Intelligent medicine adding system and method based on machine learning

Abstract

The invention discloses an intelligent dosing system and method based on machine learning, and the method comprises the steps: S1, collecting fluidized bed operation parameters and water quality characteristic data in real time, including water flow dynamic parameters, water quality index parameters and particle morphology parameters, and constructing a time sequence characteristic database; s2, constructing a dynamic dosing model based on the fluidized bed by using data in the time sequence characteristic database, and performing joint prediction on the medicament type, the dosing amount and the operation parameters; s3, according to a prediction result output by the model, controlling the agent dosage, the agent concentration, the water flow speed and the aeration intensity; and S4, monitoring the fluidized bed and dosing control parameters in real time, and dynamically correcting the fluidized bed parameters in combination with the effluent turbidity of the fluidized bed and the particle sedimentation efficiency, so as to solve the problems that the sewage treatment effect in the prior art is poor, and medicament waste, substandard treatment and secondary environmental pollution are possibly caused.

Description

Technical Field

[0001] The present invention relates to the technical field of water treatment, and in particular to the technical field of precipitation of suspended impurities, and specifically to an intelligent dosing system and method based on machine learning. Background Art

[0002] At present, the safety requirements for drinking water quality are becoming increasingly higher, and more precise dosing methods are needed to ensure that water quality meets standards.

[0003] In the field of water treatment, traditional dosing methods have problems. For example, the dosage of the agent is difficult to control accurately, resulting in agent waste or poor treatment effect; the dosing strategy cannot be adjusted in real time according to changes in water quality, and the adaptability is poor. The development of machine learning technology has provided new ideas and methods for the research of smart dosing systems. By establishing a data-driven model, intelligent optimization and control of the dosing process can be achieved.

[0004] Smart dosing methods can improve treatment efficiency and ensure the continuity and stability of the water treatment process through real-time monitoring and automatic adjustment. The current smart dosing technology mainly collects water quality data through sensors, and then uses intelligent algorithms to automatically adjust the dosage of chemicals based on these data to achieve intelligent control. However, the feedback system of existing technology cannot fully reflect the various factors in the water treatment process, resulting in inaccurate and timely adjustments to the dosing strategy.

[0005] Therefore, it is necessary to improve the intelligent dosing system and method based on machine learning in the prior art to solve the above problems. Summary of the invention

[0006] The present invention overcomes the shortcomings of the prior art and provides an intelligent dosing system and method based on machine learning, aiming to solve the problems of poor sewage treatment effect in the prior art, which may lead to waste of drugs, substandard treatment and secondary environmental pollution.

[0007] To achieve the above object, the technical solution adopted by the present invention is: a smart dosing method based on machine learning, comprising: S1. Real-time collection of fluidized bed operation parameters and water quality characteristic data, including: hydrodynamic parameters, water quality index parameters and particle morphology parameters, to build a time series characteristic database; S2. Use the data in the time series feature database to build a dynamic dosing model based on the fluidized bed to jointly predict the type of agent, dosage and operating parameters; S3. Control the dosage of the reagent, reagent concentration, water flow rate and aeration intensity according to the prediction results output by the model; S4. Real-time monitoring of fluidized bed and dosing control parameters, dynamic correction of fluidized bed parameters based on fluidized bed effluent turbidity and particle settling efficiency, to achieve optimal dosing control.

[0008] In a preferred embodiment of the present invention, a micro-hole injection device is provided at the bottom of the fluidized bed to directly inject the agent into the particle fluidized dense area, and a multi-point dosing device is provided at the water inlet and different positions of the fluidized bed; The medicine is a sustained-release medicine, which is a coated particle, and the coating layer material is a material of polylactic acid-glycolic acid copolymer, chitosan-sodium alginate composite film and mesoporous silica coating layer; According to the velocity distribution of the fluidized bed flow field, uncoated reagents are added to the bottom high-speed zone, and sustained-release reagents are added to the top low-speed zone.

[0009] In a preferred embodiment of the present invention, the drug concentration is controlled: The particle growth rate is calculated by real-time monitoring of particle size distribution, agglomerate roundness and particle morphology parameters. ,in, is the median particle size at different times, and t is the time interval; when the particle growth rate is less than the set slow growth threshold , increasing the concentration of the drug, the increased concentration and When the particle growth rate is greater than the set growth threshold , reduce the concentration of the drug, reduce the concentration and According to the change of particle growth rate, the concentration of the agent is adjusted every 10-15 minutes.

[0010] In a preferred embodiment of the present invention, the water flow rate and aeration intensity are coordinated and regulated: The aeration intensity is determined based on the dissolved oxygen concentration of the water quality index parameter and the local turbulence intensity of the hydrodynamic parameter; when the dissolved oxygen concentration Less than the lower limit of dissolved oxygen , and the local turbulence intensity Less than the lower limit of turbulence intensity When the aeration intensity is increased, ,in, is the initial aeration intensity, is the corresponding weight; when the dissolved oxygen concentration Greater than the upper limit of dissolved oxygen , and the local turbulence intensity Greater than the upper limit of turbulence intensity When the aeration intensity is reduced, ,in, is the corresponding weight.

[0011] In a preferred embodiment of the present invention, in step S2, the specific steps of constructing the dynamic dosing model are: S21, integrating historical data into a time series feature database as a model data source, and merging hydrodynamic parameters, water quality indicators, and particle morphology parameters into a 12-dimensional feature vector; S22, Dynamic joint optimization of agent delivery strategy and fluidized bed operation using Actor-Critic framework; S23. Perform transfer learning by combining pre-training with fine-tuning on actual data.

[0012] In a preferred embodiment of the present invention, the coating layer has a radial porosity gradient, and the porosity decreases from 50% to 10% from the inside to the outside.

[0013] In a preferred embodiment of the present invention, S4 includes: combining the real-time operating status of the fluidized bed, including turbidity and sedimentation efficiency, to perform dynamic optimization: The hydraulic retention time is determined based on water quality indicators and particle settling rate. The hydraulic retention time refers to the average time that sewage stays in the fluidized bed. Greater than the upper limit of turbidity , and the particle settling rate Less than the lower limit of sedimentation rate When the hydraulic retention time is increased, ,in, is the initial hydraulic retention time, is the corresponding weight; when Less than the lower limit of turbidity , and the particle settling rate Greater than the upper limit of sedimentation rate When the hydraulic retention time is reduced, ,in, is the initial hydraulic retention time, is the corresponding weight.

[0014] In a preferred embodiment of the present invention, the carrier material of the particles in the fluidized bed is magnetite and pyrite The mass ratio of the mixture is 3-5:1; the multi-level pore structure is constructed by chemical modification, including: constructing micropores by hydrothermal synthesis, constructing mesopores by sol-gel method and constructing macro pores by 3D printing technology; the pore size of the micropores is 0.3-0.4nm, the pore size of the mesopores is 5-10nm, and the pore size of the macro pores is 50-200μm.

[0015] In a preferred embodiment of the present invention, the functional groups are loaded by the particle surface grafting technology in the fluidized bed, amino groups are grafted on the surface of the carrier material by vapor deposition, and -SH groups are introduced by impregnating the particles with thioacetic acid solution.

[0016] The present invention provides a smart dosing system based on machine learning, comprising: Multimodal sensor network module, used to collect fluidized bed operation parameters and water quality characteristic data in real time; The data processing module is used to clean, synchronize and extract features from the original data and build a time series feature database; Deep reinforcement learning decision module, used to dynamically optimize the agent dosing strategy and fluidized bed operating parameters; A dosing control module is used to control the dosage of the agent, the concentration of the agent and the water flow rate according to the prediction results output by the model; The reagent dosing module is used to perform the reagent dosing operation according to the control parameters; The closed-loop feedback and adaptive module is used to collect the turbidity data at the fluidized bed outlet and the sedimentation rate data of particles in the sedimentation tank in real time and dynamically correct the operating parameters.

[0017] The present invention solves the defects existing in the background technology and has the following beneficial effects: (1) The present invention proposes a smart dosing system and method based on machine learning, which collects fluidized bed operating parameters and water quality characteristic data in real time through multimodal sensors to build a time series feature database. The deep reinforcement learning model is used to jointly predict the agent type, dosage and operating parameters, and dynamically adjusts the agent dosage, agent concentration, water flow rate and aeration intensity according to the prediction results output by the model to achieve optimal dosing control. The system monitors the turbidity and particle settling efficiency of the fluidized bed effluent in real time, and combines the Q-learning algorithm for dynamic correction to ensure the treatment effect; the present invention realizes the dual-stage release and efficient reaction of the agent through the design of the coating layer of the sustained-release agent and the coordinated dosing of the flow field velocity distribution, combined with the multi-level pore particle carrier, and improves the adsorption and sedimentation efficiency of the particles, thereby improving the efficiency and quality of sewage treatment and reducing the waste of agents and the risk of secondary pollution.

[0018] (2) The present invention realizes dual-stage release and efficient reaction of the agent by designing the coating layer of the sustained-release agent and coordinating the flow field velocity distribution, combined with a multi-level pore particle carrier. Specifically, the uncoated agent is added to the bottom high-speed zone to quickly take effect, and the sustained-release agent is added to the top low-speed zone (the porosity of the coating layer decreases from 50% to 10% from the inside to the outside). Combined with the mass transfer advantages of the carrier micropores (0.3-0.4nm), mesopores (5-10nm) and macropores (50-200μm), the agent diffusion and pollutant adsorption are promoted. In addition, multi-point delivery can ensure the uniform distribution of the agent in the fluidized bed and enhance the adsorption and removal ability of the particles for the agent. The agent adheres to the surface of the fluidized particles, forming a larger particle agglomerate sedimentation, and finally achieves separation from water. Compared with the prior art, the instantaneous dilution problem of the agent is solved, the action time is extended, and the agent waste and secondary pollution risk are further reduced.

[0019] (3) The present invention realizes adaptive optimization of dosing strategy and outputs the optimal solution through real-time monitoring of particle growth rate and dynamic adjustment of reagent concentration, combined with closed-loop feedback of Q-learning algorithm. Specifically, the growth rate is calculated according to particle size distribution and roundness. When the rate is lower than the threshold, the concentration is increased proportionally, otherwise it is reduced. The network parameters are updated by Q-learning driven by the mean square error of error function turbidity and sedimentation rate. Compared with the prior art, the problem of adjustment lag of traditional system is overcome, and the dynamic response speed of reagent concentration control and the stability of treatment effect are further improved.

[0020] (4) The present invention optimizes the matching relationship between aeration intensity and water flow velocity by coordinating the dissolved oxygen concentration and the local turbulence intensity and combining it with the dynamic correction of the hydraulic retention time. Specifically, when the dissolved oxygen is below the lower limit and the turbulence is insufficient, the aeration intensity is increased, and vice versa. At the same time, the hydraulic retention time is adjusted according to the turbidity and sedimentation rate. Compared with the prior art, this solves the contradiction that high turbulence causes particles to be difficult to settle, further improves the solid-liquid separation efficiency and reduces energy consumption.

[0021] (5) The present invention combines surface grafting technology with amino and -SH group loading and multi-level channel carrier materials to enhance the catalytic adsorption capacity of particles, adopts vapor deposition method to graft amino groups to enhance flocculant binding, introduces -SH groups through thioacetic acid impregnation to capture heavy metals, and combines micropore selective adsorption, mesopore accelerated mass transfer and macropore reduced resistance. Compared with the existing technology, it breaks through the limitation of single function of traditional carriers, further realizes the synergistic removal of complex pollutants, and significantly improves the versatility and adaptability of sewage treatment. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art are briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative work. Figure 1 is a flow chart of a preferred embodiment of the present invention; Figure 2 is a structural diagram of a sustained-release pharmaceutical preparation according to a preferred embodiment of the present invention; Figure 3 It is a schematic diagram of the device structure of a preferred embodiment of the invention; As shown in the figure: 1. Coating layer; 2. Sustained-release layer; 3. Quick-release layer; 4. Fluidized bed; 5. Granules; 6. Dosing device; 7. Water inlet. DETAILED DESCRIPTION

[0023] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0024] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0025] Application Overview: Traditional smart dosing systems mainly focus on adjusting the dosage and frequency of chemical agents according to water quality parameters. After the agents are added, a certain amount of time is needed for mixing and reaction so that the agents can fully combine with the impurities in the water to form flocs that are easy to settle. After the mixing reaction, the impurities in the water combine with the agents to form larger particles. These particles settle in the sedimentation tank under the action of gravity to achieve solid-liquid separation. However, the importance of physical and chemical processes has been ignored, which limits its ability to optimize the sewage treatment effect, resulting in the inability of the reagent to fully react with the impurities, forming flocs that are easy to settle, thereby affecting the effect of solid-liquid separation. In addition, insufficient reaction may also lead to reagent residues, causing secondary pollution to the environment.

[0026] Therefore, this application proposes a wastewater treatment solution that combines intelligent dosing with a granulation fluidized bed. It can monitor the water flow conditions and the state of the granular medium in real time, and use machine learning algorithms to find the optimal operating conditions, such as water flow rate, type and concentration of the agent, so as to achieve accurate dosing and uniform distribution of the agent, promote the formation and growth of particle agglomerates, and increase their sedimentation rate and effect. It can not only shorten the treatment time and reduce the risk of secondary pollution caused by poor sedimentation, but also further improve the efficiency and quality of sewage treatment.

[0027] Exemplary methods: like Figure 1 As shown, a smart dosing method based on machine learning includes the following steps: S1. Real-time collection of fluidized bed operation parameters and water quality characteristic data, including: hydrodynamic parameters, water quality index parameters and particle morphology parameters, to build a time series characteristic database; S2. Use the data in the time series feature database to build a dynamic dosing model based on the fluidized bed to jointly predict the type of agent, dosage and operating parameters; S3. Control the dosage of the reagent, reagent concentration, water flow rate and aeration intensity according to the prediction results output by the model; S4. Real-time monitoring of fluidized bed and dosing control parameters, and dynamic correction of fluidized bed parameters based on fluidized bed effluent turbidity and particle settling efficiency.

[0028] A granulation fluidized bed is a device that forms a fluidized state of solid particles under the action of a fluid. In the fluidized state, the particles exhibit properties similar to those of a fluid, such as fluidity and compressibility. This technology was originally used mainly in the chemical and pharmaceutical industries, but is now being transferred and applied to sewage treatment by this application. This is because the particles in the fluidized bed are in a suspended state, which increases the contact opportunities between the agent and the impurities in the sewage, which is conducive to the flocculation reaction. The limitation of the traditional dosing system is that it cannot adapt to changes in water quality in real time, while the fluidization characteristics of the granulation fluidized bed directly affect the efficiency of particle formation, thereby improving the efficiency of adsorption sedimentation and chemical reactions, and significantly improving the treatment effect.

[0029] In step S1, the fluidized bed operation parameters and water quality characteristic data are collected in real time by integrating a multimodal sensor array installed in the water treatment system, the raw data is preprocessed, and a time series characteristic database is constructed; In step S1, the multimodal sensor array includes a water quality sensor group, an optical sensor array and an acoustic sensor group integrated in the granulation fluidized bed system; the sensor layout adopts a spatial grid distribution to ensure the data coverage density in the three-dimensional space of the fluidized bed, and eliminates the limitations of a single sensor through multi-source data fusion, providing a comprehensive process characterization.

[0030] The original data include: hydrodynamic parameters, water quality index parameters and particle morphology parameters; hydrodynamic parameters include: water flow velocity, aeration intensity, bed expansion rate and local turbulence intensity; water quality index parameters include: real-time pH value, turbidity, dissolved oxygen and hardness ion concentration; particle morphology parameters include: particle size distribution, aggregate roundness, surface roughness and sedimentation rate; The raw data is preprocessed by removing outliers, filtering noise and standardizing to avoid the sensor signal drift caused by the high turbulence environment in the fluidized bed; Through the timestamp synchronization technology, the sensor data with different sampling frequencies are unified to the same time base, and the time-aligned raw data matrix is ​​constructed; the sampling frequency of the water quality sensor is 1Hz, and the sampling frequency of the optical sensor is 10Hz. The linear interpolation method is used to align the two to 10Hz; because the acoustic sensor is non-uniformly sampled, it is resampled to 10Hz through the sliding window averaging method; through the above, the raw data is aligned on the time axis and integrated into a structured matrix; The row dimension of the matrix is ​​the timestamp , a total of time points; the column dimension is the original data parameter, parameters; Multidimensional time series data matrix: , where each row corresponds to a multimodal observation at a time point, and each column corresponds to a time series of sensor parameters.

[0031] For each sensor data stream, statistics are calculated based on a sliding window length of 30s, and data points that exceed the threshold are marked as abnormal and removed; according to the fluidized bed operation principle, the hard constraints are: water flow velocity , pH , the data that does not conform to the physical feasible domain are directly eliminated; Wavelet transform combined with Kalman filter algorithm is used to separate high-frequency noise signals and retain effective characteristic frequency bands. Wavelet transform can decompose the signal into components of different frequencies, thereby effectively separating high-frequency noise; Kalman filter algorithm can estimate the state of the signal in real time based on the predicted value and observed value of the signal, further improving the filtering effect; Z-score standardization of multi-source data ,in, is the original data, is the mean of the data, is the standard deviation of the data, eliminating dimensional differences.

[0032] The standardized multimodal data is stored as a structured database to build a time series feature database; the timestamp is the primary key and the hydrodynamic parameters are 4-dimensional vectors. , representing water velocity, aeration intensity, bed expansion rate and local turbulence intensity respectively; water quality index parameters are 4-dimensional vectors , representing the real-time pH value, turbidity, dissolved oxygen and hardness ion concentration respectively; the particle morphology parameter is a 4-dimensional vector , represent the median particle size, agglomerate roundness, surface roughness and sedimentation rate of the particles, respectively.

[0033] The preprocessing in step S1 eliminates sensor noise and dimensional differences, and the time-aligned multidimensional data matrix can be used as a direct input to the neural network; In step S2, the specific steps of constructing the dynamic dosing model are: S21, integrating historical data into a time series feature database as a model data source, and merging hydrodynamic parameters, water quality indicators, and particle morphology parameters into a 12-dimensional feature vector; S22. Dynamically jointly optimize the agent delivery strategy and fluidized bed operation using the Actor-Critic framework; the strategy network outputs: discrete action (agent selection) and continuous action (dose and flow rate adjustment) value networks, evaluates state value, and integrates turbidity error, sedimentation rate error and agent cost optimization strategy through reward functions; S23, transfer learning is performed by combining pre-training with fine-tuning actual data; In step S21, by integrating historical operation data and real-time preprocessing data, a high-dimensional, multimodal time series feature database is constructed to provide diversified samples for model training and enhance the model's generalization ability for complex working conditions.

[0034] In step S22, the Actor-Critic framework is a core architecture in reinforcement learning, which combines the dual mechanisms of strategy optimization and value assessment; The input features of the input layer are 12-dimensional fused feature vectors; The network structure of the policy network Actor is a 3-layer fully connected network, i.e., input layer → 256 → 128 → output layer; the discrete action in the output action space is the selection of the agent type, and the continuous action is the addition concentration and water flow rate adjustment; Drug type prediction uses Softmax function to generate multi-category probability distribution ,in, is the activation value of the kth neuron in the output layer, corresponding to the kth drug, and selects the drug type with the highest probability as the delivery decision at the current moment; The dosage prediction uses the Tanh activation function to generate a normalized dosage adjustment amount, which is then mapped to the actual dosage range through linear transformation, thereby dynamically adjusting the dosage of the agent to adapt to water quality fluctuations; Dosage prediction ,in, The normalized value of the network output, is the predicted value of the dosage of the agent, is the dynamic adjustment amount of the drug concentration, is a normalization factor used to map the adjustment amount to a ratio within the high and low ranges, is the lower and upper limits of the dose; Operation parameter prediction is output through the continuous action output layer of the Actor network. ,in The adjustment ratio of the network output, is the current reference flow rate; The Actor network simultaneously outputs discrete action drug types and continuous action dosage and flow rate, shares underlying features through multi-task learning, and captures dependencies between parameters; The network structure of the value network Critic shares the input layer with the Actor, and the output is the state value estimate , used to evaluate the value of the current state and guide the Actor strategy update; The Critic network guides the Actor network to balance the impact of the three types of actions by evaluating the comprehensive reward. The reward function is ; is the difference, is the corresponding weight coefficient. The turbidity penalty term is the error weight term used to quantify the deviation of turbidity from the target range in the dynamic dosing model. The upper and lower limits of turbidity are set. By penalizing the situation where turbidity exceeds the standard, the model is driven to adjust parameters to return the turbidity to a safe range.

[0035] In step S23, the flow field data generated by the CFD-PBM coupling model is used to update the Actor-Critic network parameters through actual data to correct the deviation between simulation and reality; Dynamically adjust the learning rate based on the model's performance on the validation set , is the current learning rate, is the attenuation factor.

[0036] In step S2, the fusion feature library of historical data and real-time data covers various working conditions, reducing the model's dependence on a single data source, and the Actor-Critic framework realizes the coordinated control of agent delivery and fluidized bed operation.

[0037] In step S3, the dosage of the reagent, the concentration of the reagent, the water flow rate and the aeration intensity are controlled in combination.

[0038] like Figure 3As shown in the figure, due to the high turbulence of the fluidized bed, the agent may be quickly diluted, making it difficult to form an effective concentration gradient on the particle surface. Therefore, multi-point dosing devices are set at the water inlet and different positions of the fluidized bed to use the turbulent state of the fluidized bed to achieve rapid diffusion of the agent, ensuring that the agent can be evenly dispersed in the fluidized bed and fully contact with the particle medium. Specifically, a micro-hole injection device is arranged at the bottom of the fluidized bed to directly inject the reagent into the intensive fluidized area of ​​the particles, and a multi-point dosing device is arranged at the water inlet of the fluidized bed and different positions of the bed body, which increases the contact opportunity between the reagent and the particle medium and is conducive to the flocculation reaction. Dosage of drug: For diameter Meters of fluidized bed, at a distance from the bottom 6, 5, and 4 dosing points are evenly set on the circumference at the height. The dosage of the reagents at different positions is determined according to the flow field simulation. The dosage of the bottom micro-hole injection device accounts for 0.3-0.5, the dosage of the water inlet accounts for 0.1-0.2, and the dosage of the dosing points at other positions of the bed accounts for 0.3-0.6. Each point is allocated in proportion according to factors such as the height from the bottom and the flow field velocity. like Figure 2 As shown, the drug is a sustained-release drug, specifically, the sustained-release drug is a coated particle, which further realizes the sustained release of the drug after combination to prevent the drug from being quickly diluted; the coating layer material can be selected from a polymer with good biocompatibility and slow release performance, specifically a polylactic acid-glycolic acid copolymer, a chitosan-sodium alginate composite membrane and a mesoporous silica coating layer. The coating structure is designed as a core-shell structure, the inner layer is a rapid release layer, and the outer layer is a sustained-release layer, so as to realize the dual-stage release of the drug, and the coating layer has a radial pore gradient, and the porosity decreases from 50% to 10% from the inside to the outside, so as to match the turbulence intensity of different regions in the fluidized bed; According to the velocity distribution of the flow field simulated by CFD, uncoated agents are added in the bottom high-speed zone to achieve rapid onset of action, and sustained-release agents are added in the top low-speed zone to achieve the effect of prolonging the action time, thereby extending the action time of the agents and avoiding instantaneous dilution.

[0039] Drug concentration control: Specifically, the particle growth rate is calculated by real-time monitoring of particle size distribution, agglomerate roundness, and particle morphology parameters. ,in, is the median particle size at different times, and t is the time interval; when the particle growth rate is less than the set slow growth threshold , increasing the concentration of the drug, the increased concentration and When the particle growth rate is greater than the set growth threshold , reduce the concentration of the drug, reduce the concentration and According to the change of particle growth rate, the concentration of the agent is adjusted every 10-15 minutes.

[0040] Water flow rate and aeration intensity control: The aeration intensity is determined based on the dissolved oxygen concentration of the water quality index parameter and the local turbulence intensity of the hydrodynamic parameter. The dissolved oxygen concentration is the amount of dissolved oxygen. The local turbulence intensity refers to the degree of turbulence in the sewage fluidized bed, which is expressed by the root mean square value of the water flow velocity fluctuation. High turbulence intensity helps to fully mix the reagent and impurities in the water and improve the reaction efficiency. However, too high turbulence intensity may make it difficult for particles to settle. It is necessary to adjust the water flow velocity and aeration intensity to optimize the sedimentation conditions. When the dissolved oxygen concentration is Less than the lower limit of dissolved oxygen , and the local turbulence intensity Less than the lower limit of turbulence intensity When the aeration intensity is increased, ,in, is the initial aeration intensity, is the corresponding weight; when the dissolved oxygen concentration Greater than the upper limit of dissolved oxygen , and the local turbulence intensity Greater than the upper limit of turbulence intensity When the aeration intensity is reduced, ,in, is the corresponding weight.

[0041] In step S4, Online monitoring: Collect turbidity data at the outlet of the fluidized bed and sedimentation rate data of particles in the sedimentation tank in real time, compare the real-time collected turbidity and sedimentation rate data with the model prediction results, and calculate the error function; Execution feedback: For turbidity prediction and actual value , the error function can be expressed as the mean square error , where i is the predicted value of sedimentation rate at a certain time point; and actual value Using the same mean square error, we get ; The overall error function combines the errors in turbidity and sedimentation rate: , is the weight of the corresponding error; Dynamic correction: Based on the real-time monitoring data and the feedback of the error function, the Q-learning algorithm is used to update the value network parameters. By performing different parameter adjustments and learning the optimal strategy based on the processing effect, the optimal adjustment plan is finally output. Defining the state space is a multidimensional time series data matrix, including hydrodynamic parameters, water quality index parameters and particle morphology parameters, action space It is a combination of drug type, dosage and operating parameters; ,in, is the learning rate, is the discount factor, For the status Execute an action After receiving the immediate return, For the next state The maximum Q value of all possible actions a under different water quality conditions is obtained through continuous iterative updates, so that the system can learn the optimal dosing strategy under different water quality conditions.

[0042] Dynamic optimization is performed based on the real-time operating status of the fluidized bed, such as turbidity and sedimentation efficiency.

[0043] The hydraulic retention time is determined based on water quality indicators and particle settling rate. The hydraulic retention time refers to the average time that sewage stays in the fluidized bed. Greater than the upper limit of turbidity , and the particle settling rate Less than the lower limit of sedimentation rate When the hydraulic retention time is increased, ,in, is the initial hydraulic retention time, is the corresponding weight; when Less than the lower limit of turbidity , and the particle settling rate Greater than the upper limit of sedimentation rate When the hydraulic retention time is reduced, ,in, is the initial hydraulic retention time, is the corresponding weight.

[0044] The particles in the fluidized bed are composite carrier materials with surface catalytic activity. The multi-level pore structure is constructed by chemical modification and is prepared by combining surface grafting technology to load functional groups.

[0045] The carrier material of the particles is magnetite and pyrite The mass ratio of the mixture is 3-5:1; magnetite has magnetic recovery characteristics, and pyrite contains sulfur active sites, which are suitable for oxidation-reduction synergistic reactions; Multi-level pore structure construction: The ZIF-8 metal organic framework was generated on the support surface by hydrothermal synthesis, with a pore size of 0.3-0.4 nm. The micropores are small and can selectively adsorb small molecules. The hydrothermal method uses zinc nitrate and 2-methylimidazole as precursors, the temperature is 120°C, and the growth time is more than 12 hours. The sol-gel method is used to deposit mesopores with a pore size of 5-10 nm, forming a mass transfer channel that is conducive to the rapid diffusion and transmission of reactants and products. The sol-gel method uses TEOS as the silicon source and CTAB as the template to deposit mesoporous SiO2 on the surface of ZIF-8. Macroscopic pores are constructed through 3D printing with a pore size of 50-200μm. The macroscopic pores can significantly reduce the resistance of the fluid in the fluidized bed and improve the overall mass transfer efficiency. By selecting suitable carrier materials and constructing a multi-level pore structure, more efficient and environmentally friendly catalysis, adsorption and separation processes can be achieved. On the basis of constructing a multi-level pore structure, combined with surface grafting technology, surface grafting technology is a method of grafting specific functional groups or molecular segments onto the surface of a material through chemical bonding. Through surface grafting, more active sites and affinity for impurities and agents in sewage are introduced on the surface of the carrier material; The vapor deposition method is adopted, with 3-aminopropyltriethoxysilane as the modifier, and the amino groups are grafted at 150°C under nitrogen for 2-4 hours. Through the vapor deposition method, APTES can be evenly distributed on the surface of the particles and chemically bonded with the active sites on the surface of the carrier material to form a stable amino-NH2 grafted layer. The amino group has a strong electrostatic adsorption ability and can produce electrostatic adsorption with the negatively charged groups in the anionic flocculant, thereby enhancing the binding force between the particles and the flocculant and improving the flocculation effect, thereby further improving the efficiency and quality of sewage treatment.

[0046] The particles are impregnated with a 0.5M thioacetic acid solution to introduce -SH groups. The -SH groups have strong affinity and reactivity, and can react specifically with a variety of pollutants in sewage, such as forming stable sulfide precipitates with heavy metal ions, thereby achieving effective removal of heavy metal ions. At the same time, the -SH groups can also undergo addition reactions or redox reactions with heavy metal ions to form sulfide precipitates, thereby enhancing the versatility and adaptability of the particles in sewage treatment.

[0047] Through the multi-level pore structure and surface functional modification of the above-mentioned particles, the intelligent dosing system in this application can achieve a more efficient and environmentally friendly sewage treatment process, and significantly improve the efficiency and quality of sewage treatment.

[0048] The multi-level pores of the particles and the slow-release agent work together. Uncoated agents are added to the high-speed zone at the bottom of the fluidized bed, and slow-release agents are added to the low-speed zone at the top. Combined with the multi-level pore structure of the carrier, the two-stage release of the agent is achieved. The macroscopic pores reduce fluid resistance, the mesopores accelerate the diffusion of the agent, and the micropores selectively adsorb small molecular pollutants, thereby increasing the contact probability between the agent and the impurities. The CaCO3 precipitate generated by the reaction of reagents such as lime with calcium ions in water preferentially nucleates and grows at the amino sites on the surface of the carrier to form stable aggregates, which, combined with the precipitate generated by the reagent, achieves the synergistic removal of complex pollutants.

[0049] Example systems: A smart dosing system based on machine learning, comprising: Multimodal sensor network module, used to collect fluidized bed operation parameters and water quality characteristic data in real time; The data processing module is used to clean, synchronize and extract features from the original data and build a time series feature database; Deep reinforcement learning decision module, used to dynamically optimize the agent dosing strategy and fluidized bed operating parameters; A dosing control module is used to control the dosage of the agent, the concentration of the agent and the water flow rate according to the prediction results output by the model; The reagent dosing module is used to perform the reagent dosing operation according to the control parameters; The closed-loop feedback and adaptive module is used to collect the turbidity data at the fluidized bed outlet and the sedimentation rate data of particles in the sedimentation tank in real time and dynamically correct the operating parameters.

[0050] The above is based on the ideal embodiment of the present invention. Through the above description, relevant personnel can make various changes and modifications without departing from the technical concept of the present invention. The technical scope of the present invention is not limited to the content in the specification, and the technical scope must be determined according to the scope of the claims.

Claims

1. A smart dosing method based on machine learning, characterized in that: Includes steps: S1. Real-time collection of fluidized bed operation parameters and water quality characteristic data, including: hydrodynamic parameters, water quality index parameters and particle morphology parameters, to build a time series characteristic database; S2. Use the data in the time series feature database to build a dynamic dosing model based on the fluidized bed to jointly predict the type of agent, dosage and operating parameters; S3. Control the dosage of the reagent, reagent concentration, water flow rate and aeration intensity according to the prediction results output by the model; S4. Real-time monitoring of fluidized bed and dosing control parameters, combined with fluidized bed effluent turbidity and particle settling efficiency to dynamically correct fluidized bed parameters to achieve optimal dosing control.

2. The intelligent dosing method based on machine learning according to claim 1, characterized in that: A micro-hole injection device is set at the bottom of the fluidized bed to directly inject the agent into the densely fluidized particle area, and a multi-point dosing device is set at the water inlet and different positions of the fluidized bed; The medicine is a sustained-release medicine, which is a coated particle, and the coating layer material is a material of polylactic acid-glycolic acid copolymer, chitosan-sodium alginate composite film and mesoporous silica coating layer; According to the velocity distribution of the fluidized bed flow field, uncoated reagents are added to the bottom high-speed zone, and sustained-release reagents are added to the top low-speed zone.

3. The intelligent dosing method based on machine learning according to claim 1, characterized in that: Drug concentration control: The particle growth rate is calculated by real-time monitoring of particle size distribution, agglomerate roundness and particle morphology parameters. ,in, is the median particle size at different times, and t is the time interval; when the particle growth rate is less than the set slow growth threshold , increasing the concentration of the drug, the increased concentration and When the particle growth rate is greater than the set growth threshold , reduce the concentration of the drug, reduce the concentration and According to the change of particle growth rate, the concentration of the agent is adjusted every 10-15 minutes.

4. The intelligent dosing method based on machine learning according to claim 1, characterized in that: Coordinated regulation of water flow rate and aeration intensity: The aeration intensity is determined based on the dissolved oxygen concentration of the water quality index parameter and the local turbulence intensity of the hydrodynamic parameter; when the dissolved oxygen concentration Less than the lower limit of dissolved oxygen , and the local turbulence intensity Less than the lower limit of turbulence intensity When the aeration intensity is increased, ,in, is the initial aeration intensity, is the corresponding weight; when the dissolved oxygen concentration Greater than the upper limit of dissolved oxygen , and the local turbulence intensity Greater than the upper limit of turbulence intensity When the aeration intensity is reduced, ,in, is the corresponding weight.

5. The intelligent dosing method based on machine learning according to claim 1, characterized in that: In step S2, the specific steps of constructing the dynamic dosing model are: S21, integrating historical data into a time series feature database as a model data source, and merging hydrodynamic parameters, water quality indicators, and particle morphology parameters into a 12-dimensional feature vector; S22, Dynamic joint optimization of agent delivery strategy and fluidized bed operation using Actor-Critic framework; S23. Perform transfer learning by combining pre-training with fine-tuning on actual data.

6. The intelligent dosing method based on machine learning according to claim 2, characterized in that: The coating layer has a radial porosity gradient, with the porosity decreasing from 50% to 10% from the inside to the outside.

7. The intelligent dosing method based on machine learning according to claim 1, characterized in that: S4 includes: combining the real-time operating status of the fluidized bed, including turbidity and sedimentation efficiency, to perform dynamic optimization: The hydraulic retention time is determined based on water quality indicators and particle settling rate. The hydraulic retention time refers to the average time that sewage stays in the fluidized bed. Greater than the upper limit of turbidity , and the particle settling rate Less than the lower limit of sedimentation rate When the hydraulic retention time is increased, ,in, is the initial hydraulic retention time, is the corresponding weight; when Less than the lower limit of turbidity , and the particle settling rate Greater than the upper limit of sedimentation rate When the hydraulic retention time is reduced, ,in, is the initial hydraulic retention time, is the corresponding weight.

8. The intelligent dosing method based on machine learning according to claim 1, characterized in that: The carrier material of the particles in the fluidized bed is magnetite and pyrite The mass ratio of the mixture is 3-5:1; the multi-level pore structure is constructed by chemical modification, including: constructing micropores by hydrothermal synthesis, constructing mesopores by sol-gel method and constructing macro pores by 3D printing technology; the pore size of the micropores is 0.3-0.4nm, the pore size of the mesopores is 5-10nm, and the pore size of the macro pores is 50-200μm.

9. The intelligent dosing method based on machine learning according to claim 8, characterized in that: The particle surface grafting technology in the fluidized bed loads the functional groups, the amino groups are grafted on the surface of the carrier material by vapor deposition, and the -SH groups are introduced by impregnating the particles with thioacetic acid solution.

10. A smart dosing system based on machine learning, based on a smart dosing method based on machine learning according to any one of claims 1 to 9, characterized in that: include: Multimodal sensor network module, used to collect fluidized bed operation parameters and water quality characteristic data in real time; The data processing module is used to clean, synchronize and extract features from the original data and build a time series feature database; Deep reinforcement learning decision module, used to dynamically optimize the agent dosing strategy and fluidized bed operating parameters; A dosing control module is used to control the dosage of the agent, the concentration of the agent and the water flow rate according to the prediction results output by the model; The reagent dosing module is used to perform the reagent dosing operation according to the control parameters; The closed-loop feedback and adaptive module is used to collect the turbidity data at the fluidized bed outlet and the sedimentation rate data of particles in the sedimentation tank in real time and dynamically correct the operating parameters.

Images