Fault Detection Method and Detection System in Sewage Treatment Process

Abstract

The present invention provides a fault detection method in the sewage treatment process, a fault detection system in the sewage treatment process, and a storage medium. The fault detection method in the sewage treatment process includes the following steps: obtaining the process data of sewage treatment and a query sample to be predicted; determining N1 fault feature variables in the sewage treatment process according to the mechanism of the sewage treatment process and a first preset quantity N1; establishing N2 slow feature analysis models according to the process data, the fault feature variables, and a second preset quantity N2; and determining the fault of the query sample to be predicted through ensemble learning according to the N2 slow feature analysis models and the query sample to be predicted.

Description

Technical Field

[0001] The present invention relates to the technical field of sewage treatment, and in particular to a fault detection method in the sewage treatment process, a fault detection system in the sewage treatment process, and a corresponding computer-readable storage medium. Background Art

[0002] With the development of social economy, the total amount of urban sewage discharge is continuously increasing. Coupled with the increasingly serious water pollution, the per capita water resources are decreasing year by year. Urban sewage treatment is an important way to improve the recycling of water resources and is closely related to the sustainable development of water resources and environmental protection. Therefore, the urban sewage treatment process has received extensive attention. The urban sewage treatment process can be regarded as a complex non-linear system. There are various biochemical and physical reactions in this process, accompanied by large flow and load disturbances. The nature of the sewage treatment process determines that it needs to work continuously, and at the same time, the treatment standards are becoming more and more strict. In order to ensure the safe operation of the urban sewage treatment process and make the discharged products meet the sewage discharge standards, it is essential to monitor the urban sewage treatment process. In addition, once a failure occurs in the sewage treatment plant, it will cause serious personnel and economic losses. Therefore, it is of great significance to detect faults in the urban sewage treatment process as early as possible.

[0003] However, the data variables collected in large-scale processes are numerous and have a high dimension. It is difficult to monitor numerous variables by only establishing a global model using traditional process monitoring methods, and the range and degree of variables affected by different faults are also different.

[0004] In order to overcome the above-mentioned defects existing in the prior art, there is an urgent need in the art for a fault detection method in the sewage treatment process, which is used to monitor numerous variables and a dynamic and large-scale process to determine the type of faults in the sewage treatment process, so as to improve the accuracy and reliability of predicting faults in the sewage treatment process. Summary of the Invention

[0005] The following gives a brief overview of one or more aspects to provide a basic understanding of these aspects. This overview is not an exhaustive survey of all contemplated aspects, and is neither intended to identify key or decisive elements of all aspects nor to attempt to define the scope of any or all aspects. Its sole purpose is to present some concepts of one or more aspects in a simplified form as a prelude to a more detailed description to follow.

[0006] In order to overcome the above-mentioned defects existing in the prior art, the present invention provides a fault detection method in the sewage treatment process, a fault detection device in the sewage treatment process and its corresponding computer-readable storage medium, which monitor numerous variables and monitor dynamic and large-scale processes to determine the type of fault in the sewage treatment process, thereby improving the accuracy and reliability of predicting faults in the sewage treatment process.

[0007] Specifically, the above-mentioned fault detection method in the sewage treatment process provided by the first aspect of the present invention includes the following steps: obtaining the process data of sewage treatment and the query sample to be predicted; determining N1 fault feature variables in the sewage treatment process according to the mechanism of the sewage treatment process and the first preset quantity N1; establishing N2 slow feature analysis models according to the process data, the fault feature variables and the second preset quantity N2. According to the N2 slow feature analysis models and the query sample to be predicted, the fault of the query sample to be predicted is determined through ensemble learning.

[0008] Further, in some embodiments of the present invention, the step of establishing N2 slow feature analysis models according to the process data, the fault feature variables and the second preset quantity N2 includes: dividing the process data of sewage treatment into a training set and a test set. In the training set, N1 training variables are determined, and data preprocessing is performed on the N1 training variables. Slow feature modeling is performed on the preprocessed training variables.

[0009] Further, in some embodiments of the present invention, the step of performing data preprocessing on the N1 training variables includes: performing whitening processing on the N1 training variables to eliminate the correlation between the training variables.

[0010] Further, in some embodiments of the present invention, the step of determining the fault of the query sample to be predicted through ensemble learning according to the N2 slow feature analysis models and the query sample to be predicted includes: determining the corresponding slow features according to the slow feature analysis models. According to the slow features, the first statistical control limit and the second statistical control limit are determined through the kernel density estimation method, wherein the first statistical control limit is used to characterize whether the query sample has the corresponding first fault, and the second statistical control limit is used to characterize whether the query sample has the corresponding second fault. The test set is input into the slow feature analysis model, and according to the first statistical control limit and the second statistical control limit, the false alarm rate and the detection rate of the slow feature analysis model are determined. According to the slow feature analysis model, the false alarm rate corresponding to the slow feature analysis model and the accuracy rate corresponding to the slow feature analysis model, at least two base learners are determined. According to the at least two base learners, the fault of the test sample is determined.

[0011] Further, in some embodiments of the present invention, the step of inputting the test set into the slow feature analysis model to determine the false alarm rate and detection rate of the slow feature analysis model includes: According to the numerical relationship between the corresponding first statistic input according to the test set and the control limit of the first statistic, and the numerical relationship between the corresponding second statistic and the control limit of the second statistic, to determine the number of false alarm samples and normal samples in the test set. According to the number of normal samples in the process data of the sewage treatment and the number of samples misreported by the slow feature analysis model, determine the false alarm rate. According to the number of fault samples in the process data of the sewage treatment and the number of faults successfully detected by the slow feature analysis model, determine the monitoring rate.

[0012] Further, in some embodiments of the present invention, the step of determining the control limit of the first statistic and the control limit of the second statistic according to the slow feature and by kernel density estimation includes: According to the corresponding first statistic and second statistic, determine the cumulative distribution function of the first statistic and the second statistic. According to the cumulative distribution function, use the kernel density estimation method to determine the control limit of the first statistic and the control limit of the second statistic.

[0013] Further, in some embodiments of the present invention, the step of determining the fault of the test sample according to the at least two base learners includes: In response to the first statistic being greater than the control limit of the first statistic, determine that the first fault occurs in the query sample to be predicted. And in response to the second statistic being greater than the control limit of the second statistic, determine that the second fault occurs in the query sample to be predicted.

[0014] Further, in some embodiments of the present invention, the first statistic is a statistic for measuring the change in the space spanned by the slow feature, the control limit of the first statistic is a control limit of the statistic for measuring the change in the space spanned by the slow feature, the second statistic is a statistic for describing the dynamics of the slow feature during the operation process, the control limit of the second statistic is a control limit of the statistic for describing the dynamics of the slow feature during the operation process, and the step of determining the fault of the test sample according to the at least two base learners further includes: In response to the first statistic being greater than the control limit of the first statistic, determine that the steady-state characteristic fault of the sewage treatment process occurs in the query sample to be predicted. And in response to the second statistic being greater than the control limit of the second statistic, determine that the dynamic characteristic fault of the sewage treatment process occurs in the query sample to be predicted.

[0015] Further, in some embodiments of the present invention, the step of determining at least two base learners according to the slow feature analysis model, the false alarm rate corresponding to the slow feature analysis model, and the accuracy rate corresponding to the slow feature analysis model includes: performing clustering analysis on each of the slow feature analysis models to obtain a plurality of different slow feature analysis model classes. Selecting the slow feature analysis model with the lowest false alarm rate in each of the slow feature analysis model classes as the base learner.

[0016] Further, in some embodiments of the present invention, the step of determining the fault of the query sample to be predicted according to the at least two base learners further includes: performing Bayesian inference according to the fault detection results of each of the slow feature analysis models to determine a fusion statistical index. Determining the fault of the query sample to be predicted according to the numerical relationship between the fusion statistical index and a preset fusion statistical index threshold.

[0017] Further, in some embodiments of the present invention, the step of performing Bayesian inference according to the fault detection results of each of the slow feature analysis models to determine a fusion statistical index includes: determining the fusion statistical index according to the prior probabilities, confidence levels, and conditional probabilities of the normal operating state and the fault operating state of the process data of the sewage treatment.

[0018] In addition, the fault detection system in the sewage treatment process provided in the second aspect of the present invention includes a memory and a processor. The processor is connected to the memory and is configured to implement the fault detection method in the sewage treatment process provided in the first aspect of the present invention.

[0019] In addition, the computer-readable storage medium provided in the third aspect of the present invention stores computer instructions thereon. When the computer instructions are executed by a processor, the fault detection method in the sewage treatment process provided in the first aspect of the present invention is implemented. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] After reading the detailed description of the embodiments of the present disclosure in conjunction with the following drawings, the above features and advantages of the present invention can be better understood. In the drawings, the components are not necessarily drawn to scale, and components with similar related characteristics or features may have the same or similar reference numerals.

[0021] Figure 1 Shows an architecture diagram of a fault detection device in a sewage treatment process provided by some embodiments of the present invention;

[0022] Figure 2 Shows a flowchart of a fault detection method in a sewage treatment process provided by some embodiments of the present invention;

[0023] Figures 3A to 3C Shows the schematic diagram of the prediction results of the fault detection method in the sewage treatment process provided according to some embodiments of the present invention according to different statistics S 2 , T 2 , ES 2 , ET 2 . Detailed implementation manners

[0024] The following specific embodiments illustrate the implementation manners of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. Although the description of the present invention will be introduced in conjunction with the preferred embodiments, this does not mean that the features of this invention are limited to this implementation manner. On the contrary, the purpose of introducing the invention in conjunction with the implementation manner is to cover other alternatives or modifications that may extend based on the claims of the present invention. In order to provide a deep understanding of the present invention, many specific details will be included in the following description. The present invention can also be implemented without using these details. In addition, in order to avoid confusing or obscuring the key points of the present invention, some specific details will be omitted in the description.

[0025] In the description of the present invention, it should be noted that unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected, or indirectly connected through an intermediate medium, and it can be the communication inside two components. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific situations.

[0026] In addition, the "upper", "lower", "left", "right", "top", "bottom", "horizontal", and "vertical" used in the following description should be understood as the orientations shown in this section and the related drawings. This relative term is only for the convenience of description, and it does not mean that the device described needs to be manufactured or operated in a specific orientation, so it should not be understood as a limitation to the present invention.

[0027] It can be understood that although the terms "first", "second", "third", etc. can be used here to describe various components, regions, layers, and / or parts, these components, regions, layers, and / or parts should not be limited by these terms, and these terms are only used to distinguish different components, regions, layers, and / or parts. Therefore, the first component, region, layer, and / or part discussed below can be referred to as the second component, region, layer, and / or part without departing from some embodiments of the present invention.

[0028] The urban sewage treatment process can be regarded as a complex non - linear system. During this process, there are various biochemical and physical reactions, accompanied by large flow and load disturbances. The nature of the sewage treatment process determines that it needs to operate continuously, and at the same time, the treatment standards are becoming increasingly strict. To ensure the safe operation of the urban sewage treatment process and make the discharged products meet the sewage discharge standards, it is essential to monitor the urban sewage treatment process. In addition, once a failure occurs in the sewage treatment plant, it will cause serious personnel and economic losses. Therefore, it is of great significance to detect faults in the urban sewage treatment process as early as possible. However, the data collected from large - scale processes have numerous variables and high dimensions. It is difficult for traditional process monitoring methods that only establish a global model to monitor numerous variables, and the ranges and degrees of variables affected by different faults also vary.

[0029] To overcome the above - mentioned defects existing in the prior art, the present invention provides a fault detection method in a sewage treatment process, a fault detection device in a sewage treatment process, and its corresponding computer - readable storage medium, which are used to monitor numerous variables and monitor dynamic and large - scale processes to determine the fault types in the sewage treatment process, thereby improving the accuracy and reliability of predicting faults in the sewage treatment process.

[0030] In some non - restrictive embodiments, the above - mentioned fault detection method in a sewage treatment process provided by the first aspect of the present invention can be implemented by the above - mentioned fault detection device in a sewage treatment process provided by the second aspect of the present invention. Specifically, the above - mentioned fault detection device in a sewage treatment process is configured with a memory and a processor. The memory includes, but is not limited to, the above - mentioned computer - readable storage medium provided by the third aspect of the present invention, on which computer instructions are stored. The processor is connected to the memory and is configured to execute the computer instructions stored on the memory to implement the fault detection method in a sewage treatment process provided by the first aspect of the present invention.

[0031] First, please refer to Figure 1 , Figure 1 which shows the architecture diagram of a pollutant measurement device provided by some embodiments of the present invention.

[0032] Figure 1The architecture diagram of a fault detection device in a sewage treatment process provided according to some embodiments of the present invention is shown. The fault detection device in the sewage treatment process includes an internal communication bus 301, a processor 302, a read-only memory (ROM) 303, a random access memory (RAM) 304, a communication port 305, and a hard disk 307. The internal communication bus 301 can realize data communication between components of the fault detection device in the sewage treatment process. The processor 302 can make judgments and give prompts. In some embodiments, the processor 302 can be composed of one or more processors. The communication port 305 can realize data transmission and communication between the sewage index measurement device and external input / output devices. In some embodiments, the fault detection device in the sewage treatment process can send and receive information and data from the network through the communication port 305. In some embodiments, the fault detection device in the sewage treatment process can perform data transmission and communication with external input / output devices in a wired form through the input / output terminal 306. The fault detection device in the sewage treatment process also includes program storage units and data storage units in different forms, such as the hard disk 307, the read-only storage (ROM) 303, and the random access memory (RAM) 304, which can store various data files used for computer processing and / or communication, as well as possible program instructions executed by the processor 302. The processor 302 executes these instructions to implement the main part of the method. The result processed by the processor 302 is transmitted to an external output device through the communication port 305 and displayed on the user interface of the output device.

[0033] The working principle of the above-mentioned fault detection device in the sewage treatment process will be described below in conjunction with some embodiments of the fault detection method in the sewage treatment process. Those skilled in the art can understand that these embodiments of the communication method are only some non-limiting implementation manners provided by the present invention, aiming to clearly show the main concept of the present invention and provide some specific solutions convenient for the public to implement, rather than limiting all functions or all working modes of the fault detection device in the sewage treatment process. Similarly, the fault detection device in the sewage treatment process is also a non-limiting implementation manner provided by the present invention and does not limit the execution subject of each step in these fault detection methods in the sewage treatment process.

[0034] Please refer to Figure 2 , Figure 2 which shows the flowchart of the fault detection method in the sewage treatment process provided according to some embodiments of the present invention.

[0035] As Figure 2As shown in step S1, during the process of detecting faults in the sewage treatment process, the fault detection method for the sewage treatment process can first obtain the process data of the sewage treatment and the query samples to be predicted. After that, as Figure 2 shown in step S2, the present invention can determine N1 fault feature variables in the sewage treatment process according to the mechanism of the sewage treatment process and the first preset quantity N1. Then, as Figure 2 shown in step S3, the present invention can establish N2 slow feature analysis models according to the process data, the fault feature variables, and the second preset quantity N2. Then, the present invention can determine the faults of the query samples to be predicted through ensemble learning according to the N2 slow feature analysis models and the query samples to be predicted.

[0036] Optionally, in some embodiments of the present invention, those skilled in the art can obtain the process data of the sewage treatment by installing sensors for measuring the process data of the above sewage treatment, and establish a database of the process data of the sewage treatment including the historical data.

[0037] Those skilled in the art can understand that the sensors for the process data of the above sewage treatment are only a non-limiting implementation manner provided by the present invention, aiming to measure the process data of the above sewage treatment rather than limiting the protection scope of the present invention.

[0038] Further, after obtaining the process data of the sewage treatment and the query samples to be predicted. The present invention can determine N1 fault feature variables in the above sewage treatment process according to experience, the mechanism of the sewage treatment process, and the first preset quantity N1.

[0039] Further, after determining the fault feature variables in the above sewage treatment process, the process data of the sewage treatment can be divided into a training set and a test set.

[0040] Specifically, in some embodiments of the present invention, the present invention can collect samples (a total of 672) of the sewage treatment process running for 7 days under normal weather as the training data X tr , and the test data is samples of the sewage treatment process running for 14 days under normal weather. Starting from the seventh day, that is, the 673rd sample, 3 kinds of faults are introduced until the end. The three kinds of faults are: Fault 1: The maximum growth rate of autotrophic bacteria in the aerobic reaction tank has a step change from 0.5 to 0.3; Fault 2: The oxygen transfer coefficient of bioreactor 4 drops to half of the original; Fault 3: The SNO_2 sensor of bioreactor 2 has an offset of +0.5. A total of 1344 samples X v .

[0041] Further, after dividing the training set and the test set, the present invention can determine N1 training variables in the training set and perform data preprocessing on the N1 training variables.

[0042] Specifically, the data preprocessing can be: performing whitening processing on the N1 training variables to eliminate the correlation between the training variables. The whitening processing can be performed by singular value decomposition (SVD) to obtain the expected value B = <xx of the covariance matrix representing the input vector x Τ > t , and performing singular value decomposition on it:

[0043] B = UΛU Τ , where Λ is a diagonal matrix composed of eigenvalues, and U is a matrix composed of the corresponding eigenvectors.

[0044] Obtain the whitened data z:

[0045] z = Λ -1 / 2 U Τ x = Qx

[0046] where the whitened data z satisfies cov(z) = <zz Τ > = I p , I p is the identity matrix, and Q = Λ -1 / 2 U Τ is the whitening matrix.

[0047] Further, after performing data preprocessing on the above training variables, the present invention can perform slow feature modeling on the preprocessed training variables to establish N2 slow feature analysis models.

[0048] Specifically, the SFA method (slow feature analysis method) is a linear method, which assumes that there is a correlation between the input variables and the output variables. Therefore, the extracted slow feature s j (t) can be regarded as a linear combination of a series of input signals The expression of the extracted slow feature is

[0049] s(t) = Wx(t)

[0050] where W = [w1, w2,... w p Τ is the weight matrix;

[0051] The optimization objective of the SFA method (slow feature analysis method) is simplified to obtain a matrix P that satisfies the s = Pz transformation, and at the same time satisfies ss Τ = I p ​, Correspondingly, in order to extract slow features, perform singular value decomposition on the covariance matrix of the first derivative of the whitened data z above:

[0052]

[0053] Thus, obtain the coefficient matrix W and the slow feature s. The expressions of the coefficient matrix W and the slow feature are respectively:

[0054] W = PQ

[0055] s = Wx

[0056] Furthermore, after establishing N2 slow feature analysis models, the present invention can determine the faults of the query sample to be predicted according to the above N2 slow feature analysis models and the above query sample to be predicted through ensemble learning. The present invention can determine the corresponding slow features according to the slow feature analysis model. Then, according to this slow feature, determine the first statistic control limit and the second statistic control limit through the kernel density estimation method, wherein the first statistic control limit is used to characterize whether the query sample has the corresponding first fault, and the second statistic control limit is used to characterize whether the query sample has the corresponding second fault. Then, input the above test set into this slow feature analysis model, and determine the false alarm rate and detection rate of this slow feature analysis model according to this first statistic control limit and this second statistic control limit. Determine at least two base learners according to the slow feature analysis model, the false alarm rate corresponding to the slow feature analysis model, and the accuracy rate corresponding to the slow feature analysis model. Then, determine the faults of the test sample according to the at least two base learners.

[0057] Specifically, the present invention can use the statistic T 2 as the first statistic, S 2 as the second statistic, as the first statistic control limit, as the second statistic control limit. The present invention can construct N SFA models, obtain the corresponding transformation matrices and use kernel density estimation to obtain the control limits (thresholds) of the statistics and

[0058] The expressions of the statistic T 2 and S 2 are respectively:

[0059] T 2 = s Τ s

[0060]

[0061] wherein, Ω is the covariance matrix of .

[0062] Accordingly, the statistic T 2 and S 2 The control limits are obtained by calculating their cumulative distribution functions. When the cumulative distribution reaches θ (θ is usually set to 95% or 99%), the corresponding value is used as the threshold and Kernel Density Estimation (KDE), as a non-parametric method, is often used to infer the distribution characteristics of data. Its mathematical description is:

[0063]

[0064] where y is the data to be estimated, y i is the observed value of the process data, n is the number of samples, and h is the smoothing parameter. There are many choices for the kernel function K. In this embodiment, the most commonly used Gaussian function is selected:

[0065]

[0066] After training is completed, when new query data arrives, the statistic is calculated. If the first statistic T 2 is greater than the control limit of the first statistic, it is considered that the first fault has occurred in the process. Otherwise, it is operating under normal conditions. Similarly, when the second statistic S 2 is greater than the control limit of the second statistic it is considered that the second fault has occurred in the process. Otherwise, it is operating under normal conditions

[0067] Preferably, in some embodiments of the present invention, the above-mentioned first statistic is a statistic that measures the change in the space spanned by slow features, the above-mentioned first statistic control limit is a statistic control limit that measures the change in the space spanned by slow features, the above-mentioned second statistic is a statistic that describes the dynamics of slow features during the operation process, the above-mentioned second statistic control limit is a statistic control limit that describes the dynamics of slow features during the operation process, the above-mentioned first fault is a steady-state characteristic fault of the process, and the above-mentioned second fault is a dynamic characteristic fault of the process.

[0068] Further, in some embodiments of the present invention, after inputting the above test set into the slow feature analysis model to determine the faults of the test samples, the present invention can determine the false alarm rate and detection rate of the above slow feature analysis model. The present invention can determine the number of false alarm samples and normal samples in the test set according to the numerical relationship between the first statistic corresponding to the input of the above test set and the control limit of the first statistic, and the numerical relationship between the corresponding second statistic and the control limit of the second statistic. According to the number of normal samples in the process data of the above sewage treatment and the number of samples misreported by the slow feature analysis model, the false alarm rate is determined. After that, the present invention can determine the monitoring rate according to the number of fault samples in the process data of the sewage treatment and the number of samples with faults successfully detected by the slow feature analysis model.

[0069] Specifically, the present invention can perform fault detection on each SFA model based on the test set X v to obtain the false alarm rate FAR and the detection rate FDR. Among them, the calculation formulas for the false alarm rate (False Alarm Rate, FAR) and the detection rate (Fault Detection Rate, FDR) are respectively:

[0070]

[0071]

[0072] where, N n and N f respectively represent the number of normal samples and fault samples, and N fa and N fd respectively represent the number of misreported samples and the number of samples with faults successfully detected.

[0073] Further, after determining the slow feature analysis model, the false alarm rate corresponding to the slow feature analysis model, and the accuracy rate, the present invention can determine at least two base learners according to the slow feature analysis model, the false alarm rate corresponding to the slow feature analysis model, and the accuracy rate corresponding to the slow feature analysis model. The present invention can perform clustering analysis on each of the above slow feature analysis models to obtain multiple different slow feature analysis model classes. After that, the slow feature analysis model with the lowest false alarm rate in each of the above slow feature analysis model classes is selected as the base learner.

[0074] Specifically, the present invention can use the DHCA algorithm to cluster the above N2 SFA (Slow Feature Analysis) models, and select the model with the lowest false alarm rate in each class as the base learner. The specific steps of the DHCA algorithm are as follows: Step 1: Set the parameters of the DHCA algorithm. The DHCA algorithm is solved iteratively, and the termination condition of clustering needs to be set. The termination condition is to stop when a maximum number of clusters is set manually or the intra-class similarity reaches a threshold; Step 2: Place all objects in one class, that is, the top layer of the hierarchy; Step 3: For each SFA model, find two objects with the farthest distance as the centers of the new classes, and then assign the remaining points to the clusters with the closer class centers respectively; Step 4: Perform the classification in step 503 for each class generated in step 3, and repeat until the termination condition is reached.

[0075] Further, after selecting the base learner, the present invention can use the test set X for each base learner v to perform tests and calculate the statistics T 2 and S 2 .

[0076] T 2 and S 2 The statistics are as follows:

[0077] where i represents the i-th base learner, and T is the number of base learners.

[0078] Further, after selecting the base learner and determining the first statistic T 2 corresponding to each base learner and the second statistic S 2 after that, the present invention can perform Bayesian inference based on the fault detection results of each of the above slow feature analysis models to determine the fusion statistical index. After that, the present invention can determine the fault of the query sample to be predicted according to the numerical relationship between the fusion statistical index and the preset fusion statistical index threshold. Here, the above fusion statistical index can be determined according to the prior probability of the normal operation state, the prior probability of the fault operation state, the confidence level, and the conditional probability of the process data of the above sewage treatment.

[0079] Specifically, the present invention can use Bayesian inference to fuse the fault detection results of different SFA models into the comprehensive statistical indexes ET 2 and ES 2 to determine the fault of the query sample to be predicted. The present invention can define the prior probabilities of the normal and fault operation states of the process as and and set their values as the confidence levels α and (1 - α). Then, obtain the conditional probabilities and The conditional probabilities and The expressions are as follows:

[0080]

[0081]

[0082] Wherein, and are the monitoring statistic and the corresponding control limit (threshold) of the i-th base learner;

[0083] Derived according to Bayes' theorem The calculation formula of is:

[0084]

[0085] Wherein, represents the new probability, and the conditional probabilities and and the prior probability are respectively

[0086] are and

[0087] Therefore, the monitoring statistic can be converted into the probability of failure occurrence, and the expression of the probability of failure occurrence is:

[0088]

[0089] Thus, the expression of the final fusion statistical index is:

[0090]

[0091] Similarly, the expression of another fusion statistical index is:

[0092]

[0093] The thresholds of the fusion statistical indices ET 2 and ES 2 are both set to (1-α), that is, the prior probability of the process occurring in a faulty condition. When the comprehensive statistical index ET 2 or ES 2 is higher than its corresponding control limit (threshold), it is considered that the steady-state characteristic or the dynamic characteristic of the process has failed. Otherwise, it is considered that the process is operating under normal conditions. In this way, the present invention can monitor numerous variables and monitor dynamic and large-scale processes to determine the type of failure in the sewage treatment process, thereby improving the accuracy and reliability of predicting failures in the sewage treatment process

[0094] Optionally, the control limit (threshold) of the above fusion statistical index may be a threshold set according to the mechanism of sewage treatment and human experience.

[0095] Please refer to Figures 3A to 3C , Figures 3A to 3C shows the schematic diagram of the predicted results of the fault detection method in the sewage treatment process according to some embodiments of the present invention according to different statistics S 2 , T 2 , ES 2 , ET 2 .

[0096] Optionally, in some embodiments of the present invention, the present invention can perform fault detection on the BSM1 benchmark simulation model for the sewage treatment process. The BSM1 benchmark simulation model is a sewage treatment process model proposed by the European Union's Cooperation in Science and Technology (COST) and is widely used in the research of sewage treatment processes by personnel and institutions around the world. It includes two models, namely the activated sludge process model and the double exponential sedimentation model. The BSM1 benchmark simulation model mainly focuses on removing carbon and nitrogen in sewage, so that the treated wastewater can reach the secondary discharge standard. The present invention can classify the faults of this model into two categories. The first type of fault is that the maximum growth rate of autotrophic bacteria in the first aerobic reactor has a step change, decreasing from 0.5 to 0.3. Since the life activities of autotrophic bacteria are weakened, it affects the complex biochemical reactions in the urban sewage treatment process, thereby changing the state of the entire system. The second type of fault is that the oxygen transfer coefficient of the biological reaction tank 4 drops to half of the original. The oxygen transfer coefficient has a great relationship with the dissolved oxygen concentration in the reaction tank, and the dissolved oxygen concentration is an important element in the reaction of the urban sewage treatment process, which is related to the sewage treatment effect. As Figures 3A to 3C shown, using the above method and combining ET 2 and ES 2 of the fusion unified index has the best detection accuracy. For this model, using the fault detection method for the sewage treatment project provided by the first aspect of the present invention has better detection performance than the traditional slow feature analysis method and can detect faults in a timely manner.

[0097] Although the above methods are illustrated and described as a series of actions for simplicity of explanation, it should be understood and appreciated that these methods are not limited by the order of the actions, because according to one or more embodiments, some actions may occur in a different order and / or concurrently with other actions not illustrated and described herein but understood by those skilled in the art.

[0098] Those skilled in the art will further appreciate that the various illustrative logical blocks, modules, circuits, and algorithm steps described in connection with the embodiments disclosed herein can be implemented as electronic hardware, computer software, or a combination of both. To clearly illustrate this interchangeability of hardware and software, the various illustrative components, blocks, modules, circuits, and steps are described above in terms of their functionality. Whether such functionality is implemented as hardware or software depends upon the particular application and the design constraints imposed on the overall system. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.

[0099] Although the sewage index measurement system described in the above embodiments can be implemented by a combination of software and hardware. However, it can be understood that the sewage index measurement system can also be implemented in software or hardware. For hardware implementation, the sewage index measurement system can be implemented in one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), processors, controllers, microcontrollers, microprocessors, other electronic devices for performing the above functions, or a selected combination of the above devices. For software implementation, the sewage index measurement system can be implemented by independent software modules such as procedures and functions running on a general-purpose chip, where each module performs one or more of the functions and operations described herein.

[0100] The steps of a method or algorithm described in connection with the embodiments disclosed herein can be embodied directly in hardware, in a software module executed by a processor, or in a combination of both. The software modules can reside in RAM memory, flash memory, ROM memory, EPROM memory, EEPROM memory, registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor such that the processor can read from, and write to, the storage medium. In an alternative, the storage medium may be integrated into the processor. The processor and the storage medium may reside in an ASIC. The ASIC may reside in a user terminal. In an alternative, the processor and the storage medium may reside as discrete components in a user terminal.

[0101] In one or more exemplary embodiments, the described functionality may be implemented in hardware, software, firmware, or any combination thereof. If implemented in software as a computer program product, the functions may be stored on or transmitted via a computer-readable medium as one or more instructions or code. The computer-readable medium includes both computer storage media and communication media including any medium that facilitates transfer of a computer program from one place to another. The storage media may be any available media that can be accessed by a computer. By way of example and not limitation, such computer-readable media can comprise RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer. Any connection is properly termed a computer-readable medium. For example, if the software is transmitted from a web site, server, or other remote source using a coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, and microwave, then the coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, and microwave are included in the definition of medium. As used herein, the disk and disc include compact disc (CD), laser disc, optical disc, digital versatile disc (DVD), floppy disk, and Blu-ray disc where the disk typically reproduces the data magnetically, while the disc reproduces the data optically with a laser. Combinations of the above should also be included within the scope of computer-readable media.

[0102] The foregoing description of the disclosure has been provided to enable any person skilled in the art to make or use the disclosure. Various modifications to the disclosure will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other variations without departing from the spirit or scope of the disclosure. Thus, the disclosure is not intended to be limited to the examples and designs described herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A fault detection method in a sewage treatment process, characterized in that, The method includes the following steps: Obtain the process data of sewage treatment and the query samples to be predicted; Determine N1 fault feature variables in the sewage treatment process according to the mechanism of the sewage treatment process and the first preset quantity N1; Divide the process data of the sewage treatment into a training set and a test set, determine N1 training variables in the training set, and perform data preprocessing on the N1 training variables; Perform slow feature modeling on the preprocessed training variables to establish N2 slow feature analysis models according to the process data, the fault feature variables, and the second preset quantity N2; Determine the corresponding slow features according to the slow feature analysis models; Determine the first statistic control limit and the second statistic control limit according to the slow features and by using the kernel density estimation method, wherein the first statistic control limit is used to characterize whether the query sample has the corresponding first fault, and the second statistic control limit is used to characterize whether the query sample has the corresponding second fault; Determine the number of false alarm samples and normal samples in the above test set according to the numerical relationship between the corresponding first statistic and the first statistic control limit and the numerical relationship between the corresponding second statistic and the second statistic control limit input by the test set; Determine the false alarm rate according to the number of normal samples in the process data of the sewage treatment and the number of samples misreported by the slow feature analysis model; Determine the detection rate according to the number of fault samples in the process data of the sewage treatment and the number of samples with faults successfully detected by the slow feature analysis model; Determine at least two base learners according to the slow feature analysis model, the false alarm rate corresponding to the slow feature analysis model, and the detection rate corresponding to the slow feature analysis model; and Determine the faults of the test samples according to the at least two base learners.

2. The fault detection method according to claim 1, wherein The step of performing data preprocessing on the N1 training variables includes: Perform whitening processing on the N1 training variables to eliminate the correlation between the training variables.

3. The fault detection method according to claim 1, wherein The step of determining the first statistic control limit and the second statistic control limit according to the slow features and by using kernel density estimation includes: Determine the cumulative distribution functions of the first statistic and the second statistic according to the corresponding first statistic and second statistic; and Determine the first statistic control limit and the second statistic control limit by using the kernel density estimation method according to the cumulative distribution functions.

4. The fault detection method according to claim 3, characterized in that The step of determining the faults of the test samples according to the at least two base learners includes: In response to the first statistic being greater than the first statistic control limit, determine that the test sample has the first fault; and In response to the second statistic being greater than the second statistic control limit, determine that the test sample has the second fault.

5. The fault detection method according to claim 4, characterized in that The first statistic is a statistic that measures the change in the space spanned by slow features, the first statistic control limit is a statistic control limit that measures the change in the space spanned by slow features, the second statistic is a statistic that describes the dynamics of slow features during the operation process, the second statistic control limit is a statistic control limit that describes the dynamics of slow features during the operation process, and the step of determining the fault of the test sample to be predicted according to the at least two base learners further includes: In response to the first statistic being greater than the first statistic control limit, determining that the query sample to be predicted has a steady-state characteristic fault in the sewage treatment process; and In response to the second statistic being greater than the second statistic control limit, determining that the query sample to be predicted has a dynamic characteristic fault in the sewage treatment process.

6. The fault detection method according to claim 4, wherein The step of determining at least two base learners according to the slow feature analysis model, the false alarm rate corresponding to the slow feature analysis model, and the accuracy rate corresponding to the slow feature analysis model includes: Performing clustering analysis on each of the slow feature analysis models to obtain multiple different slow feature analysis model classes; and Selecting the slow feature analysis model with the lowest false alarm rate in each of the slow feature analysis model classes as the base learner.

7. The fault detection method according to claim 6, wherein The step of determining the fault of the query sample to be predicted according to the at least two base learners further includes: Performing Bayesian inference based on the fault detection results of each of the slow feature analysis models to determine a fusion statistical index; and Determining the fault of the query sample to be predicted according to the numerical relationship between the fusion statistical index and a preset fusion statistical index threshold.

8. The fault detection method according to claim 7, characterized in that The step of performing Bayesian inference based on the fault detection results of each of the slow feature analysis models to determine a fusion statistical index includes: Determining the fusion statistical index according to the prior probabilities, confidence levels, and conditional probabilities of the normal operation state and the fault operation state of the process data of the sewage treatment.

9. A fault detection system in a sewage treatment process, comprising: A memory; And A processor, the processor is connected to the memory and is configured to implement the fault detection method in the sewage treatment process according to any one of claims 1 to 8.

10. A computer-readable storage medium having computer instructions stored thereon, characterized in that, When the computer instructions are executed by the processor, the fault detection method in the sewage treatment process according to any one of claims 1 to 8 is implemented.

Images