Intelligent dosing control system and method for sewage treatment plant

Abstract

This invention relates to the field of wastewater treatment technology, and more particularly to an intelligent dosing control system and method for wastewater treatment plants. The system includes a multi-level water quality monitoring unit, a data processing unit, a dynamic water quality prediction unit, a strategy generation unit, and an execution unit. The method includes: real-time acquisition of influent, process point, and effluent water quality and flow data; data cleaning, time alignment based on hydraulic retention time, and feature engineering processing to construct a multi-dimensional feature vector; inputting the feature vector into a memory network model to predict future effluent water quality; the strategy generation unit generating a feedforward control quantity based on the deviation between the predicted value and the set value, and generating a feedback control quantity based on the deviation between the actual effluent value and the set value to generate the optimal dosing command; and the execution unit adjusting the dosing amount according to the command. This invention achieves advanced prediction and precise control of effluent water quality, significantly reducing chemical consumption and improving effluent compliance rates.

Description

Technical Field

[0001] This invention relates to the field of wastewater treatment technology, and more specifically, to an intelligent dosing control system and method for wastewater treatment plants. Background Technology

[0002] Wastewater treatment plants, as core infrastructure for water pollution control, directly impact ecological safety and public health through their treatment efficiency and effluent quality. Chemical dosing is a crucial step in the wastewater treatment process. By adding various agents such as coagulants, flocculants, and disinfectants to the treated water, suspended solids, colloids, and harmful substances can be removed, ensuring that the effluent meets national or local discharge standards. Currently, with increasingly stringent wastewater treatment standards and greater fluctuations in influent water quality and quantity, higher demands are placed on the accuracy, real-time performance, and stability of chemical dosing. The performance of the chemical dosing control system has become one of the core factors determining the operational efficiency and treatment effectiveness of wastewater treatment plants.

[0003] Currently, the vast majority of wastewater treatment plants still rely on traditional feedback control for chemical dosing. This involves using various monitoring devices to monitor effluent water quality indicators (such as COD, BOD, SS, TP, etc.) in real time or at set intervals. The test results are then compared with preset standards, and the chemical dosage is adjusted based on the deviation to achieve water quality compliance. This control method can basically meet basic treatment needs under normal water quality fluctuation scenarios and is characterized by its simple structure, low cost, and ease of implementation, making it widely used in the wastewater treatment industry.

[0004] Specifically, in wastewater treatment processes, water flows from the dosing point through multiple treatment units such as reaction tanks, sedimentation tanks, and filtration tanks, and the hydraulic retention time at the effluent is typically 2-8 hours, and even longer in some complex wastewater treatment plants. This means that the water quality indicators detected at the effluent actually reflect the effect of the dosing several hours prior, resulting in a significant time lag in the feedback and adjustment signals. When there are sudden changes in the influent water quality or quantity (such as a surge in suspended solids (SS) during heavy rains or a sharp increase in COD due to the influx of industrial wastewater), this lag leads to a severe disconnect in the dosing and adjustment of chemicals, failing to respond promptly to changes in water quality. If the concentration of pollutants in the influent suddenly increases, the delayed feedback signal will result in insufficient chemical dosage, failing to effectively remove pollutants in a short time, leading to effluent exceeding standards and facing environmental fines and the risk of shutdown. Conversely, if the concentration of pollutants in the influent decreases, the delayed adjustment will lead to excessive chemical dosage, significantly increasing chemical consumption costs, potentially causing secondary pollution, increasing the difficulty and cost of sludge disposal, and clogging treatment equipment, further increasing operation and maintenance costs. Furthermore, traditional feedback control modes rely solely on water quality data from a single monitoring point at the outlet, failing to fully consider the comprehensive impact of multiple factors such as influent water quality, quantity, temperature, and pH on the effectiveness of chemical dosing. This crude adjustment method makes it difficult to achieve precise matching of chemical dosage, further exacerbating the difficulty of achieving water quality standards and increasing operational cost pressures. It cannot meet the refined, efficient, and energy-saving operational needs of wastewater treatment plants.

[0005] Therefore, existing chemical dosing control systems and methods for wastewater treatment plants, limited by the traditional "outlet detection-feedback adjustment" model, cannot resolve the core contradiction between control lag caused by hydraulic retention time and real-time water quality treatment. This results in low dosing accuracy, slow response speed, and a tendency for effluent to exceed standards or for chemicals to be wasted. Furthermore, they neglect the combined effects of multiple influencing factors, have imperfect control logic, and are ill-suited to complex fluctuations in influent water quality and quantity. Consequently, they fail to meet the urgent needs of the wastewater treatment industry for water quality compliance, energy conservation, and refined operation in the context of high-quality development. Therefore, developing an intelligent chemical dosing control system and method that can effectively solve the control lag problem, achieve precise real-time control of chemical dosing, and is adaptable to the actual operating scenarios of wastewater treatment plants has become a pressing technical challenge in the wastewater treatment field, possessing significant practical importance and application value. Summary of the Invention

[0006] The technical problem this invention aims to solve lies in the contradiction between the lag in control and the real-time requirements of water treatment. Most systems rely on feedback adjustments based on the detection results of the effluent water quality. From the time the reagent is added to the time the effect is detected at the effluent, there is a process delay of several hours or even longer (hydraulic retention time). To address the above-mentioned deficiencies in the prior art, this invention provides an intelligent dosing control system and method for wastewater treatment plants.

[0007] The technical solution adopted by this invention to solve its technical problem is: on the one hand An intelligent dosing control system for wastewater treatment plants includes: A multi-level water quality monitoring unit is used to collect water quality parameters and flow data in real time at the inlet point, at least one process point, and the outlet water in the wastewater treatment process. The data processing unit is communicatively connected to the multi-level water quality monitoring unit and is used to clean, align, and perform feature engineering on the collected data to construct a multi-dimensional feature vector for prediction. The dynamic water quality prediction unit is communicatively connected to the data processing unit, and is used to receive the multidimensional feature vector and output the predicted effluent water quality value after one hydraulic retention time. The strategy generation unit is communicatively connected to the effluent monitoring point in the data processing unit and the multi-level water quality monitoring unit. It is used to generate a feedforward control quantity based at least on the deviation between the predicted effluent water quality value and the target set value, and to generate a feedback control quantity based on the deviation between the actual measured value at the effluent point and the target set value. The two quantities are then fused to generate a dosing control command. An execution unit, communicatively connected to the strategy generation unit, is used to adjust the dosage of the agent according to the dosage control command.

[0008] Preferably, the multi-level water quality monitoring unit includes: The first online water quality analyzer and the first flow meter are installed at the water inlet point; A second online water quality analyzer is installed at the outlet of the key biological treatment process section; The third online water quality analyzer is installed at the final outlet.

[0009] Preferably, the data processing unit includes a data cleaning module, a data synchronization module, and a feature construction module; the data cleaning module is used to filter and process outliers in the raw monitoring data; the data synchronization module is used to timestamp the influent data with the effluent data at the corresponding future time according to the hydraulic delay time of the process flow; the feature construction module is used to calculate derived features, including influent load, load change rate, moving average, and process section removal rate, based on the cleaned and aligned data.

[0010] Preferably, the prediction model used by the dynamic water quality prediction unit is a time series neural network model. The dynamic water quality prediction unit includes one of a long short-term memory network, a gated recurrent unit, or a temporal convolutional network. The dynamic water quality prediction unit is trained using historical data and has a mechanism to update parameters online based on prediction errors.

[0011] Preferably, the strategy generation unit includes a feedforward control channel, a feedback control channel, a rule base, and an optimization algorithm module; the feedforward control channel takes the predicted value output by the dynamic water quality prediction unit as its core input; the feedback control channel takes the real-time measurement of the effluent point in the multi-level water quality monitoring unit as its core input; the rule base stores fuzzy control rules or expert rules based on process experience; the optimization algorithm module is used to fuse and continuously optimize the feedforward control quantity and the feedback control quantity with the goal of minimizing reagent consumption or operating costs, and generate the optimal dosing setting sequence.

[0012] Preferably, the execution and drive unit includes a frequency converter-driven dosing pump or an electric regulating valve.

[0013] on the other hand A smart dosing control method for wastewater treatment plants, applied to any of the systems described above, includes the following steps: S1. Real-time collection of water quality parameters and flow data at the inlet point, at least one process point, and the outlet point in the wastewater treatment process; S2. Perform data cleaning, time alignment, and feature engineering on the collected data to construct a multi-dimensional feature vector for prediction; S3. Input the multidimensional feature vector into the dynamic water quality prediction model to obtain the predicted effluent water quality value after one hydraulic retention time. S4. Calculate the feedforward control quantity based at least on the deviation between the predicted effluent water quality value and the target set value, and calculate the feedback control quantity based on the deviation between the actual measured value at the effluent point and the target set value. Combine the two and perform optimization calculations to generate the current dosing control command. S5. Adjust the operating status of the dosing device according to the dosing control command; S6. Repeat steps S1 to S5 periodically, and update the dynamic water quality prediction model online based on the error between the predicted and actual values.

[0014] Preferably, S4 specifically includes: S41. Compare the predicted water output value with the target set value to calculate the basic feedforward control quantity; S42. Compare the current actual water output value with the target set value to calculate the feedback correction amount; S43. Call the preset expert rule base and make preliminary adjustments to the basic feedforward control quantity and feedback correction quantity in combination with the current system operating status; S44. With the constraints of achieving the predicted water discharge standard and minimizing costs over the next few control cycles, an optimization algorithm is used to perform rolling optimization on the initially adjusted dosing sequence to obtain the optimal dosing control command for the current moment.

[0015] Preferably, the triggering condition for online updating of the dynamic water quality prediction model in S6 is: the model prediction error continuously exceeds a set threshold for a predetermined time length.

[0016] The beneficial effects of this invention are as follows: 1. This application employs an Attention-LSTM network as the core prediction model, combined with rigorous data cleaning, precise time alignment based on hydraulic retention time, and multi-dimensional feature engineering (constructing derived features such as influent load and load change rate). This enables in-depth mining of the nonlinear and long-delay dynamic characteristics of the wastewater treatment process, thereby achieving accurate and advanced prediction of future effluent quality. The introduction of the attention mechanism allows the model to adaptively focus on key time steps in historical data that have a greater impact on the current prediction, further improving the accuracy and robustness of the prediction and providing a reliable basis for feedforward control.

[0017] 2. This application combines Model Predictive Control (MPC) algorithm with a preliminary fuzzy rule fusion strategy. The MPC algorithm, based on a built-in lightweight water quality prediction model, aims to minimize the deviation from effluent standards, reagent dosage costs, and operational change rates across multiple future time domains within each control cycle. It then continuously solves for the optimal control sequence while satisfying the physical constraints of the actuators. This global rolling optimization strategy, compared to traditional single-point PID control or simple feedforward-feedback combinations, can more effectively balance control performance and operating costs, avoid excessive reagent dosage, and achieve significant energy savings and consumption reduction.

[0018] 3. This application introduces an online incremental update mechanism triggered by prediction error in the dynamic water quality prediction unit. When the model prediction error (such as the mean absolute percentage error within the sliding window) continuously exceeds a preset threshold, the system automatically fine-tunes the model weights using the latest collected data without requiring complete retraining. This enables the prediction model to adapt in real time to slow or sudden changes in actual operating conditions such as fluctuations in influent water quality, changes in water temperature, and changes in sludge activity, maintaining high prediction accuracy and thus ensuring the long-term effectiveness of the control system throughout its entire lifecycle.

[0019] 4. This application adopts a multi-level, multi-strategy integrated control architecture. The feedforward control channel acts proactively based on predicted values ​​to anticipate impending water quality shocks; the feedback control channel corrects in real time based on actual effluent deviations, eliminating steady-state errors and unmodeled disturbances; and the fuzzy inference of the rule base can make reasonable nonlinear adjustments based on expert experience under extreme operating conditions or sensor anomalies, providing a better initial solution for the MPC optimizer. This significantly enhances the overall stability and anti-interference capability of the system.

[0020] 5. This application constructs a complete automated closed loop from data acquisition, processing, prediction, decision-making to execution. The system can autonomously sense changes in operating conditions, autonomously make optimization decisions, and autonomously execute control commands, and has the ability to learn and iteratively upgrade itself. This not only significantly reduces the need for manual intervention by operators and their reliance on experience, but also provides strong technical support for the refined management, energy conservation, and stable compliance of wastewater treatment plants, representing a key technological path for realizing the digital transformation and intelligent upgrading of wastewater treatment plants. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the present invention will be further described below in conjunction with the accompanying drawings and embodiments. The drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort: Figure 1 This is a schematic diagram of the components of the intelligent dosing control system for wastewater treatment plants according to an embodiment of this application.

[0022] Figure 2 This is a schematic diagram of the steps of an intelligent dosing control method for wastewater treatment plants according to an embodiment of this application. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, a clear and complete description will be provided below in conjunction with the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the protection scope of the present invention.

[0024] Example 1 Preferred embodiments of the present invention, for example Figure 1 As shown, an intelligent dosing control system for wastewater treatment plants includes a multi-level water quality monitoring unit, a data processing unit, a dynamic water quality prediction unit, a strategy generation unit, and an execution unit.

[0025] The multi-stage water quality monitoring unit is used to collect water quality parameters and flow data in real time at the influent point, at least one process point, and the effluent point in the wastewater treatment process. The multi-stage water quality monitoring unit includes, but is not limited to, a first online water quality analyzer (for monitoring COD, NH3-N, TP, pH, etc.) and a first flow meter installed at the influent point, a second online water quality analyzer installed at the outlet of the key biological treatment process section, and a third online water quality analyzer installed at the final effluent outlet. The multi-stage water quality monitoring unit continuously collects data at a frequency of 1-5 minutes per data collection and uploads it to the central control system in real time via industrial Ethernet or a wireless network.

[0026] The data processing unit communicates with multiple levels of water quality monitoring units to transform raw monitoring data into high-quality, standardized data suitable for modeling and control. It performs data cleaning, event alignment, and feature engineering to construct multi-dimensional feature vectors for prediction. The data processing unit includes a data cleaning module, a data synchronization module, and a feature construction module. The data cleaning module filters and handles outliers in the raw monitoring data. It uses a median-based Hampel filter to remove random noise caused by sensor interference. Simultaneously, it uses a 3σ-based outlier detection method to identify and remove outliers (such as signal zero jumps and out-of-range values). For temporarily missing data, it uses forward imputation or linear interpolation to complete the data. The data synchronization module timestamps the influent data with the corresponding future effluent data based on the hydraulic delay time of the wastewater treatment plant's process flow. The feature construction module calculates derived features, including influent load, load change rate, moving average, and process removal rate, based on the cleaned and aligned data.

[0027] The dynamic water quality prediction unit is communicatively connected to the data processing unit. It receives multi-dimensional feature vectors and outputs the predicted effluent water quality value after one hydraulic retention time. The prediction model used by the dynamic water quality prediction unit is a time series neural network model. The dynamic water quality prediction unit includes one of the following: long short-term memory network, gated recurrent unit, or temporal convolutional network. The dynamic water quality prediction unit is trained using historical data and has a mechanism to update parameters online based on prediction errors.

[0028] The strategy generation unit is communicatively connected to the data processing unit and the effluent monitoring points in the multi-level water quality monitoring unit. It generates feedforward control variables based on the deviation between the predicted effluent water quality value and the target setpoint, and generates feedback control variables based on the deviation between the actual measured value at the effluent point and the target setpoint. These are then fused to generate a dosing control command. The strategy generation unit includes a feedforward control channel, a feedback control channel, a rule base, and an optimization algorithm module. The feedforward control channel uses the predicted value output by the dynamic water quality prediction unit as its core input; the feedback control channel uses the real-time measurements from the effluent points in the multi-level water quality monitoring unit as its core input; and the rule base stores fuzzy control rules or expert rules based on process experience.

[0029] The optimization algorithm module is used to fuse and continuously optimize the feedforward and feedback control variables to generate the optimal dosing sequence, aiming to minimize reagent consumption or operating costs. As an optional embodiment, the optimization algorithm module employs Model Predictive Control (MPC). The core of MPC is: at the current moment, combining feedforward and feedback information, with the objective function of minimizing the predicted effluent quality and cumulative operating costs over several future control cycles (e.g., one hour divided into four 15-minute control steps), it solves online for the optimal control sequence within a finite time domain, issuing control commands only at the current moment and repeating the process at the next moment. Through this continuous optimization, the optimal allocation of control variables is achieved while satisfying complex constraints (such as pump speed limits).

[0030] The execution unit, communicatively connected to the strategy generation unit, receives and parses dosing control commands, converting them into physical actions to precisely adjust the dosage. Depending on the application scenario, the execution unit may include a frequency converter-driven dosing pump or an electric regulating valve with a positioner, achieving stepless adjustment of the dosage.

[0031] Example 2 A smart dosing control method for wastewater treatment plants includes the following steps: S1. Real-time collection of water quality parameters and flow data at the inlet, at least one process point, and the outlet of the wastewater treatment process; construction of a real-time monitoring data stream covering the entire process.

[0032] S2. The collected data is processed by clearing, time alignment and feature engineering to construct a multi-dimensional feature vector for prediction. Specifically, this includes removing abnormal fluctuation data, aligning the influent and effluent data with timestamps based on the hydraulic residence time calculated by the hydraulic model, and constructing a multi-dimensional feature vector containing information such as influent load and load change rate.

[0033] S3. Input the multidimensional feature vector into the dynamic water quality prediction model to obtain the predicted effluent water quality after one hydraulic retention time. S4. Calculate the feedforward control quantity based at least on the deviation between the predicted effluent water quality value and the target set value, and calculate the feedback control quantity based on the deviation between the actual measured value at the effluent point and the target set value. Combine the two and perform optimization calculations to generate the current dosing control command. S41. Compare the predicted water output value with the target set value to calculate the basic feedforward control quantity; S42. Compare the current actual water output value with the target set value to calculate the feedback correction amount; S43. Call the preset expert rule base and, in conjunction with the current system operating status, make preliminary adjustments to the basic feedforward control quantity and feedback correction quantity; initially superimpose the basic feedforward quantity and the feedback correction quantity. At the same time, call the preset expert rule base and, in conjunction with the current system operating status, dynamically adjust and constrain the preliminary fusion quantity.

[0034] S44. With the constraints of achieving effluent quality standards and minimizing costs over several future control cycles, an optimization algorithm is used to perform rolling optimization on the initially adjusted dosing sequence to obtain the optimal dosing control command for the current moment. The optimization algorithm is a Model Predictive Control (MPC) algorithm. The initially adjusted dosing sequence is used as the initial solution and input into the MPC optimizer. This optimizer uses the dual constraints of achieving effluent quality standards (predicted by a built-in simplified water quality model or neural network model) and minimizing chemical dosing costs over several future control cycles as its objectives. Under the condition of satisfying the physical constraints of the equipment, it solves for the optimal dosing sequence in the next control time domain. According to the rolling optimization principle of MPC, only the first value in this sequence is output as the dosing control command for the current moment.

[0035] S5. The execution unit adjusts the frequency of the dosing pump or the valve opening according to the dosing control command, thereby changing the operating status of the dosing device.

[0036] S6. Repeat steps S1 to S5 periodically to achieve continuous closed-loop control. Simultaneously, the system monitors the performance of the prediction model in real time. When the model's prediction error continuously exceeds a set threshold for a predetermined time period, an online model update mechanism is triggered. The model is then fine-tuned using the latest cached data to adapt to the slow, time-varying operating conditions, ensuring the system's long-term stability and control accuracy.

[0037] It should be understood that those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.

Claims

1. An intelligent dosing control system for wastewater treatment plants, characterized in that, include: A multi-level water quality monitoring unit is used to collect water quality parameters and flow data in real time at the inlet point, at least one process point, and the outlet water in the wastewater treatment process. The data processing unit is communicatively connected to the multi-level water quality monitoring unit and is used to clean, align, and perform feature engineering on the collected data to construct a multi-dimensional feature vector for prediction. The dynamic water quality prediction unit is communicatively connected to the data processing unit, and is used to receive the multidimensional feature vector and output the predicted effluent water quality value after one hydraulic retention time. The strategy generation unit is communicatively connected to the effluent monitoring point in the data processing unit and the multi-level water quality monitoring unit. It is used to generate a feedforward control quantity based at least on the deviation between the predicted effluent water quality value and the target set value, and to generate a feedback control quantity based on the deviation between the actual measured value at the effluent point and the target set value. The two quantities are then fused to generate a dosing control command. An execution unit, communicatively connected to the strategy generation unit, is used to adjust the dosage of the agent according to the dosage control command.

2. The intelligent dosing control system for wastewater treatment plants according to claim 1, characterized in that, The multi-level water quality monitoring unit includes: The first online water quality analyzer and the first flow meter are installed at the water inlet point; A second online water quality analyzer is installed at the outlet of the key biological treatment process section; The third online water quality analyzer is installed at the final outlet.

3. The intelligent dosing control system for wastewater treatment plants according to claim 1, characterized in that, The data processing unit includes a data cleaning module, a data synchronization module, and a feature construction module; the data cleaning module is used to filter and process outliers in the raw monitoring data; the data synchronization module is used to align the influent data with the effluent data at the corresponding future time according to the hydraulic delay time of the process flow. The feature construction module is used to calculate derived features, including influent load, load change rate, moving average, and process section removal rate, based on the cleaned and aligned data.

4. The intelligent dosing control system for wastewater treatment plants according to claim 1, characterized in that, The prediction model used by the dynamic water quality prediction unit is a time series neural network model. The dynamic water quality prediction unit includes one of a long short-term memory network, a gated recurrent unit, or a temporal convolutional network. The dynamic water quality prediction unit is trained using historical data and has a mechanism to update parameters online based on prediction errors.

5. The intelligent dosing control system for wastewater treatment plants according to claim 1, characterized in that, The strategy generation unit includes a feedforward control channel, a feedback control channel, a rule base, and an optimization algorithm module. The feedforward control channel takes the predicted value output by the dynamic water quality prediction unit as its core input. The feedback control channel takes the real-time measurement of the effluent point in the multi-level water quality monitoring unit as its core input. The rule base stores fuzzy control rules or expert rules based on process experience. The optimization algorithm module is used to fuse and continuously optimize the feedforward control and feedback control variables with the goal of minimizing reagent consumption or operating costs, thereby generating the optimal dosing sequence.

6. The intelligent dosing control system for wastewater treatment plants according to claim 1, characterized in that, The execution and drive unit includes a frequency converter-driven dosing pump or an electric regulating valve.

7. A smart dosing control method for wastewater treatment plants, applied to the system described in any one of claims 1-6, characterized in that, Includes the following steps: S1. Real-time collection of water quality parameters and flow data at the inlet point, at least one process point, and the outlet point in the wastewater treatment process; S2. Perform data cleaning, time alignment, and feature engineering on the collected data to construct a multi-dimensional feature vector for prediction; S3. Input the multidimensional feature vector into the dynamic water quality prediction model to obtain the predicted effluent water quality value after one hydraulic retention time. S4. Calculate the feedforward control quantity based at least on the deviation between the predicted effluent water quality value and the target set value, and calculate the feedback control quantity based on the deviation between the actual measured value at the effluent point and the target set value. Combine the two and perform optimization calculations to generate the current dosing control command. S5. Adjust the operating status of the dosing device according to the dosing control command; S6. Repeat steps S1 to S5 periodically, and update the dynamic water quality prediction model online based on the error between the predicted and actual values.

8. The intelligent dosing control method for wastewater treatment plants according to claim 7, characterized in that, S4 specifically includes: S41. Compare the predicted water output value with the target set value to calculate the basic feedforward control quantity; S42. Compare the current actual water output value with the target set value to calculate the feedback correction amount; S43. Call the preset expert rule base and make preliminary adjustments to the basic feedforward control quantity and feedback correction quantity in combination with the current system operating status; S44. With the constraints of achieving the predicted water discharge standard and minimizing costs over the next few control cycles, an optimization algorithm is used to perform rolling optimization on the initially adjusted dosing sequence to obtain the optimal dosing control command for the current moment.

9. The intelligent dosing control method for wastewater treatment plants according to claim 7, characterized in that, The trigger condition for online updating of the dynamic water quality prediction model in S6 is: the model prediction error continuously exceeds the set threshold for a predetermined time length.

Images