Sewage treatment equipment operation monitoring system and method based on artificial intelligence

Abstract

The invention discloses a sewage treatment equipment operation monitoring system and method based on artificial intelligence, and relates to the technical field of sewage treatment equipment monitoring. The invention relates to a sewage treatment equipment monitoring system based on artificial intelligence. The system comprises a sensor module, a feature fusion module, a self-adaptive multi-modal feature fusion fault prediction model and a real-time monitoring and early warning module. The sensor module is used for carrying out processing and data transmission on the data collected by the equipment; the feature fusion module is used for carrying out feature selection and extraction on the transmitted data and carrying out feature fusion; the adaptive multi-modal feature fusion fault prediction model is used for automatically adjusting a feature fusion weight according to real-time data and historical data and carrying out fault probability calculation on equipment; and the real-time monitoring and early warning module is used for monitoring whether the equipment has an abnormal working state or not through the analyzed real-time data and fault prediction, and giving an early warning to the system according to the abnormal state condition of the system.

Description

Technical Field

[0001] The present invention relates to the technical field of sewage treatment equipment operation monitoring, and in particular to a sewage treatment equipment monitoring system and method based on artificial intelligence. Background Art

[0002] With the increasing awareness of environmental protection, sewage treatment equipment plays an important role in the field of environmental protection. However, traditional sewage treatment monitoring systems often rely on manual inspections and manual data recording, which is inefficient; and traditional monitoring systems usually lack effective fault diagnosis tools, making it difficult to accurately determine the cause of equipment failures; at present, many monitoring systems have not yet fully integrated artificial intelligence technology, and the level of intelligence is not high, which cannot achieve intelligent data analysis and decision support.

[0003] Therefore, exploring and introducing a more intelligent and efficient sewage treatment monitoring system is of great significance to the current sewage treatment monitoring equipment. Summary of the invention

[0004] The purpose of the present invention is to provide a sewage treatment equipment operation monitoring system based on artificial intelligence, and an adaptive multimodal feature fusion fault prediction model, which can predict the fault probability when the monitoring equipment is running to solve the problems raised in the prior art.

[0005] To achieve the above-mentioned object, the present invention provides the following technical solutions: an artificial intelligence-based sewage treatment equipment operation monitoring system, the system comprising a sensor module, a feature fusion module, an adaptive multimodal feature fusion fault prediction model and a real-time monitoring and early warning module;

[0006] The sensor module is used to analyze and process the data collected by the equipment, monitor the on-site environment and transmit the data; the feature fusion module is used to select and extract features of the transmitted data and process data from multiple modes; the adaptive multimodal feature fusion fault prediction model is used to automatically adjust the feature fusion weights according to real-time data and historical data, and calculate the failure probability of the equipment; the real-time monitoring and early warning module is used to monitor whether the equipment is in an abnormal working state through the analysis of real-time data and fault prediction, and issue an early warning to the system for abnormal system conditions.

[0007] According to the above technical solution, the sensor for collecting the operation data of the sewage treatment equipment includes a data acquisition unit, an environmental monitoring unit and a data transmission unit;

[0008] The data acquisition unit is used to automatically collect various physical and chemical parameters in the sewage treatment process, and collect parameters such as pH value, temperature, dissolved oxygen content, conductivity, liquid level, and water quality composition; the environmental monitoring unit is used to monitor the environmental parameters of the sewage treatment site, including influencing factors such as harmful gas concentration, humidity, and temperature, to ensure the safety and efficiency of the sewage treatment process; the data transmission unit is used to transmit the collected data to the central processing system, which can be a wired or wireless transmission method.

[0009] According to the above technical solution, the feature fusion module for processing multimodal data includes a multimodal data processing unit, a feature selection unit and a feature extraction unit;

[0010] The multimodal data processing unit is used to process data from different modes; the feature selection unit is used to perform feature selection from the data in the data processing unit; then the feature extraction unit extracts parameters that affect the operating state of the equipment (i.e., pH value, temperature, dissolved oxygen content, conductivity, liquid level, water quality components) from the feature selection unit; the feature extraction unit is used to extract features for monitoring and analysis from the original data, including time domain features, frequency domain features, statistical features, and other feature values ​​for feature fusion;

[0011] According to the above technical solution, the training module for training the adaptive multimodal feature fusion fault prediction model includes an adaptive feature fusion weight adjustment unit and a fault probability prediction unit;

[0012] The adaptive adjustment feature fusion weight function is used to automatically adjust the feature fusion weight according to the real-time data and historical data provided by the feature fusion module, and process multimodal data from different sensors. The need to adjust the fusion weight is based on the importance of the extracted features and the sum of the extracted feature importance scores, where the sum of the feature importance scores is ensured to be 1 after normalizing the weights, and the importance score is calculated by the system based on the degree of influence of each feature on the model performance. The formula for adjusting the feature fusion weight is:

[0013]

[0014] Among them, W i is the fusion weight of the i-th feature, S i is the importance score of the i-th feature, S j is the term to be summed, where j is a variable among all the features extracted by the feature extraction unit, used to traverse each feature value in all the features, N is the total number of features, and S j The summation upper limit is used to normalize the weights to ensure that the sum is 1;

[0015] The accuracy of fault prediction is improved by feature fusion, so as to predict the fault probability and predict the possible faults of the equipment in advance. The calculation method of fault probability is as follows:

[0016] P = σ[1 / (1+e -Z )]

[0017] Z=X·W+b

[0018] Where P is the probability of equipment failure given the equipment feature vector, and e is the base of the natural logarithm;

[0019] In the second equation, Z is a column vector, which contains the result of the linear combination, which is the fused feature vector, and its dimension is usually (m,times1), where m is the number of rows (i.e. the number of samples) of the input matrix X;

[0020] X is an (m,n) matrix, where m represents the number of data samples (temperature, pressure, vibration) collected from multiple sensors, and n represents the number of features (i.e. pH, temperature, dissolved oxygen content, conductivity, liquid level, water quality composition). Each row represents a data sample and each column represents a feature;

[0021] W is a (n,p) matrix, where p is the number of output nodes. In this context, each column represents a possible output, and the weights W represent the contribution of different features to each output node.

[0022] b is a (p, times1) column vector, called the bias term or intercept term, each element of which corresponds to a column in the W matrix and is used to adjust the activation threshold of each output node;

[0023] After obtaining the fused feature vector Z, substitute it into the first equation and use the σ function to convert the fused feature vector into a fault probability, which ranges from 0 to 1 and represents the probability of a fault occurring.

[0024] According to the above technical solution, the module for real-time monitoring of equipment status and early warning includes a status monitoring unit, a fault detection unit and an early warning and alarm unit;

[0025] The status monitoring function is used to monitor the working status of the sewage treatment equipment in real time; the fault detection function is used to monitor whether the equipment is in an abnormal working state by analyzing real-time data and fault prediction; the early warning and alarm function is used to send out a warning or alarm signal when it is detected that the equipment status is abnormal or the parameters exceed the preset threshold.

[0026] The purpose of the present invention is to provide a sewage treatment equipment monitoring method based on artificial intelligence, and an adaptive multimodal feature fusion fault prediction model, which can predict the fault probability when the monitoring equipment is running to solve the problems raised in the prior art.

[0027] To achieve the above object, the present invention provides the following technical solution: a sewage treatment equipment operation monitoring method based on artificial intelligence, the method steps comprising the following contents:

[0028] S1. Use sensors to collect the operating data of sewage equipment, monitor the environmental parameters of the sewage treatment site, monitor the operating status of the sewage treatment equipment, and finally transmit the collected data to the central processing system; S2. Use the multimodal data processing function to process the collected data, and then use the feature selection unit to select features from the data in the data processing unit; the feature extraction unit is used to extract parameters that affect the operating status of the equipment (i.e., pH value, temperature, dissolved oxygen content, conductivity, liquid level, water quality components) from the feature selection unit; finally, the feature extraction unit extracts features for monitoring and analysis from the original data, including time domain features, frequency domain features, statistical features, and other feature values ​​for feature fusion;

[0029] S3. The extracted feature data is used with artificial intelligence algorithms to automatically adjust the feature fusion weights according to the real-time data and historical data provided by the feature fusion module, and multimodal data from different sensors are processed to improve the accuracy of fault prediction through feature fusion; thereby predicting the probability of faults and predicting possible equipment failures in advance;

[0030] S4. When the above steps are running normally, the real-time monitoring unit monitors whether the equipment is in abnormal working state by analyzing real-time data and fault prediction. When it is detected that the equipment state is abnormal or the parameters exceed the preset threshold, the early warning and alarm function will send a warning or alarm signal to the system to notify the operation and maintenance personnel to handle it.

[0031] The sensor collects data on the operation of sewage equipment including the following steps: first, various physical and chemical parameters in the sewage treatment process are collected, such as pH value, temperature, dissolved oxygen content, conductivity, liquid level, and water quality components; then the environmental monitoring unit monitors the environmental parameters of the sewage treatment site, including influencing factors such as harmful gas concentration, humidity, and temperature, to ensure the safety and efficiency of the sewage treatment process; finally, the data transmission unit is used to transmit the collected data to the central processing system, which can be a wired or wireless transmission method.

[0032] The multimodal data processing function is used to process the collected data, firstly, feature selection is performed from the data in the data processing unit; then the feature extraction unit extracts the parameters affecting the operation state of the equipment (i.e., pH value, temperature, dissolved oxygen content, conductivity, liquid level, water quality components) from the feature selection unit; finally, the feature extraction unit extracts features for monitoring and analysis from the original data, including time domain features, frequency domain features, statistical features, and these feature values ​​are subjected to feature fusion;

[0033] The method of using an algorithm to adjust the feature fusion weight to predict the fault probability includes the following steps:

[0034] S3.1 automatically adjusts the feature fusion weight according to the real-time data and historical data provided by the feature fusion module, and processes multimodal data from different sensors to improve the accuracy of fault prediction through feature fusion; the need to adjust the fusion weight is based on the importance of the extracted features and the sum of the extracted feature importance scores, where the sum of the feature importance scores is ensured to be 1 after normalizing the weights, and the importance score is calculated by the system based on the degree of influence of each feature on the model performance. The formula for adjusting the feature fusion weight is:

[0035]

[0036] Among them, W i is the fusion weight of the i-th feature, S i is the importance score of the i-th feature, S j is the term to be summed, where j is a variable among all the features extracted by the feature extraction unit, used to traverse each feature value in all the features, N is the total number of features, and S j The summation upper limit is used to normalize the weights to ensure that the sum is 1;

[0037] S3.2 performs fault probability prediction to predict possible equipment failures in advance; the prediction formula for fault probability is:

[0038] P = σ[1 / (1+e -Z )]

[0039] Z=X·W+b

[0040] Where P is the probability of equipment failure given the equipment feature vector, and e is the base of the natural logarithm;

[0041] In the second equation, Z is a column vector, which contains the result of the linear combination, which is the fused feature vector, and its dimension is usually (m,times1), where m is the number of rows (i.e. the number of samples) of the input matrix X;

[0042] X is an (m,n) matrix, where m represents the number of data samples (temperature, pressure, vibration) collected from multiple sensors, and n represents the number of features (i.e. pH, temperature, dissolved oxygen content, conductivity, liquid level, water quality composition). Each row represents a data sample and each column represents a feature;

[0043] W is a (n,p) matrix, where p is the number of output nodes. In this context, each column represents a possible output, and the weights W represent the contribution of different features to each output node.

[0044] b is a (p, times1) column vector, called the bias term or intercept term, each element of which corresponds to a column in the W matrix and is used to adjust the activation threshold of each output node;

[0045] After obtaining the fused feature vector Z, substitute it into the first equation and use the σ function to convert the fused feature vector into a fault probability, which ranges from 0 to 1 and represents the probability of a fault occurring.

[0046] The analysis of real-time data and fault prediction data, and monitoring of equipment for early warning include the following steps: When the above steps are operated normally, the real-time monitoring unit monitors whether the equipment has abnormal working conditions by analyzing real-time data and fault predictions. When an abnormal equipment state is detected or a parameter exceeds a preset threshold, the early warning and alarm unit will send an early warning or alarm signal to the system to notify the operation and maintenance personnel to handle it.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] 1. Improved prediction accuracy: Adaptive adjustment of fusion weights can dynamically adjust the weights of different data sources and models based on real-time data and equipment operation data, thereby improving the accuracy of fault prediction;

[0049] 2. Improved real-time monitoring capabilities: The system can monitor equipment status in real time and quickly analyze the collected data to achieve real-time early warning of faults;

[0050] 3. Reduced operation and maintenance costs: Accurate fault prediction can reduce unnecessary equipment inspection and maintenance, thereby reducing operation and maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 This is a system architecture diagram of an artificial intelligence-based sewage treatment equipment operation monitoring system of the present invention.

[0052] Figure 2 The present invention is a flowchart of a method for monitoring the operation of sewage treatment equipment based on artificial intelligence. DETAILED DESCRIPTION

[0053] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0054] Example: Figure 1-Figure 2 As shown, the present invention provides a technical solution, a sewage treatment equipment operation monitoring system based on artificial intelligence, the system includes a sensor module, a feature fusion module, an adaptive multimodal feature fusion fault prediction model and a real-time monitoring and early warning module;

[0055] The sensor module is used to analyze and process the data collected by the equipment, monitor the on-site environment and transmit the data; the feature fusion module is used to select and extract features of the transmitted data and process data from multiple modes; the adaptive multimodal feature fusion fault prediction model is used to automatically adjust the feature fusion weights according to real-time data and historical data, and calculate the failure probability of the equipment; the real-time monitoring and early warning module is used to monitor whether the equipment is in an abnormal working state through the analysis of real-time data and fault prediction, and issue an early warning to the system for abnormal system conditions.

[0056] Suppose there is currently a sewage treatment monitoring and operation system, which consists of a sensor module, a feature fusion module, an adaptive multimodal feature fusion fault prediction model and a real-time monitoring and early warning module; each module has its own function, the first of which is the sensor module: this module includes temperature sensors, humidity sensors, pressure sensors, pH sensors and other types of sensors, which are used to collect the working data and environmental data of the sewage treatment equipment in real time; the sensor module also has a data transmission unit, which can send the collected data to the data processing center.

[0057] After the data from the sensor module is transmitted to the system, the feature fusion module receives the data from the sensor module and preprocesses the data, including cleaning and normalization, and then performs feature selection and extraction; the feature fusion module can process multimodal data from different types of sensors and extract features for fault prediction.

[0058] After feature extraction is completed, the data is transmitted to the adaptive multimodal feature fusion fault prediction training model, which is the core of the system. It uses artificial intelligence algorithms to automatically adjust feature fusion weights based on real-time data and historical data. The adjustment steps are as follows:

[0059]

[0060] Among them, W i is the fusion weight of the i-th feature, S i is the importance score of the i-th feature, S j is the term to be summed, where j is a variable among all the features extracted by the feature extraction unit, used to traverse each feature value in all the features, N is the total number of features, and S j The summation upper limit is used to normalize the weights to ensure that the sum is 1;

[0061] Improve the accuracy of fault prediction by fusion of features, thereby improving the accuracy of fault prediction; the model can calculate the failure probability of the equipment and continuously optimize the prediction performance;

[0062] Assume that the system collects the following data sets:

[0063] X (feature matrix): contains 1 sample m, i.e. m = 1 and 3 features n, i.e. n = 3;

[0064] W (weight matrix): initially set to [0.5, 0.3, 0.2];

[0065] b (bias term): initially set to 0.1;

[0066] The feature matrix X is as follows:

[0067] X=[1,2,3]

[0068] Assume that the true label vector Y (indicating whether the device fails) is:

[0069] Y=[0,0,1,1]

[0070] According to the given prediction formula of failure probability:

[0071] P = σ[1 / (1+e -Z )]

[0072] Z=X·W+b

[0073] Calculate Z: For each sample, calculate the Z value:

[0074] Z=X·W+b

[0075] For the first sample:

[0076] Z1=(1·0.5+2·0.3+3·0.2)+0.1

[0077] =0.5+0.6+0.6+0.1

[0078] =1.8

[0079] Apply activation function σ: Use function σ to convert Z value into probability P:

[0080] P = σ[1 / (1+e -Z )]

[0081] Assume that the value of the σ function is 1

[0082] For the above sample:

[0083] P1=1 / (1+e^ (-1.8) )

[0084] ≈1 / (1+0.1653)

[0085] ≈1 / 1.1653

[0086] ≈0.8611

[0087] Get the probability P vector:

[0088] P≈0.8611

[0089] Results Analysis

[0090] Since the output range of the σ function is between 0 and 1, where a value close to 1 indicates a high probability of failure, and a value close to 0 indicates a low probability of failure, according to the above calculation results, P = 0.8611, which is close to 1, and is judged to be a high probability of failure.

[0091] In order to improve the model, we can use the gradient descent algorithm to automatically adjust the weight W and bias term b according to the prediction error, so that the model's prediction is closer to the true label. Through multiple iterations, it is expected that the model can better learn the relationship between features and failure probability, thereby improving prediction accuracy.

[0092] After calculating the prediction results, the real-time monitoring and early warning module will analyze the real-time data and prediction results to monitor whether the equipment is in an abnormal state. When the equipment status is detected to be abnormal or the parameters exceed the preset threshold, the early warning and alarm function will automatically send a warning or alarm signal to the system to notify the operation and maintenance personnel to handle it.

[0093] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A sewage treatment equipment operation monitoring system based on artificial intelligence, characterized in that: It includes sensor module, feature fusion module, adaptive multi-modal feature fusion fault prediction module and real-time monitoring and early warning module; The sensor module is used to analyze and process the data collected by the equipment, monitor the on-site environment and transmit the data; the feature fusion module is used to select and extract features of the transmitted data and process data from multiple modes; the adaptive multimodal feature fusion fault prediction model is used to automatically adjust the feature fusion weights according to real-time data and historical data, and calculate the failure probability of the equipment; the real-time monitoring and early warning module is used to monitor whether the equipment is in an abnormal working state through the analysis of real-time data and fault prediction, and issue an early warning to the system for abnormal system conditions.

2. According to claim 1, a sewage treatment equipment operation monitoring system based on artificial intelligence is characterized by: The sensor module further includes a data acquisition unit, an environmental monitoring unit and a data transmission unit; the data acquisition unit is used to automatically collect various physical and chemical parameters in the sewage treatment process; the environmental monitoring unit is used to monitor the environmental parameters of the sewage treatment site; the data transmission unit is used to transmit the collected data to the central processing system.

3. The artificial intelligence-based sewage treatment equipment operation monitoring system according to claim 1 is characterized by: The feature fusion module further includes a multimodal data processing unit, a feature selection unit and a feature extraction unit; the multimodal data processing unit is used to process data from different modalities; the feature selection unit is used to perform feature selection from the data in the data processing unit; and the feature extraction unit is used to extract parameters from the feature selection unit for feature fusion.

4. The artificial intelligence-based sewage treatment equipment operation monitoring system according to claim 1 is characterized in that: The adaptive multimodal feature fusion fault prediction module further includes an adaptive feature fusion weight adjustment unit and a fault probability prediction unit; the adaptive feature fusion weight adjustment unit is used to use an artificial intelligence algorithm to automatically adjust the feature fusion weight according to the real-time data and historical data provided by the feature fusion module, and process multimodal data from different sensors. The need to adjust the fusion weight is based on the importance of the extracted features and the sum of the extracted feature importance scores, wherein the sum of the feature importance scores is ensured to be 1 after normalizing the weights, and the importance score is calculated by the system according to the degree of influence of each feature on the model performance. The formula for adjusting the feature fusion weight is: Among them, W i is the fusion weight of the i-th feature, S i is the importance score of the i-th feature, S j is the term to be summed, where j is a variable among all the features extracted by the feature extraction unit, used to traverse each feature value in all the features, N is the total number of features, and S j The summation upper limit is used to normalize the weights to ensure that the sum is 1; The accuracy of fault prediction is improved by feature fusion; the fault probability prediction unit can predict possible equipment failures in advance; the prediction formula for fault probability is: P=σ[1 / (1+e -Z )] Z=X·W+b Where P is the probability of equipment failure given the equipment feature vector, and e is the base of the natural logarithm; In the second equation, Z is a column vector, which contains the result of the linear combination, which is the fused feature vector, and its dimension is usually (m,times1), where m is the number of rows (i.e. the number of samples) of the input matrix X; X is an (m,n) matrix, where m represents the number of data samples collected from multiple sensors and n represents the number of features. Each row represents a data sample and each column represents a feature; W is a (n,p) matrix, where p is the number of output nodes. In this context, each column represents a possible output, and the weights W represent the contribution of different features to each output node. b is a (p, times1) column vector, called the bias term or intercept term, each element of which corresponds to a column in the W matrix and is used to adjust the activation threshold of each output node; After obtaining the fused feature vector Z, substitute it into the first equation and use the σ function to convert the fused feature vector into a fault probability, which ranges from 0 to 1 and represents the probability of a fault occurring.

5. The artificial intelligence-based sewage treatment equipment operation monitoring system according to claim 1 is characterized by: The real-time monitoring device and early warning status module further includes a status monitoring unit, a fault detection unit and an early warning and alarm unit; the status monitoring unit is used to monitor the working status of the sewage treatment equipment in real time; the fault detection unit is used to monitor whether the equipment is in an abnormal working state by analyzing real-time data and fault prediction; The early warning and alarm unit is used to automatically send out an early warning or alarm signal when the device status is abnormal or the parameters exceed the preset threshold.

6. A method for monitoring the operation of sewage treatment equipment based on artificial intelligence, characterized in that: The method steps include the following: S1. Use sensors to collect the operating data of sewage equipment, monitor the environmental parameters of the sewage treatment site, monitor the operating status of the sewage treatment equipment, and finally transmit the collected data to the central processing system; S2, processing the collected data using a multimodal data processing function, and then using a feature selection unit to perform feature selection from the data in the data processing unit; the feature extraction unit is used to extract features from the feature selection unit for feature fusion; S3. The extracted feature data is used with artificial intelligence algorithms to automatically adjust the feature fusion weights according to the real-time data and historical data provided by the feature fusion module, and multimodal data from different sensors are processed to improve the accuracy of fault prediction through feature fusion; thereby predicting the probability of faults and predicting possible equipment failures in advance; S4. When the above steps are running normally, the real-time monitoring unit monitors whether the equipment is in abnormal working state by analyzing real-time data and fault prediction. When it is detected that the equipment state is abnormal or the parameters exceed the preset threshold, the early warning and alarm function will send a warning or alarm signal to the system to notify the operation and maintenance personnel to handle it.

7. The method for monitoring the operation of sewage treatment equipment based on artificial intelligence according to claim 7 is characterized in that: The sensor's data collection, monitoring and transmission include the following steps: using sensors to collect operating data of sewage equipment, monitoring environmental parameters at the sewage treatment site, monitoring the operating status of the sewage treatment equipment, and finally transmitting the collected data to a central processing system.

8. The method for monitoring the operation of sewage treatment equipment based on artificial intelligence according to claim 7 is characterized in that: The multimodal data processing of the collected data, the selection and extraction of characteristic values ​​include the following steps: firstly, processing the data from different modes, and performing feature selection from the data in the data processing unit; then the feature extraction unit performs feature extraction from the feature selection unit; the feature extraction unit is used to extract features for monitoring and analysis from the original data, including time domain features, frequency domain features, statistical features, and these characteristic values ​​are subjected to feature fusion.

9. The method for monitoring the operation of sewage treatment equipment based on artificial intelligence according to claim 7 is characterized in that: The method of automatically adjusting the feature fusion weight according to the real-time data and the historical data and calculating the probability of equipment failure using the fault probability prediction function includes the following steps: first, automatically adjusting the feature fusion weight according to the real-time data and the historical data provided by the feature fusion module, and processing the multimodal data from different sensors, and adjusting the fusion weight according to the importance of the extracted features and the sum of the extracted feature importance scores, wherein the sum of the feature importance scores is ensured to be 1 after the normalization weights, and the importance scores are calculated by the system according to the degree of influence of each feature on the model performance, and the formula for adjusting the feature fusion weight is: Among them, W i is the fusion weight of the i-th feature, S i is the importance score of the i-th feature, S j is the term to be summed, where j is a variable among all the features extracted by the feature extraction unit, used to traverse each feature value in all the features, N is the total number of features, and S j The summation upper limit is used to normalize the weights to ensure that the sum is 1; The accuracy of fault prediction is improved by feature fusion, so as to predict the fault probability and predict the possible faults of the equipment in advance. The calculation method of fault probability is as follows: P=σ[1 / (1+e -Z )] Z=X·W+b Where P is the probability of equipment failure given the equipment feature vector, and e is the base of the natural logarithm; In the second equation, Z is a column vector, which contains the result of the linear combination, which is the fused feature vector, and its dimension is usually (m,times1), where m is the number of rows (i.e. the number of samples) of the input matrix X; X is an (m,n) matrix, where m represents the number of data samples (temperature, pressure, vibration) collected from multiple sensors, and n represents the number of features (i.e. pH, temperature, dissolved oxygen content, conductivity, liquid level, water quality composition). Each row represents a data sample and each column represents a feature; W is a (n,p) matrix, where p is the number of output nodes. In this context, each column represents a possible output, and the weights W represent the contribution of different features to each output node. b is a (p, times1) column vector, called the bias term or intercept term, each element of which corresponds to a column in the W matrix and is used to adjust the activation threshold of each output node; After obtaining the fused feature vector Z, substitute it into the first equation and use the σ function to convert the fused feature vector into a fault probability, which ranges from 0 to 1 and represents the probability of a fault occurring.

10. The method for monitoring the operation of sewage treatment equipment based on artificial intelligence according to claim 7, characterized in that: The real-time monitoring and equipment fault warning through data analysis include the following steps: by analyzing the transmitted real-time data, monitoring the operating status of the equipment, and detecting whether the equipment is in an abnormal working state based on the fault prediction given by the training model; when the equipment status is abnormal or the parameters exceed the preset threshold, the early warning and alarm unit will send a warning or alarm signal to the system.

Images