, stroke length and / or motor speed of the metering pump.

[0047] The chemical dosing module may include a chemical tank, a metering pump, valves, and a chemical delivery pipeline. The chemical to be added may be one or more of the following: coagulant, flocculant, disinfectant, alkali, acid, or phosphorus removal agent.

[0048] In the above modules, the data preprocessing module, the reagent demand calculation module, and the signal conversion module are deployed on the embedded edge controller. The embedded edge controller interacts with the PLC control module via serial communication or industrial Ethernet. In this embodiment, the data input and control signal output are completed in the Raspberry Pi. The reagent dosing module completes the precise dosing of reagents by controlling the start and stop frequency, stroke, or flow rate of the metering pump.

[0049] Example 2: See Figure 2 Data preprocessing and drug requirement calculation module

[0050] The multi-source data acquisition module transmits the real-time detected water quality parameters to the data preprocessing module. The data preprocessing module sequentially performs missing value imputation, outlier removal, and standardization on the acquired data. Preferably, the Z-score standardization method is used to normalize monitoring data of different dimensions, ensuring that all input parameters have a uniform data scale. The preprocessed data serves as the input parameter for the dosage prediction module. , where x ∗This represents the data after normalization. This represents the input sample data. Represents the sample mean. This represents the sample standard deviation.

[0051] The drug demand calculation module includes an input layer, a fuzzification layer, a fuzzy rule layer, a normalization layer, and an output layer.

[0052] The input layer receives water quality parameters and reagent-related data from the wastewater treatment process. Where n represents the number of input parameters, and n is an integer greater than or equal to 1. These represent the water quality parameters of the influent and effluent during the wastewater treatment process, namely pH value, temperature, color, turbidity, DO, COD, total phosphorus, total nitrogen, influent and effluent flow rate, and return flow rate.

[0053] The fuzzification layer uses a Gaussian membership function to convert input variables into corresponding membership degrees. In the formula: , Let be the width and center of the membership function of the j-th fuzzy set of the i-th input variable.

[0054] The fuzzy rule layer achieves comprehensive analysis of multiple input factors through fuzzy rule matching, performing rule matching based on the membership degree of each input variable. In the formula It is the fuzzy segmentation number of the i-th input.

[0055] The normalization layer is used to calculate the normalized value of the applicability of each rule and to normalize the trigger strength of each rule. .

[0056] The output layer outputs the target drug dosage y based on the normalized rule weights. ,in, For connection weights, For the applicability of the rule, r is 1 to... Integers between [a certain range].

[0057] Example 3: See Figure 3 Training of fuzzy neural network prediction models

[0058] The training samples for the dosage prediction module consist of three parts: first, historical operational data of wastewater treatment plant chemical dosing, including influent and effluent water quality parameters as well as the type and dosage of chemicals; second, expert experience data in the field of wastewater treatment, which obtains corresponding chemical dosing schemes by setting different water quality parameters and target effluent effects; and third, online data continuously accumulated during actual operation after the system is built.

[0059] The dosage prediction module can use the RBF learning algorithm to adjust the network connection weights, membership function center values, and width to improve the model's fitting and adaptive capabilities for complex working conditions. The error cost function can be... Parameter updates can be achieved using gradient descent combined with momentum factor, where the update expression can be written as: In the formula, For learning rate, Momentum factor This represents the number of iteration steps.

[0060] Example 4: Conversion of target drug dosage to control signal and feedback correction

[0061] The signal conversion module first calculates the target reagent mass flow rate M based on the target treatment concentration and the influent flow rate. In the formula, M is the mass concentration of the reagent added, mg / h; C is the target concentration, mg / L; and Q is the influent flow rate, L / h.

[0062] Meanwhile, the mass flow rate and volumetric flow rate of the reagent can be numerically converted: In the formula, The target reagent volumetric flow rate, The density of the drug solution, This represents the effective mass fraction of the drug solution.

[0063] Secondly, establish the calibration relationship between the metering pump flow rate and the PWM. For nonlinear relationships, piecewise fitting or table lookup methods can be used to generate the corresponding duty cycle or equivalent driving command linearity; for linear intervals, a linear mapping can be directly used, which can be expressed as: .

[0064] At the same time, the duty cycle is limited: In the formula, D is the PWM duty cycle (0~100%), and k and b are calibration parameters.

[0065] Finally, based on the controller's PWM register range, the actual PWM output value is converted: In the formula, PWM max This is the full-scale value of the register.

[0066] Example 5: PLC Control and Troubleshooting

[0067] The PLC control unit controls the operating frequency, stroke length, and motor speed of the metering pump based on the modified drive signal. When the system identifies that the connected metering pump is an electromagnetic diaphragm pump, the operating frequency control mode can be preferred. When the system identifies that the connected metering pump has an adjustable stroke mechanism, the stroke length control mode can be preferred. When the system is equipped with a frequency converter, the motor speed control mode can be preferred.

[0068] When the sensor malfunctions, the sample reliability is insufficient, the model output jumps beyond the limit, or the actuator feedback is abnormal, the system switches to the safe dosing mode and calls the pre-stored safe dosing curve; after the data returns to normal and the feedback stabilizes, it switches back to the collaborative dosing control mode.

[0069] The above description is merely a preferred embodiment of the present invention, used to illustrate differentiated revision ideas, and is not intended to limit the scope of protection. Equivalent substitutions or conventional changes made by those skilled in the art without departing from the concept of the present invention should fall within the scope of protection of the present invention.

Claims

1. A smart dosing control system and method for wastewater treatment oriented towards metering pump constraints, characterized in that: The system includes a multi-source data acquisition module, a data preprocessing module, a reagent demand calculation module, a signal conversion module, a PLC control module, and a reagent dosing module. The multi-source data acquisition module collects influent and effluent water quality parameters from the wastewater treatment unit. The data preprocessing module, connected to the multi-source data acquisition module, processes the collected water quality parameters, handling missing values, removing outliers, and determining reliability to obtain input data. The reagent demand calculation module performs inference calculations on the input data based on a fuzzy neural network model, outputting the target reagent dosage. The signal conversion module generates a control signal based on the target reagent dosage, reagent parameters, and metering pump calibration relationships, including the metering pump's minimum stable start-up threshold, start-up dead zone, stroke hysteresis, and duty cycle change rate limits, to compensate and correct the initial drive signal. The PLC control module outputs the final drive command based on the control signal. The reagent dosing module is used to add reagents to the wastewater treatment unit under the action of drive commands.

2. The intelligent dosing control system for wastewater treatment oriented towards metering pump constraints as described in claim 1, characterized in that: The parameters collected by the multi-source data acquisition module include one or more of the following: pH value, temperature, color, turbidity, DO, COD, total phosphorus, total nitrogen, influent and effluent flow rate, reflux flow rate, liquid level in the drug tank, outlet pressure of the metering pump, and actual number of dosing pulses. The collected data parameters are detected, stored, and then transmitted to the data preprocessing module for preprocessing.

3. The intelligent dosing control system for wastewater treatment oriented towards metering pump constraints according to claim 1, characterized in that: The data preprocessing module includes missing value imputation, outlier removal, time alignment, and sample credibility evaluation. When the sample credibility is lower than a preset threshold, the sample is removed, downweighted, or switched to a safety control mode. The processing method is Z-score standardization. in This indicates the processed data. This represents the input sample data. Represents the sample mean. This represents the sample standard deviation.

4. The intelligent dosing control system for wastewater treatment oriented towards metering pump constraints according to claim 1, characterized in that: The reagent demand calculation module employs a fuzzy neural network model with process rule initialization and residual correction mechanisms. The model outputs the target reagent dosage rather than directly outputting metering pump control parameters. This module includes an input layer, a fuzzification layer, a fuzzy rule layer, a normalization layer, and an output layer. The input layer is used to input data, including water quality parameters and reagent dosage data. The fuzzification layer uses a Gaussian membership function to fuzzify the input variables. The fuzzy rule layer is used to perform fuzzy rule matching. The normalization layer is used to normalize the rule applicability calculation. The output layer is used to output the target reagent dosage. The first layer is the input layer. The input values ​​are water quality parameters and reagent dosage data during the wastewater treatment process, represented as follows: Where n represents the number of input parameters, and n is an integer greater than or equal to 1. These represent the water quality parameters of the influent and effluent during the wastewater treatment process, namely pH value, temperature, color, turbidity, DO, COD, total phosphorus, total nitrogen, influent and effluent flow rates, return flow rate, and type and dosage of added chemicals. The input-output relationship of this node is as follows: n The second layer is the fuzzification layer. Each node in this layer corresponds to a linguistic variable value. Its function is to convert the input deterministic quantity into a fuzzy vector. The input variable is transformed into a corresponding membership degree through the membership function defined on its fuzzy subset. A Gaussian function is selected as the membership function. The membership degree of each input component corresponding to different fuzzy linguistic values ​​is as follows: In the formula: , The width and center of the membership function of the j-th fuzzy set of the i-th input variable; The third layer is the fuzzy rule layer. Each node in this layer represents a fuzzy rule. Fuzzy rule matching is achieved through connections with the fuzzification layer, enabling fuzzy computation between nodes. The output of each node is the product of all its signals, i.e.: In the formula It is the fuzzy segmentation number of the i-th input; The fourth layer performs normalized calculations, using the following formula: The fifth layer is the output layer. Its function is to convert the fuzzy output into a sharp output, and to output the PID parameters. in, For connection weights, For the applicability of the rule, r is 1 to... Integers between [a certain range].

5. The intelligent dosing control system for wastewater treatment oriented towards metering pump constraints according to claim 1, characterized in that: The training samples for the fuzzy neural network model in the reagent demand calculation module include historical wastewater treatment operation data, expert experience data, and continuously updated online data during system operation. An initial rule set is established based on the influent load range, target effluent index range, and reagent type. An error-driven adaptive learning algorithm is used to update the network connection weights, membership function center values, and width parameters. Simultaneously, the rule weights or correction coefficients are adjusted online based on the post-dosing deviation residuals. The parameter that the fuzzy neural network needs to learn is the center value of the membership function in the second layer. and width and the fifth layer network connection weights The learning algorithm chosen is the RBF algorithm, which is essentially a multi-layer feedforward network with arbitrary nonlinear mapping capabilities from input to output. The learning algorithm is key to optimizing the weight coefficients of the fuzzy neural network. By using the learning algorithm to change the width and center of the network weights and membership functions, the network's adaptive control process is achieved. The RBF neural network learning algorithm is characterized by adjusting the connection weights through the learning process. The central value of the membership function and width The formula for calculating the error cost function is: In the formula, , For the ideal output and actual output of each iteration, Indicates control error; The algorithm for learning network weights is as follows: The learning methods for the center vector and width vector of the fuzzy layer are as follows: If momentum factor is considered, the weights of the output layer are: In the formula, For the iterative steps of the network; For learning rate, ; To learn the momentum factor, .

6. The intelligent dosing control system for wastewater treatment oriented towards metering pump constraints according to claim 1, characterized in that: The signal conversion module is used to convert the target agent dosage into the target agent mass flow rate or volume flow rate, and generate an initial drive signal based on the calibration curve of the metering pump, piecewise linear fitting relationship or lookup table relationship, including feedforward-feedback correction. Calculation of drug dosage: In the formula, M is the mass concentration of the reagent added, mg / h; C is the target concentration, mg / L; and Q is the influent flow rate, L / h. Conversion between mass flow rate and volumetric flow rate for reagent concentration: In the formula, The target reagent volumetric flow rate, The density of the drug solution, This represents the effective mass fraction of the drug solution. Establish the calibration relationship between the metering pump flow rate and the PWM. When dealing with nonlinear relationships, piecewise linearity, lookup table methods, and other conversion fitting techniques are used. For linear approximations, it can be expressed as: or Duty cycle limit is: In the formula, D is the PWM duty cycle (0~100%), and k and b are calibration parameters; Based on the range of the controller's PWM register, it can be converted into the actual PWM output value: In the formula, The value is the full-scale value of the register; The feedforward-feedback correction refers to a composite correction of the initial drive signal based on effluent quality feedback, actual dosage feedback, and execution status feedback, and a switch to a safe dosage mode when there are sensor, model, or actuator malfunctions. The PWM control signal is also corrected in a closed loop based on feedback signals after chemical dosing. These feedback signals include one or more of the following: flow rate feedback, pH feedback, conductivity feedback, turbidity feedback, and chemical dosing effect feedback. The PWM is then corrected in real time based on these feedback signals. In the formula, Calculated from the target dosage. Corrections are made based on feedback from water quality parameters such as flow meter readings, pH, and conductivity.

7. The intelligent dosing control system for wastewater treatment oriented towards metering pump constraints according to claim 1, characterized in that: The PLC control module can select one or more control modes, including action frequency control mode, stroke length control mode and motor speed control mode, and automatically switch or manually specify the control mode according to the metering pump type to adjust the actual dosage of the agent dosing module.

8. A wastewater treatment intelligent dosing control system oriented towards metering pump constraints according to claim 1, characterized in that: The data preprocessing module, the reagent demand calculation module, and the signal conversion module are deployed on an embedded edge controller (Raspberry Pi), which interacts with the PLC control module via serial communication or industrial Ethernet.

9. A smart dosing control method for wastewater treatment oriented towards metering pump constraints, characterized in that: Includes the following steps: S1 collects influent and effluent water quality parameters, flow parameters, reagent dosing and operation feedback parameters during the sewage treatment process. After the multi-source data acquisition module collects the parameters, it is connected to the GPIO port of the Raspberry Pi to realize data collection from the water body. S2, the collected water quality data is transmitted to the data preprocessing module of the Raspberry Pi for preprocessing. Time alignment, missing value compensation, outlier filtering and confidence judgment are performed on the collected data to obtain estimated input data; S3, input the estimated parameters into the fuzzy neural network prediction model and output the target dosage of the drug; S4 converts the target dosage into an initial drive signal based on the drug solution density, effective mass fraction, metering pump calibration relationship, and control cycle. S5 compensates and corrects the initial drive signal for the minimum stable pump start threshold, pump start dead zone, stroke hysteresis and duty cycle change rate limit. S6 sends the corrected drive signal to the PLC control unit to control the operating frequency, stroke length and / or motor speed of the metering pump to realize the addition of the agent. S7, based on water quality feedback and execution feedback after addition, performs online correction of the estimation model and compensation parameters.

10. The intelligent dosing control method for wastewater treatment oriented towards metering pump constraints according to claim 1, characterized in that, Step S7 is followed by an anomaly handling step. When the effluent index changes abruptly, the sample reliability is insufficient, or the actuator feedback exceeds the limit, the pre-stored safety dosing curve is invoked, and the dosing control mode is switched back after the situation returns to normal. At the same time, water quality treatment effect data and the final dosage are collected and used as online update samples to supplement the training database of the dosage prediction module for subsequent model optimization.

Images