Wastewater desalination treatment automatic control method and system based on internet of things

Abstract

The application relates to the technical field of data processing, and discloses a wastewater desalination treatment automatic control method and system based on an Internet of Things. The method comprises the following steps: collecting a membrane surface image, an interfacial potential, an infrared spectrum, a transmembrane pressure difference, a feed flow and a flocculation image; calculating a salt gradient attention feature vector according to a salt concentration difference value and a Berkeley number; inputting the membrane pollution time sequence memory network combined with multi-modal data to obtain a pollution development rate and a flocculant dosage ratio; extracting a floc particle size parameter to obtain a flocculation effect score; and constructing a multi-objective function optimization solution to obtain a control instruction. By introducing a salt gradient attention mechanism and a membrane pollution time sequence memory network, combining visual image recognition and multi-modal data fusion, the problems of inaccurate membrane pollution state evaluation, low intelligent level of a control algorithm and insufficient multi-target collaborative optimization capability in the prior art are solved, and the automation level and processing efficiency of wastewater desalination treatment are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, and in particular to a wastewater desalination treatment automatic control method and system based on Internet of Things. BACKGROUND

[0002] Wastewater desalination treatment is an important link of industrial wastewater treatment. Reverse osmosis technology is widely used in high-salinity wastewater treatment in chemical and petrochemical industries due to its high desalination performance. The traditional wastewater desalination treatment automatic control method mainly relies on threshold control or PID control of conventional process parameters such as pH value, conductivity, turbidity, flow rate, temperature, and pressure. By setting upper and lower limits of parameters, the method triggers dosing, cleaning, and other operations. This control method is simple and intuitive, but the control logic is fixed and difficult to adapt to the complex changes of wastewater quality. In recent years, the development of Internet of Things technology has provided technical support for the intelligent upgrading of wastewater treatment systems. By deploying multiple types of sensors to collect process parameters in real time, using cloud platforms for data storage and analysis, remote monitoring and data visualization of the wastewater treatment process have been achieved.

[0003] However, the existing wastewater desalination treatment automatic control method based on Internet of Things still has obvious deficiencies. First, the existing method mainly relies on indirect parameters (such as transmembrane pressure difference and conductivity) to infer membrane fouling state, lacks direct observation means for membrane surface pollution morphology and chemical composition, resulting in inaccurate membrane pollution type identification and weak predictive maintenance capability. Second, the existing control algorithm mostly adopts single-parameter independent control strategy, which fails to fully consider the coupling effect of multiple factors such as concentration polarization, membrane fouling, and flocculation effect in the wastewater desalination process, resulting in insufficient control accuracy and adaptability. Third, the existing system lacks intelligent multi-objective collaborative optimization mechanism, and it is difficult to balance between membrane pollution rate control, energy consumption reduction, and reagent consumption optimization, leading to low system operation efficiency. SUMMARY

[0004] The present application provides a wastewater desalination treatment automatic control method and system based on Internet of Things, which introduces salt gradient attention mechanism and membrane pollution time series memory network, combines visual image recognition and multi-modal data fusion, solves the problems of inaccurate membrane pollution state evaluation, low intelligent level of control algorithm, and insufficient multi-objective collaborative optimization capability in the prior art, and improves the automation level and processing efficiency of wastewater desalination treatment.

[0005] In a first aspect, the present application provides a wastewater desalination treatment automatic control method based on Internet of Things, which comprises:

[0006] Step S1: collecting membrane surface image, interfacial potential time series data, infrared spectrum data, transmembrane pressure difference, feed flow rate, and flocculation reaction tank image of reverse osmosis membrane;

[0007] Step S2: dividing the membrane surface image into grid regions, calculating salt gradient attention weights according to the concentration difference between the membrane surface salt concentration and the salt concentration on the permeation side and the local Beckley number, and obtaining a salt gradient attention feature vector;

[0008] Step S3: splicing the salt gradient attention feature vector with the infrared spectrum data and the interfacial potential time series data, inputting a membrane fouling time series memory network combined with the transmembrane pressure difference and the feed flow, and obtaining a membrane fouling development rate and an optimal coagulant dosage ratio;

[0009] Step S4: extracting the particle size parameters of the flocs in the coagulation reaction tank image, performing weighted calculation on the particle size parameters, and obtaining a coagulation effect score;

[0010] Step S5: constructing a target function by weighted sum of the deviation square term of the membrane fouling development rate, the energy consumption term, the change rate square term of the optimal coagulant dosage ratio, and the deviation square term of the coagulation effect score, and solving to obtain a feed flow set value, a coagulant dosage set value, and a membrane cleaning trigger signal under the constraint condition.

[0011] In a second aspect, the application provides a wastewater desalination treatment automatic control system based on the Internet of Things, which comprises:

[0012] The acquisition module is configured to acquire a membrane surface image, interfacial potential time series data, infrared spectrum data, a transmembrane pressure difference, a feed flow, and a coagulation reaction tank image of a reverse osmosis membrane.

[0013] The calculation module is configured to divide the membrane surface image into grid regions, calculate salt gradient attention weights according to the concentration difference between the membrane surface salt concentration and the salt concentration on the permeation side and the local Beckley number, and obtain a salt gradient attention feature vector.

[0014] The input module is configured to splice the salt gradient attention feature vector with the infrared spectrum data and the interfacial potential time series data, input a membrane fouling time series memory network combined with the transmembrane pressure difference and the feed flow, and obtain a membrane fouling development rate and an optimal coagulant dosage ratio.

[0015] The extraction module is configured to extract the particle size parameters of the flocs in the coagulation reaction tank image, perform weighted calculation on the particle size parameters, and obtain a coagulation effect score.

[0016] The solving module is configured to construct a target function by weighted sum of the deviation square term of the membrane fouling development rate, the energy consumption term, the change rate square term of the optimal coagulant dosage ratio, and the deviation square term of the coagulation effect score, and solve to obtain a feed flow set value, a coagulant dosage set value, and a membrane cleaning trigger signal under the constraint condition.

[0017] In a third aspect, an Internet of Things-based automatic control device for wastewater desalination treatment is provided, comprising a memory and at least one processor, the memory storing instructions; the at least one processor calling the instructions in the memory to cause the Internet of Things-based automatic control device for wastewater desalination treatment to perform the Internet of Things-based automatic control method for wastewater desalination treatment described above.

[0018] In a fourth aspect, a computer-readable storage medium is provided, the computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the Internet of Things-based automatic control method for wastewater desalination treatment described above.

[0019] In the technical scheme provided in the present application, by deploying a multi-modal sensor network in the wastewater desalination treatment system, the membrane surface image, the interfacial potential time series data, the infrared spectrum data, the transmembrane pressure difference, the feed flow and the flocculation reaction tank image of the reverse osmosis membrane are collected, the membrane pollution state is comprehensively perceived, the limitation of relying on indirect parameters to speculate the membrane pollution state in the prior art is overcome, and a data foundation is laid for subsequent intelligent control decision. In particular, by dividing the membrane surface image into grid regions and calculating the salt gradient attention weight according to the concentration difference between the membrane surface salt concentration and the permeation side salt concentration and the local Berkeley number, the present application creatively solves the problem of poor reliability of pollution information in different regions of the membrane surface caused by concentration polarization, so that the regions with good flow state and light concentration polarization contribute more to the pollution state evaluation, and the accuracy of membrane pollution identification is significantly improved. After the salt gradient attention feature vector is spliced with the infrared spectrum data and the interfacial potential time series data, the transmembrane pressure difference and the feed flow are input into the membrane pollution time series memory network, the visual topographic features, chemical composition information and electrochemical dynamic characteristics are fully fused, the information limitation of single modal data is broken through, and the accurate identification of different pollution modes such as organic pollution, inorganic scaling and biofilm pollution is realized, providing a basis for formulating differentiated control strategies. By extracting the particle size parameters of the flocs in the flocculation reaction tank image and performing weighted calculation to obtain the flocculation effect score, the present application realizes the quantitative evaluation of the flocculation pretreatment effect, changes the status of relying on manual visual judgment in the traditional method, so that the adjustment of the flocculant dosage is more accurate and timely, effectively reducing the waste of reagents and the risk of membrane pollution.

[0020] The target function is constructed by weighting and summing the membrane pollution development rate deviation term, the energy consumption term, the optimal flocculant dosage ratio change rate term and the flocculation effect score deviation term, and the feed flow rate set value, the flocculant dosage set value and the membrane cleaning trigger signal are obtained by solving under the constraint condition, the multi-objective collaborative optimization of membrane pollution control, energy consumption optimization, stable flocculant dosage and pretreatment effect guarantee is realized, and the problem of low system operation efficiency caused by single objective control in the prior art is overcome. The salt gradient attention mechanism differentiates the pollution characteristics of different regions on the membrane surface by weighting the degree of concentration polarization, so that the model can focus on the region with the most reliable pollution information, and the negative influence of the interference information of the region with serious concentration polarization on the overall evaluation is avoided. Compared with the general attention mechanism, the attention mechanism designed for the concentration polarization phenomenon in the wastewater desalination process has stronger field adaptability and higher pollution identification accuracy. The membrane pollution time sequence memory network identifies the pollution mode probability vector according to the absorption peak intensity in the infrared spectrum data, and adjusts the forgetting gate weight based on the pollution mode probability vector, so that the network can adaptively adjust the memory behavior according to the dynamic characteristics of different pollution types. The slow accumulation characteristics of organic pollution, the burst characteristics of inorganic scaling and the periodic fluctuation characteristics of biofilm pollution are fully learned and utilized, which significantly improves the prediction accuracy of the membrane pollution development rate and the optimal flocculant dosage ratio, and provides reliable prediction information for model predictive control. The optimal control sequence containing multiple time steps is solved by the sequence quadratic programming algorithm, and the control instructions of the current time step are executed by using the rolling time domain optimization strategy. The present application realizes the forward-looking prediction of the future system state change trend and the dynamic response to real-time disturbance, and has stronger robustness and better control performance than the traditional feedback control, effectively reduces the membrane pollution rate, prolongs the service life of the membrane, reduces the energy consumption and the consumption of flocculants, and improves the overall automation level of the wastewater desalination treatment system. BRIEF DESCRIPTION OF DRAWINGS

[0021] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0022] Figure 1 An embodiment of the wastewater desalination treatment automatic control method based on the Internet of Things in the present application is shown in the figure.

[0023] Figure 2 An embodiment of the wastewater desalination treatment automatic control system based on the Internet of Things in the present application is shown in the figure.

[0024] Figure 3is a structural schematic block diagram of a wastewater desalination treatment automatic control device based on Internet of Things in an embodiment of the present application. DETAILED DESCRIPTION

[0025] The present application provides a wastewater desalination treatment automatic control method and system based on Internet of Things. The terms "first", "second", "third", "fourth" and the like (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily have to be used to describe a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "include" or "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not have to be limited to only those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0026] For ease of understanding, the specific process of the embodiments of the present application is described below. Please refer to Figure 1 One embodiment of the wastewater desalination treatment automatic control method based on Internet of Things in the present application includes:

[0027] Step S1: collecting membrane surface image, interfacial potential time series data, infrared spectrum data, transmembrane pressure difference, feed flow and flocculation reaction tank image of reverse osmosis membrane;

[0028] Step S2: dividing the membrane surface image into grid regions, calculating salt gradient attention weight according to the concentration difference between the salt concentration of the membrane surface and the salt concentration of the permeation side and the local Berkeley number, and obtaining a salt gradient attention feature vector;

[0029] Step S3: splicing the salt gradient attention feature vector with the infrared spectrum data and the interfacial potential time series data, inputting the transmembrane pressure difference and the feed flow into the membrane pollution time series memory network, obtaining the membrane pollution development rate and the optimal flocculant dosage ratio;

[0030] Step S4: extracting the particle size parameters of the flocs in the flocculation reaction tank image, performing weighted calculation on the particle size parameters, and obtaining a flocculation effect score;

[0031] Step S5: weighting and summing the deviation square term of the membrane pollution development rate, the energy consumption term, the change rate square term of the optimal flocculant dosage ratio, and the deviation square term of the flocculation effect score to construct a target function, and solving the feed flow set value, the flocculant dosage set value and the membrane cleaning trigger signal under the constraint condition.

[0032] It can be understood that the execution subject of the present application can be a wastewater desalination treatment automatic control system based on Internet of Things, and can also be a terminal or a server, and the specific implementation is not limited herein. The server is taken as an example for description in the embodiments of the present application.

[0033] Specifically, pH sensors, conductivity sensors, turbidity sensors, flow sensors, temperature sensors, and pressure sensors are deployed at the inlet of the reverse osmosis membrane module, which constitute the basic sensing layer of the wastewater desalination system. The pH sensor measures the pH value of the inlet water by the glass electrode method, and the pH value affects the charge distribution on the membrane surface and the dissolution state of the pollutants. The conductivity sensor reflects the total concentration of dissolved salts by measuring the ion conductivity in water, and the higher the conductivity value, the higher the salt concentration, which is directly related to the desalination load of the membrane. The turbidity sensor measures the content of suspended particulate matter in water by the scattering light method, and the turbidity value reflects the concentration of colloids and particulate matter in the inlet water, which is one of the main sources of membrane pollution. The flow sensor is installed on the inlet pipeline and measures the water volume passing through the pipeline per unit time by electromagnetic induction or ultrasonic principle. The flow data is used to calculate the membrane flux and optimize the dosing ratio. The temperature sensor measures the inlet water temperature, which affects the viscosity of water and the diffusion coefficient of solutes, and in turn affects the permeability of the membrane and the pollution rate. The pressure sensor measures the inlet water pressure, which can calculate the transmembrane pressure difference combined with the outlet water pressure. The transmembrane pressure difference is the core driving force parameter of the reverse osmosis process. An industrial camera is arranged on the surface of the reverse osmosis membrane, which is equipped with a macro lens and a ring light source. The camera has a resolution of 1920x1080 pixels and a frame rate of 15 frames per second. The industrial camera focuses on the membrane surface through the lens, and the light source uniformly illuminates the membrane surface to eliminate shadows. The image sensor converts the optical signal into a digital image. The membrane surface image can directly display the distribution characteristics of the pollution layer. The clean membrane surface presents a uniform light texture, and the pollution area presents dark patches or stripes. Organic pollution usually forms a viscous brown or yellow pollution layer, inorganic scaling forms white or gray-white crystal deposits, and biofilm forms green or gray-green adhesive layers. An interfacial potential online analyzer is installed on the concentrated water side of the membrane module. The interfacial potential, also known as the Zeta potential, reflects the potential difference between the membrane surface and the solution. The measurement principle is to apply an electric field near the membrane surface. Charged particles migrate under the action of the electric field. The particle migration speed is measured by a laser Doppler velocimeter, and the interfacial potential is calculated according to the speed and electric field strength. When the membrane surface adsorbs negatively charged organic pollutants, the interfacial potential becomes more negative. When the membrane surface deposits positively charged metal ions or calcium and magnesium salts, the interfacial potential moves towards the positive direction. The time sequence curve of the interfacial potential can reveal the accumulation process of the pollutants and the transition of the pollution type. The Fourier transform infrared spectrum probe emits broadband infrared light to irradiate the membrane surface. The chemical bonds on the membrane surface absorb infrared light of specific wave numbers. The unabsorbed light is received by the detector, and the time domain signal is converted to the frequency domain spectrum by Fourier transform. The infrared spectrum data is a vector, and each element corresponds to the absorption intensity of a wave number point. The amide bond in protein produces a strong absorption peak at 1650 cm -1 , the silicon-oxygen bond in silicate produces an absorption peak at 1100 cm -1 , and the carbon-hydrogen bond in lipid produces an absorption peak at 2920 cm-1 The intensity of these characteristic absorption peaks can be analyzed to quantitatively identify the degree of organic contamination, inorganic scaling and biofilm contamination on the membrane surface. A flocculation image acquisition device is installed in the flocculation pretreatment unit to take pictures of the liquid surface of the flocculation reactor from a vertical overhead angle, with a resolution of 1280x720 pixels and a frame rate of 10 frames per second. After the addition of flocculants in the flocculation reactor, the fine suspended particles in the wastewater are aggregated into larger flocs through mechanisms such as electric neutralization, adsorption bridging and net capture sweeping. The flocs are separated by gravity settling. The images taken by the flocculation image acquisition device show the morphology and distribution of the flocs. Good flocculation effect is characterized by large floc size, uniform distribution and compact structure. The flocs appear as dark areas with clear boundaries in the images, and the water between the flocs appears transparent and light-colored. When the flocculation effect is poor, a large number of small and dispersed particles are shown in the images, the water is turbid, and the floc boundaries are unclear. By collecting images of the flocculation reactor in real time and extracting floc size parameters and morphological characteristics through image processing algorithms, visual evidence is provided for optimizing the dosage of flocculants and evaluating the pretreatment effect.

[0034] In a specific embodiment, step S1 comprises:

[0035] An acid-base sensor, a conductivity sensor, a turbidity sensor, a flow sensor, a temperature sensor and a pressure sensor are installed at the inlet of the reverse osmosis membrane module to collect the inlet acid-base value, conductivity value, turbidity value, feed flow, inlet temperature and inlet pressure;

[0036] An industrial camera is arranged on the surface of the reverse osmosis membrane to take pictures of the membrane surface of the reverse osmosis membrane;

[0037] An interfacial potential online analyzer and a Fourier transform infrared spectrum probe are installed on the concentrated water side of the membrane module to collect interfacial potential time series data and infrared spectrum data;

[0038] A flocculation image acquisition device is installed in the flocculation pretreatment unit to take pictures of the flocculation process in the flocculation reactor from a vertical overhead angle to obtain images of the flocculation reactor.

[0039] Specifically, pH sensors, conductivity sensors, turbidity sensors, flow sensors, temperature sensors, and pressure sensors are deployed at the inlet of the reverse osmosis membrane module, which constitute the basic sensing layer of the wastewater desalination system. The pH sensor measures the pH value of the inlet water by the glass electrode method, and the pH value affects the charge distribution on the membrane surface and the dissolution state of the pollutants. The conductivity sensor reflects the total concentration of dissolved salts by measuring the ion conductivity in water, and the higher the conductivity value, the higher the salt concentration, which is directly related to the desalination load of the membrane. The turbidity sensor measures the content of suspended particulate matter in water by the scattering light method, and the turbidity value reflects the concentration of colloids and particulate matter in the inlet water, which is one of the main sources of membrane pollution. The flow sensor is installed on the inlet pipeline and measures the water volume passing through the pipeline per unit time by electromagnetic induction or ultrasonic principle. The flow data is used to calculate the membrane flux and optimize the dosing ratio. The temperature sensor measures the inlet water temperature, which affects the viscosity of water and the diffusion coefficient of solutes, and in turn affects the permeability of the membrane and the pollution rate. The pressure sensor measures the inlet water pressure, which can calculate the transmembrane pressure difference combined with the outlet water pressure. The transmembrane pressure difference is the core driving force parameter of the reverse osmosis process. An industrial camera is arranged on the surface of the reverse osmosis membrane, which is equipped with a macro lens and a ring light source. The camera has a resolution of 1920x1080 pixels and a frame rate of 15 frames per second. The industrial camera focuses on the membrane surface through the lens, and the light source uniformly illuminates the membrane surface to eliminate shadows. The image sensor converts the optical signal into a digital image. The membrane surface image can directly display the distribution characteristics of the pollution layer. The clean membrane surface presents a uniform light texture, and the pollution area presents dark patches or stripes. Organic pollution usually forms a viscous brown or yellow pollution layer, inorganic scaling forms white or gray-white crystal deposits, and biofilm forms green or gray-green adhesive layers. An interfacial potential online analyzer is installed on the concentrated water side of the membrane module. The interfacial potential, also known as the Zeta potential, reflects the potential difference between the membrane surface and the solution. The measurement principle is to apply an electric field near the membrane surface. Charged particles migrate under the action of the electric field. The particle migration speed is measured by a laser Doppler velocimeter, and the interfacial potential is calculated according to the speed and electric field strength. When the membrane surface adsorbs negatively charged organic pollutants, the interfacial potential becomes more negative. When the membrane surface deposits positively charged metal ions or calcium and magnesium salts, the interfacial potential moves towards the positive direction. The time sequence curve of the interfacial potential can reveal the accumulation process of the pollutants and the transition of the pollution type. The Fourier transform infrared spectrum probe emits broadband infrared light to irradiate the membrane surface. The chemical bonds on the membrane surface absorb infrared light of specific wave numbers. The unabsorbed light is received by the detector, and the time domain signal is converted to a frequency domain spectrum by Fourier transform. The infrared spectrum data is a vector, and each element corresponds to the absorption intensity of a wave number point. The amide bond in protein produces a strong absorption peak at 1650 cm -1 , the silicon-oxygen bond in silicate produces an absorption peak at 1100 cm -1 , and the carbon-hydrogen bond in lipid produces an absorption peak at 2920 cm-1 The flocculation image acquisition device is installed in the flocculation pretreatment unit, and the liquid surface of the flocculation reaction tank is photographed from a vertical top-down angle. The photographing resolution is 1280*720 pixels, and the frame rate is 10 frames per second. After the flocculant is added into the flocculation reaction tank, the fine suspended particles in the wastewater are aggregated to form larger flocs through mechanisms such as electric neutralization, adsorption bridging and net capture sweeping. The flocs are separated and settled under the action of gravity. The images taken by the flocculation image acquisition device show the morphology and distribution of the flocs. Good flocculation effect is manifested in that the flocs have large particle size, uniform distribution and compact structure. In the images, the flocs appear as dark regions with clear boundaries, and the water between the flocs appears as light transparent state. When the flocculation effect is poor, a large number of fine and dispersed particles are shown in the images, the water body is turbid, and the boundaries of the flocs are unclear.

[0040] In a specific embodiment, step S2 comprises:

[0041] The membrane surface image is subjected to contrast-limited adaptive histogram equalization processing and edge detection processing, and the processed membrane surface image is divided into a plurality of grid regions;

[0042] According to the measurement values of the conductivity sensors corresponding to each grid region, the membrane surface salt concentration of each grid region is calculated by inverse calculation based on the Nernst-Planck diffusion equation, and the permeation side salt concentration and the feed salt concentration are obtained;

[0043] According to the local permeation flux and the mass transfer coefficient of each grid region, the local Peclet number of each grid region is calculated, and the concentration difference ratio of the membrane surface salt concentration to the permeation side salt concentration and the feed salt concentration, and the ratio of the local Peclet number to the critical Reynolds number are substituted into the exponential function and the normalization function to calculate the salt gradient attention weight of each grid region;

[0044] After the convolution feature map of each grid region is subjected to element-wise multiplication operation with the corresponding salt gradient attention weight and summed, a salt gradient attention feature vector is obtained.

[0045] Specifically, the film surface image is subjected to contrast-limited adaptive histogram equalization processing, which divides the image into small regions, and performs histogram equalization on each small region. By limiting the contrast enhancement amplitude, the noise is avoided from being excessively enlarged. The processed image enhances the local detail contrast while maintaining the overall brightness distribution, making the pollution texture and deposit boundary of the film surface clearer. Then, the enhanced image is subjected to edge detection processing. The Canny operator is used for edge detection. The operator first uses a Gaussian filter to smooth the image to remove noise, then calculates the amplitude and direction of the image gradient, and then retains the local maximum points of the gradient amplitude through non-maximum suppression. Finally, a double-threshold method is used to determine strong edges and weak edges. The strong edges are directly retained, and the weak edges are only retained when they are connected to the strong edges. The result of edge detection is a binary image, with edge pixels marked as white and non-edge pixels marked as black. These edges correspond to the outlines and boundaries of the contaminants on the film surface. The processed film surface image is divided into multiple grid regions. The division method is to divide the image into 8 segments in the horizontal and vertical directions respectively, forming 64 grid regions, each with a size of 240x135 pixels. The salt concentration on the film surface is calculated according to the conductivity sensor measurement value of each grid region. The conductivity sensor measures the overall feed solution conductivity, while the salt concentration of each region on the film surface has spatial differences due to concentration polarization. Concentration polarization refers to the phenomenon that the concentration of solute near the membrane surface is higher than that of the bulk solution during membrane separation. The Nernst-Planck diffusion equation is used to back-calculate the salt concentration on the film surface of each grid region. The Nernst-Planck equation describes the mass transfer process of ions under the action of concentration gradient and electric field. In reverse osmosis membrane separation, solute is carried to the membrane surface by convection and back to the bulk solution from the membrane surface by diffusion. When reaching steady state, the convection flux equals the diffusion flux. The salt concentration on the film surface is equal to the feed salt concentration multiplied by the concentration polarization coefficient, which is equal to 1 plus the ratio of permeation flux to mass transfer coefficient. The permeation flux is calculated from the pressure sensor and flow sensor data, and the mass transfer coefficient is calculated according to the Sherwood number empirical correlation. The Sherwood number is a dimensionless mass transfer coefficient, which is proportional to the power product of Reynolds number and Schmidt number. The Reynolds number represents the ratio of inertial force to viscous force in fluid flow, and the Schmidt number represents the ratio of momentum diffusion rate to mass diffusion rate. Through this series of calculations, the salt concentration on the film surface of each grid region is obtained, as well as the permeate side salt concentration and the feed salt concentration. The permeate side salt concentration is calculated from the conductivity sensor measurement value on the product water side, and the feed salt concentration is calculated from the conductivity sensor measurement value on the feed side. The local Peclet number is calculated according to the local permeation flux and mass transfer coefficient of each grid region. The Peclet number is a dimensionless number that represents the relative strength of convective mass transfer and diffusive mass transfer, defined as the product of permeation flux, concentration polarization boundary layer thickness, and solute diffusion coefficient. The larger the Peclet number, the stronger the convective effect and the more serious the concentration polarization.The thickness of the concentration polarization boundary layer is estimated using the interfacial potential gradient at the membrane surface. The interfacial potential exhibits a gradient distribution within the concentration polarization boundary layer, and the boundary layer thickness is inversely proportional to the rate of change of the interfacial potential. The ratio of the concentration difference between the membrane surface salt concentration and the permeate side salt concentration to the feed salt concentration, and the ratio of the local Bekeleton number to the critical Reynolds number, are substituted into an exponential function and a normalization function to calculate the salt gradient attention weight for each grid region. The input to the exponential function is the negative ratio of the Bekeleton number to the critical Reynolds number. The exponential function value decreases as the ratio increases, indicating that regions with more severe concentration polarization have lower weights because the image features of these regions are more significantly affected by concentration polarization, resulting in lower reliability. The ratio of the concentration difference to the feed salt concentration reflects the relative increase in the membrane surface salt concentration. This ratio is multiplied by the exponential function value to obtain the original weight for each grid region. The original weights are then processed by a normalization function so that the sum of the weights of all grid regions equals 1. The normalization method divides the original weight of each grid region by the sum of the original weights of all grid regions. The convolutional feature maps of each grid region are multiplied element-wise with their corresponding salt gradient attention weights. The convolutional feature maps are high-dimensional feature representations obtained by extracting features from the membrane surface image using a convolutional neural network. Each grid region corresponds to a feature vector, and each dimension of the feature vector represents a certain abstract feature of that region. The element-wise multiplication operation multiplies each dimension value of the feature vector by its corresponding salt gradient attention weight. Regions with larger weights have their features enhanced, while regions with smaller weights have their features suppressed. The weighted feature vectors of each grid region are then summed. This summation operation adds the feature vectors of all 64 grid regions along their corresponding dimensions to obtain a comprehensive feature vector, called the salt gradient attention feature vector, with a dimension of 512. This feature vector integrates membrane surface contamination information from all grid regions and differentiates the contribution of different regions based on the degree of concentration polarization, ensuring that regions with good flow patterns and less concentration polarization contribute more to the final feature.

[0046] In one specific embodiment, step S3 includes:

[0047] Infrared spectral data is mapped and processed through a fully connected layer to obtain an infrared spectral embedding vector. Interface potential time series data is input into a single-layer long short-term memory network for encoding and processing to obtain an interface potential time series encoding vector.

[0048] The salt gradient attention feature vector, infrared spectral embedding vector, and interface potential temporal coding vector are concatenated to obtain a multimodal fusion feature vector.

[0049] The multimodal fusion feature vector is normalized and spliced ​​with the transmembrane pressure difference, feed flow rate, flocculant dosage and membrane flux at the historical time step to obtain the time-series input vector;

[0050] The time sequence input vector is input into the membrane pollution time sequence memory network, an organic pollution, inorganic scaling and biofilm formation pollution mode probability vector is recognized according to the absorption peak intensity in the infrared spectrum data, the forgetting gate weight of the membrane pollution time sequence memory network is adjusted based on the pollution mode probability vector, and the membrane pollution development rate and the optimal flocculant adding proportion are output after being processed by the double-layer long short-term memory network.

[0051] Specifically, the infrared spectrum data is processed by a full connection layer mapping, the infrared spectrum data is a vector containing absorption intensity values of multiple wave number points, and the vector dimension depends on the scanning resolution of the infrared spectrometer. The scanning resolution is 4cm -1 to 4000cm -1 -1 -1The resolution scanning of the infrared spectrum obtains 900 data points, forming a 900-dimensional infrared spectrum vector. The fully connected layer is a basic structure in the neural network, and each neuron of the input layer has a connection weight with each neuron of the output layer. The mapping process of the fully connected layer is to multiply the 900-dimensional infrared spectrum vector with the weight matrix and add the bias vector, then perform a nonlinear transformation through the activation function, and output a 128-dimensional infrared spectrum embedding vector. The size of the weight matrix is 900x128. The size of the bias vector is 128. The activation function uses the ReLU function, which maps negative values to 0. The positive values remain unchanged. This mapping compresses and abstracts the original absorption intensity information of the infrared spectrum into a more compact feature representation. The interfacial potential time series data is input into a single-layer long short-term memory network for encoding processing. The interfacial potential time series data contains the interfacial potential measurements of the past 10 time points. Each time point is 1 minute apart. Forming a time series sequence of length 10. Long short-term memory network is a recurrent neural network. It is specifically designed for processing time series data. A single-layer long short-term memory network contains an input gate, a forget gate, an output gate, and a memory cell. The input gate controls the degree of new information entering the memory cell. The forget gate controls the degree of old information being forgotten in the memory cell. The output gate controls the degree of information in the memory cell output to the hidden state. The interfacial potential time series is input into the long short-term memory network in chronological order. Each input of the interfacial potential value of a time point. The long short-term memory network updates its internal memory cell and hidden state. The input gate calculates the activation value according to the current input value and the hidden state at the previous time. The forget gate calculates the forgetting proportion according to the same input. The update of the memory cell is the old memory multiplied by the forgetting proportion plus the new input multiplied by the input gate activation value. The output gate calculates the output activation value according to the memory cell and the current input. The hidden state is equal to the memory cell after activation by the hyperbolic tangent function multiplied by the output gate activation value. After processing 10 time points, the final hidden state is used as the interfacial potential time series encoding vector. The dimension is 64. This encoding vector contains the trend information and dynamic characteristics of the interfacial potential over time. The salt gradient attention feature vector, the infrared spectrum embedding vector, and the interfacial potential time series encoding vector are concatenated. The concatenation operation is to connect the three vectors head-to-tail in the dimension direction. The dimension of the salt gradient attention feature vector is 512. The dimension of the infrared spectrum embedding vector is 128. The dimension of the interfacial potential time series encoding vector is 64. After concatenation, a 704-dimensional multi-modal fusion feature vector is obtained. The multi-modal fusion feature vector contains the spatial pollution distribution information of the membrane surface image, the chemical composition information of the infrared spectrum, and the electrochemical dynamic information of the interfacial potential. The three modal information complements each other. The multi-modal fusion feature vector is normalized and concatenated with the historical time step transmembrane pressure difference, feed flow, flocculant dosage, and membrane flux. The historical time step takes the past 10 time points. Each time point contains 4 operating parameters. Forming 40 historical operating parameter values.Normalization involves subtracting the historical minimum value from each parameter value and then dividing by the difference between the historical maximum and minimum values. The normalized parameter values ​​range from 0 to 1. Normalization eliminates the influence of differences in the dimensions and numerical ranges of different parameters. The 40 normalized historical operating parameter values ​​are concatenated with a 704-dimensional multimodal fusion feature vector to obtain a 744-dimensional time-series input vector. This time-series input vector is then fed into a membrane fouling time-series memory network. Based on the absorption peak intensity in the infrared spectral data, probability vectors for fouling patterns such as organic fouling, inorganic scaling, and biofilm formation are identified. The identification method involves extracting infrared spectral data at 1650 cm⁻¹. -1 1100cm -1 and 2920cm -1 The absorption peak intensities at three characteristic wavenumbers are normalized to a sum of 1. These normalized intensities are used as the probabilities of organic fouling, inorganic fouling, and biofilm formation, respectively, forming a fouling pattern probability vector. The forgetting gate weights of the membrane fouling temporal memory network are adjusted based on this probability vector. The adjustment method involves multiplying the fouling pattern probability vector by a modulation coefficient and adding it to the bias term of the forgetting gate. The modulation coefficient is set to 0.3. When the probability of a certain fouling pattern is high, the activation value of the forgetting gate increases accordingly. This enhances the retention of historical information by the memory units, allowing the network to adjust its memory behavior according to the characteristics of different fouling patterns. After processing by the bilayer long short-term memory network, the membrane fouling development rate and the optimal flocculant dosage are output. The first layer of the bilayer long short-term memory network contains 256 hidden units, and the second layer contains 128 hidden units. The output of the first layer serves as the input of the second layer. The cascading of the two layers allows the network to learn more complex temporal patterns. The network's output layer contains two parallel branches. The first branch outputs the membrane fouling development rate through a fully connected layer. The membrane fouling development rate represents the rate at which the transmembrane pressure difference increases over time. The second branch outputs the optimal flocculant dosage through the fully connected layer. The optimal flocculant dosage represents the mass of flocculant required per unit suspended solids concentration.

[0052] In one specific embodiment, step S4 includes:

[0053] Image enhancement and edge detection processing are performed on the images of the flocculation reaction tank to identify the floc outline and fill the internal voids;

[0054] The equivalent diameter, roundness, and fractal dimension of each floc are calculated based on the geometric parameters of the floc profile.

[0055] The particle size distribution of flocs in the flocculation reaction tank image was statistically analyzed, and the volume-weighted average particle size and particle size uniformity index were calculated.

[0056] The volume-weighted average particle size, particle size uniformity index, and roundness are normalized and then summed according to the weighting coefficients to obtain the flocculation effect score.

[0057] Specifically, the flocculation reaction tank image is subjected to image enhancement processing, and the image enhancement adopts a contrast-limited adaptive histogram equalization method. The method divides the image into several small block regions, and performs histogram equalization on each small block. By limiting the enhancement amplitude of the contrast, the noise is prevented from being excessively enlarged. The contrast of the flocculation and the water background in the enhanced image is enhanced, and the edge and texture details of the flocculation are clearer. The enhanced image is subjected to edge detection processing, and the edge detection adopts a Canny operator. The operator first uses a Gaussian filter to perform smoothing processing on the image to reduce the influence of noise, then calculates the gradients of the image in the horizontal and vertical directions, the pixel points with large gradient amplitudes correspond to the edge positions, the local maximum points of the gradient are retained through non-maximum suppression, and finally a double-threshold method is used to distinguish strong edges and weak edges. The strong edges are directly marked as the flocculation boundary, and the weak edges are only retained when they are connected with the strong edges. The result of the edge detection is the flocculation contour line. After recognizing the flocculation contour, morphological closing operation is performed to fill the internal cavities of the flocculation. The closing operation includes two steps of expansion and corrosion. The expansion operation uses a structure element to slide on the image, and marks all positions with any foreground pixels in the covered region of the structure element as foreground. The expansion makes the flocculation contour expand outward, and the small cavities in the contour are filled. The corrosion operation is opposite to the expansion. The structure element is used to slide on the expanded image, and only when the covered region of the structure element is all foreground pixels, the foreground is retained. The corrosion makes the expanded boundary return to the original position. After the closing operation, the flocculation region with complete internal filling is obtained. The equivalent diameter, circularity and fractal dimension of each flocculation are calculated according to the geometric parameters of the flocculation contour. The geometric parameters are extracted from the pixel coordinates of the flocculation contour. The projection area of the flocculation is equal to the number of pixels contained in the contour multiplied by the actual area of a single pixel. The actual area of a single pixel is calibrated according to the resolution and shooting distance of the camera. The perimeter of the flocculation is equal to the total length of the contour line, which is composed of a series of adjacent pixel points. The distance between adjacent pixel points is accumulated to obtain the perimeter. The equivalent diameter is defined as the diameter of a circle with the same projection area as the flocculation. The calculation method is to multiply the projection area by 4, divide by pi, and then take the square root. The circularity reflects the degree to which the shape of the flocculation approaches a circle. The calculation method is to multiply the projection area by 4 times pi, divide by the square of the perimeter. The circularity ranges from 0 to 1. When the circularity is equal to 1, the flocculation is a perfect circle. The smaller the circularity, the more irregular the shape of the flocculation. The fractal dimension reflects the looseness of the structure of the flocculation. The fractal dimension is calculated by the box counting method. The flocculation image is covered with square grids of different sizes, and the number of grids required to cover the flocculation boundary is counted. The grid size and the number of grids show a linear relationship in the double logarithmic coordinate system. The slope of the linear relationship is the fractal dimension. The larger the fractal dimension, the more compact the structure of the flocculation. The smaller the fractal dimension, the more porous the structure of the flocculation.The particle size distribution of the flocs in the image of the statistical flocculation reaction tank is counted. The particle size distribution counting method is to sort all the flocs in the image according to the equivalent diameter from small to large, calculate the volume of each floc, and calculate the volume of the floc according to the cube of the equivalent diameter, multiply by pi and then divide by 6 to obtain the total volume. This calculation assumes that the floc is spherical. The total volume is obtained by adding the volumes of all the flocs. The volume-weighted average particle size is calculated by taking the equivalent diameter corresponding to 50% of the cumulative volume. The cumulative volume of 50% indicates that the total volume of flocs smaller than the particle size is half of the total volume. The particle size uniformity index reflects the dispersion degree of the particle size distribution. The calculation method is to subtract the particle size corresponding to 10% of the cumulative volume from the particle size corresponding to 90% of the cumulative volume, and then divide the difference by the volume-weighted average particle size. The smaller the particle size uniformity index, the more concentrated and uniform the particle size distribution. The larger the particle size uniformity index, the more dispersed the particle size distribution. The volume-weighted average particle size, the particle size uniformity index and the circularity are normalized. The normalization method is to subtract the minimum expected value of each parameter value from the maximum expected value and then divide by the difference between the maximum expected value and the minimum expected value. The expected range of the volume-weighted average particle size is set to 1mm to 5mm. The expected range of the particle size uniformity index is set to 0.2 to 1.2. The expected range of the circularity is set to 0.5 to 1.0. The value range of the three normalized parameters is 0 to 1. The flocculation effect score is calculated by weighted summation according to the weight coefficients. The weight coefficients are volume-weighted average particle size 0.5, particle size uniformity index 0.3 and circularity 0.2. The particle size uniformity index is taken as 1 minus the normalized value before weighted summation, because the smaller the particle size uniformity index, the better the flocculation effect. The result of weighted summation is the flocculation effect score. The value range of the flocculation effect score is 0 to 1. The higher the score, the better the flocculation effect. When the score is greater than or equal to 0.75, it is determined that the flocculation effect is good. The corresponding flocs have the characteristics of large particle size, uniform distribution and compact structure.

[0058] In a specific embodiment, step S5 comprises:

[0059] The square deviation term of the membrane fouling development rate and the target membrane fouling rate is calculated. The pumping energy consumption is calculated according to the transmembrane pressure difference and the feed flow rate. When the membrane cleaning trigger condition is met, the membrane cleaning energy consumption is calculated to obtain the energy consumption term.

[0060] The square deviation term of the optimal flocculant dosage ratio between adjacent time steps and the target flocculation effect score is calculated.

[0061] The square deviation term of the membrane fouling development rate, the energy consumption term, the square term of the change rate of the optimal flocculant dosage ratio and the square deviation term of the flocculation effect score are weighted and summed according to the preset weight coefficients to construct the objective function.

[0062] Under the constraints of the upper limit of the trans-membrane pressure difference, the range of the feed flow rate, the range of the flocculant dosage and the lower limit of the salt rejection rate, the sequential quadratic programming algorithm is used to optimize and solve the objective function, to obtain the set value of the feed flow rate, the set value of the flocculant dosage and the membrane cleaning trigger signal.

[0063] Specifically, a squared deviation term of the membrane fouling development rate and a target membrane fouling rate is calculated, the membrane fouling development rate represents the growth rate of the transmembrane pressure difference with time, the unit is bar per hour, the value is predicted by the membrane fouling time series memory network, the target membrane fouling rate is set to 0.05 bar per hour, indicating that the membrane fouling rate is expected to be controlled at a low level, the calculation of the squared deviation term is to subtract the target membrane fouling rate from the actual predicted membrane fouling development rate, and then square it, the squaring operation makes the deviation positive or negative, and the penalty value is positive, the greater the absolute value of the deviation, the greater the penalty value. The pumping energy consumption is calculated according to the transmembrane pressure difference and the feed flow rate, the transmembrane pressure difference is measured by a pressure sensor, the unit is bar, the feed flow rate is measured by a flow sensor, the unit is cubic meters per hour, the pumping energy consumption is equal to the transmembrane pressure difference multiplied by the feed flow rate, then divided by 3600, and then divided by the pump efficiency, 3600 is the coefficient for converting hours to seconds, the pump efficiency is 0.75, the unit of pumping energy consumption is kilowatt-hour, this calculation reflects the energy consumption required to maintain the operation of the reverse osmosis membrane, the higher the transmembrane pressure difference or the greater the feed flow rate, the higher the pumping energy consumption. When the membrane cleaning trigger condition is met, the membrane cleaning energy consumption is calculated, the membrane cleaning trigger condition includes the transmembrane pressure difference exceeding the set threshold or the membrane flux falling by more than the set proportion, when the cleaning is triggered, the membrane cleaning energy consumption is 15 kilowatt-hours, which includes the power consumption of the cleaning liquid circulating pump, the heat energy consumption of heating the cleaning liquid and the auxiliary equipment energy consumption during the cleaning process, when the cleaning is not triggered, the membrane cleaning energy consumption is 0, the pumping energy consumption and the membrane cleaning energy consumption are added to obtain the energy consumption term. The squared rate of change term of the optimal flocculant dosage ratio between adjacent time steps is calculated, the optimal flocculant dosage ratio is output by the membrane fouling time series memory network, the unit is milligrams of flocculant per milligram of suspended solids, the time interval of adjacent time steps is 1 minute, the rate of change is equal to the optimal flocculant dosage ratio of the current time step minus the optimal flocculant dosage ratio of the previous time step, the unit of the rate of change is milligrams of flocculant per milligram of suspended solids per minute, the squared rate of change term is the square of the rate of change, which is used to punish the sharp fluctuations of the flocculant dosage, frequent and large changes in the flocculant dosage will lead to unstable flocculation effect, increase the waste of reagents and the complexity of operation. The squared deviation term of the flocculation effect score and a target flocculation effect score is calculated, the flocculation effect score is output by the flocculation image processing module, the value range is 0 to 1, the target flocculation effect score is set to 0.75, indicating that the flocculation effect is expected to be at a good level, the calculation of the squared deviation term is to subtract the target flocculation effect score from the actual flocculation effect score.The deviation square term of the membrane pollution development rate, the energy consumption term, the variation rate square term of the optimal coagulant dosage ratio, and the deviation square term of the flocculation effect score are weighted and summed according to preset weight coefficients, the weight coefficients are respectively 100 for the membrane pollution development rate deviation term, 50 for the energy consumption term, 200 for the optimal coagulant dosage ratio variation rate term, and 80 for the flocculation effect score deviation term, these weight coefficients reflect the relative importance of different optimization objectives, the weight of the membrane pollution rate control and the coagulant dosage stability is higher, indicating that more attention is paid to the prevention of membrane pollution and the stability of operation, the four weighted values are added to construct the objective function, the smaller the value of the objective function, the closer the system operating state to the ideal state. Under the constraints of the transmembrane pressure upper limit, the feed flow range constraint, the coagulant dosage range constraint, and the salt retention rate lower limit constraint, the transmembrane pressure upper limit constraint is set to be that the transmembrane pressure does not exceed 4.5 bar, exceeding this pressure will increase the risk of membrane damage, the feed flow range constraint is set to be that the feed flow is between 20 cubic meters per hour and 50 cubic meters per hour, too low flow will lead to low production efficiency, too high flow will lead to too strong membrane surface scouring, the coagulant dosage range constraint is set to be that the dosage is between 0 mg / L and 80 mg / L, a dosage of 0 means no coagulant is added, and the upper limit of the dosage avoids excessive dosage causing reagent waste and secondary pollution, the salt retention rate lower limit constraint is set to be that the salt retention rate is not less than 98.5%, the salt retention rate is calculated by the conductivity of the product water side and the feed side, the salt retention rate is equal to 1 minus the product water side salt concentration divided by the feed side salt concentration, then multiplied by 100%, this constraint ensures that the product water quality meets the desalination requirements. The objective function is optimized and solved by using the sequential quadratic programming algorithm, the sequential quadratic programming algorithm is a numerical method for solving nonlinear constraint optimization problems, the algorithm first linearizes the nonlinear objective function and constraint conditions at the current iteration point by Taylor expansion, converts the nonlinear optimization problem into a quadratic programming sub-problem, the objective function of the quadratic programming sub-problem is the quadratic approximation of the original objective function, and the constraint condition is the linear approximation of the original constraint condition, the search direction is obtained by solving the quadratic programming sub-problem, the step size is determined by one-dimensional search along the search direction, the control variables are updated, and the above process is repeated until the change of the objective function is less than the convergence threshold or the maximum number of iterations is reached, the convergence threshold is set to be that the change of the objective function is less than 0.01%, and the maximum number of iterations is set to be 50. The optimal control sequence including the feed flow, the coagulant dosage, the cross-flow velocity, and the membrane cleaning trigger signal is obtained by optimization and solving, the control sequence includes control decisions for multiple time steps in the future, a rolling horizon optimization strategy is adopted to execute only the control instructions of the current time step, the feed flow set value, the coagulant dosage set value, and the membrane cleaning trigger signal corresponding to the first time step in the optimal control sequence are extracted, these control instructions are issued to the field execution mechanism through the Internet of Things communication protocol.

[0064] In a specific embodiment, under the constraints of transmembrane pressure upper limit, feed flow rate range, flocculant dosage range and salt rejection lower limit, the objective function is optimized and solved by using a sequential quadratic programming algorithm to obtain the feed flow rate set value, flocculant dosage set value and membrane cleaning trigger signal, including:

[0065] The transmembrane pressure change trend is corrected based on the membrane fouling development rate, and a state space model is constructed combined with historical operation data. The state variables of the state space model include transmembrane pressure, membrane flux, membrane surface salt concentration and flocculation effect score, and the control variables include feed flow rate, flocculant dosage, cross-flow speed and cleaning trigger signal.

[0066] The state space model, objective function and constraint conditions are input into the sequential quadratic programming algorithm, the objective function is linearized and quadratic approximation processed, and a quadratic programming subproblem is constructed.

[0067] The quadratic programming subproblem is iteratively solved, and the iteration is terminated when the change of the objective function is less than the convergence threshold or the iteration number reaches the upper limit, to obtain an optimal control sequence containing multiple time steps.

[0068] The optimal control sequence is extracted using a rolling horizon optimization strategy to obtain the control instructions of the current time step, including the feed flow rate set value, flocculant dosage set value, cross-flow speed set value and membrane cleaning trigger signal.

[0069] Specifically, the trans-membrane pressure difference trend is corrected based on the membrane fouling development rate, which represents the growth rate of the trans-membrane pressure difference over time, predicted by the membrane fouling time series memory network. The future trend of the trans-membrane pressure difference is calculated by adding the current trans-membrane pressure difference value to the membrane fouling development rate multiplied by the predicted time step. The purpose of the correction is to incorporate the dynamic impact of membrane fouling on the trans-membrane pressure difference into the state space model, enabling the model to more accurately describe the evolution of system state. The state space model is constructed based on historical operation data, including trans-membrane pressure difference, membrane flux, membrane surface salt concentration, flocculation effect score, and corresponding control inputs over a period of time. The state space model is a mathematical model that describes dynamic systems, including state equations and output equations. The state equations describe the evolution of system state over time, and the output equations describe the relationship between observations and state variables. The state variables of the state space model include trans-membrane pressure difference, membrane flux, membrane surface salt concentration, and flocculation effect score, which describe the operating state of the wastewater desalination system. The trans-membrane pressure difference reflects the degree of membrane fouling, the membrane flux reflects the permeability of the membrane, the membrane surface salt concentration reflects the degree of concentration polarization, and the flocculation effect score reflects the pretreatment effect. Control variables include feed flow rate, flocculant dosage, cross-flow velocity, and cleaning trigger signal. The feed flow rate affects the hydraulic load and permeation flux of the membrane, the flocculant dosage affects the suspended solids concentration in the feed water and the membrane fouling rate, the cross-flow velocity affects the shear force on the membrane surface and the scouring effect of pollutants, and the cleaning trigger signal determines whether to start the membrane cleaning program. The state space model is obtained offline from historical operation data using the subspace identification method. The subspace identification method extracts the state sequence of the system from input-output data by constructing a Hankel matrix, and then estimates the parameters of the system matrix and the control matrix using the least squares method. The system matrix describes the coupling relationship between state variables and their own dynamic characteristics, and the control matrix describes the influence of control inputs on state variables. The state space model, objective function, and constraint conditions are input into the sequential quadratic programming algorithm. The objective function is the weighted sum of the membrane fouling development rate deviation term, energy consumption term, flocculant dosage change rate term, and flocculation effect score deviation term. The constraint conditions include the upper limit constraint of the trans-membrane pressure difference, the range constraint of the feed flow rate, the range constraint of the flocculant dosage, and the lower limit constraint of the salt rejection rate. The sequential quadratic programming algorithm performs linearization and quadratic approximation on the objective function. Linearization is a first-order Taylor expansion of the objective function at the current working point, approximating the original objective function with a linear function. Quadratic approximation is an extension of linearization, adding a second-order term to approximate the original objective function with a quadratic function. The second-order coefficient matrix of the quadratic function is updated iteratively using the quasi-Newton method, which gradually approximates the true Hessian matrix using the change information of the gradient of the objective function. The constraint conditions are also linearized by performing a first-order Taylor expansion of the nonlinear constraints at the current working point to obtain linear constraints. The linearized objective function and constraint conditions form a quadratic programming subproblem.The quadratic programming sub-problem is solved iteratively, and the solution of the quadratic programming sub-problem adopts an active set method or an interior point method. The active set method starts from a vertex of the feasible region and searches for the optimal solution along the boundary of the feasible region. The interior point method starts from the interior of the feasible region, and gradually approaches the optimal solution by incorporating the constraint conditions into the objective function through an obstacle function. Each iteration obtains a search direction and a step size, updates the control variables along the search direction, and recalculates the objective function value and the constraint violation degree. When the change in the objective function is less than the convergence threshold or the number of iterations reaches the upper limit, the iteration is terminated. The convergence threshold indicates that the improvement in the objective function is already very small, and further iteration will have limited improvement on the result. The maximum number of iterations limits the calculation time of the algorithm and avoids infinite loops. After the iteration is terminated, an optimal control sequence containing multiple time steps in the future is obtained. The optimal control sequence is a vector sequence, each vector corresponds to a control decision at a time step, and contains the feed flow rate, the flocculant dosage, the cross-flow velocity and the cleaning trigger signal at the time step. The rolling horizon optimization strategy is used to extract the control instructions at the current time step in the optimal control sequence. The core idea of the rolling horizon optimization strategy is to re-plan the control sequence for the future at each sampling time, and only execute the control instruction at the first time step. At the next sampling time, the optimization is performed again. This strategy can continuously correct the control decision according to the real-time feedback of the system state, and enhance the robustness of the system to disturbances and uncertainties. The set value of the feed flow rate, the set value of the flocculant dosage, the set value of the cross-flow velocity and the membrane cleaning trigger signal corresponding to the first time step in the optimal control sequence are extracted. These control instructions are issued to the field execution mechanism through the Internet of Things communication protocol for execution.

[0070] The wastewater desalination treatment automatic control method based on the Internet of Things in the embodiments of the present application is described above, and the wastewater desalination treatment automatic control system based on the Internet of Things in the embodiments of the present application is described below. Please refer to Figure 2 The wastewater desalination treatment automatic control system based on the Internet of Things in the embodiments of the present application includes one embodiment:

[0071] The acquisition module is configured to acquire the membrane surface image, the interfacial potential time series data, the infrared spectrum data, the transmembrane pressure difference, the feed flow rate and the flocculation reaction tank image of the reverse osmosis membrane.

[0072] The calculation module is configured to divide the membrane surface image into grid regions, calculate the salt gradient attention weight according to the concentration difference between the salt concentration on the membrane surface and the salt concentration on the permeation side and the local Berkeley number, and obtain a salt gradient attention feature vector.

[0073] The input module is configured to splice the salt gradient attention feature vector with the infrared spectrum data and the interfacial potential time series data, input the membrane pollution time series memory network in combination with the transmembrane pressure difference and the feed flow rate, and obtain the membrane pollution development rate and the optimal flocculant dosage ratio.

[0074] The extraction module is configured to extract a particle size parameter of the floc in the flocculation reaction tank image, and to obtain a flocculation effect score by performing weighted calculation on the particle size parameter.

[0075] The solving module is configured to construct a target function by performing weighted summation on a square deviation term of the membrane pollution development rate, a square term of the variation rate of the optimal flocculant addition ratio, and a square deviation term of the flocculation effect score, and to obtain the feed flow rate setting value, the flocculant addition amount setting value, and the membrane cleaning trigger signal under the constraint condition.

[0076] The above Figure 2 The wastewater desalination treatment automatic control system based on the Internet of Things in the embodiment of the application is described in detail from the perspective of modular functional entities, and the wastewater desalination treatment automatic control device based on the Internet of Things in the embodiment of the application is described in detail from the perspective of hardware processing.

[0077] Referring to Figure 3 The embodiment of the application also provides a wastewater desalination treatment automatic control device based on the Internet of Things, which can be a server, and the internal structure of the server can be as shown in Figure 3 The wastewater desalination treatment automatic control device based on the Internet of Things includes a processor, a memory, a display screen, an input device, a network interface, and a database connected through a system bus. The processor of the computer is configured to provide computing and control capabilities. The memory of the wastewater desalination treatment automatic control device based on the Internet of Things includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The database of the wastewater desalination treatment automatic control device based on the Internet of Things is configured to store corresponding data in the embodiment. The network interface of the wastewater desalination treatment automatic control device based on the Internet of Things is configured to communicate with an external terminal through a network connection. The computer program is executed by the processor to implement the above method.

[0078] Those skilled in the art can understand Figure 3 The structure shown in the embodiment of the application is only a block diagram of part of the structure related to the scheme of the application, and does not constitute a limitation on the wastewater desalination treatment automatic control device based on the Internet of Things to which the scheme of the application is applied.

[0079] The application also provides a computer readable storage medium, which can be a non-volatile computer readable storage medium or a volatile computer readable storage medium. The computer readable storage medium stores instructions, and when the instructions are run on a computer, the computer performs the steps of the wastewater desalination treatment automatic control method based on the Internet of Things.

[0080] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described system, system and unit can refer to the corresponding processes in the foregoing method embodiments, and will not be described here.

[0081] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application or the whole or part of the technical solutions that essentially contribute to the prior art can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a number of instructions for causing a wastewater desalination treatment automatic control device based on the Internet of Things (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various program code storage media.

[0082] The above embodiments are only used to illustrate the technical solutions of the present application, but not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An automatic control method for wastewater desalination treatment based on the Internet of Things, characterized by, The method comprises: Step S1: collecting the membrane surface image, interfacial potential time series data, infrared spectrum data, transmembrane pressure difference, feed flow and flocculation reaction tank image of the reverse osmosis membrane; Step S2: dividing the membrane surface image into grid regions, calculating the salt gradient attention weight according to the concentration difference value of the membrane surface salt concentration and the salt concentration on the permeation side and the local Beckley number, and obtaining the salt gradient attention feature vector, including: performing contrast limited adaptive histogram equalization processing and edge detection processing on the membrane surface image, and dividing the processed membrane surface image into a plurality of grid regions; according to the conductivity sensor measurement value corresponding to each grid region, combining the Nernst-Planck diffusion equation to calculate the membrane surface salt concentration of each grid region, and obtaining the salt concentration on the permeation side and the feed salt concentration; according to the local permeation flux and mass transfer coefficient of each grid region, the local Beckley number of each grid region is calculated, the ratio of the concentration difference value of the membrane surface salt concentration and the salt concentration on the permeation side to the feed salt concentration, and the ratio of the local Beckley number to the critical Reynolds number are substituted into the exponential function and the normalization function to calculate the salt gradient attention weight of each grid region; the convolution feature map of each grid region is multiplied element by element with the corresponding salt gradient attention weight and summed to obtain the salt gradient attention feature vector; Step S3: splicing the salt gradient attention feature vector with the infrared spectrum data and the interfacial potential time series data, inputting the transmembrane pressure difference and the feed flow into the membrane fouling time series memory network to obtain the membrane fouling development rate and the optimal flocculant dosage ratio, including: mapping the infrared spectrum data through a fully connected layer to obtain an infrared spectrum embedding vector, and inputting the interfacial potential time series data into a single-layer long short-term memory network for coding processing to obtain an interfacial potential time series coding vector; the salt gradient attention feature vector, the infrared spectrum embedding vector and the interfacial potential time series coding vector are spliced to obtain a multi-modal fusion feature vector; the multi-modal fusion feature vector is normalized and spliced with the transmembrane pressure difference, the feed flow, the flocculant dosage and the membrane flux of the historical time step to obtain a time series input vector; the time series input vector is input into the membrane fouling time series memory network, the absorption peak intensity in the infrared spectrum data is recognized to obtain a pollution mode probability vector of organic pollution, inorganic scaling and biofilm formation, the forgetting gate weight of the membrane fouling time series memory network is adjusted based on the pollution mode probability vector, and the membrane fouling development rate and the optimal flocculant dosage ratio are output after double-layer long short-term memory network processing; Step S4: extracting the particle size parameters of the flocs in the flocculation reaction tank image, performing weighted calculation on the particle size parameters to obtain a flocculation effect score; Step S5: constructing a target function by weighted sum of the deviation square term of the membrane fouling development rate, the energy consumption term, the change rate square term of the optimal flocculant dosage ratio and the deviation square term of the flocculation effect score, and solving the feed flow set value, the flocculant dosage set value and the membrane cleaning trigger signal under the constraint condition. 2.The IoT-based automatic control method for wastewater desalination treatment according to claim 1, wherein, The step S1 comprises: Install pH sensor, conductivity sensor, turbidity sensor, flow sensor, temperature sensor and pressure sensor at the water inlet end of the reverse osmosis membrane module to collect the pH value, conductivity value, turbidity value, feed flow, water temperature and water pressure; Arrange an industrial camera on the surface of the reverse osmosis membrane to shoot the membrane surface image of the reverse osmosis membrane; Install an interfacial potential online analyzer and a Fourier transform infrared spectrum probe on the concentrated water side of the membrane module to collect interfacial potential time series data and infrared spectrum data; Install a flocculation image acquisition device in the flocculation pretreatment unit to shoot the flocculation process in the flocculation reaction tank at a vertical overhead angle and obtain the flocculation reaction tank image. 3.The IoT-based automatic control method for wastewater desalination treatment according to claim 1, wherein, The step S4 comprises: Perform image enhancement processing and edge detection processing on the flocculation reaction tank image, identify the flocculation profile and fill the internal cavities; Calculate the equivalent diameter, circularity and fractal dimension of each flocculation according to the geometric parameters of the flocculation profile; Statistically analyze the particle size distribution of the flocculation in the flocculation reaction tank image, calculate the volume-weighted average particle size and the particle size uniformity index; After normalization processing of the volume-weighted average particle size, the particle size uniformity index and the circularity, weight sum the three indexes according to the weight coefficients to obtain the flocculation effect score. 4.The IoT-based automatic control method for wastewater desalination treatment according to claim 1, wherein, The step S5 comprises: Calculate the deviation square term of the membrane pollution development rate and the target membrane pollution rate, calculate the pumping energy consumption according to the transmembrane pressure difference and the feed flow, calculate the membrane cleaning energy consumption when the membrane cleaning trigger condition is met to obtain the energy consumption term; Calculate the change rate square term of the optimal flocculant dosage ratio between adjacent time steps, and calculate the deviation square term of the flocculation effect score and the target flocculation effect score; Weight sum the deviation square term of the membrane pollution development rate, the energy consumption term, the change rate square term of the optimal flocculant dosage ratio and the deviation square term of the flocculation effect score according to the preset weight coefficients to construct a target function; Under the constraints of the upper limit of the transmembrane pressure difference, the feed flow range, the flocculant dosage range and the lower limit of the salt retention rate, the target function is optimized and solved by using a sequential quadratic programming algorithm to obtain the feed flow set value, the flocculant dosage set value and the membrane cleaning trigger signal. 5.The IoT-based automatic control method for wastewater desalination treatment according to claim 4, wherein, The sequential quadratic programming algorithm is used to optimize and solve the target function under the constraints of the upper limit of the transmembrane pressure difference, the feed flow range, the flocculant dosage range and the lower limit of the salt retention rate to obtain the feed flow set value, the flocculant dosage set value and the membrane cleaning trigger signal, which comprises: Based on the membrane pollution development rate, correct the transmembrane pressure difference trend, and construct a state space model combined with historical operation data, the state variables of the state space model include the transmembrane pressure difference, the membrane flux, the membrane surface salt concentration and the flocculation effect score, and the control variables include the feed flow, the flocculant dosage, the cross-flow speed and the cleaning trigger signal; Input the state space model, the target function and the constraint conditions into the sequential quadratic programming algorithm, linearize and quadraticly approximate the target function to construct a quadratic programming subproblem; Solving the quadratic programming sub-problem iteratively, terminating iteration when the change of objective function is below a convergence threshold or the number of iterations reaches an upper limit, and obtaining an optimal control sequence including multiple time steps; Extracting control instructions of a current time step in the optimal control sequence using a rolling horizon optimization strategy, and obtaining a feed flow rate set value, a flocculant dosage set value, a cross-flow speed set value, and a membrane cleaning trigger signal.

6. An automatic control system for wastewater desalination treatment based on the Internet of Things, characterized by, The application discloses an Internet of Things (IoT)-based automatic control system for wastewater desalination treatment. The application discloses an IoT-based automatic control system for wastewater desalination treatment. The application discloses an IoT-based automatic control system for wastewater desalination treatment. The application discloses an IoT-based automatic control system for wastewater desalination treatment. The application discloses an IoT-based automatic control system for wastewater desalination treatment. The input module is used for splicing the salt gradient attention feature vector with the infrared spectrum data and the interfacial potential time series data, combining the transmembrane pressure difference and the feed flow to input the membrane fouling time series memory network, and obtaining a membrane fouling development rate and an optimal coagulant dosage ratio, and includes: mapping and processing the infrared spectrum data through a fully connected layer to obtain an infrared spectrum embedding vector, and inputting the interfacial potential time series data into a single-layer long short-term memory network for coding processing to obtain an interfacial potential time series coding vector; splicing and processing the salt gradient attention feature vector, the infrared spectrum embedding vector and the interfacial potential time series coding vector to obtain a multi-modal fusion feature vector; normalizing and splicing the multi-modal fusion feature vector with the transmembrane pressure difference, the feed flow, the coagulant dosage and the membrane flux of the historical time step to obtain a time series input vector; inputting the time series input vector into the membrane fouling time series memory network, identifying an organic pollution, inorganic scaling and biological membrane formation pollution mode probability vector according to the absorption peak intensity in the infrared spectrum data, adjusting the forgetting gate weight of the membrane fouling time series memory network based on the pollution mode probability vector, and outputting the membrane fouling development rate and the optimal coagulant dosage ratio after double-layer long short-term memory network processing; The extraction module is used for extracting a particle size parameter of the floc in the coagulation reaction tank image, performing weighted calculation on the particle size parameter, and obtaining a coagulation effect score; The solving module is used for weighting and summing a deviation square term of the membrane fouling development rate, an energy consumption term, a change rate square term of the optimal coagulant dosage ratio and a deviation square term of the coagulation effect score to construct a target function, and solving to obtain a feed flow set value, a coagulant dosage set value and a membrane cleaning trigger signal under a constraint condition.

7. A wastewater desalination treatment automatic control device based on the Internet of Things, characterized in that, The computer program is stored in the memory and can be run on the processor, and the processor implements the automatic control method of wastewater desalination treatment based on the Internet of Things when executing the computer program.

8. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is stored in the memory and can be run on the processor, and the processor implements the automatic control method of wastewater desalination treatment based on the Internet of Things when executing the computer program.

Images