An artificial intelligence control system for sewage treatment

Abstract

The present invention discloses an artificial intelligence control system for sewage treatment, wherein the intelligent perception layer is used to collect the inlet water quality parameters, core section water quality parameters and outlet water quality parameters of the sewage treatment system, and establish a water quality intelligent prediction model based on the perception data, and output the predicted core section water quality parameters and outlet water quality parameters through the intelligent prediction model; the intelligent decision-making layer is used to establish a sewage treatment model that can reflect the physical, chemical and biological reaction mechanisms in the sewage treatment process, and integrate expert knowledge and experience into the sewage treatment model; the intelligent execution layer is used to obtain the results of multi-objective optimization and decision analysis of the intelligent decision-making layer, and generate operation instructions by applying the model prediction control algorithm; the operation instructions are sent to the execution device to execute the operation. The artificial intelligence control system for sewage treatment of the present invention has the advantages of being able to improve the production and operation efficiency of sewage treatment plants, reduce environmental risks, reduce resource waste, and improve the level of digitization and intelligence.

Description

Technical Field

[0001] The present invention relates to the intersection of sewage treatment technology and artificial intelligence, and in particular to an artificial intelligence control system for sewage treatment. Background Art

[0002] Wastewater treatment plants are critical infrastructure. The function of the infrastructure is to treat and process wastewater from cities and industrial areas to protect the environment and public health. However, traditional wastewater treatment plant operations face a series of challenges and problems that limit their efficiency and sustainability. These problems include untimely data collection, lack of scientific basis for decision-making, and imprecise execution control. These problems have led to inefficient operation of many wastewater treatment plants, inadequate resource utilization, and may even have adverse effects on the environment.

[0003] The development of artificial intelligence (AI) technology has provided the possibility to improve the production and operation of sewage treatment plants and increase their operating efficiency. At present, artificial intelligence technology can be applied to various aspects of sewage treatment plants, including data collection, water quality analysis and prediction, decision support and automated control. By introducing artificial intelligence technology into the production and operation of sewage treatment plants, real-time data collection and analysis can be achieved, the accuracy and timeliness of water quality prediction and abnormal warning can be improved, a scientific basis can be provided for decision-making, and precise execution control can be achieved.

[0004] In the existing technology, artificial intelligence has achieved some important results and application cases in the field of sewage treatment. For example, the water quality monitoring system based on machine learning and deep learning can analyze and predict a large amount of water quality data, helping operators to detect water quality abnormalities in a timely manner and take corresponding measures. The intelligent decision support system can integrate various data and knowledge resources of sewage treatment plants, conduct decision analysis, optimize scheduling and risk assessment, and improve the efficiency and accuracy of production operations. Automated control algorithms and intelligent execution strategies can achieve precise water system control and regulation, optimize operating costs and save resources.

[0005] However, the existing intelligent water management system still has some limitations. First, due to the limited ability of data integration and analysis, data acquisition and analysis are difficult, resulting in low data utilization and value. The data collection and processing methods of traditional sewage plants are relatively backward, making it difficult to achieve real-time monitoring and analysis of comprehensive parameters, full sections, and full processes, resulting in inaccurate grasp of key information such as sewage quality and treatment process, and inability to detect abnormal situations and make predictions and warnings in a timely manner.

[0006] Secondly, the current decision-making model of the sewage treatment process lacks flexibility, and the decision-making and execution effects are difficult to optimize, and cannot adapt to the characteristics and needs of different sewage treatment plants. The decision-making process of traditional sewage treatment plants mostly relies on manual experience and simple models, and cannot fully consider the influence of multiple factors and complex dynamic changes, resulting in poor decision-making results. In addition, there is a lack of intelligent control and adjustment methods in the execution process, which cannot achieve precise control and optimized operation.

[0007] Thirdly, the current sewage treatment process has a low level of digitization and intelligence, and the accuracy of the control algorithm needs to be improved. The production and operation of traditional sewage treatment plants mostly rely on manual operation and experience judgment, and lack of systematic digital and intelligent support. This leads to problems such as low management efficiency, low decision-making accuracy, and high operating costs, and cannot meet complex production and operation requirements.

[0008] Therefore, based on a series of technical problems in the production and operation of traditional sewage treatment plants, such as low levels of digitization and intelligence, difficulties in data acquisition and analysis, and difficulty in optimizing decision-making and execution effects, the production and operation of sewage treatment plants requires a new intelligent digital management method to comprehensively solve the above problems in order to improve production and operation efficiency, optimize resource utilization and reduce environmental risks. Summary of the invention

[0009] The present invention is to avoid the shortcomings of the above-mentioned existing technologies and provide an artificial intelligence control system for sewage treatment to improve the efficiency of production and operation of sewage treatment plants, reduce environmental risks, reduce resource waste, and improve the digitization and intelligence level of sewage treatment plants.

[0010] The present invention adopts the following technical solutions to solve the technical problems.

[0011] An artificial intelligence control system for sewage treatment of the present invention is characterized in that it includes an intelligent perception layer, an intelligent decision-making layer and an intelligent execution layer;

[0012] The intelligent perception layer is used to collect the inlet water quality parameters Qj, the core section water quality parameters Qh and the outlet water quality parameters Qc of the sewage treatment system, and establish a water quality intelligent prediction model based on the perception data, and output the predicted core section water quality parameters Qh and outlet water quality parameters Qc through the intelligent prediction model;

[0013] The intelligent decision-making layer is used to establish a sewage treatment model that can reflect the physical, chemical and biological reaction mechanisms in the sewage treatment process, integrate expert knowledge and experience into the sewage treatment model, and then optimize the treatment process and improve efficiency. Through the integrated knowledge and experience, multi-objective optimization and decision analysis are carried out according to different operation goals;

[0014] The intelligent execution layer is used to obtain the results of multi-objective optimization and decision analysis of the intelligent decision layer, generate operation instructions by applying intelligent control; send the operation instructions to the execution device to execute the operation; during the execution process, monitor the execution results and provide real-time feedback on the execution results, compare and analyze the execution results with the expected goals, and make adjustments and optimizations based on the results of the comparison and analysis.

[0015] The technical features of the artificial intelligence control system for sewage treatment of the present invention are also:

[0016] Furthermore, the working process of the intelligent perception layer includes the following steps:

[0017] Step 11: Collect the sensory data of the sewage treatment system and pre-process the sensory data; the sensory data includes the inlet water quality parameter Qj, the core section water quality parameter Qh and the outlet water quality parameter Qc;

[0018] Step 12: Identify the data type of the perception data, extract the features of the perception data, and perform sensitivity analysis on the features of the extracted perception data;

[0019] Step 13: Apply machine learning algorithms to model and train water quality intelligent prediction models based on the perception data;

[0020] Step 14: Calculate the structural parameters of the water quality intelligent prediction model, and finally output the prediction results of the perception data according to the water quality intelligent prediction model.

[0021] Furthermore, in step 11, the process of preprocessing the perception data includes the following steps:

[0022] Step 111: sensing data cleaning step; data cleaning includes removing noise, processing missing values ​​and outliers;

[0023] Step 112: sensing data normalization step; data normalization includes data normalization or data standardization;

[0024] Step 113: feature engineering step of perception data;

[0025] Step 114: Perception data alignment step.

[0026] Furthermore, in step 12, the type of the perception data preprocessed in step 11 is determined, and a sample library is established; a sensitivity analysis is performed on the feature information of the type of the extracted perception data to obtain an importance score of the feature information, and based on the importance score of the feature information, features with an importance score below a threshold are eliminated.

[0027] Furthermore, in step 13, the machine learning algorithm is one or more of a decision tree algorithm, a support vector machine algorithm, a random forest algorithm or a neural network algorithm.

[0028] Furthermore, in step 14, the structure of the water quality intelligent prediction model includes the number of layers, the number of nodes, the learning rate and the loss function.

[0029] Furthermore, the working process of the intelligent decision-making layer includes the following steps:

[0030] Step 21: Establish a sewage treatment mechanism model;

[0031] Step 22: Integrate expert knowledge and experience into the sewage treatment mechanism model;

[0032] Step 23: Optimize the processing process and improve production operation efficiency.

[0033] Furthermore, in step 21, the sewage treatment mechanism model is used to understand and describe the physical, chemical and biological reaction mechanisms in the sewage treatment process.

[0034] Furthermore, in step 22, expert knowledge and expert experience are converted into computable rules and models, and integrated into the sewage treatment mechanism model.

[0035] Furthermore, in step 23, in the optimization process of constructing the sewage treatment mechanism model, a multi-objective optimization weighted sum method is used to set weights and objective functions for multi-objective optimization and decision analysis.

[0036] Z=∑(W i *F i (C1,C2,...,C n )),i∈{1,2,...,n} (11)

[0037] In formula (11), Z is the total optimization target value, which represents the weighted sum of the water quality prediction target parameters to be minimized; W i is the objective function F i (C1,C2,...,C n ) related weight, indicating the weight of the i-th water quality parameter; F i (C1,C2,...,C n ) is the i-th optimization objective function, which is the water quality parameters C1, C2, ..., C n Complex function of C i Represents the specified water quality parameter, i represents the index of the parameter, ranging from 1 to n; n represents the number of objectives of the optimization problem.

[0038] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0039] The present invention discloses an artificial intelligence control system for sewage treatment, wherein the intelligent perception layer is used to collect inlet water quality parameters, core section water quality parameters and outlet water quality parameters of the sewage treatment system, and establish a water quality intelligent prediction model based on the perception data, and output the predicted core section parameters and outlet water quality parameters through the intelligent prediction model; the intelligent decision-making layer is used to establish a sewage treatment model that can reflect the physical, chemical and biological reaction mechanisms in the sewage treatment process, and integrate expert knowledge and experience into the sewage treatment model to optimize the treatment process and improve efficiency; the intelligent execution layer is used to obtain the results of multi-objective optimization and decision analysis of the intelligent decision-making layer, generate operation instructions by applying the model prediction control algorithm; and send the operation instructions to the execution device to execute the operation.

[0040] The artificial intelligence control system for sewage treatment of the present invention provides an efficient management method by applying digitization and intelligence to the production and operation of sewage treatment plants. It integrates advanced data collection, prediction, decision-making and execution technologies, makes the production and operation of sewage treatment plants concrete, and leads them to develop in the direction of digitization and intelligence. The application of this system can improve the efficiency of production and operation of sewage treatment plants, reduce environmental risks, and reduce resource waste, making an important contribution to the realization of artificial intelligence production and operation goals of sewage treatment plants.

[0041] The artificial intelligence control system for sewage treatment of the present invention has the advantages of being able to improve the efficiency of production and operation of sewage treatment plants, reduce environmental risks, reduce resource waste, and improve the digitization and intelligence level of sewage treatment plants. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 The present invention is a framework diagram of an artificial intelligence control system for sewage treatment.

[0043] Figure 2 It is a structural schematic diagram of an intelligent prediction module model of an artificial intelligence control system for sewage treatment of the present invention.

[0044] The present invention will be further described below through specific implementation modes in conjunction with the accompanying drawings. DETAILED DESCRIPTION

[0045] See also Figure 1-2 , an artificial intelligence control system for sewage treatment of the present invention is characterized by comprising an intelligent perception layer, an intelligent decision-making layer and an intelligent execution layer;

[0046] The intelligent perception layer is used to collect the inlet water quality parameters Qj, the core section water quality parameters Qh and the outlet water quality parameters Qc of the sewage treatment system, and establish a water quality intelligent prediction model based on the perception data, and output the predicted core section water quality parameters Qh and outlet water quality parameters Qc through the intelligent prediction model;

[0047] Among them, Qj includes influent COD, influent NH3, influent TN, influent TP, influent flow rate, influent PH, etc., which can be represented by a one-dimensional vector Qj = [Qj1, Qj2, Qj3, Qj4, Qj5, Qj6], where Qj1, Qj2, Qj3, Qj4, Qj5 and Qj6 are used to represent influent COD, influent NH3, influent TN, influent TP, influent flow rate, influent PH and other parameters respectively.

[0048] Qh includes section COD, section NH3, section TN, section TP, section PAC flow, section PAM flow, section DO, section air volume, section MLSS, section O3, section SS, section secondary sodium flow, section carbon source flow, etc.; it can be expressed by a one-dimensional vector Qh=[Qh1, Qh2, Qh3, Qh4, Qh5, Qh6, Qh7, Qh8, Qh9, Qh10, Qh11, Qh12, Qh1 3]; among them, Qh1, Qh2, Qh3, Qh4, Qh5, Qh6, Qh7, Qh8, Qh9, Qh10, Qh11, Qh12, and Qh13 are respectively used to indicate parameters such as section COD, section NH3, section TN, section TP, section PAC flow, section PAM flow, section DO, section air volume, section MLSS, section O3, section SS, section secondary sodium flow, and section carbon source flow.

[0049] Qc includes effluent COD, effluent NH3, effluent TN, effluent TP, effluent DO, etc., which can be represented by a one-dimensional vector Qc=[Qc1, Qc2, Qc3, Qc4, Qc5]; among which Qc1, Qc2, Qc3, Qc4 and Qc5 are used to represent parameters such as effluent COD, effluent NH3, effluent TN, effluent TP, effluent DO, etc., respectively.

[0050] The intelligent decision-making layer is used to establish a sewage treatment model that can reflect the physical, chemical and biological reaction mechanisms in the sewage treatment process, integrate expert knowledge and experience into the sewage treatment model, and then optimize the treatment process and improve efficiency. Through the integrated knowledge and experience, multi-objective optimization and decision analysis are carried out according to different operation goals;

[0051] The intelligent execution layer is used to obtain the results of multi-objective optimization and decision analysis of the intelligent decision layer, generate operation instructions by applying intelligent control; send the operation instructions to the execution device to execute the operation; during the execution process, monitor the execution results and provide real-time feedback on the execution results, compare and analyze the execution results with the expected goals, and make adjustments and optimizations based on the results of the comparison and analysis.

[0052] The artificial intelligence control system for sewage treatment of the present invention takes the production and operation of sewage treatment plants as its goal, combines the development needs of artificial intelligence water control technology and the actual operation of sewage treatment plants, and proposes an efficient intelligent information management method.

[0053] First, the artificial intelligence control system for sewage treatment of the present invention realizes the real-time and comprehensiveness of data collection through the intelligent sensing module. Advanced sensors and monitoring equipment are used to monitor the key parameters and indicators of the sewage treatment plant, such as water quality, flow, pressure, etc. in real time. Through data collection and processing technology, the collected data is integrated, cleaned and transformed to form a high-quality data set, providing a reliable basis for subsequent analysis and decision-making.

[0054] Secondly, the system's intelligent prediction module uses machine learning and statistical analysis methods to model and predict the operation data of the sewage plant. By analyzing historical data and real-time data, an accurate water quality prediction model and anomaly detection algorithm are established. This can timely predict and identify possible abnormal situations, and provide corresponding early warning information to help operators take timely and effective measures to avoid potential environmental risks.

[0055] Thirdly, the intelligent decision-making module of the system combines mechanism models, expert knowledge and experience to provide scientific decision-making support for production operations. Based on the characteristics and needs of the sewage treatment plant, a flexible decision-making model has been established, which can conduct personalized decision analysis and optimize scheduling according to actual conditions. At the same time, the module also integrates expert knowledge and experience, provides professional advice and guidance, and helps operators make accurate and effective decisions.

[0056] Finally, the intelligent execution module of the system realizes precise control and regulation of the water system. The intelligent control and execution strategy is adopted to realize the digitization and intelligence of the production and operation process of the sewage treatment plant. This can accurately control and adjust the key parameters in the sewage treatment process, optimize the production and operation costs, save resources, and ensure the stable operation and efficient operation of the sewage treatment plant. By introducing this sewage treatment artificial intelligence control system, the production and operation of the sewage treatment plant has been substantially improved. The data collection and prediction functions of the system improve the quality and timeliness of the data, and provide a reliable basis for decision-making and execution. The application of the intelligent decision-making module enables sewage treatment plant managers to make accurate decisions based on scientific models and expert knowledge, and realize optimized operation scheduling and risk assessment. The introduction of the intelligent execution module makes the sewage treatment process digital and intelligent, realizes precise control and regulation, and improves the stability and operation efficiency of the system.

[0057] In specific implementation, the working process of the intelligent perception layer includes the following steps:

[0058] Step 11: Collect the sensory data of the sewage treatment system and pre-process the sensory data; the sensory data includes the inlet water quality parameter Qj, the core section water quality parameter Qh and the outlet water quality parameter Qc;

[0059] Step 12: Identify the data type of the perception data, extract the features of the perception data, and perform sensitivity analysis on the features of the extracted perception data;

[0060] Step 13: Apply machine learning algorithms to model and train water quality intelligent prediction models based on the perception data;

[0061] Step 14: Calculate the structural parameters of the water quality intelligent prediction model, and finally output the prediction results of the perception data according to the water quality intelligent prediction model.

[0062] In specific implementation, in step 11, the process of preprocessing the perception data includes the following steps:

[0063] In the present invention, based on the collection of perception data by using an automatic sampling detection system, relevant data processing software and manual preprocessing experience are used to preprocess the perception data.

[0064] Step 111: sensing data cleaning step; data cleaning includes removing noise, processing missing values ​​and outliers;

[0065] The integrity and accuracy of the perception data are ensured by removing noise from the perception data, handling missing values ​​and outliers.

[0066] Step 112: sensing data normalization step; data normalization includes data normalization or data standardization;

[0067] Data normalization refers to the normalized processing of data, including standardization, normalization, centering and logarithmization. In the present invention, the method of perceptual data normalization or perceptual data standardization is adopted to make different features have the same scale and weight, so as to avoid certain features from having too much influence on the prediction results. Normalization is to convert the characteristic values ​​of the sample to the same dimension, map the perceptual data to the interval [0, 1] or [-1, 1], and is only determined by the extreme values ​​of the variables. Standardization is to process the perceptual data according to the columns of the feature matrix, which is converted into a standard normal distribution by the method of z-score standardization, which is related to the overall sample distribution, and each sample point can have an impact on the standardization.

[0068] Step 113: feature engineering step of perception data;

[0069] The system of the present invention adopts a feature extraction method of feature engineering to extract key features from the perception data. Feature engineering is the process of extracting features from raw data using domain knowledge. It can help us reduce redundant information, select useful features, and thus improve model performance. Feature extraction uses the Pearson correlation analysis method to measure the linear correlation between the various features of the perception data, determine the strength and direction of the linear relationship between the feature and the target variable, and the formula for feature extraction is shown in the following formula (1).

[0070]

[0071] In formula (1), X i and Y i Respectively represent the values ​​of water quality parameter X and water quality parameter Y of the i-th data point in the sample; and Respectively represent the mean of water quality parameter X and water quality parameter Y; n represents the total number of sample data points.

[0072] The feature engineering step is used to extract key features from the perceived data and perform dimensionality reduction or select useful features to reduce redundant information. Feature engineering or feature extraction or feature discovery is the process of extracting features (characteristics, properties, attributes) from raw data using domain knowledge.

[0073] Step 114: Perception data alignment step.

[0074] In the present invention, time stamps and spatial information are used to align the perception data in time and space. In the process of perception data collection and storage, timestamps are used to mark the time of each data point. Then, by comparing timestamps, it is ensured that data from different data sources or collected at different times can be aligned in the time dimension; spatial coordinate information or identifiers are used to align data collected at different spatial locations.

[0075] The purpose of aligning data in time and space is to ensure the consistency and continuity of the data.

[0076] In specific implementation, in step 12, the type of the perception data preprocessed in step 11 is determined, and a sample library is established; sensitivity analysis is performed on the feature information of the type of extracted perception data to obtain the importance score of the feature information, and based on the importance score of the feature information, features with importance scores below a threshold are eliminated.

[0077] The types of perception data include water quality data types, such as ammonia nitrogen, total nitrogen, total phosphorus concentration and pH value, etc.; flow data types, such as inlet flow and outlet flow, etc.; biological data concentration, such as activated sludge concentration and microbial growth rate, etc. Sensitivity analysis is performed on the characteristic information of the extracted perception data types, and the Gain method built into XGBoost is used to calculate the average gain of each feature for splitting. The Gain method considers the impact of each split on reducing impurity, and weights the impact of all splits to obtain the importance score of the feature information, analyze the sensitivity of the feature information, and reflect the contribution of the feature to the model prediction.

[0078] By relatively measuring the feature importance score, the sensitivity of the feature information is analyzed. Based on the set threshold, features with importance lower than the threshold are eliminated, and beneficial indicators with higher feature contribution are added.

[0079]

[0080] In formula (2), k represents a node; T represents the number of all trees; N(t) represents the number of non-leaf nodes in the t-th tree; β(t,i) represents the partitioning feature of the i-th non-leaf node in the t-th tree, β(.)∈1,2,...,K; I(.) is the indicator function; ,H γ(t,i) They represent the sum of the first-order derivative and the second-order derivative of all samples falling on the i-th non-leaf node of the t-th tree; and Respectively represent the sum of the first-order derivatives on the left and right nodes of the ith non-leaf node on the tth tree; H γ(t,i,L) and H γ(t,i,R) They represent the sum of the second-order derivatives on the left and right nodes of the i-th non-leaf node on the t-th tree respectively; λ is the hyperparameter of the regularization term.

[0081] In step 12, the type of perception data is determined according to the specific needs and problems of the intelligent control system, and the corresponding sample library is established. By analyzing the data features preprocessed in step 11, the characteristic information of the type of relevant data type is extracted, and sensitivity analysis is performed, and the features with little impact on the prediction results are eliminated, and beneficial indicators are added to improve the accuracy and efficiency of the intelligent prediction model and enhance the discrimination of data types, which serves as the main data source for intelligent perception.

[0082] In specific implementation, in step 13, the machine learning algorithm is one or more of a decision tree algorithm, a support vector machine algorithm, a random forest algorithm or a neural network algorithm.

[0083] In step 13, a machine learning algorithm is applied to model and train the perception data. According to the type of prediction problem and data characteristics, an appropriate machine learning algorithm is selected, such as decision tree, support vector machine, random forest, neural network, etc. By learning and optimizing the training data, a suitable network structure is designed, a water quality intelligent prediction model is constructed, feature extraction is performed, and adjustments and improvements are made according to the training effect of the water quality intelligent prediction model.

[0084] In order to improve the efficiency and reliability of the method, the present invention adopts the GA_XGBoost model as the basic framework. First, the fitness function and optimization parameters are set according to the problem to be solved; secondly, it performs parameter encoding and creates an initial population containing optimization parameters, sets the population size, the number of individuals retained in each generation, and the parameter optimization range; finally, the best result of each iteration is obtained through the elite selection strategy. The present invention uses the genetic algorithm GA (Genetic Algorithm) to optimize the four parameters of GA_XGBoost, selects the best individuals through the elite retention strategy, and the meaning of the GA_XGBoost optimization parameters is shown in Table 1.

[0085]

[0086]

[0087] GA_XGBoost model fitness function setting: with the true value y i With the predicted value The error is used as the fitness function F t , to determine the model's fitting effect.

[0088]

[0089] In formula (3), y i is the true value of the water quality parameter inputted by i; is the average value of all values ​​of the water quality parameter input for the ith time; is the predicted value of the target water quality parameter.

[0090] The standard mathematical model of genetic algorithm GA can be expressed as:

[0091] GA={C,F t ,P0,N,S l , C r , M u , T e} (4)

[0092] In formula (4), C is the encoding method of water quality parameters, which means that different attributes of water quality parameters (such as influent COD, effluent COD, etc.) are encoded into a form that can be operated by the algorithm; S l is the selection operator of genetic operation, which means selecting the combination of water quality parameters with the best prediction effect for the next generation of evolution; C r is the crossover operator of genetic operation, which represents the new combination of water quality parameters generated by combining the characteristics of different water quality parameters; M u is the mutation operator of genetic operation, which means introducing random changes in water quality parameters and exploring new water quality parameter combinations; F t is the fitness function, which measures the degree of fit between the prediction results of the water quality prediction model and the actual observed values; P0 is the initial population, which represents the initial combination of water quality parameters; N is the population size, which represents the number of water quality parameter combinations in each generation; T e It is the end condition of the genetic operator, indicating that the condition for stopping iteration has been reached in water quality prediction.

[0093] The calculation process of the optimal parameters of XGBoost includes the following steps.

[0094] (1) GA_XGBoost model related parameter settings, model iteration number M, initial population number N, crossover and mutation probability P C , P m .

[0095] (2) Set the search range of the GA_XGBoost model optimization parameters.

[0096] (3) Encode the target problem, that is, encode the water quality parameters appropriately for genetic algorithm optimization.

[0097] (4) Population initialization: N initial populations C(t) are randomly generated, and each individual represents a set of water quality parameters.

[0098] (5) Calculate the individual fitness of the population C(t). The fitness function can be defined based on the water quality prediction error.

[0099] (6) Using the selection operator S lSelect a new population C(t1).

[0100] (7) Based on the new population C(t1), through the crossover operator C r With probability P C Generate a new population C(t2), retain the best individual generated, and do not participate in the next round of crossover operation.

[0101] (8) Based on the new population C(t2), using the mutation operator M u With a certain probability P m Generate a new population C(t3), retain the best individual generated, and do not participate in the next round of crossover operation.

[0102] (9) When the genetic operator end condition T is reached, the ideal fitness value is found and the evolution stops. The fitness of individuals is screened by the fitness function, and then the best individual is selected and decoded to obtain the optimal water quality parameter combination of XGBoost, realize the calculation of the optimal solution of the GA_XGBoost model, and improve the accuracy of the water quality prediction model.

[0103] The model learning process of XGBoost is as follows: (1) First, fit the first weak learner to the entire water quality parameter input data space; (2) Then fit the second model to these residuals to overcome the shortcomings of the weak learner; (3) Repeat the fitting process several times until the stopping criterion is met; (4) The final water quality parameter prediction of the model is obtained by summing the predictions of each learner.

[0104] The XGBoost model is designed to prevent overfitting and optimize computing resources. The principle is obtained by simplifying the objective function, which allows the combination of water quality parameter prediction terms and regularization terms while maintaining the best computing speed. Compared with traditional machine learning algorithms, the objective function of XGBoost can be expressed as a loss function + regularization term.

[0105]

[0106] In formula (5), y i represents the true value of the i-th water quality parameter sample, n represents the sum of the number of water quality parameter samples imported into the k-th tree, and k represents the sum of the trees.

[0107] Represents the loss function, which measures the actual value y of the true water quality parameter label i Predicted values ​​of water quality parameters The difference between. represents the regularization term. XGBoost minimizes obj to obtain the optimal At the same time, the error rate and complexity of the model are minimized.

[0108] After the tth iteration:

[0109]

[0110]

[0111] Expand equation (7) into the second-order Taylor expansion:

[0112]

[0113]

[0114] In formulas (8) and (9), g i represents the first-order derivative, h i represents the second-order derivative.

[0115] The regularization term is solved and defined as follows (10):

[0116]

[0117] In formula (10), γ and λ represent the regularization coefficients, T represents the number of leaf nodes, and w tm Represents the value of the mth leaf node of the weak learner at the tth iteration.

[0118] The present invention takes into account the complexity of water quality prediction problems and the diversity of data. In order to build a water quality intelligent prediction model, XGBoost is selected as the basic machine learning framework. XGBoost performs well in solving regression problems, can effectively process large-scale data sets, and has high interpretability and stability. XGBoost is used to evaluate the importance of features in sensory data, extract key features in sensory data, reduce the complexity of the model, and improve the generalization ability of the model. The training of the XGBoost model is an iterative process. Each round of iteration will fit a new weak learner, which focuses on correcting the errors of the previous round of models. Through multiple rounds of iterations, the model is continuously improved and a powerful integrated model is finally formed. A genetic algorithm is used to optimize the parameter configuration of the XGBoost model so that the model can better adapt to the water quality prediction problem. The genetic algorithm automatically searches for the best combination of water quality parameters, and designs a fitness function according to the error between the actual value and the predicted value of the real water quality parameter label to ensure that the optimization of the parameters is consistent with the actual water quality data. The genetic algorithm selects the best result of each iteration through an elite retention strategy to improve the model performance.

[0119] The constructed and optimized intelligent water quality prediction model can be applied to actual water quality prediction problems. The model uses the key features of the perception data and combines information such as water quality parameters to predict water quality. Because the model has been trained and optimized, it has a high degree of accuracy and adaptability and can provide reliable water quality prediction results based on actual conditions.

[0120] During specific implementation, in step 14, the structure of the water quality intelligent prediction model includes the number of layers, the number of nodes, the learning rate and the loss function.

[0121] In step 14, the structure and parameters of the water quality intelligent prediction model are designed. According to the specific prediction problem and algorithm selection, the model parameters such as the number of layers, the number of nodes, the learning rate, and the appropriate loss function and optimization algorithm are determined. Through training and optimization, a water quality intelligent prediction model that can accurately predict the perception data is obtained, and the water quality intelligent prediction model is used to finally output the prediction results including multiple water quality parameters such as core section MLSS, core section O3, effluent COD, effluent TN, effluent NH3, effluent TP, and effluent flow.

[0122] In specific implementation, the working process of the intelligent decision-making layer includes the following steps:

[0123] Step 21: Establish a sewage treatment mechanism model;

[0124] Step 22: Integrate expert knowledge and experience into the sewage treatment mechanism model;

[0125] Step 23: Optimize the processing process and improve production operation efficiency.

[0126] The intelligent decision-making layer is the core part of the artificial intelligence control system for sewage treatment. It is based on mechanism models, expert knowledge and experience, and provides scientific decision-making support for the sewage treatment process. The work of the intelligent decision-making layer includes three parts: application of mechanism models, integration of expert knowledge and experience, and optimization of the treatment process and improvement of production and operation efficiency.

[0127] During specific implementation, in step 21, the sewage treatment mechanism model is used to understand and describe the physical, chemical and biological reaction mechanisms in the sewage treatment process.

[0128] Sewage treatment mechanism models, such as the ASM (Activated Sludge Model) sewage treatment model, the system of the present invention can understand and describe the physical, chemical and biological reaction mechanisms in the sewage treatment process. By modeling and simulating the sewage treatment process, the system can predict the sewage treatment effect under different operating conditions and assist in decision-making analysis. The mechanism model provides theoretical support for the sewage treatment process and helps the system understand the complexity and interrelationships in the operation process.

[0129] During specific implementation, in step 22, expert knowledge and expert experience are converted into computable rules and models, and integrated into the sewage treatment mechanism model.

[0130] In step 22, the system of the present invention integrates expert knowledge and experience in the field of sewage treatment, including the practical experience of operators, professional knowledge of process optimization, etc. Through the expert system method, expert knowledge and experience are converted into computable rules and models, making them a reliable reference in the decision-making process. The integration of expert knowledge and experience can provide more accurate and reliable decision support, helping the system to quickly respond to changes and abnormal situations in the sewage treatment process.

[0131] During specific implementation, in step 23, during the optimization process of constructing the sewage treatment mechanism model, a multi-objective optimization weighted sum method is used to set weights and objective functions for multi-objective optimization and decision analysis.

[0132] Z=∑(W i *F i (C1,C2,...,C n )),i∈{1,2,...,n} (11)

[0133] In formula (11), Z is the total optimization target value, which represents the weighted sum of the water quality prediction target parameters to be minimized; W i is the objective function F i (C1,C2,...,C n ) related weight, indicating the weight of the i-th water quality parameter; F i (C1,C2,...,C n ) is the i-th optimization objective function, which is the water quality parameters C1, C2, ..., C n Complex function of C i Represents the specified water quality parameter, i represents the index of the parameter, ranging from 1 to n; n represents the number of objectives of the optimization problem.

[0134] In step 23, the system optimizes the treatment process and improves production and operation efficiency according to the sewage treatment objectives and operational requirements. By setting different weights and objective functions, the system can perform multi-objective optimization and decision analysis according to different operational objectives. The optimization process enables the system to comprehensively consider multiple factors such as economic benefits, environmental impact, and resource utilization to maximize comprehensive benefits.

[0135] Through the above steps 21 to 23, the intelligent decision-making module organically combines the mechanism model, expert knowledge and experience, and the optimization process, and provides scientific decision support through data analysis, model reasoning and optimization algorithms. The system can make decision-making operations such as operation adjustment, abnormal situation processing, and optimization parameter setting based on real-time perception data, prediction results, and operation requirements. The application of the intelligent decision-making module can improve the stability, efficiency, and sustainability of the sewage treatment process, and provide accurate and reliable decision support for the production and operation of the sewage treatment plant.

[0136] In the intelligent execution layer of the present invention, intelligent execution is realized based on artificial intelligence technology and control algorithm, including the following steps:

[0137] Step 31: Obtain the decision results and parse them; use the rule engine to extract key information from the decision results, obtain the decision results from the intelligent decision module, use the keyword extraction method in text processing technology combined with the rule engine analysis to parse and process the results, and perform data conversion based on the parsed information to ensure the accuracy and availability of the results. Convert the decision results into understandable forms, such as operating instructions, control parameters, or action sequences, to prepare for subsequent execution steps.

[0138] Step 32: Generate operation instructions according to the decision results; Generate corresponding operation instructions according to the decision results. Consider the system constraints, minimize the control amount D(t), and achieve precise intelligent control. Apply intelligent control to generate operation instructions, generate optimal operation instructions through optimization algorithms, and apply the generated operation instructions to the sewage treatment system to adjust water quality control parameters (such as aeration air volume, agent addition, carbon source addition, reflux ratio, etc.) to achieve changes in control target requirements, so as to achieve precise control of the sewage treatment process. Intelligent control comprehensively considers system constraints, target optimization and model prediction, and can realize intelligent execution process under the premise of ensuring system stability.

[0139] D(t)=K p ·(R(t)-O(t)) (12)

[0140] In formula (12), K p is the proportional gain, which is used to adjust the sensitivity of the control quantity D(t); R(t) is the target value, that is, the target value of the system water quality parameter, such as the target value of the effluent COD, effluent NH3 and other concentrations; O(t) is the predicted water quality parameter value, such as the actual predicted value of the effluent COD, effluent NH3 and other concentrations.

[0141] Step 33: Execute the operation instruction using the control system; Execute the generated operation instruction using the control system. By communicating and interacting with the interface of the device or system, the operation instruction is sent and the device is controlled to perform the corresponding operation, such as opening, closing, adjusting, etc., to realize the intelligent execution process. When applying intelligent control, the generation of the operation instruction can be based on the results of intelligent control optimization to ensure that the sewage treatment process is adjusted and controlled according to the optimized strategy.

[0142] Step 34: Monitor the execution process and provide real-time feedback on the execution results; Monitor the execution process and provide real-time feedback on the execution results. Through sensors, monitoring equipment or system feedback, data and status information during the execution process, including execution progress, execution results, environmental parameters, etc., are obtained to understand the execution status in real time. In particular, when applying intelligent control, the feedback information of the actual system can be monitored and compared with the results predicted by the model to evaluate the control effect and system performance.

[0143] Step 35: Feedback adjustment and optimization based on the execution results. Feedback adjustment and optimization are performed based on the execution results. The execution results are compared and analyzed with the expected targets. The feedback amount CO is used to adjust the key water quality parameters in the sewage treatment system so that the predicted water quality parameter result values ​​gradually tend to the target actual values. Model parameters, constraints, optimization targets, etc. are adjusted and optimized based on the feedback information. In particular, in intelligent control, feedback information can be used to correct model predictions and re-optimize operating instructions based on actual sewage treatment system performance and water quality parameter indicators. Feedback optimization and adjustment are used to continuously optimize the system's operating effects to ensure that the sewage treatment system maintains water quality parameters within the required range under changing conditions.

[0144] CO=K P *Eor+K i +K d *∫Eor dt+Kd*d(Eor) / dt (13)

[0145] In formula (13), K P is the proportional coefficient, which is used to adjust the control amount according to the current error and represents the strength of the proportional action; Eor is the control error, which represents the error between the current water quality parameter prediction value and the current water quality parameter target actual value; K i K is the integral coefficient, which is used to consider the influence of error accumulation over time and represents the strength of the integral effect; d is the differential coefficient, which is used to consider the influence of the speed of error change over time and indicates the intensity of the differential effect; ∫Eor dt is the integral of the error over time, which considers the accumulation of the error; d(Eor) / dt is the rate of change of the error over time, which considers the trend of the error.

[0146] The artificial intelligence control system for sewage treatment of the present invention can realize intelligent management and optimization of the intelligent control system through the technical solutions of the three core parts of intelligent perception, intelligent decision-making and intelligent execution, improve the efficiency of water resource utilization, reduce operating costs, and provide decision makers with scientific and accurate decision-making basis.

[0147] An artificial intelligence control system for sewage treatment of the present invention combines sewage treatment technology with artificial intelligence technology, and is composed of three core parts: intelligent perception, intelligent decision-making, and intelligent execution. In terms of intelligent perception, it is used to collect and obtain data in real time, conduct water quality analysis and prediction, and timely discover abnormal situations to provide warning information, so as to realize full-parameter, full-section, and full-process prediction and warning by combining data collection, intelligent prediction, and intelligent early warning. In terms of intelligent decision-making, it is used to integrate various data and knowledge resources, conduct decision analysis, optimize scheduling and risk assessment, and realize intelligent water affairs decision support by combining mechanism models, expert knowledge and experience, and optimized processing processes. In terms of intelligent execution, it is used to automate control algorithms and intelligent execution strategies by combining intelligent control and execution strategies, realize accurate control and regulation of water affairs systems, optimize operating costs and save resources. The present invention combines the development needs of artificial intelligence water affairs technology and the actual operation of sewage treatment plants to realize the digitization and intelligence of sewage treatment plant production and operation, and provides an efficient intelligent digital management method to improve operational efficiency, reduce costs, and help sewage treatment plants realize digital intelligent production and operation.

[0148] The artificial intelligence control system for sewage treatment of the present invention has the following characteristics.

[0149] 1. The artificial intelligence control system for sewage treatment of the present invention can realize intelligent sewage treatment production and operation, including sewage treatment process optimization, operation monitoring and abnormal warning. Through intelligent algorithms and data analysis, the system can improve sewage treatment efficiency, reduce energy consumption and chemical usage, and ensure that the effluent meets environmental standards.

[0150] 2. The artificial intelligence control system for sewage treatment of the present invention has intelligent monitoring and real-time early warning functions, and can monitor key indicators such as the influent water quality, sludge concentration, and equipment operation status of the sewage treatment plant. The system can promptly detect abnormal situations, such as equipment failure or abnormal water quality, and provide early warning information to help sewage treatment plant managers take quick measures to prevent accidents or reduce their impact.

[0151] 3. The artificial intelligence control system for sewage treatment of the present invention supports intelligent sewage treatment process control. Through real-time monitoring data and advanced control algorithms, the system can automatically adjust sewage treatment process parameters, such as aeration volume, sludge return ratio, etc., to optimize the treatment effect and improve the stability and reliability of the treatment system.

[0152] 4. The artificial intelligence control system for sewage treatment of the present invention provides intelligent decision support. Based on historical data and prediction models, the system can generate reasonable operation suggestions and optimization plans, such as sludge treatment plans, chemical dosage, etc. This helps sewage plant managers to formulate scientific decision-making strategies and improve the efficiency of sewage treatment and resource utilization.

[0153] 5. The artificial intelligence control system for sewage treatment of the present invention has the characteristics of energy saving, emission reduction and environmental protection. By optimizing sewage treatment process and energy utilization, the system can reduce energy consumption and carbon emissions in the sewage treatment process and reduce the negative impact on the environment. At the same time, the system can monitor the discharge water quality and environmental parameters, detect abnormal conditions in time, and take measures to reduce the discharge of pollutants.

[0154] The artificial intelligence control system for sewage treatment of the present invention provides a centralized management platform and security mechanism to ensure the security and reliability of sewage treatment data. The system supports authority management, limits data access and operation permissions, and reduces the risk of data leakage and manipulation. At the same time, the system provides a user interface and report display to help sewage plant managers understand the sewage treatment process in real time and conduct system management and decision analysis. The present invention provides an efficient management method by applying digitization and intelligence to the production and operation of sewage plants. It integrates advanced data collection, prediction, decision-making and execution technologies to concretize the production and operation of sewage plants and lead them to develop in the direction of digitization and intelligence. The application of this system can improve the efficiency of the production and operation of sewage plants, reduce environmental risks, and reduce resource waste, making an important contribution to the realization of the production and operation goals of sewage plants based on artificial intelligence.

[0155] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

[0156] In addition, it should be understood that although the present specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.

Claims

1. An artificial intelligence control system for sewage treatment, characterized in that: It includes intelligent perception layer, intelligent decision-making layer and intelligent execution layer; The intelligent perception layer is used to collect the inlet water quality parameters Qj, the core section water quality parameters Qh and the outlet water quality parameters Qc of the sewage treatment system, and establish a water quality intelligent prediction model based on the perception data, and output the predicted core section water quality parameters Qh and outlet water quality parameters Qc through the intelligent prediction model; The intelligent decision-making layer is used to establish a sewage treatment model that can reflect the physical, chemical and biological reaction mechanisms in the sewage treatment process, integrate expert knowledge and experience into the sewage treatment model, and then optimize the treatment process and improve efficiency. Through the integrated knowledge and experience, multi-objective optimization and decision analysis are carried out according to different operation goals; The working process of the intelligent decision-making layer includes the following steps: Step 21: Establish a sewage treatment mechanism model; Step 22: Integrate expert knowledge and experience into the sewage treatment mechanism model; Step 23: Optimize the processing process and improve production operation efficiency; In the process of optimizing the sewage treatment mechanism model, the weighted sum method of multi-objective optimization is used to set weights and objective functions for multi-objective optimization and decision analysis; （11） In formula (11), Z is the total optimization target value, which represents the weighted sum of the water quality prediction target parameters to be minimized; With the objective function The relevant weight represents the weight of the i-th water quality parameter; is the i-th optimization objective function, which is the water quality parameter Complex functions of represents the specified water quality parameter, i represents the index of the parameter, ranging from 1 to n; n represents the number of objectives of the optimization problem; The intelligent execution layer is used to obtain the results of multi-objective optimization and decision analysis of the intelligent decision layer, generate operation instructions by applying intelligent control, and send the operation instructions to the execution device to execute the operation; During the execution process, the execution results are monitored and fed back in real time. The execution results are compared and analyzed with the expected goals, and adjustments and optimizations are made based on the results of the comparison and analysis.

2. According to claim 1, an artificial intelligence control system for sewage treatment is characterized in that: The working process of the intelligent perception layer includes the following steps: Step 11: Collect the sensory data of the sewage treatment system and pre-process the sensory data; the sensory data includes the inlet water quality parameter Qj, the core section water quality parameter Qh and the outlet water quality parameter Qc; Step 12: Identify the data type of the perception data, extract the features of the perception data, and perform sensitivity analysis on the features of the extracted perception data; Step 13: Apply machine learning algorithms to model and train water quality intelligent prediction models based on the perception data; Step 14: Calculate the structural parameters of the water quality intelligent prediction model, and finally output the prediction results of the perception data according to the water quality intelligent prediction model.

3. The artificial intelligence control system for sewage treatment according to claim 2 is characterized in that: In step 11, the process of preprocessing the perception data includes the following steps: Step 111: sensing data cleaning step; data cleaning includes removing noise, processing missing values ​​and outliers; Step 112: sensing data normalization step; data normalization includes data normalization or data standardization; Step 113: feature engineering step of perception data; Step 114: Perception data alignment step.

4. The artificial intelligence control system for sewage treatment according to claim 2 is characterized in that: In step 12, the type of the perception data preprocessed in step 11 is determined, and a sample library is established; sensitivity analysis is performed on the feature information of the type of the extracted perception data to obtain the importance score of the feature information, and based on the importance score of the feature information, features with importance scores below a threshold are removed.

5. The artificial intelligence control system for sewage treatment according to claim 2 is characterized in that: In step 13, the machine learning algorithm is one or more of a decision tree algorithm, a support vector machine algorithm, a random forest algorithm or a neural network algorithm.

6. The artificial intelligence control system for sewage treatment according to claim 2 is characterized in that: In step 14, the structure of the water quality intelligent prediction model includes the number of layers, the number of nodes, the learning rate and the loss function.

7. The artificial intelligence control system for sewage treatment according to claim 1 is characterized in that: In step 21, the sewage treatment mechanism model is used to understand and describe the physical, chemical and biological reaction mechanisms in the sewage treatment process.

8. The artificial intelligence control system for sewage treatment according to claim 1 is characterized in that: In step 22, expert knowledge and expert experience are converted into computable rules and models, and integrated into the sewage treatment mechanism model.

Images