Intelligent management and control method for advanced oxidation water treatment process

Abstract

The invention belongs to the technical field of intelligent water plant operation, particularly discloses an intelligent management and control method for an advanced oxidation water treatment process, and aims to solve the problems of old technology, insufficient monitoring, low resource utilization and the like of a traditional water plant. According to the method, real-time monitoring and intelligent adjustment of the water treatment process are realized through technologies such as high-precision sensing, Internet of Things, machine learning and automatic control. The method specifically comprises the steps of data acquisition, uploading, cleaning, visualization, processing, issuing and the like, water quality data are acquired by utilizing various sensors, encrypted and uploaded to a big data center through an edge gateway, and the data are cleaned by adopting a dynamic box graph method. Meanwhile, a three-dimensional visual digital twinborn model is constructed to display real-time data, and machine learning models such as MLR, SVR, RF and LSTM are used to predict the effluent quality and generate an adjustment instruction. And a Bayesian algorithm is adopted to finely adjust and optimize hyper-parameters, so that accurate adjustment and intelligent management of process parameters are realized. According to the method, the treatment efficiency and precision are remarkably improved, the operation cost is reduced, the effluent quality is ensured to be stable and reach the standard, and remarkable progress and new breakthrough are brought to the water treatment industry.

Description

Technical Field

[0001] The present invention relates to an intelligent control method for an advanced oxidation water treatment process, and belongs to the technical field of intelligent water plant operation. Background Art

[0002] With the acceleration of urbanization, the demand for water resources continues to increase, but the problem of water pollution is also becoming more serious.

[0003] Traditional water plants face multiple limitations in the water treatment process. The first is that the equipment is old and the process is single. Many water plants still rely on traditional physical and chemical treatment methods, which are difficult to cope with complex water quality changes. In addition, traditional water treatment systems usually lack real-time monitoring and intelligent adjustment capabilities, resulting in slow response and low treatment efficiency when water quality changes suddenly, and cannot meet increasingly stringent water quality standards. As an emerging water treatment technology, advanced oxidation process (AOP) has attracted widespread attention for its efficient removal capacity and wide applicability. However, how to improve its treatment efficiency, reduce energy consumption and achieve real-time monitoring is an urgent problem to be solved. The rapid development of big data and machine learning technology has provided a new solution for the water treatment industry. Through real-time monitoring, data analysis and intelligent regulation, not only can the efficiency of water treatment be improved, but also the precise management of water quality can be achieved to ensure that the final effluent meets environmental protection standards. Therefore, it is particularly important to develop an advanced oxidation water treatment process based on intelligent control.

[0004] Secondly, the operation and management of water plants often rely on manual monitoring, and operators may make errors in data analysis and judgment. This not only affects the treatment effect, but also increases operating costs. In addition, since the water treatment process involves many variables (such as water source water quality, treatment process parameters, etc.), manual intervention can easily lead to unstable treatment results, which poses a hidden danger to the safety of water quality. Therefore, it is particularly important to introduce intelligent management and control technology. By combining big data analysis, Internet of Things technology and artificial intelligence, the intelligent management and control system can achieve real-time monitoring and intelligent adjustment of the water treatment process. This technology can automatically identify changes in water quality, predict treatment effects, and optimize and adjust when necessary, greatly improving the efficiency and stability of water treatment. In addition, intelligent management and control can also reduce labor costs and improve resource utilization efficiency, thereby achieving sustainable development of water plants while ensuring water quality safety. This innovative management method not only improves the refinement of water treatment, but also provides an effective solution to future water resource challenges.

[0005] In order to solve the problems faced by traditional water plants, such as outdated technology, insufficient monitoring, low resource utilization, and excessive reliance on personnel experience in operation, researchers have developed a variety of intelligent management and control methods for water plants. Patent CN220485416U invented a distributed water plant intelligent disinfection device, which is used in the disinfection process of the water plant, and can adjust the disinfection dosage in real time to ensure the accurate concentration of the disinfectant in the produced water. However, the device only mechanically adjusts the disinfectant dosage through the water consumption data, and does not take into account the impact of water quality fluctuations on the dosage, and there are still deficiencies in data collection. Patent CN117073762A introduces a big data collection method for intelligent water plants, which receives various water quality index collection instructions issued by the big data center through sensors, and uploads the data to the background server using wireless transmission technology. The server analyzes the data and forms processing instructions and sends them to the mobile terminal of the maintenance and inspection personnel, who complete the task processing by the maintenance and inspection personnel. The patent does not mention the specific big data algorithm used, and does not achieve full process automation, which still requires high labor costs. Patent CN114563988B invented a water plant intelligent PAC dosing method based on the random forest algorithm, which uses the random forest algorithm to build a PAC dosing prediction model, and realizes intelligent prediction and control of PAC dosage through big data samples and multiple machine learning algorithms. This patent is limited to the random forest algorithm and does not mention other more advanced machine learning algorithms or integrated learning methods. The data presentation method is not clear and does not have excellent user-friendliness and ease of use. Patent CN113433910A invented a water treatment plant intelligent dosing control system and method based on digital twins. The system includes a physical dosing room, a digital twin of the dosing room, data acquisition equipment, a network transmission system, a three-dimensional visualization system, and a management platform. This method relies on the integrity and accuracy of data acquisition, but does not mention specific data cleaning methods.

[0006] In order to make up for the above defects, the present invention provides an intelligent management and control method for advanced oxidation water treatment process, which assists water plants in better operation through the coordinated use of high-precision sensing, Internet of Things, machine learning and automated control technologies. Summary of the invention

[0007] The purpose of the present invention is to provide an intelligent control method for an advanced oxidation water treatment process, which can automatically adjust the treatment conditions of the advanced oxidation process according to the water supply conditions and treatment requirements, and solve the problems of complex operation, difficult control and high operating cost in the application of the advanced oxidation process in existing water plants. The method is characterized in that it includes the following steps:

[0008] Step 1: Data collection: Use a variety of sensors to monitor the water treatment system in real time. The big data center regularly issues data collection instructions to control instruments, valves and sensors to complete data collection. The collected indicator data includes water temperature, flow, pH, turbidity, UV254 , DO, COD, TOC, SUVA, ammonia nitrogen, total nitrogen and total phosphorus, etc.;

[0009] Step 2: Data upload: The sensor collects data on time and uploads it to the big data center through the edge gateway. Encryption protocols are used during the transmission process to ensure data security. All collected indicator data are stored in the historical database to improve the data samples of the database.

[0010] Step 3: Data cleaning: Use a variety of statistical methods to automatically detect and remove abnormal data, including box plot method, 3σ principle, missing value filling, redundant data removal, data normalization processing technology and correlation analysis. Taking the box plot method as an example, the inner limit of the box plot is Q a -αQr to Q b +αQr, where Q a For the lower quartile, Q b is the upper quartile, Q r is the interquartile range, a takes a value of 1.0-2.0, and then collects data for statistical analysis to solve the upper and lower quartiles, and the interquartile range, and then uses the inner limit as Q a -αQr to Q b +αQr method eliminates data to ensure data reliability;

[0011] Step 4: Data visualization: Use multi-source data from drinking water treatment plants, sewage treatment plants, and industrial wastewater treatment plants and stations to build a 3D visualized digital twin model, monitor and manage uploaded data in real time, and provide an intuitive process status view;

[0012] Step 5: Model construction, using a variety of machine learning models to conduct in-depth analysis of the data. The input values ​​of the model include various water quality indicators of the influent and advanced oxidation process parameters, and the output values ​​are various water quality indicators of the effluent, to establish a reliable prediction model;

[0013] Step 6: Algorithm optimization: Use optimization algorithms to fine-tune model parameters to improve prediction accuracy. The model generates specific water treatment process parameter adjustment instructions by identifying data patterns and trends, and establishes a closed-loop feedback mechanism to update model parameters using real-time data to achieve adaptive adjustment.

[0014] Step 7: Data is sent down. The generated processing instructions will be sent down to the edge control gateway and distributed to each valve and control device to adjust process parameters, implement hierarchical control, prioritize the management of key operating parameters, and ensure the system is flexible and efficient.

[0015] In the step 5, the models used include MLR model, SVR model, RF model and LSTM model.

[0016] In step 6, Bayesian optimization is used to fine-tune and optimize the hyperparameters of the model. The hyperparameters of the SVR model include c and gamma, the hyperparameters of the RF model include n_estimators, min_samples_splits and min_samples_leaf, and the hyperparameters of the LSTM model include batch_size, Learn rate, epochs, Hidden size and seed. The model is trained with actual monitoring data, and the training data accounts for 80% of the total data. The remaining 20% ​​is used as test data. The result of model fitting reaches the training set R 2 ≥0.84, test set R 2 ≥0.90.

[0017] The beneficial effects of the present invention are: combining high-precision sensing, Internet of Things, machine learning and automated control technologies, significantly improving the processing efficiency and accuracy of the advanced oxidation water treatment process, reducing operating costs, and ensuring that the effluent water quality is stable and up to standard. At the same time, through the three-dimensional visualization digital twin model, the management level and response speed are improved, and the management difficulty is reduced. In addition, the method has also promoted the development of advanced oxidation technology, bringing significant progress and new breakthroughs to the water treatment industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 is a flow chart of the method of the present invention;

[0019] Figure 2 This is a flow chart of the test result evaluation of the present invention; DETAILED DESCRIPTION

[0020] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods, but the present invention is not limited to the following embodiments.

[0021] like Figure 1 In the specific embodiment shown, we can see an intelligent control process of advanced oxidation water treatment process, which can be applied to the advanced oxidation deep treatment process of industrial wastewater. It uses high-precision sensing, Internet of Things, machine learning and automatic control technologies to control process parameters through intelligent control to maintain efficient and stable treatment performance of the advanced oxidation process. The specific control process is data collection → data upload → data cleaning → 3D visualization → model construction → algorithm optimization → data distribution.

[0022] In the above specific process, digital thermometer, electromagnetic flowmeter, ultraviolet water quality analyzer, pH sensor, turbidity meter, UV 254Detectors, DO sensors, online COD detectors, etc., collect water quality data in real time. The collected data is encrypted by the edge gateway and uploaded to the big data center and stored in the historical database. The uploaded data is cleaned using the dynamic box plot method to remove outliers. The current water volume, water quality data and structure operation status are displayed in the constructed three-dimensional visual digital twin model of the structure. The MLR, SVR, RF and LSTM models are trained according to historical data, and the model hyperparameters are optimized using the Bayesian algorithm. The current water quality indicators and ozone oxidation process parameters are input to predict the effluent water quality indicators and generate process parameter adjustment instructions. These processing instructions are then sent to the edge control gateway and further distributed to each valve and control device to achieve precise adjustment of process parameters.

[0023] like Figure 2 In the specific embodiment shown, we can see an evaluation process of an intelligent management and control system for an electro-flocculation coupled catalytic ozone oxidation water treatment process. The wastewater pH value range is 7.2-8.23, the SS concentration range is 86-130 mg / L, the COD concentration range is 156-187 mg / L, and the conductivity is 5.88-8.25 mS / cm. The models selected are MLR, SVR, RF and LSTM models. The input values ​​of the model include influent COD concentration, current density, ozone concentration, catalyst filling rate and reaction time, and the output value is the effluent COD concentration. The Bayesian method is used to optimize the five hyperparameters of batch-size, Lr, epochs, hidden-size and seed. The optimal hyperparameter values ​​are batch-size of 8, Lr of 0.004, epochs of 110, hidden-size of 6 and seed of 65. The process parameters are adjusted by reverse prediction. LSTM has a high prediction accuracy, R 2 The LSTM model provides a process parameter control scheme with a current density of 10-15 mA / cm. 2 ; Ozone concentration is 40-50 mg / L; catalyst filling rate is 10-20%; contact time is 90-105 min; ozone aeration intensity is 100-500 mL / min; electrode plate spacing is 2-4 cm, and the COD removal rate can reach 64.87-69.84%.

[0024] The specific embodiments described above are only preferred implementations of the present invention, but other aspects and implementations will be obvious to those skilled in the art. Several modifications and improvements may be made without departing from the principles of the present invention, all of which fall within the scope of protection of this application.

Claims

1. A smart control method for an advanced oxidation water treatment process, characterized in that: The following steps are involved: Step 1: Data collection: Use a variety of sensors to monitor the water treatment system in real time. The big data center regularly issues data collection instructions to control instruments, valves and sensors to complete data collection. The collected indicator data includes water temperature, flow, pH, turbidity, UV 254 , DO, COD, TOC, SUVA, ammonia nitrogen, total nitrogen and total phosphorus, etc.; Step 2: Data upload: The sensor collects data on time and uploads it to the big data center through the edge gateway. Encryption protocols are used during the transmission process to ensure data security. All collected indicator data are stored in the historical database to improve the data samples of the database. Step 3: Data cleaning: Use a variety of statistical methods to automatically detect and remove abnormal data, including box plot method, 3σ principle, missing value filling, redundant data removal, data normalization processing technology and correlation analysis. Taking the box plot method as an example, the inner limit of the box plot is Q a -αQr to Q b +αQr, where Q a For the lower quartile, Q b is the upper quartile, Q r is the interquartile range, a takes a value of 1.0-2.0, and then collects data for statistical analysis to solve the upper and lower quartiles, and the interquartile range, and then uses the inner limit as Q a -αQr to Q b +αQr method eliminates data to ensure data reliability; Step 4: Data visualization: Use multi-source data from drinking water treatment plants, sewage treatment plants, and industrial wastewater treatment plants and stations to build a 3D visualized digital twin model, monitor and manage uploaded data in real time, and provide an intuitive process status view; Step 5: Model construction, using a variety of machine learning models to conduct in-depth analysis of the data. The input values ​​of the model include various water quality indicators of the influent and advanced oxidation process parameters, and the output values ​​are various water quality indicators of the effluent, to establish a reliable prediction model; Step 6: Algorithm optimization: Use optimization algorithms to fine-tune model parameters to improve prediction accuracy. The model generates specific water treatment process parameter adjustment instructions by identifying data patterns and trends, and establishes a closed-loop feedback mechanism to update model parameters using real-time data to achieve adaptive adjustment. Step 7: Data is sent down. The generated processing instructions will be sent down to the edge control gateway and distributed to each valve and control device to adjust process parameters, implement hierarchical control, prioritize the management of key operating parameters, and ensure the system is flexible and efficient.

2. The intelligent control method of an advanced oxidation water treatment process according to claim 1, characterized in that: In the step 5, the models used include MLR model, SVR model, RF model and LSTM model.

3. The intelligent control method of an advanced oxidation water treatment process according to claim 1, characterized in that: In step 6, Bayesian optimization is used to fine-tune and optimize the hyperparameters of the model. The hyperparameters of the SVR model include c and gamma, the hyperparameters of the RF model include n_estimators, min_samples_splits and min_samples_leaf, and the hyperparameters of the LSTM model include batch_size, Learn rate, epochs, Hidden size and seed. The model is trained with actual monitoring data, and the training data accounts for 80% of the total data. The remaining 20% ​​is used as test data. The result of model fitting reaches the training set R 2 ≥0.84, test set R 2 ≥0.90.

Images