An online monitoring method, device and storage medium for a sewage treatment process

Abstract

The application discloses an online monitoring method and device for a sewage treatment process and a storage medium. An offline data is used to establish an auto-encoder one-class support vector machine model, then the collected field data is preprocessed, and the prediction score of each real-time observation sample is calculated. The prediction score is compared with a preset threshold. If the prediction score value does not exceed the control limit, it indicates that the working condition is normal. On the contrary, if the prediction score value exceeds the control limit, it can be determined that the sewage treatment process has failed at the observation time. The application fully considers the characteristics of the sewage treatment process with a large number of observation variables, and obtains the sparse characteristics of the observation variables through the feature extraction technology of the auto-encoder, thereby solving the problem that the online monitoring method based on the one-class support vector machine cannot process high-dimensional data. Meanwhile, the application of the one-class support vector machine solves the problem of unbalanced data samples in the sewage treatment process. The application can be widely applied to the technical field of online fault monitoring.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of fault online monitoring, and in particular to an online monitoring method and device for a sewage treatment process and a storage medium. BACKGROUND

[0002] With the increasing speed of urbanization, the discharge of urban sewage has significantly increased. In order to cope with the increasing urban sewage, it has become one of the needs of social development to build a sewage treatment plant with high automation. Under such a background, it is necessary to effectively monitor the working state of the sewage treatment plant to ensure its normal operation and make the sewage meet the specified discharge standard.

[0003] However, the sewage treatment process is affected by many physical, biological and chemical factors, and has the characteristics of strong nonlinearity, dynamics, strong coupling and the like, making it difficult to model. Therefore, a data-driven model, support vector machine, is applied without prior biochemical knowledge. However, the data collected in the sewage treatment process lacks label information, and the data has strong unbalanced characteristics, requiring the use of unsupervised learning method, one-class support vector machine. However, due to inefficient feature engineering and optimization complexity, the biggest problem of one-class support vector machine is that it cannot handle large and high-dimensional data sets, resulting in that one-class support vector machine is not ideal in the application of high-dimensional fault detection scenarios. In order to ensure the safety and stability of the sewage treatment process operation system, it is necessary to establish an effective process monitoring scheme to timely monitor abnormal phenomena. SUMMARY

[0004] To at least partially solve one of the technical problems existing in the prior art, the purpose of the present application is to provide an online monitoring method and device for a sewage treatment process and a storage medium.

[0005] The technical solution adopted by the present application is:

[0006] An online monitoring method for a sewage treatment process,

[0007] comprising the following steps:

[0008] determining observation variables and setting a sampling interval, collecting observation data in normal working conditions as a training set, pre-processing the training set to obtain a pre-processed training set;

[0009] determining the structure of an auto-encoder network according to the characteristics of the training set, and setting a loss function;

[0010] training the auto-encoder using the pre-processed training set to obtain the optimal parameters of the network, and then using the encoder in the auto-encoder to extract the characteristics of the training set to obtain a sparse feature set of the training set;

[0011] Based on the characteristics of sparse feature sets, the kernel function type and corresponding hyperparameters of a class of support vector machines are determined. A class of support vector machines is trained using sparse feature sets, and the prediction scores of samples corresponding to sparse features are obtained.

[0012] Box plots were used to analyze the data distribution characteristics of the predicted score set in order to determine the control limits for fault monitoring.

[0013] Real-time observation data is collected as test samples. After preprocessing the test set, sparse features of the test samples are extracted using an autoencoder, and a type of support vector machine is used to predict the scores of the sparse features.

[0014] To determine if a malfunction has occurred in the wastewater treatment process: Subtract the predicted score of the observed sample at that moment from the control limit obtained from the box plot. If the predicted score does not exceed the control limit, the operation is considered normal, and the next sampling is performed. If the predicted score exceeds the control limit, a malfunction is determined to have occurred at that observation moment, and an alarm is issued. The larger the value exceeding the control limit, the more severe the malfunction.

[0015] Furthermore, the preprocessing of the training set includes:

[0016] Using the standard deviation standardization method, the data matrix observed from the wastewater treatment plant is preprocessed according to the following formula:

[0017]

[0018] Where, X∈R m×n It is the original observation data matrix, R m×n Let X be an m x n real matrix, where each row of X represents an observed sample and each column represents an observed variable, μ x and σ x X represents the sample mean and sample standard deviation of each observed variable in X, and X' represents the preprocessed training set.

[0019] Furthermore, the preprocessing of the training set includes:

[0020] Using range standardization, the data matrix observed from the wastewater treatment plant is preprocessed according to the following formula:

[0021]

[0022] Where, X∈R m×n It is the original observation data matrix, R m×n Let X be an m x n real matrix, where each row of X represents an observed sample and each column represents an observed variable. Let X' represent the preprocessed training set. Let X(i,j) and X'(i,j) represent the elements at the i-th row and j-th column positions of X and X', respectively. Let X(j) be the element at the i-th row and j-th column of X and X', respectively. maxX(j) min Let X represent the maximum and minimum values ​​in the j-th column of X, respectively.

[0023] Furthermore, the step of determining the autoencoder network structure based on the features of the training set and setting the loss function includes:

[0024] The number of hidden layers, as well as the number of neurons in the input, hidden, and output layers, are determined based on the preprocessed training set X'. For low-dimensional and small-sample datasets, fewer hidden layers are used, which reduces computational cost while still effectively reconstructing the input. In particular, for high-dimensional and large-sample datasets, the number of hidden layers can be appropriately stacked to expand the autoencoder model's ability to reconstruct data.

[0025] Based on the preprocessed training set X', a loss function and convergent model are selected. During training, the autoencoder loss function is chosen as the L1 regularized loss function, expressed as follows:

[0026]

[0027] Where x' i,j Let x represent a variable in a specific observation sample of the preprocessed training set. * i,j This represents the output reconstructed by the autoencoder based on the input, where m is the total number of observation samples in the training set and n is the total number of variables in the observation samples.

[0028] Furthermore, the structure of the autoencoder depends on the preprocessed training set X'. For the preprocessed dataset X', the encoder (based on W)... h The linear transformation of the vector (W) outputs the hidden feature S, which is then passed through a decoder (based on W). x (linear transformation) to reconstruct X * An autoencoder network is represented as follows:

[0029] s = f(W h x'+b n )

[0030] x * =g(W x s+b s )

[0031] The autoencoder in this invention specifically refers to a stacked autoencoder, which is composed of multiple stacked autoencoders. It is a neural network consisting of multiple layers of autoencoders, with the output of the previous layer serving as the input of the next layer. For low-dimensional and small-sample datasets, fewer hidden layers are used, which reduces computational cost while still reconstructing the input effectively. In particular, for high-dimensional and large-sample datasets, the number of hidden layers can be appropriately stacked to expand the autoencoder model's ability to reconstruct data.

[0032] Furthermore, the expression for this type of support vector machine is:

[0033] f(x) = w T φ(x)-ρ

[0034] The objective equation for this type of support vector machine is:

[0035]

[0036] w T φ(x i )≥ρ-ξ i ,ξ i >0

[0037] Where φ is the feature projection function that projects the input vector onto the high-dimensional feature space; w is the general vector of the decision hyperplane perpendicular to the hyperplane; ρ is the intercept term; and ξ... i represents the slack variable, v is the upper bound hyperparameter set for the outlier score, and n represents the vector dimension.

[0038] Furthermore, by applying a Lagrange transform to the objective equation and using the kernel function technique for the dot product operation, the decision function becomes:

[0039]

[0040] Where, α i It is a Lagrange multiplier, k(x i ,x)=φ(x i ) T φ(x) is the kernel function. The kernel function can be preferably chosen as a radial basis function, or it can be chosen as a linear kernel function.

[0041]

[0042] K(x,z) = x·z

[0043] Here, γ represents a free parameter, and x and z represent two vectors respectively.

[0044] Furthermore, the step of using box plot analysis to predict the data distribution characteristics of the score set to determine the control limits for fault monitoring includes:

[0045] Assume the upper and lower quartiles of the predicted score set are Q3 and Q, respectively. 1` Then the control limit is expressed as:

[0046] L = 2.5Q³ - 1.5Q¹

[0047] Furthermore, determining whether a malfunction has occurred in the wastewater treatment process includes:

[0048] During the online monitoring phase, the difference between the predicted score P of a support vector machine and the control limit L determined by the box plot is calculated:

[0049] E=PL

[0050] According to the theory of box plots, when P is less than 1.5(Q3-Q1), a minor fault is determined to have occurred, and when P is greater than 3(Q3-Q1), a serious fault is determined to have occurred.

[0051] Furthermore, the sparse features S output by the autoencoder during the encoding process are set as the input to a support vector machine. In the offline training phase, the model is trained using the preprocessed training set to obtain the optimal parameters of the autoencoder. Then, the model extracts the compact features from the training set and feeds them into the support vector machine. In the online monitoring phase, the autoencoder extracts compact features only from a single real-time acquired data sample. This extraction process occurs during the autoencoder's encoding process and does not involve the decoding process.

[0052] Furthermore, the preprocessing method for individual observation data samples acquired in real time is consistent with the preprocessing method for the training set; when the training set uses the standard deviation standardization method, individual data samples are standardized using the mean and variance of the training set; when the training set uses the range standardization method, if the value of a variable in an individual data sample is greater than or less than the extreme value, the variable value should be set to 1 or 0 accordingly.

[0053] Another technical solution adopted in this invention is:

[0054] An online monitoring device for wastewater treatment processes includes:

[0055] At least one processor;

[0056] At least one memory for storing at least one program;

[0057] When the at least one program is executed by the at least one processor, the at least one processor implements the method described above.

[0058] Another technical solution adopted in this invention is:

[0059] A computer-readable storage medium storing a processor-executable program, which, when executed by a processor, performs the method described above.

[0060] The beneficial effects of this invention are as follows: This invention fully considers the numerous observation variables in the wastewater treatment process and uses an autoencoder feature extraction technique to obtain sparse features of these observation variables, thus solving the problem that online monitoring methods based on a certain type of support vector machine cannot handle high-dimensional data. Simultaneously, the application of a certain type of support vector machine effectively addresses the problem of imbalanced data samples in the wastewater treatment process. Attached Figure Description

[0061] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following description is provided with accompanying drawings of the relevant technical solutions in the embodiments of the present invention or the prior art. It should be understood that the accompanying drawings described below are only for the purpose of clearly illustrating some embodiments of the technical solutions of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0062] Figure 1 This is a flowchart of an online monitoring method for a type of support vector machine with an autoencoder in a wastewater treatment process, as described in an embodiment of the present invention.

[0063] Figure 2 This is a fault detection diagram of 25 randomly observed variables in Embodiment 1 of the present invention;

[0064] Figure 3 This is a fault detection diagram of 25 randomly observed variables in Embodiment 2 of the present invention;

[0065] Figure 4 This is a fault detection diagram of 25 randomly observed variables in Embodiment 3 of the present invention;

[0066] Figure 5 This is a fault detection diagram of all 38 observed variables in Embodiment 4 of the present invention. Detailed Implementation

[0067] The embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention. The step numbers in the following embodiments are set only for ease of explanation, and there is no limitation on the order between the steps. The execution order of each step in the embodiments can be adaptively adjusted according to the understanding of those skilled in the art.

[0068] In the description of this invention, it should be understood that the orientation descriptions, such as up, down, front, back, left, right, etc., are based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limiting this invention.

[0069] In the description of this invention, "several" means one or more, "more than" means two or more, "greater than," "less than," and "exceeding" are understood to exclude the stated number, while "above," "below," and "within" are understood to include the stated number. The use of "first" and "second" in the description is merely for distinguishing technical features and should not be construed as indicating or implying relative importance, or implicitly indicating the number of indicated technical features, or implicitly indicating the order of the indicated technical features.

[0070] In the description of this invention, unless otherwise explicitly defined, terms such as "set up," "install," and "connect" should be interpreted broadly, and those skilled in the art can reasonably determine the specific meaning of the above terms in this invention in conjunction with the specific content of the technical solution.

[0071] Example 1

[0072] like Figure 1 As shown, this embodiment provides an online monitoring method for a class of support vector machines (SVMs) using an autoencoder in a wastewater treatment process. The method first trains an autoencoder-based SVM monitoring model (the monitoring model is an online fault monitoring model connecting the autoencoder and the SVM) using offline data from the wastewater treatment plant under normal operating conditions. Then, it preprocesses the collected real-time field data, extracts the compact features of each real-time observation sample using the autoencoder, and feeds these features into the SVM to obtain a predicted score. If the predicted score exceeds the control limit, a fault is determined to have occurred. Specifically, the method includes the following steps:

[0073] A1. Determine the observed variables and set the same sampling interval. Use the data matrix collected from the wastewater treatment plant under normal operating conditions as the training set, and perform data preprocessing according to the following formula:

[0074]

[0075] Where, X∈R m×n It is the original observation data matrix, R m×n Let X be an m x n real matrix. Each row of X represents an observation sample, and each column represents an observation variable, including total input flow rate, pH value, etc. x and σ x X represents the sample mean and sample standard deviation of each observed variable in X, and X' represents the preprocessed training set.

[0076] A2. Construct the autoencoder network structure and set the loss function during autoencoder training as follows:

[0077]

[0078] A3. Train an autoencoder using the preprocessed training set X', and use the encoder to decompose the preprocessed training set X' into a low-dimensional feature S:

[0079] s = f(W h x'+b n )

[0080] A4. Set the hyperparameters v of a class of support vector machines, and choose the kernel function as follows:

[0081]

[0082] A class of support vector machines is trained using the tightly packed features S encoded in the autoencoder to obtain the optimal parameters.

[0083] A5. Use box plots to analyze the distribution characteristics of the predicted fractional data, and thereby determine the control limits for fault monitoring:

[0084] L = 2.5Q³ - 1.5Q¹

[0085] A6. Collect real-time observation data as test samples. After preprocessing the test set, use an autoencoder to extract sparse features of the test samples and use a support vector machine to predict the scores of the sparse features.

[0086] A7. To determine whether a malfunction has occurred in the wastewater treatment process, the predicted score P of the observed sample at that moment is subtracted from the control limit L obtained from the box plot:

[0087] E=PL

[0088] If the predicted score does not exceed the control limit, the operating condition is normal, and new samples can continue to be collected. Conversely, if the predicted score exceeds the control limit, a fault is determined to have occurred at that observation time, and an alarm is issued. The larger the value exceeding the control limit, the more severe the fault.

[0089] In this embodiment, wastewater treatment process data provided by a wastewater treatment plant was used for verification. The dataset was provided by the Machine Learning Library at the University of California, Irvine (UCI) and contained 527 samples with a daily sampling interval. Some sensor data were missing in 84 influent samples and 72 effluent samples. There were a total of 38 observed variables, the meaning of which is shown in Table 1. Among the 527 samples, there were 14 fault samples with 9 fault types; the remaining samples could be considered normal samples.

[0090] Table 1 Experimental Data Observation Variables

[0091]

[0092]

[0093] like Figure 1 As shown, the general process of the above-mentioned wastewater treatment plant fault monitoring method is as follows: 25 observation variables are randomly selected, and the observation data is divided into two parts: the samples from the first 263 days are used as the training set to build the model, and the samples from the last 264 days are used as the test set to verify the effectiveness of the proposed method. The specific steps are as follows:

[0094] B1. Preprocess the training data using the standard deviation standardization method;

[0095] B2. Construct an autoencoder network structure with 25 input layer neurons, five hidden layer neurons with 32, 16, 8, 16, and 32 neurons respectively, and 25 output layer neurons. The encoder part outputs 8-dimensional features. Set 1000 training epochs and use the following loss function during autoencoder training:

[0096]

[0097] B3. Train an autoencoder using the preprocessed training set X', and use the encoder to decompose the preprocessed training set X' into 8-dimensional features S:

[0098] s = f(W h x'+b n )

[0099] B4. Set the hyperparameter v = 0.1 for a class of support vector machines, and choose the following kernel function:

[0100]

[0101] A support vector machine is trained using the tightly encoded features S obtained from the autoencoder to obtain the optimal values ​​of the parameters Y.

[0102] B5. Use box plots to analyze the distribution characteristics of the predicted fractional data, thereby determining the control limits for fault monitoring;

[0103] L = 2.5Q3 - 1.5Q1 = 1.0007115

[0104] B6. Collect real-time observation data as test samples. After preprocessing the test set, use an autoencoder to extract sparse features of the test samples and use a support vector machine to predict the scores of the sparse features.

[0105] B7. To determine whether a malfunction has occurred in the wastewater treatment process, the predicted score P of the observed sample at that moment is subtracted from the control limit L obtained from the box plot:

[0106] E=PL

[0107] If the predicted score does not exceed the control limit, the operating condition is normal, and new samples can continue to be collected. Conversely, if the predicted score exceeds the control limit, a fault is determined to have occurred at that observation time, and an alarm is issued. The larger the value exceeding the control limit, the more severe the fault.

[0108] The advantage of autoencoders in extracting features lies in their effective use of hidden information from normal data points; they are trained to minimize reconstruction error. Therefore, during testing, they can reconstruct normal data points with a very low error rate. However, autoencoders cannot identify faults and cannot reconstruct them correctly. In such cases, the reconstruction error is higher than normal. Therefore, the extracted sparse features can be used in a class of support vector machines for fault detection.

[0109] from Figure 2 Experimental results show that this scheme has excellent monitoring effects on various faults in wastewater treatment plants. During the online monitoring phase, faults such as secondary sedimentation tank settling, solid overload, and rainstorm-induced faults occurred successively in the wastewater treatment process. The trained model was able to detect all of them, achieving a fault recall rate of 96%. Meanwhile, due to the excellent feature representation of normal data, the scheme's false alarm rate was only 4.4%, and the fault diagnosis accuracy was 95.45%. Computation time is a crucial factor in evaluating classifier performance, and reducing computation time is also very important. Using an autoencoder to extract sparse features significantly reduces the computational load of the first-class support vector machine, greatly improving the speed of online monitoring. Comparing different types of faults, the difference between rainstorm-related faults and control limits was significantly smaller than that of the other two types, consistent with the experience that rainstorm-related faults are less severe than other types, providing a basis for staff to judge the severity of faults.

[0110] Example 2

[0111] Using experimental data different from that in Example 1, 25 additional variables that were not exactly the same were randomly selected. The parameters set during the modeling process were the same as those in Example 1. The difference was that the calculated control limit was changed to 0.61.

[0112] from Figure 3Experimental results show that this scheme has excellent monitoring effects on various faults in wastewater treatment plants. During the online monitoring phase, faults such as secondary sedimentation tank settling, solids overload, and rainstorm-induced faults occurred successively in the wastewater treatment process. The trained model was able to detect all of them, with a fault recall rate of 96%. Meanwhile, due to the excellent feature representation of normal data, the false alarm rate of this scheme was only 5.2%, and the fault diagnosis accuracy rate was 94.7%. Comparing different types of faults, the difference between the faults caused by rainstorms and the control limits was significantly smaller than that of the other two types of faults, consistent with the experience that rainstorm-related faults are less severe than other types of faults, providing a basis for staff to judge the severity of faults.

[0113] Example 3

[0114] Using experimental data different from that in Example 1, 25 additional variables that were not exactly the same were randomly selected. The parameters set during the modeling process were the same as those in Example 1. The difference was that the calculated control limit was changed to 0.61.

[0115] from Figure 4 Experimental results show that this scheme has excellent monitoring effects on various faults in wastewater treatment plants. During the online monitoring phase, faults such as secondary sedimentation tank settling, solids overload, and rainstorm-induced faults occurred successively in the wastewater treatment process. The trained model was able to detect all of these faults, achieving a fault recall rate of 100%. Meanwhile, due to the excellent feature representation of normal data, the false alarm rate of the above scheme was 13.2%, and the fault diagnosis accuracy rate was 87.5%. Comparing different types of faults, the difference between the rainstorm-related faults and the control limits was significantly smaller than that of the other two types of faults, consistent with the experience that rainstorm-related faults are less severe than other types of faults, providing a basis for staff to judge the severity of faults.

[0116] Example 4

[0117] Using experimental data different from the three embodiments mentioned above, all 38 observed variables were used as experimental data, and the number of input and output neurons was changed to 38. The parameters set in the modeling process were the same as in Embodiment 1. The difference was that the calculated control limit was changed to 1.20.

[0118] from Figure 5Experimental results show that the above-mentioned scheme has a good monitoring effect on various faults in wastewater treatment plants. During the online monitoring phase, faults such as secondary sedimentation tank settling, solid overload, and rainstorm-induced faults occurred successively in the wastewater treatment process. The trained model was able to detect all of them, with a fault recall rate of 100%. Meanwhile, due to the excellent feature representation of normal data, the false alarm rate of the above-mentioned scheme was low at 2.4%, and the fault diagnosis accuracy rate was 97.73%. Comparing different types of faults, the difference between the rainstorm-related faults and the control limits was significantly smaller than that of the other two types of faults, consistent with the experience that rainstorm-related faults are less severe than other types of faults, providing a basis for staff to judge the severity of faults.

[0119] The above embodiments demonstrate that the online fault monitoring model based on an autoencoder-based support vector machine developed in this invention exhibits more satisfactory performance compared to traditional online fault monitoring models. Table 2 lists the online monitoring performance of the autoencoder-based support vector machine developed in this invention for different numbers and types of input variables. The statistical results in Table 2 show that using all observed variables as training data has certain advantages, enabling the model to achieve a lower false alarm rate and higher monitoring accuracy. However, using a single-class support vector machine alone cannot effectively model high-dimensional training data, necessitating the use of an autoencoder, a feature extraction technique. Comparing the results in Table 2, the autoencoder-based support vector machine online monitoring model demonstrates better performance.

[0120] Table 2 Comparison of fault detection results between principal component analysis and feature extraction.

[0121]

[0122] In summary, compared with existing technologies, the present invention has at least the following advantages and beneficial effects:

[0123] 1. This invention effectively reduces the computational load of the fault monitoring process through feature extraction methods, while identifying and refining the fault information contained in the data samples.

[0124] 2. This invention effectively solves the problems of insufficient fault data and imbalanced label data in the sewage treatment process by learning a classification hyperplane of a support vector machine using normal working data samples, and realizes fault monitoring in the sewage treatment process.

[0125] 3. This invention can significantly reduce the false alarm rate and false alarm rate when detecting faults in the sewage treatment process of sewage treatment plants. Therefore, it can reduce the workload of sewage treatment plant staff, while also enabling timely detection of faults, allowing sufficient time for the maintenance of sewage treatment plant equipment and saving maintenance costs.

[0126] 4. The prediction score of the support vector machine used in this invention varies with different types of faults and the severity of the faults. Therefore, the difference between the prediction score and the control limit can reflect the severity of the fault, thus providing reference information for on-site personnel to make maintenance strategies.

[0127] This embodiment provides an online monitoring device for wastewater treatment processes, including:

[0128] At least one processor;

[0129] At least one memory for storing at least one program;

[0130] When the at least one program is executed by the at least one processor, the at least one processor implements Figure 1 The method shown.

[0131] This embodiment provides an online monitoring device for a wastewater treatment process. It can execute an online monitoring method for a wastewater treatment process provided in the method embodiment of the present invention, and can execute any combination of the implementation steps of the method embodiment, possessing the corresponding functions and beneficial effects of the method.

[0132] This application also discloses a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium and execute the computer instructions, causing the computer device to perform... Figure 1 The method shown.

[0133] This embodiment also provides a storage medium storing instructions or programs that can execute the online monitoring method for wastewater treatment provided in the method embodiment of the present invention. When the instructions or programs are run, any combination of implementation steps of the method embodiment can be executed, and the method has the corresponding functions and beneficial effects.

[0134] In some alternative embodiments, the functions / operations mentioned in the block diagrams may not occur in the order shown in the operation diagrams. For example, depending on the functions / operations involved, two consecutively shown blocks may actually be executed substantially simultaneously, or the blocks may sometimes be executed in reverse order. Furthermore, the embodiments presented and described in the flowcharts of this invention are provided by way of example to provide a more comprehensive understanding of the technology. The disclosed methods are not limited to the operations and logic flows presented herein. Alternative embodiments are contemplated in which the order of various operations is altered and sub-operations described as part of a larger operation are executed independently.

[0135] Furthermore, although the invention has been described in the context of functional modules, it should be understood that, unless otherwise stated, one or more of the described functions and / or features may be integrated into a single physical device and / or software module, or one or more functions and / or features may be implemented in a separate physical device or software module. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding the invention. Rather, given the properties, functions, and internal relationships of the various functional modules in the apparatus disclosed herein, the actual implementation of the module will be understood within the scope of conventional skill of an engineer. Therefore, those skilled in the art can implement the invention as set forth in the claims using ordinary techniques without excessive experimentation. It is also understood that the specific concepts disclosed are merely illustrative and not intended to limit the scope of the invention, which is determined by the full scope of the appended claims and their equivalents.

[0136] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0137] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0138] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0139] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0140] In the foregoing description of this specification, references to terms such as "one embodiment," "another embodiment," or "some embodiments" indicate that a specific feature, structure, material, or characteristic described in connection with an embodiment or example is included in at least one embodiment or example of the present invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0141] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

[0142] The above is a detailed description of the preferred embodiments of the present invention. However, the present invention is not limited to the above embodiments. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention. All such equivalent modifications or substitutions are included within the scope defined by the claims of this application.

Claims

1. An online monitoring method for wastewater treatment processes, characterized in that, Includes the following steps: Determine the observation variables and set the sampling interval. Use the observation data collected under normal operating conditions as the training set. Preprocess the training set to obtain the preprocessed training set. The autoencoder network structure is determined based on the features of the training set, and the loss function is set accordingly; The autoencoder is trained using the preprocessed training set to obtain the optimal parameters of the network. Then, the encoder in the autoencoder is used to extract the features of the training set to obtain a sparse feature set of the training set. Based on the characteristics of sparse feature sets, the kernel function type and corresponding hyperparameters of a class of support vector machines are determined. A class of support vector machines is trained using sparse feature sets, and the prediction scores of samples corresponding to sparse features are obtained. Box plots were used to analyze the data distribution characteristics of the predicted score set in order to determine the control limits for fault monitoring. To determine whether a malfunction has occurred in the wastewater treatment process: collect observation data at the sampling interval as test samples, and calculate the difference between the predicted score of the test samples and the control limit obtained from the box plot; if the predicted score does not exceed the control limit, the operating condition is determined to be normal, and the next sampling is performed. If the predicted score exceeds the control limit, it is determined that a fault occurred during the collection of the test sample, and an alarm is issued. The expression for this type of support vector machine is: The objective equation for this type of support vector machine is: in, It is a feature projection function that projects the input vector onto a high-dimensional feature space; It is a general vector of the decision hyperplane perpendicular to the hyperplane. It is the intercept term. Represents slack variables. This is a hyperparameter that sets the upper limit for outlier scores. Represents the vector dimension.

2. The online monitoring method for a wastewater treatment process according to claim 1, characterized in that, The preprocessing of the training set includes: Using the standard deviation standardization method, the data matrix observed from the wastewater treatment plant is preprocessed according to the following formula: in, It is the original observation data matrix. express m OK n A column of real matrices, Each row represents an observed sample, and each column represents an observed variable. and express The sample mean and sample standard deviation of each observed variable in the sample. This represents the preprocessed training set.

3. The online monitoring method for a wastewater treatment process according to claim 1, characterized in that, The preprocessing of the training set includes: Using range standardization, the data matrix observed from the wastewater treatment plant is preprocessed according to the following formula: in, It is the original observation data matrix. express m OK n A column of real matrices, Each row represents an observed sample, and each column represents an observed variable. This represents the preprocessed training set. , They represent and The Line 1 The element at the column position, , They represent The The maximum and minimum values ​​in the column.

4. The online monitoring method for a wastewater treatment process according to claim 1, characterized in that, The step of determining the autoencoder network structure based on training set features and setting the loss function includes: Based on the preprocessed training set Determine the number of hidden layers in the stacked autoencoder, as well as the number of neurons in the input, hidden, and output layers; Based on the preprocessed training set The loss function and convergence model are selected, and during training, the autoencoder loss function is chosen as follows: The regularization loss function is expressed as: in This represents a variable in a specific observation sample within the preprocessed training set. This represents the output reconstructed by the autoencoder from the input. m The total number of observed samples in the training set. n This represents the total number of variables in the observed sample.

5. The online monitoring method for a wastewater treatment process according to claim 1, characterized in that, The control limit is expressed as: in, and These are the upper and lower quartiles of the predicted score set, respectively.

6. The online monitoring method for a wastewater treatment process according to claim 5, characterized in that, The determination of whether a malfunction has occurred in the wastewater treatment process includes: During the online monitoring phase, the predicted scores of a class of support vector machines are... Control limits determined by box plot Difference: According to the theory of box plots, when Less than A minor fault was determined to have occurred. Greater than A serious malfunction has been determined.

7. The online monitoring method for a wastewater treatment process according to claim 6, characterized in that, The preprocessing method for individual observation data samples acquired in real time is consistent with the preprocessing method for the training set; When the training set uses standard deviation standardization, individual data samples are standardized using the mean and variance of the training set; when the training set uses range standardization, if a variable value in an individual data sample is greater than or less than the extreme value, the variable value should be set to 1 or 0 accordingly.

8. An online monitoring device for wastewater treatment processes, characterized in that, include: At least one processor; At least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the method of any one of claims 1-7.

9. A computer-readable storage medium storing a processor-executable program, characterized in that, The processor-executable program, when executed by the processor, is used to perform the method as described in any one of claims 1-7.

Images