Automatic monitoring system and method for sewage treatment by vibrating membrane bioreactor

Abstract

The present invention relates to the technical field of sewage treatment and its automatic control, specifically to an automatic monitoring system and method for sewage treatment by a vibrating membrane bioreactor; the method steps are as follows: by real-time monitoring of key indicators, and performing standardization processing and interpolation on sensor data to ensure the accuracy of the data; adopting an information code generation and verification mechanism to ensure data integrity and security; combining a physical model with a deep learning model to predict the sewage treatment efficiency, and analyzing the spatio-temporal correlation through a spatio-temporal convolutional network; optimizing the system operation parameters through a deep reinforcement learning algorithm, adjusting the vibrating membrane frequency, amplitude, aeration power and chemical dosage to achieve the best treatment effect; and improving the sewage treatment efficiency through the real-time adjustment of the vibrating membrane and aeration equipment; displaying the real-time treatment status, monitoring the sewage removal rate, and triggering an alarm mechanism in case of abnormalities. The present invention takes into account data security and system optimization while ensuring the treatment effect.

Description

Technical Field

[0001] The present invention relates to the technical field of sewage treatment and its automatic control, and particularly to an automatic monitoring system and method for sewage treatment by a vibrating membrane bioreactor. Background Art

[0002] With the continuous acceleration of the industrialization and urbanization processes, the sewage discharge volume has been increasing year by year globally. Substances such as organic pollutants and nitrogen and phosphorus compounds in sewage pose a serious threat to the water ecosystem. Traditional sewage treatment technologies have certain limitations in terms of treatment efficiency, energy consumption, and operating costs, and it is difficult to meet the increasingly stringent environmental protection requirements. As a new sewage treatment technology, the vibrating membrane bioreactor has the advantages of efficient separation, biodegradation, and membrane cleaning integration. However, during its operation, it is affected by factors such as pollutant concentration fluctuations and membrane fouling, resulting in difficulty in stably maintaining the treatment effect.

[0003] The Chinese invention patent with the publication number CN118131631B discloses a dynamic optimization scheduling system and method for sewage treatment. Among them, the dynamic optimization scheduling system includes a data acquisition module, a data processing module, a scheduling decision module, an optimization execution module, an optimization monitoring module, and a remote terminal. Various data in the sewage treatment process are collected in real time through the data acquisition module, the data processing module processes and analyzes the data, the scheduling decision module determines the best scheduling plan according to the processed data and in combination with a dynamic scheduling algorithm, and the optimization execution module controls the operation of the sewage treatment equipment according to the scheduling plan. The scheduling plan in the sewage treatment process can be adjusted and optimized in real time, and the scheduling analysis execution process and the sewage treatment performance status can be analyzed to ensure the subsequent sewage treatment efficiency and treatment effect while reducing the treatment cost, significantly reducing the management difficulty of management personnel, and having a high degree of intelligence and automation.

[0004] Most of the existing control methods rely on empirical rules and lack an intelligent dynamic optimization mechanism, making it difficult to achieve the operation goals of high efficiency and low consumption. To address the above problems, the present invention combines cutting-edge technologies such as sensor networks, big data analysis, artificial intelligence, and deep reinforcement learning, and proposes an automatic monitoring method for sewage treatment by a vibrating membrane bioreactor. By real-time monitoring of key index data, intelligent analysis of the operation status, and dynamic optimization of treatment parameters, it realizes the refined management of the sewage treatment process and the global performance optimization, providing technical support and theoretical basis for the practical application of the vibrating membrane bioreactor in complex water treatment scenarios. Summary of the Invention

[0005] The object of the present invention is to propose an automatic monitoring system and method for sewage treatment by a vibrating membrane bioreactor in view of the problems existing in the background art.

[0006] Technical solution of the present invention: An automatic monitoring method for sewage treatment in a vibrating membrane bioreactor, comprising the following specific implementation steps:

[0007] S1. Perform initialization settings, calibrate the sensors, set the initial operating parameters, set the initial values of the vibrating membrane amplitude and frequency, and load historical data;

[0008] S2. A sensor data denoising method based on wavelet transform and Kalman filter. First, use wavelet decomposition to remove high-frequency noise, then suppress low-frequency drift through Kalman filter to obtain denoised data. Subsequently, perform Z-score normalization on the denoised data, and use linear interpolation method to supplement missing data. Then, combine random numbers and hash functions to generate information codes;

[0009] S3. Verify the integrity and rationality of the data by parsing the information code, and perform intelligent analysis and optimization of the sewage treatment efficiency: a hybrid model combining physical model and deep learning. Based on reaction kinetics, establish a physical model. At the same time, use the Transformer model to process time-series data to generate hybrid prediction values, and optimize the model output through a spatio-temporal convolutional network. And use a deep reinforcement learning algorithm to dynamically adjust the operating parameters, including the vibrating membrane frequency, amplitude, aeration power, and chemical dosage;

[0010] S4. Adjust the amplitude and frequency in real time according to the optimized vibrating membrane frequency and the optimized amplitude, and adjust the operating parameters of the aeration equipment and the chemical dosing equipment according to the optimized aeration power and the optimized chemical dosage;

[0011] S5. Display the real-time processing status data, sewage parameters, and vibrating membrane operation data, and calculate the sewage removal rate. If the sewage removal rate is less than the set threshold, it is considered that the vibrating membrane is seriously polluted, trigger the alarm mechanism, and provide the cause of the failure and the recommended solution.

[0012] Preferably, the implementation process of the sensor data denoising method based on wavelet transform and Kalman filter is as follows:

[0013] S21. Analyze the sensor noise, and set the sensor measurement value x measured (t)=x true (t)+n H (t)+n L (t);

[0014] Among them, x measured (t) represents the sensor data collected at time t; x true (t) represents the true value; n H (t) and n L (t) represent high-frequency noise and low-frequency noise respectively;

[0015] S22. Perform multi-level wavelet decomposition on the sensor measurement value x measured (t):

[0016] ;

[0017] Where D j (t) represents the wavelet detail component of the j-th layer, corresponding to the high-frequency part in the signal; A J (t) represents the wavelet approximation component of the J-th layer, corresponding to the low-frequency part in the signal; J represents the number of wavelet decomposition layers;

[0018] S23. Perform threshold processing on the detail component D j (t) to suppress high-frequency noise. The soft threshold method is used, and the formula is as follows:

[0019] ;

[0020] ;

[0021] Where represents the detail component after threshold processing; λ' represents the threshold, usually determined by the noise standard deviation σ n and the number of data points N: sign(D j (t)) is used to retain the sign of the original detail component D j (t); sign() represents the sign function;

[0022] S24. Reconstruct the processed detail component and the approximation component A J (t) to obtain the preliminary denoised signal ;

[0023] S25. Establish a state space model: ;

[0024] Where x k represents the state of the true signal at time k; z k represents the observed value, that is ; F represents the state transition matrix; H represents the observation matrix; w k represents the process noise; v k represents the observation noise;

[0025] S26. Kalman filtering is divided into two steps: prediction and update:

[0026] Prediction: ;

[0027] Update: ;

[0028] Where Indicates the predicted state; Indicates the updated state; P k|k-1 and P k|k respectively represent the predicted and updated covariance matrices; K k represents the Kalman gain; Q and R respectively represent the covariance matrices of the process noise and the observation noise;

[0029] S27. After wavelet transform and Kalman filtering, the denoised signal is obtained .

[0030] Preferably, the generation process of the information code is as follows:

[0031] S31. Select a random number k ∈ [1, q - 1], and calculate the first-order auxiliary information code parameter fi1 = (Cg × k mod p) mod q;

[0032] where q is a large prime number of 160 bits, p is a large prime number of 1024 bits, and q|p - 1 is satisfied; Cg is the auxiliary information code, Cg = [r(p - 1) / q] mod p, and Cg > 1 is satisfied; r is a number selected in the group ; q|p - 1 means that q is a factor of p - 1, that is, p - 1 is divisible by q;

[0033] S32. Calculate the second-order auxiliary information code parameter fi2 = {(k - 1) × [H(BX t ) + Pg × fi1]} mod q;

[0034] where H() represents the hash function; BX t represents the binary string of the processed key index data X t , BX t =COD real ||BOD real ; COD real represents the COD data in the key index data X t ; BOD real represents the BOD data of the key index data X t ; || represents the concatenation operation; Pg represents a randomly selected information code generation parameter, and accordingly calculate the information code parsing parameter Pa = Cg × Pg mod p, and make {p, q, Cg, Pa} public, and Pg is an integer in [1, q - 1];

[0035] S33. If fi2 = 0 and fi2 - 1 mod q does not exist, then return to S31;

[0036] S34. Generate the information code IC = {fi1, fi2}.

[0037] Preferably, the verification process for verifying data integrity and rationality by parsing the information code is as follows:

[0038] S41. Extract parameters {prime number p, prime number q, auxiliary information code Cg, information code parsing parameter Pa};

[0039] S42. Determine whether fi1 and fi2 are integers between [1, q - 1]. If so, calculate the parsing code Ca = fi2 - 1 mod q; if not, alert the user;

[0040] S43. Calculate the following auxiliary parsing parameters:

[0041] Pp1 = [H(BX t ) × Ca] mod q;

[0042] Pp2 = (fi1 × Ca) mod q;

[0043] Wherein, H() represents a hash function; BX t represents the binary string of the key index data X t for verification, BX t = COD real || BOD real ; COD real represents the COD data in the key index data X t for verification; BOD real represents the BOD data in the key index data X t for verification; || represents a concatenation operation;

[0044] S44. Calculate the check factor Fa = (Cg × Pp1 × Pa × Pp2 mod p) mod q;

[0045] S45. If Fa = fi1, the verification passes; if not, alert the user.

[0046] Preferably, the analysis and optimization process for intelligent analysis and optimization of sewage treatment efficiency is as follows:

[0047] S51. Based on the key index data X t , construct an adaptive hybrid model AHM, integrating a physical model and a deep learning model, predict the key index data of sewage treatment, and output time series data y t ;

[0048] S52. Introduce a spatio-temporal correlation analysis model, combine the time series data y t output by the hybrid model with the spatial distribution information of the sensors, and output the spatio-temporal analysis result;

[0049] S53. Optimize the operating parameters of sewage treatment using a deep reinforcement learning algorithm based on the spatio-temporal analysis results to achieve the optimal global goal.

[0050] Preferably, the output process of the time series data y t is as follows:

[0051] S61. Establish a physical model based on the process mechanism according to the reaction kinetics formula:

[0052] ;

[0053] In the formula, y phy,t represents the physical prediction value, that is, the predicted concentration value of the target pollutant at a certain time point t; k represents the reaction rate constant; C reactant,t represents the reactant concentration; n represents the reaction order;

[0054] S62. Use the deep learning model Transformer to process the time series data:

[0055] ;

[0056] In the formula, f NN () represents the neural network prediction function; X t represents the input data vector, that is, the sensor time series; θ represents the neural network parameters;

[0057] S63. Combine the physical model and the neural network model, and generate a mixed prediction value y t :

[0058] ;

[0059] In the formula, α and β represent the weight coefficients, which are dynamically adjusted through Bayesian optimization.

[0060] Preferably, the process of combining the time series data y t output by the hybrid model with the spatial distribution information of the sensors is as follows:

[0061] S71. Define the spatial correlation intensity matrix A between sensors, based on the actual sewage flow path and sensor layout:

[0062] ;

[0063] In the formula, A ij represents the correlation intensity between sensors i and j; d ij represents the physical distance between sensors; σ' represents the adjustment parameter;

[0064] S72. Combine the time series and spatial distribution characteristics to construct a spatio-temporal convolutional network fST , namely, a hybrid model including temporal attention and spatial convolution:

[0065] ;

[0066] In the formula, represents the spatio-temporal analysis result; X t represents the processed key index data.

[0067] Preferably, the optimization process of optimizing the operation parameters of sewage treatment by using the deep reinforcement learning algorithm is as follows:

[0068] S81. Define the state, action, and reward functions:

[0069] State s t : The current sewage treatment state, key index data, including: chemical oxygen demand, biological oxygen demand, COD, and BOD;

[0070] Action a t : The adjusted operation parameter vector, including: vibration membrane frequency, amplitude, aeration power, and chemical agent dosage;

[0071] Reward function R:

[0072] ;

[0073] In the formula, Eff(x) represents the treatment efficiency; Energy(x) represents the energy consumption; Cost(x) represents the operation cost;

[0074] S82. Use the deep deterministic policy gradient algorithm to generate the optimization policy:

[0075] ;

[0076] In the formula, π represents the optimization policy function; represents the probability of selecting action a t under state s t ; E[] represents the expected value, which is used to quantify the average value of the rewards that may be obtained after selecting an action under the current state by a certain policy; represents the expected value of the reward R obtained when performing action a t under the current state s t ; θ' represents the policy parameter, which is optimized by gradient update.

[0077] The technical solution of the present invention: An automatic monitoring system for sewage treatment in a vibrating membrane bioreactor, which is used to execute the above-mentioned automatic monitoring method for sewage treatment in a vibrating membrane bioreactor, includes:

[0078] The diaphragm control module is used to adjust the amplitude, frequency and operation duration of the diaphragm, and optimize the sewage filtration efficiency;

[0079] The multi-parameter sensor network is used to monitor the parameters in sewage, including chemical oxygen demand (COD) and biochemical oxygen demand (BOD);

[0080] The data processing and transmission module is used to collect and process sensor data, and transmit the data to the intelligent analysis and control module by using industrial Internet of Things technology;

[0081] The intelligent analysis and control module is used to dynamically adjust the working parameters of the diaphragm based on artificial intelligence algorithms;

[0082] The user interface and alarm module is used to provide a remote operation interface, and realize the visual monitoring of the processing status and the alarm of abnormal status.

[0083] Compared with the prior art, the above technical solutions of the present invention have the following beneficial technical effects:

[0084] The present invention designs an automatic monitoring system and method for sewage treatment by a membrane bioreactor. Through the combination of multi-layer data processing, dynamic optimization and intelligent control technologies, the efficiency, accuracy and economy of sewage treatment are significantly improved, and it has wide application value and remarkable technical advantages:

[0085] (1) Realize real-time monitoring and optimization of the whole process of sewage treatment: Through the distributed multi-parameter sensor network, key index data (such as COD, BOD) in the sewage treatment process are collected in real time, and combined with data processing methods such as data standardization, missing data cleaning and interpolation, to ensure the integrity and accuracy of key data, providing a reliable basis for system optimization;

[0086] (2) Ensure data integrity and reliability: The proposed information code generation and verification mechanism is based on multi-layer hash function and random number generation technology to encrypt and verify the processed data; This mechanism can effectively prevent data from being tampered with during transmission, ensuring the integrity and rationality of key data;

[0087] (3) Improve the efficiency and prediction accuracy of sewage treatment: The system constructs an adaptive hybrid model (AHM) combining physical models and deep learning models, dynamically adjusts the weights to generate prediction results, and further explores the global relevance of sewage treatment through a spatio-temporal analysis model; This method takes into account the accuracy of the process mechanism and the adaptive ability of deep learning, improving the accuracy of prediction and the global optimization effect;

[0088] (4) Realize the intelligent optimization of sewage treatment parameters: Introduce the deep reinforcement learning (DRL) algorithm. By defining the state, action, and reward functions, dynamically optimize the frequency, amplitude, aeration power, and chemical dosage of the vibrating membrane according to the comprehensive goals of treatment efficiency, energy consumption, and cost, ensuring the high efficiency and economy of the system operation;

[0089] (5) Provide flexible and reliable closed-loop control: Realize the closed-loop control of the sewage treatment process by adjusting the vibrating membrane control module, aeration equipment, and chemical dosing equipment in real time with the optimized parameters, which can quickly respond to the changes in different sewage concentrations and operating conditions, ensuring the stability and high efficiency of the treatment process;

[0090] (6) Enhance the visualization and alarm capabilities of the system: Real-time display the treatment status, key indicators, and operating parameters, and calculate the sewage removal rate; When the treatment efficiency is lower than the set threshold, trigger an alarm in a timely manner and provide cause analysis and solution suggestions, improving the safety and user experience of the system;

[0091] (7) Energy conservation, consumption reduction, and environmental protection benefits: Realize the refined control of sewage treatment energy consumption through the dynamic adjustment of parameters such as the frequency, amplitude, and aeration power of the vibrating membrane, and minimize the operating cost to the greatest extent on the premise of ensuring the treatment effect, with significant energy-saving and environmental protection effects. Brief Description of the Drawings

[0092] Figure 1 It is the system architecture diagram of an automatic monitoring system for sewage treatment by a vibrating membrane bioreactor proposed by the present invention;

[0093] Figure 2 It is the method flow chart of an automatic monitoring method for sewage treatment by a vibrating membrane bioreactor proposed by the present invention. Detailed Embodiments

[0094] Example 1, as Figure 1 shown, the automatic monitoring system for sewage treatment by a vibrating membrane bioreactor proposed by the present invention includes: a vibrating membrane control module, a multi-parameter sensor network, a data processing and transmission module, an intelligent analysis and control module, and a user interface and alarm module.

[0095] The vibrating membrane control module is used to adjust the amplitude, frequency, and operation duration of the vibrating membrane to optimize the sewage filtration efficiency;

[0096] The multi-parameter sensor network monitors the parameters in the sewage, including but not limited to COD (chemical oxygen demand), BOD (biochemical oxygen demand);

[0097] The data processing and transmission module collects and processes the sensor data, and uses the industrial Internet of Things (IIoT) technology to transmit the data to the intelligent analysis and control module;

[0098] The intelligent analysis and control module dynamically adjusts the working parameters of the vibration membrane based on artificial intelligence algorithms;

[0099] The user interface and alarm module provides a remote operation interface to achieve visual monitoring of the processing status and alarm for abnormal status.

[0100] Embodiment 2, as Figure 2 shown, the automatic monitoring method for sewage treatment of the vibration membrane bioreactor proposed by the present invention is applied to the automatic monitoring system for sewage treatment of the vibration membrane bioreactor proposed in Embodiment 1, and its specific implementation steps are as follows:

[0101] S1. System initialization, start the multi-parameter sensor network and the vibration membrane control module, calibrate all sensors, including but not limited to COD and BOD, set the initial operating parameters, including but not limited to the amplitude A 初始 and frequency f 初始 of the vibration membrane, and load historical data.

[0102] S2. Real-time monitor the key index data in the sewage treatment process through the distributed multi-parameter sensor network, including but not limited to: chemical oxygen demand COD real , biological oxygen demand BOD real , and perform the following processing through the data processing and transmission module:

[0103] S21. Based on the joint denoising method of wavelet transform and Kalman filter, remove high-frequency noise through wavelet transform and suppress low-frequency drift through Kalman filter, realizing comprehensive noise filtering, ensuring the accuracy and reliability of sensor data, and providing a solid data foundation for the efficient operation of the system. The specific implementation process is as follows:

[0104] S2101. Analyze the sensor noise, and set the sensor measurement value x measured (t) = x true (t) + n H (t) + n L (t);

[0105] Among them, x measured (t) represents the sensor data collected at time t; x true (t) represents the true value; n H (t) and n L (t) respectively represent high-frequency noise (introduced by including but not limited to electrical interference and environmental vibration, with a higher frequency) and low-frequency noise (caused by sensor aging or environmental changes, with a lower frequency);

[0106] S2102. Perform multi-layer wavelet decomposition on the sensor measurement value x measured (t):

[0107] ;

[0108] In the formula, D j (t) represents the wavelet detail component of the j-th layer, corresponding to the high-frequency part in the signal; A J (t) represents the wavelet approximation component of the J-th layer, corresponding to the low-frequency part in the signal; J represents the number of wavelet decomposition layers, and an appropriate number of layers is selected according to the frequency characteristics of the signal;

[0109] Perform threshold processing on the detail component D j (t) to suppress high-frequency noise. The soft threshold method is adopted, and the formula is as follows:

[0110] ;

[0111] ;

[0112] In the formula, represents the detail component after threshold processing; λ' represents the threshold, usually determined by the noise standard deviation σ n and the number of data points N: sign(D j (t)) is used to retain the sign (positive or negative) of the original detail component D j (t), so as to ensure the directionality of the data after noise reduction (that is, positive numbers remain positive and negative numbers remain negative); sign() represents the sign function;

[0113] Accordingly, the processed detail component and the approximation component A J (t) are reconstructed to obtain the preliminary denoised signal ;

[0114] S2103. Establish a state space model: ;

[0115] Among them, x k represents the state of the true signal at time k; z k represents the observation value, that is, ; F represents the state transition matrix; H represents the observation matrix; w k represents the process noise; v k represents the observation noise;

[0116] Kalman filtering is divided into two steps: prediction and update:

[0117] Prediction: ;

[0118] Update: ;

[0119] In the formula, represents the predicted state; Indicates the updated state; P k|k-1 and P k|k respectively represent the predicted and updated covariance matrices; K k represents the Kalman gain, which is used to balance the weights of prediction and observation; Q and R respectively represent the covariance matrices of process noise and observation noise;

[0120] Accordingly: After wavelet transform and Kalman filtering, the denoised signal is obtained ;

[0121] S22. Since the data collected by different sensors have different units and the dimensionality differences may affect the subsequent analysis results, data standardization processing is required. After standardization, the data have the same scale, enabling comparison and fusion between different data. The standardization formula uses the Z-score standardization method, that is:

[0122] ;

[0123] In the formula, x represents the denoised signal ; μ represents the historical mean corresponding to the real-time data x collected by the sensor, that is, the average value of the real-time data x collected by the sensor within a certain time range; σ represents the historical standard deviation corresponding to the real-time data x collected by the sensor; x 标准化 represents the standardized sensor data;

[0124] S23. To ensure the integrity of the data, the missing data are cleaned and interpolated, and the linear interpolation method is used for supplementation;

[0125] Assuming that the data change is linear, interpolation is performed through adjacent valid data points. The formula of the linear interpolation method is:

[0126] ;

[0127] In the formula, x t represents the interpolated value at time t for supplementation; x t-1 and x t+1 respectively represent the standardized data at times t - 1 and t + 1;

[0128] S24. To ensure that the intelligent analysis and control module can promptly detect the missing or unreasonable key index data, an information code IC is generated for the processed key index data, and its generation process is as follows:

[0129] S2401. Select a random number k ∈ [1, q - 1], and calculate the first-order auxiliary information code parameter fi1 = (Cg × k mod p) mod q;

[0130] Among them, q is a large prime number of 160 bits, p is a large prime number of 1024 bits, and q|p - 1 is satisfied; Cg is an auxiliary information code, Cg = [r(p - 1) / q] mod p; r is a number selected in the group and Cg > 1 is satisfied; accordingly, the information code generation parameter Pg is randomly selected, the information code parsing parameter Pa = Cg × Pg mod p is calculated, and {p, q, Cg, Pa} is disclosed to the intelligent analysis and control module. Pg is an integer in [1, q - 1]; q|p - 1 means that q is a factor of p - 1, that is, p - 1 is divisible by q;

[0131] S2402. Calculate the second-order auxiliary information code parameter fi2 = {(k - 1) × [H(BX t ) + Pg × fi1]} mod q;

[0132] Among them, H() represents a hash function; BX t represents the binary string of the processed key index data X t , BX t = COD real ||BOD real ; COD real represents the COD data in the key index data X t ; BOD real represents the BOD data of the key index data X t ; || represents a concatenation operation;

[0133] S2403. If fi2 = 0 and fi2 - 1 mod q does not exist, then return to S2401;

[0134] S2404. Generate the information code IC = {fi1, fi2};

[0135] S25. Transmit {IC, X t} to the intelligent analysis and control module.

[0136] S3. The intelligent analysis and control module performs intelligent analysis and optimization of the sewage treatment efficiency through a data-driven dynamic analysis and optimization strategy, combined with the actual operation data of the vibrating membrane bioreactor. The specific implementation process is as follows:

[0137] S31. Parse the information code IC = {fi1, fi2}, and verify the integrity and rationality of the processed data X t . The verification process is as follows:

[0138] S3101. Extract the parameters {prime number p, prime number q, auxiliary information code Cg, information code parsing parameter Pa};

[0139] S3102. Determine whether fi1 and fi2 are integers between [1, q - 1]. If so, calculate the parsing code Ca = fi2 - 1 mod q; if not, alert the user interface and the alarm module.

[0140] S3103. Calculate the following auxiliary parsing parameters:

[0141] Pp1 = [H(BX t ) × Ca] mod q;

[0142] Pp2 = (fi1 × Ca) mod q;

[0143] where H() represents the hash function; BX t represents the binary string of the received key index data X t , BX t = COD real || BOD real ; COD real represents the COD data in the received key index data X t ; BOD real represents the BOD data in the received key index data X t ; || represents the concatenation operation;

[0144] S3104. Calculate the verification factor Fa = (Cg × Pp1 × Pa × Pp2 mod p) mod q;

[0145] S3105. If Fa = fi1, the verification passes, and execute step S32; if not, alert the user interface and the alarm module.

[0146] S32. Based on the received key index data X t , construct an adaptive hybrid model AHM, and comprehensively utilize the advantages of physical models and deep learning models to predict the key index data of sewage treatment (including but not limited to COD, BOD). The specific implementation process is as follows:

[0147] S3201. Establish a physical model based on the process mechanism according to the reaction kinetics formula:

[0148] ;

[0149] In the formula, y phy,t represents the physical prediction value, that is, the concentration prediction value of the target pollutant (including but not limited to COD, BOD) at a certain time point t; k represents the reaction rate constant; C reactant,t represents the reactant concentration; n represents the reaction order;

[0150] S3202. Use the deep learning model Transformer to process time series data:

[0151] ;

[0152] where f NN () represents the neural network prediction function; X t represents the input data vector, i.e., the sensor time series; θ represents the neural network parameters;

[0153] S3203. Combine the physical model and the neural network model, and generate a hybrid prediction value y by dynamically adjusting the weights t :

[0154] ;

[0155] where α and β represent the weight coefficients, which are dynamically adjusted by Bayesian optimization;

[0156] S33. To improve the global accuracy of the prediction, introduce a spatio-temporal correlation analysis model, combine the time series data y output by the hybrid model t with the spatial distribution information of the sensors, and further explore the global correlation in the sewage treatment process. The specific implementation process is as follows:

[0157] S3301. Define the spatial correlation intensity matrix A between sensors, based on the actual sewage flow path and sensor layout:

[0158] ;

[0159] where A ij represents the correlation intensity between sensors i and j; d ij represents the physical distance between sensors; σ' represents the adjustment parameter;

[0160] S3302. Combine the time series and spatial distribution characteristics to construct a spatio-temporal convolutional network f ST , that is, a hybrid model including time attention and spatial convolution:

[0161] ;

[0162] where represents the spatio-temporal analysis result; X t represents the processed data;

[0163] S34. Based on the spatio-temporal analysis result, use the deep reinforcement learning (DRL) algorithm to optimize the operating parameters of sewage treatment and achieve the global objective optimization. The specific implementation process is as follows:

[0164] S3401. Define the state, action, and reward functions:

[0165] State s t : The current sewage treatment state, key index data (including but not limited to COD, BOD);

[0166] Action a t : The adjusted operation parameter vector (including but not limited to the vibration membrane frequency or amplitude);

[0167] Reward function R:

[0168] ;

[0169] In the formula, Eff(x) represents the treatment efficiency; Energy(x) represents the energy consumption; Cost(x) represents the operation cost;

[0170] S3402. Use the Deep Deterministic Policy Gradient (DDPG) algorithm to generate an optimized policy:

[0171] ;

[0172] In the formula, π represents the optimized policy function; represents the probability of selecting action a t under state s t ; E[] represents the expected value, which is used to quantify the average reward that a certain policy may obtain after selecting an action in the current state; represents the expected value of the reward R obtained when performing action a t under the current state s t ; θ' represents the policy parameter, which is updated and optimized through the gradient;

[0173] S35. According to the optimized policy output by reinforcement learning, dynamically adjust the system operation parameters and continuously optimize through closed-loop control. The specific implementation process is as follows:

[0174] S3501. Dynamically set the frequency f vib and amplitude A vib of the vibration membrane to improve the treatment efficiency:

[0175] ;

[0176] ;

[0177] In the formula, f vib,opt represents the optimized vibration membrane frequency; f base represents the default vibration membrane frequency; Δf represents the frequency adjustment value, which is generated by the optimized action output by reinforcement learning; A vib,opt represents the optimized amplitude; A baseA represents the default diaphragm amplitude; ΔA represents the amplitude adjustment value, which is generated by the optimized action output by reinforcement learning;

[0178] S3502. Adjust the aeration power P in real time air to balance energy consumption and dissolved oxygen concentration:

[0179] ;

[0180] In the formula, P vib,opt represents the optimized aeration power; P base represents the default aeration power; ΔP represents the aeration power adjustment value;

[0181] S3503. Dynamically adjust the optimized M according to the sewage concentration chem :

[0182] ;

[0183] In the formula, M chem,opt represents the optimized chemical agent dosage, that is, the amount of chemical agent to be added under the current sewage concentration to ensure that the chemical reaction of sewage treatment reaches the optimal effect; k chem represents the chemical reaction coefficient, that is, the amount of chemical agent required for each unit concentration of reactants; C reactant represents the concentration of reactants (including but not limited to COD, BOD) in the sewage;

[0184] S36. Transmit the optimized diaphragm frequency f vib,opt and the optimized amplitude A vib,opt and the optimized aeration power P vib,opt as well as the optimized chemical agent dosage M chem to the diaphragm control module and the user interface and alarm module.

[0185] S4. The diaphragm control module adjusts the amplitude and frequency in real time according to the optimized diaphragm frequency f vib,opt and the optimized amplitude A vib,opt , and adjusts the operating parameters of the aeration equipment and the chemical agent dosing equipment according to the optimized aeration power P vib,opt and the optimized chemical agent dosage M chem to ensure the coordination and efficiency of the sewage treatment process.

[0186] S5. The user interface and alarm module display the real-time processing status data, sewage parameters and diaphragm operation data, and calculate the sewage removal rate Rr:

[0187] ;

[0188] In the formula, COD realRepresents the COD data in the real-time processing status data; BOD real Represents the BOD data in the real-time processing status data; COD initial Represents the initial COD value, i.e., the COD concentration when the sewage enters the reactor; BOD initial Represents the initial BOD value, i.e., the BOD concentration when the sewage enters the reactor;

[0189] If |Rr| is less than the set threshold, it is considered that the vibration membrane is seriously polluted, and the alarm mechanism is triggered. The alarm is notified through multiple channels, including but not limited to text messages, APP push, and audible and visual signals, and the cause of the failure and a recommended solution are provided.

[0190] The embodiments of the present invention have been described in detail above in conjunction with the accompanying drawings. However, the present invention is not limited thereto. Various changes can be made within the scope of knowledge of those skilled in the art without departing from the gist of the present invention.

Claims

1. An automatic monitoring method for sewage treatment in a vibrating membrane bioreactor, characterized in that, It includes the following specific implementation steps: S1. Perform initialization settings, calibrate the sensor, set initial operating parameters, set the initial values of the vibration membrane amplitude and frequency, and load historical data; S2. For the sensor data denoising method based on wavelet transform and Kalman filter, first use wavelet decomposition to remove high-frequency noise, then suppress low-frequency drift through Kalman filter to obtain denoised data. Subsequently, perform Z-score normalization on the denoised data, and use linear interpolation method to supplement missing data. After that, combine random numbers and hash functions to generate information codes; S3. Verify the integrity and rationality of the data by parsing the information code, and perform intelligent analysis and optimization of the sewage treatment efficiency: Combine a hybrid model of physical model and deep learning. Establish a physical model based on reaction kinetics, and at the same time use the Transformer model to process time-series data to generate hybrid prediction values, and optimize the model output through a spatio-temporal convolutional network. And use a deep reinforcement learning algorithm to dynamically adjust operating parameters, including the vibration membrane frequency, amplitude, aeration power, and chemical dosage; The output process of the hybrid prediction value is as follows: A1. Establish a physical model based on the process mechanism according to the reaction kinetics formula: ； where y phy,t represents the physical prediction value, i.e., the predicted concentration of the target pollutant at a certain time point t; k represents the reaction rate constant; C reactant,t represents the reactant concentration; n represents the reaction order; A2. Use the deep learning model Transformer to process time-series data: ； where f NN () represents the neural network prediction function; X t represents the input data vector, i.e., the sensor time series; θ represents the neural network parameters; A3. Combine the physical model and the neural network model to generate a hybrid prediction value y by dynamically adjusting the weights t : ； In the formula, α and β represent weight coefficients, which are dynamically adjusted through Bayesian optimization; S4. Adjust the amplitude and frequency in real time according to the optimized vibration membrane frequency and the optimized amplitude, and adjust the operating parameters of the aeration equipment and chemical dosing equipment according to the optimized aeration power and the optimized chemical dosage; S5. Display real-time processing status data, sewage parameters, and vibration membrane operation data, and calculate the sewage removal rate. If the sewage removal rate is less than the set threshold, it is considered that the vibration membrane is seriously polluted, trigger the alarm mechanism, and provide the cause of the failure and a recommended solution.

2. The automatic monitoring method for sewage treatment by a vibrating membrane bioreactor according to claim 1, wherein, The implementation process of the sensor data denoising method based on wavelet transform and Kalman filter is as follows: S21. Analyze the sensor noise and set the sensor measurement value x measured (t) = x true (t) + n H (t) + n L (t); where x measured (t) represents the sensor data collected at time t; x true (t) represents the true value; n H (t) and n L (t) represent high-frequency noise and low-frequency noise respectively; S22. Perform multi-level wavelet decomposition on the sensor measurement value x measured (t): ； where D j (t) represents the wavelet detail component of the j-th layer, corresponding to the high-frequency part in the signal; A J (t) represents the wavelet approximation component of the J-th layer, corresponding to the low-frequency part in the signal; J represents the number of wavelet decomposition layers; S23. Threshold the detail component D j (t) of each layer to suppress high-frequency noise. The soft thresholding method is used, and the formula is as follows: ； ； In the formula, represents the detail component after threshold processing; λ' represents the threshold, which is determined by the noise standard deviation σ n and the number of data points N: sign(D j (t)) is used to retain the sign of the original detail component D j (t); sign() represents the sign function; S24. Accordingly, reconstruct the processed detail component and the approximation component A J (t) to obtain a preliminary denoised signal ; S25. Establish a state space model: ; where, x k represents the state of the true signal at time k; z k represents the observed value, i.e., ; F represents the state transition matrix; H represents the observation matrix; w k represents the process noise; v k represents the observation noise; S26. Kalman filter is divided into two steps: prediction and update: Prediction: ; Update: ; wherein, represents the predicted state; represents the updated state; P k|k-1 and P k|k respectively represent the covariance matrices of prediction and update; K k represents the Kalman gain; Q and R respectively represent the covariance matrices of process noise and observation noise; S27. After wavelet transform and Kalman filtering, the denoised signal is obtained .

3. The automatic monitoring method for sewage treatment by a vibrating membrane bioreactor according to claim 1, characterized in that, The generation process of the information code is as follows: S31. Select a random number k∈[1, q - 1], and calculate the first-order auxiliary information code parameter fi1=(Cg×k mod p) mod q; Among them, q is a large prime number of 160 bits, p is a large prime number of 1024 bits, and q|p - 1 is satisfied; Cg is an auxiliary information code, Cg = [r(p - 1) / q] mod p, and Cg > 1 is satisfied; r is a number selected in the group ; q|p - 1 means that q is a factor of p - 1, that is, p - 1 can be divided evenly by q. S32. Calculate the second-order auxiliary information code parameter fi2 = {(k - 1) × [H(BX t ) + Pg × fi1]} mod q; Among them, H() represents the hash function; BX t represents the binary string of the processed key metric data X t , BX t =COD real ||BOD real ; COD real represents the COD data in the key metric data X t ; BOD real represents the BOD data of the key metric data X t ; || represents the concatenation operation; Pg represents a randomly selected information code generation parameter, based on which the information code parsing parameter Pa = Cg × Pg mod p is calculated, and {p, q, Cg, Pa} is made public, where Pg is an integer in [1, q - 1]; S33. If fi2 = 0, then fi2 - 1 mod q does not exist, then return to S31; S34. Generate the information code IC={fi1, fi2}.

4. The automatic monitoring method for sewage treatment by a vibrating membrane bioreactor according to claim 3, characterized in that, The verification process of verifying the integrity and rationality of the data by parsing the information code is as follows: S41. Extract the parameters {prime number p, prime number q, auxiliary information code Cg, information code parsing parameter Pa}; S42. Judge whether fi1 and fi2 are integers between [1, q - 1]. If so, calculate the parsing code Ca = fi2 - 1 mod q; if not, alert the user; S43. Calculate the following auxiliary parsing parameters: Pp1 = [H(BX t ) × Ca] mod q; Pp2=(fi1×Ca) mod q; Among them, H() represents a hash function; BX t represents the binary string of the key index data X for verification t of, BX t = COD real || BOD real ; COD real represents the COD data in the key index data X for verification t ; BOD real represents the BOD data of the key index data X for verification t ; || represents a concatenation operation; S44. Calculate the check factor Fa=(Cg×Pp1×Pa×Pp2 mod p) mod q; S45. If Fa = fi1, the verification passes; if not, alert the user.

5. The automatic monitoring method for sewage treatment by a vibrating membrane bioreactor according to claim 1, characterized in that, The analysis and optimization process for intelligent analysis and optimization of sewage treatment efficiency is as follows: S51. Based on the key indicator data X t , construct an adaptive hybrid model AHM, integrating a physical model and a deep learning model, to predict the key indicator data of sewage treatment and output time series data y t ; S52. Introduce a spatio-temporal correlation analysis model, and combine the time series data y output by the hybrid model with the spatial distribution information of the sensors to output spatio-temporal analysis results; t ​ S53. Based on the spatio-temporal analysis results, use the deep reinforcement learning algorithm to optimize the operating parameters of sewage treatment to achieve the optimal global goal.

6. The automatic monitoring method for sewage treatment by a vibrating membrane bioreactor according to claim 5, characterized in that, The time series data y output by the hybrid model t The combination process of combining with the spatial distribution information of the sensors is as follows: S71. Define the spatial correlation intensity matrix A between sensors, based on the actual sewage flow path and sensor layout: ； where A ij represents the association strength between sensors i and j; d ij represents the physical distance between sensors; σ' represents the adjustment parameter; S72. Construct a spatio-temporal convolutional network f by combining the characteristics of time series and spatial distribution ST , that is, a hybrid model including time attention and spatial convolution: ； In the formula, represents the spatio-temporal analysis result; X t represents the processed key index data.

7. The automatic monitoring method for sewage treatment by a vibrating membrane bioreactor according to claim 5, characterized in that, The optimization process for using the deep reinforcement learning algorithm to optimize the operating parameters of sewage treatment is as follows: S81. Define the state, action, and reward functions: State s t : The current sewage treatment state, key indicator data, including: Chemical Oxygen Demand and Biochemical Oxygen Demand (COD and BOD); Action a t : The adjusted operating parameter vector includes: diaphragm vibration frequency, amplitude, aeration power, and chemical dosage; Reward function R: ； In the formula, Eff(x) represents the treatment efficiency; Energy(x) represents the energy consumption; Cost(x) represents the operating cost; S82. Use the deep deterministic policy gradient algorithm to generate the optimization strategy: ； Where, π represents the optimized policy function; represents the probability of selecting action a t in state s t ; E[] represents the expected value, which is used to quantify the average value of the rewards that may be obtained after selecting an action under a certain policy in the current state; represents the expected value of the reward R obtained when t performing action a t in the current state s; θ' represents the policy parameter, which is optimized by gradient update.

8. An automatic monitoring system for sewage treatment by a vibrating membrane bioreactor, which is used to execute the automatic monitoring method for sewage treatment by a vibrating membrane bioreactor according to any one of claims 1 to 7, characterized in that, Including: The vibration membrane control module is used to adjust the amplitude, frequency, and operating duration of the vibration membrane to optimize the sewage filtration efficiency; The multi-parameter sensor network is used to monitor the parameters in the sewage, including chemical oxygen demand (COD) and biochemical oxygen demand (BOD); The data processing and transmission module is used to collect and process sensor data and transmit the data to the intelligent analysis and control module using industrial Internet of Things technology; The intelligent analysis and control module is used to dynamically adjust the working parameters of the vibration membrane based on artificial intelligence algorithms; The user interface and alarm module is used to provide a remote operation interface to achieve visual monitoring of the treatment status and alarm for abnormal status.

Images