Multistage AO sewage treatment real-time monitoring and fault diagnosis system based on Internet of Things

Abstract

The invention discloses a multi-stage AO sewage treatment real-time monitoring and fault diagnosis system based on the Internet of Things, and relates to the field of the Internet of Things, and the system is characterized in that multi-dimensional data of an anaerobic tank, an anoxic tank, an aerobic tank and a secondary sedimentation tank are collected in real time through a multi-stage AO process monitoring module, are encrypted through an Internet of Things data transmission module, and then are uploaded to a cloud; the edge calculation module performs noise filtering and data completion to generate a preliminary regulation and control instruction; the fault diagnosis module combines a deep convolutional neural network and a gradient boosting tree algorithm to realize accurate fault classification and traceability; parameters such as aeration, stirring and reflux ratio are automatically adjusted according to a diagnosis result, and an alarm is pushed through a multi-stage early warning mechanism; and the cloud end adopts distributed storage to support multi-dimensional retrieval, and optimizes model parameters based on historical data. The method has the advantages that accurate fault diagnosis and cooperative regulation are achieved through edge calculation and an intelligent algorithm, a closed loop is formed in combination with cloud management and remote operation and maintenance, the process stability and the fault processing efficiency are improved, and the operation and maintenance cost is reduced.

Description

Technical Field

[0001] This invention relates to the field of the Internet of Things (IoT), and in particular to a real-time monitoring and fault diagnosis system for multi-level AO wastewater treatment based on the IoT. Background Technology

[0002] With the rapid development of industrialization and urbanization, wastewater treatment has become a key link in environmental protection and water resource recycling. Multi-stage AO (anaerobic-aerobic) processes are widely used in wastewater treatment plants due to their high efficiency in nitrogen and phosphorus removal. However, their operation involves complex biochemical reactions and multiple control parameters, making them susceptible to factors such as influent load, dissolved oxygen, and sludge concentration, which can lead to fluctuations in treatment efficiency or even system failures.

[0003] Current wastewater treatment monitoring and control systems on the market are ill-suited to the demands of precise and intelligent operation. Most systems monitor only a single dimension, lack adaptive sampling mechanisms, and are vulnerable to fluctuations in water quality and quantity, leading to data inaccuracies. Data transmission often employs a single architecture, making it susceptible to interference, packet loss, and delays. Furthermore, insufficient encryption and protocol conversion capabilities result in poor data security and compatibility. In data processing and fault diagnosis, there is a lack of real-time edge computing preprocessing capabilities, relying on traditional algorithms or manual judgment. This leads to inaccurate fault feature extraction, vague classification and location, and difficulty in predicting fault signs in advance. Control execution is often a crude, unified control approach, lacking regional collaborative control capabilities and exhibiting delayed response times. Simultaneously, operation and maintenance management relies on on-site manual inspections, lacking remote control and intelligent maintenance capabilities. Data from multiple platforms is independent and fragmented, hindering comprehensive analysis and iterative model optimization throughout the entire process. This results in low overall operational efficiency, high maintenance costs, and a significant risk to environmental compliance. Summary of the Invention

[0004] To improve the existing system, a real-time monitoring and fault diagnosis system for multi-level AO wastewater treatment based on the Internet of Things (IoT) is provided. This system relies on IoT to achieve accurate monitoring and safe data transmission of the entire multi-level AO wastewater treatment process. Through edge computing and intelligent algorithms, it achieves accurate fault diagnosis and collaborative control. Combined with cloud management and remote operation and maintenance, it forms a closed loop, which greatly improves process stability and fault handling efficiency, and reduces operation and maintenance costs.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] Multi-stage AO process monitoring module: Deployed in the anaerobic tank, anoxic tank, aerobic tank, and secondary sedimentation tank of the multi-stage AO wastewater treatment process, including multi-dimensional sensing components and adaptive sampling units;

[0007] The IoT data transmission module adopts a three-level transmission architecture of "edge node-gateway-cloud", which gathers and pre-filters data from various sensors nearby, encrypts the transmitted data using encryption algorithms, and converts the private data protocol into the MQTT standard protocol.

[0008] Edge computing data processing module: Removes random noise from sensor data, intelligently completes missing data based on the material balance principle of multi-level AO process, and generates preliminary control instructions;

[0009] Fault diagnosis module: It extracts features from preprocessed data and historical operation data using deep convolutional neural network algorithm, combines gradient boosting tree algorithm to accurately classify fault types, builds a correlation model between faults and causes, and generates diagnostic results;

[0010] Control and execution module: Connects to the execution equipment of multi-stage AO process, and performs aeration control, stirring control, reflux ratio control and sludge discharge control based on preliminary control commands and diagnostic results;

[0011] Early warning push module: Based on the diagnostic results of the fault diagnosis module, multi-level early warnings are set, with each level corresponding to different early warning thresholds and processing priorities. Early warning information is pushed through various push methods.

[0012] Cloud-based data management module: It adopts a distributed storage architecture to securely store all data, performs multi-dimensional retrieval based on time, process, and fault type, builds a historical data mining model, and optimizes the parameters and control strategy thresholds of the fault diagnosis model by analyzing historical operation data and fault handling data.

[0013] Remote operation and maintenance module: View the running status, fault diagnosis results and early warning information of each module in real time through the web and APP, remotely issue operation and maintenance instructions, and generate equipment maintenance plans based on equipment data.

[0014] Preferably, the multi-level AO process monitoring module specifically includes:

[0015] Multi-dimensional sensing component unit: includes dissolved oxygen sensor, COD sensor, ammonia nitrogen sensor, total nitrogen sensor, sludge concentration sensor, pH sensor and tank level sensor, each sensor is equipped with PTFE composite anti-fouling membrane;

[0016] Adaptive sampling unit: integrates a flow sensor and an electric lifting sampling mechanism. It sets a fluctuation threshold based on real-time flow data. When the flow fluctuation exceeds ±10% or the water quality parameter fluctuation exceeds ±5%, it automatically increases the sampling frequency. The electric lifting sampling rod mechanism is driven by a stepper motor to locate the sampling depth.

[0017] Preferably, the IoT data transmission module specifically includes:

[0018] Edge node communication unit: It adopts a dual-mode integrated design of LoRa and NB-IoT to gather and initially filter the data from various sensors nearby, and automatically switches to NB-IoT signal when LoRa signal is interrupted;

[0019] Gateway processing unit: performs hardware encryption on data uploaded by edge nodes, with the key dynamically generated through the device's unique identifier, and converts the sensor's private protocol into the MQTT standard protocol in real time;

[0020] Cloud communication unit: It adopts dual-mode redundant transmission of 4G / 5G and optical fiber, and performs integrity verification of transmitted data based on CRC-32 algorithm. If the verification fails, a retransmission mechanism is triggered.

[0021] Preferably, the edge computing data processing module specifically includes:

[0022] Data preprocessing unit: A sliding window filtering algorithm is used to remove random noise from sensor data. The window length is dynamically adjusted according to data fluctuations. System error correction is performed by comparing the standard reference values ​​of the calibration curve parameters of each sensor model in real time. Data completion is performed based on the parameters of adjacent processing units.

[0023] Edge decision computing unit: The edge decision model is built based on a lightweight neural network algorithm. The model structure is optimized for scenarios with limited computing power at the edge, and redundant network layers are removed.

[0024] Control command generation unit: It has a built-in fault symptom recognition rule library. It inputs preprocessed real-time data into the edge decision model to obtain simple fault symptoms, and generates preliminary control commands by matching the fault symptom recognition rule library.

[0025] Preferably, the fault diagnosis module specifically includes:

[0026] Feature extraction unit: Extracts features from the preprocessed data transmitted by the edge computing data processing module, including water quality parameter fluctuation coefficient, equipment operating parameter deviation value, and process unit response delay. It removes interference features through signal smoothing processing and retains key features strongly correlated with faults based on the mutual information entropy principle.

[0027] Fault Classification and Localization Unit: A fusion model of deep convolutional neural network and gradient boosting tree is used to classify faults. A fault-cause association network is constructed through knowledge graph to locate the source of the fault and generate diagnostic results containing fault type, scope of impact and handling steps.

[0028] Diagnostic model iteration and storage unit: caches fault data and diagnostic results from the past month, receives model parameters optimized from the cloud based on historical data in real time, and performs incremental updates to the model.

[0029] Preferably, the control execution module specifically includes:

[0030] Aeration control unit: It adopts an industrial-grade vector frequency converter and a high-efficiency aeration head array. It receives control commands through a 4-20mA analog signal. Each aerobic tank is divided into zones with independent aeration branches. The aeration volume of each zone is controlled by the branch solenoid valve.

[0031] The stirring control unit is a submersible low-speed impeller mixer with a propeller-type impeller. The impeller speed is adjusted to control the impeller intensity.

[0032] Reflux ratio control unit: integrates an electromagnetic flow sensor and a variable frequency reflux pump. The flow sensor collects the reflux flow of nitrification liquid and sludge in real time, and the reflux pump adjusts the pump speed to achieve dynamic control of the reflux ratio.

[0033] Sludge discharge control unit: The sludge discharge action is triggered by real-time data from the sludge concentration sensor. When the sludge concentration exceeds the set threshold, the opening degree is increased; when it is below the threshold, the opening degree is decreased.

[0034] Preferably, the early warning push module specifically includes:

[0035] Multi-level early warning judgment unit: Based on the diagnosis results of the fault diagnosis module, the degree of parameter exceedance of the edge computing data processing module, and combined with the operating threshold benchmarks of each unit of the multi-level AO process, the first, second and third level early warning levels are determined;

[0036] Multi-channel push execution unit: Construct a multi-channel redundant push architecture that integrates GSM SMS, 4G communication, audible and visual alarms and industrial Ethernet, and customize push content and push frequency based on the identity of the recipient;

[0037] Early warning ledger and feedback unit: Records the early warning time, level, triggering reason, push target, reception status, and processing result in real time to form a traceable early warning ledger. Operation and maintenance personnel can provide feedback on the early warning processing progress through the APP.

[0038] Preferably, the cloud data management module specifically includes:

[0039] Distributed data storage unit: It adopts a hybrid architecture of "distributed file storage + relational database + time series database". The distributed file storage adopts a layered design of SSD and mechanical hard disk. The relational database stores equipment information and structured data of operation and maintenance ledger. The time series database stores water quality and process parameter time series data.

[0040] Data processing and retrieval unit: Data cleaning optimizes data quality by removing outliers, standardizing data formats, and deduplicating duplicate data; data retrieval supports multi-dimensional combined retrieval based on time, process unit, data type, and fault type.

[0041] Model optimization unit: By analyzing historical operating data and fault handling records, it optimizes the parameter thresholds of the fault diagnosis module and the control strategy of the control execution module, and generates an optimized parameter package to push to the edge.

[0042] Preferably, the remote operation and maintenance module specifically includes:

[0043] Command interaction unit: Built-in hierarchical permission management module, which divides permissions into three levels: administrator, maintenance personnel and operator. Only the administrator can issue maintenance commands. After the command is issued, the execution status is fed back in real time. If the execution fails, a retry mechanism and anomaly alarm are triggered.

[0044] Equipment maintenance management unit: Pre-stores model parameters, maintenance cycle and spare parts information of aeration blowers and return pumps, and generates periodic maintenance plans based on equipment runtime, failure frequency and load rate data;

[0045] Remote upgrade and maintenance support unit: It adopts differential upgrade technology to remotely update the firmware of core modules, only transmitting the difference data of the upgrade package, and has a built-in local maintenance interface for on-site debugging and emergency fault handling.

[0046] Compared with the prior art, the advantages of the present invention are:

[0047] Accurate monitoring of the entire process is achieved through multi-dimensional sensing and adaptive sampling, while a three-tiered encrypted transmission architecture ("edge-gateway-cloud") ensures secure and stable data transmission. Real-time data preprocessing and initial control are implemented using edge computing, coupled with a deep convolutional neural network and gradient boosting tree fusion algorithm for accurate fault diagnosis and source localization. Multi-level early warning push and multi-unit collaborative control capabilities enable rapid fault response and process parameter optimization. Cloud-based distributed storage and historical data mining support continuous model optimization, while a remote operation and maintenance module enables full-process visualized management and intelligent maintenance. This comprehensive system forms a closed loop of "monitoring-transmission-processing-diagnosis-control-operation and maintenance," significantly improving wastewater treatment process stability, fault handling efficiency, reducing operation and maintenance costs, and facilitating the intelligent upgrade of wastewater treatment. Attached Figure Description

[0048] Figure 1 This is a schematic diagram of the system proposed in this invention;

[0049] Figure 2 This is a diagram of the multi-level AO process monitoring module proposed in this invention;

[0050] Figure 3 This is a diagram of the IoT data transmission module proposed in this invention;

[0051] Figure 4 This is a diagram of the edge computing data processing module proposed in this invention;

[0052] Figure 5 This is a diagram of the fault diagnosis module proposed in this invention;

[0053] Figure 6 This is a diagram of the control execution module proposed in this invention;

[0054] Figure 7 This is a diagram of the early warning push module proposed in this invention;

[0055] Figure 8 This is a diagram of the cloud data management module proposed in this invention;

[0056] Figure 9 This is a diagram of the remote operation and maintenance module proposed in this invention. Detailed Implementation

[0057] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0058] See Figure 1 As shown, the IoT-based multi-level AO wastewater treatment real-time monitoring and fault diagnosis system includes:

[0059] Multi-stage AO process monitoring module: Deployed in the anaerobic tank, anoxic tank, aerobic tank, and secondary sedimentation tank of the multi-stage AO wastewater treatment process, including multi-dimensional sensing components and adaptive sampling units;

[0060] The IoT data transmission module adopts a three-level transmission architecture of "edge node-gateway-cloud", which gathers and pre-filters data from various sensors nearby, encrypts the transmitted data using encryption algorithms, and converts the private data protocol into the MQTT standard protocol.

[0061] Edge computing data processing module: Removes random noise from sensor data, intelligently completes missing data based on the material balance principle of multi-level AO process, and generates preliminary control instructions;

[0062] Fault diagnosis module: It extracts features from preprocessed data and historical operation data using deep convolutional neural network algorithm, combines gradient boosting tree algorithm to accurately classify fault types, builds a correlation model between faults and causes, and generates diagnostic results;

[0063] Control and execution module: Connects to the execution equipment of multi-stage AO process, and performs aeration control, stirring control, reflux ratio control and sludge discharge control based on preliminary control commands and diagnostic results;

[0064] Early warning push module: Based on the diagnostic results of the fault diagnosis module, multi-level early warnings are set, with each level corresponding to different early warning thresholds and processing priorities. Early warning information is pushed through various push methods.

[0065] Cloud-based data management module: It adopts a distributed storage architecture to securely store all data, performs multi-dimensional retrieval based on time, process, and fault type, builds a historical data mining model, and optimizes the parameters and control strategy thresholds of the fault diagnosis model by analyzing historical operation data and fault handling data.

[0066] Remote operation and maintenance module: View the running status, fault diagnosis results and early warning information of each module in real time through the web and APP, remotely issue operation and maintenance instructions, and generate equipment maintenance plans based on equipment data.

[0067] See Figure 2 As shown, the multi-level AO process monitoring module specifically includes:

[0068] Multi-dimensional sensing component unit: includes dissolved oxygen sensor, COD sensor, ammonia nitrogen sensor, total nitrogen sensor, sludge concentration sensor, pH sensor and tank level sensor, each sensor is equipped with PTFE composite anti-fouling membrane;

[0069] Adaptive sampling unit: integrates a flow sensor and an electric lifting sampling mechanism. It sets a fluctuation threshold based on real-time flow data. When the flow fluctuation exceeds ±10% or the water quality parameter fluctuation exceeds ±5%, it automatically increases the sampling frequency. The electric lifting sampling rod mechanism is driven by a stepper motor to locate the sampling depth.

[0070] Specifically, based on the multi-stage AO process flow procedure number, sensors are positioned and installed at key locations such as the anaerobic tank inlet, the middle of the tank, the anoxic tank mixing zone and effluent zone, the aerobic tank aeration zone and mixing zone, the secondary sedimentation tank inlet, and the sludge return outlet. Based on the wastewater treatment plant's design flow rate and historical water quality fluctuation data, basic sampling parameters are preset. The initial sampling frequency for the anaerobic and anoxic tanks is set to 5 minutes / time, for the aerobic tank (due to rapid water quality changes) it is set to 2 minutes / time, and for the secondary sedimentation tank it is set to 10 minutes / time. The sampling depth is divided into three layers, and initially, it operates in a three-layer alternating sampling mode. Water quality fluctuation thresholds and flow rate fluctuation thresholds are set.

[0071] The sensors collect water quality and process parameters at each location in real time, and record the collection time and location information. The adaptive sampling unit compares the current water quality and flow data with the preset threshold in real time. If the fluctuation of a certain unit parameter exceeds the threshold, the sampling cycle of that unit is automatically shortened, and the full-depth continuous sampling mode is started to cover the 0-100% water depth of the pool. If the parameters are stable for more than 30 minutes, the initial sampling parameters are restored.

[0072] See Figure 3As shown, the IoT data transmission module specifically includes:

[0073] Edge node communication unit: It adopts a dual-mode integrated design of LoRa and NB-IoT to gather and initially filter the data from various sensors nearby, and automatically switches to NB-IoT signal when LoRa signal is interrupted;

[0074] Gateway processing unit: performs hardware encryption on data uploaded by edge nodes, with the key dynamically generated through the device's unique identifier, and converts the sensor's private protocol into the MQTT standard protocol in real time;

[0075] Cloud communication unit: It adopts dual-mode redundant transmission of 4G / 5G and optical fiber, and performs integrity verification of transmitted data based on CRC-32 algorithm. If the verification fails, a retransmission mechanism is triggered.

[0076] Specifically, the LoRa and NB-IoT dual-mode communication module serves as an edge node, which is connected one-to-one with each sensor of the multi-level AO process monitoring module. The edge node collects the raw data transmitted by each sensor according to a preset polling cycle. The collection order follows the process sequence of anaerobic tank, anoxic tank, aerobic tank, and secondary sedimentation tank. The collected data undergoes preliminary filtering to remove abnormal data that exceeds the sensor's range and filter out invalid data with a signal strength lower than -110dBm.

[0077] Edge nodes aggregate the encapsulated data frames to the gateway device in the factory area via LoRa communication. The gateway initiates the AES-256 encryption process, first establishing a security key with the edge nodes and the cloud through a key negotiation mechanism, then performing block encryption on the data frames, and then initiating protocol conversion to parse the private communication protocol of the sensor data, extract the core data fields, and re-encapsulate them according to the MQTT protocol specification.

[0078] The gateway transmits MQTT format data to the edge computing nodes within the factory area via LoRa communication. A timeout retransmission mechanism is set during the transmission process, with a timeout period of 3 seconds and a maximum of 3 retransmissions. If three consecutive transmissions fail, the system automatically switches to the NB-IoT communication link. For long-distance transmission scenarios within the factory area, a LoRa repeater is added between the gateway and the edge computing nodes. The repeater is powered by solar energy to ensure uninterrupted link operation in areas without external power supply.

[0079] See Figure 4 As shown, the edge computing data processing module specifically includes:

[0080] Data preprocessing unit: A sliding window filtering algorithm is used to remove random noise from sensor data. The window length is dynamically adjusted according to data fluctuations. System error correction is performed by comparing the standard reference values ​​of the calibration curve parameters of each sensor model in real time. Data completion is performed based on the parameters of adjacent processing units.

[0081] Edge decision computing unit: The edge decision model is built based on a lightweight neural network algorithm. The model structure is optimized for scenarios with limited computing power at the edge, and redundant network layers are removed.

[0082] Control command generation unit: It has a built-in fault symptom recognition rule library. It inputs preprocessed real-time data into the edge decision model to obtain simple fault symptoms, and generates preliminary control commands by matching the fault symptom recognition rule library.

[0083] Specifically, after receiving encrypted data from the IoT data transmission module, AES-256 decryption and data frame parsing are first performed to extract the original sensor data, acquisition time, and location information. A sliding window filtering algorithm is used to remove noise from the original data. Data of the same parameter from five consecutive sampling periods are used as a window, and outliers deviating from the mean by ±3 times the standard deviation are removed. The mean of the remaining data is taken as the filtered data. Based on the sensor calibration coefficient table and temperature compensation curve, system error correction is performed on the filtered data. Combining the historical calibration data of each sensor with the current operating temperature, the correction range is controlled within ±5%.

[0084] The filtered and calibrated data undergoes integrity checks. If data is missing, intelligent completion is performed using a combination of trend fitting and interval interpolation, based on the material balance principle of the multi-stage AO process and the data trends of adjacent sampling points. For example, if dissolved oxygen data is missing in the aerobic tank for a certain period, the reasonable value for the missing period is calculated by combining the dissolved oxygen change trends of the previous three sampling points and the operating parameters of the aeration equipment during the same period. After completion, the data integrity is verified to ensure that the data integrity rate of a single batch is not less than 99%. Batches of data that fail the verification are marked and retransmitted.

[0085] Based on the pre-processed accurate data, a lightweight neural network decision model is invoked to perform real-time analysis of the operating status of each processing unit. The analysis focuses on whether water quality parameters exceed preset thresholds, whether parameter fluctuation rates are abnormal, and the matching degree between process parameters and equipment operating parameters. It also identifies simple fault signs such as excessively high sludge concentration and insufficient dissolved oxygen. If an abnormality is detected, a preliminary control instruction is immediately generated, which includes the location of the abnormality, the fault type, the preliminary adjustment direction, and the adjustment range.

[0086] See Figure 5 As shown, the fault diagnosis module specifically includes:

[0087] Feature extraction unit: Extracts features from the preprocessed data transmitted by the edge computing data processing module, including water quality parameter fluctuation coefficient, equipment operating parameter deviation value, and process unit response delay. It removes interference features through signal smoothing processing and retains key features strongly correlated with faults based on the mutual information entropy principle.

[0088] Fault Classification and Localization Unit: A fusion model of deep convolutional neural network and gradient boosting tree is used to classify faults. A fault-cause association network is constructed through knowledge graph to locate the source of the fault and generate diagnostic results containing fault type, scope of impact and handling steps.

[0089] Diagnostic model iteration and storage unit: caches fault data and diagnostic results from the past month, receives model parameters optimized from the cloud based on historical data in real time, and performs incremental updates to the model.

[0090] Specifically, a deep convolutional neural network feature extraction model is invoked to extract features from the preprocessed multi-source data. The extraction dimensions include absolute parameter values, rate of change, fluctuation amplitude, and cross-parameter correlation, such as extracting abnormal features of COD concentration change rate > 50 mg / L per hour and correlation features between dissolved oxygen and aeration power. The extracted real-time feature vectors are then matched with standard feature vectors in the fault feature library using a cosine similarity algorithm. A similarity threshold of ≥ 85% is set as a preliminary successful match, and the feature vectors are marked as suspected faults. Feature vectors that fail to match are temporarily stored in the diagnostic buffer.

[0091] For suspected faults that are initially matched, a gradient boosting tree classification model is called for accurate classification. The fault type and specific fault level are determined by combining the weight ratio of multi-dimensional feature vectors. After classification, fault type labels are generated. A fault-cause association model based on knowledge graph technology is used to locate the fault cause. The knowledge graph covers four association dimensions: fault type, influencing factors, equipment association, and process link association, and the influence weight of each cause is quantified.

[0092] See Figure 6 As shown, the control execution module specifically includes:

[0093] Aeration control unit: It adopts an industrial-grade vector frequency converter and a high-efficiency aeration head array. It receives control commands through a 4-20mA analog signal. Each aerobic tank is divided into zones with independent aeration branches. The aeration volume of each zone is controlled by the branch solenoid valve.

[0094] The stirring control unit is a submersible low-speed impeller mixer with a propeller-type impeller. The impeller speed is adjusted to control the impeller intensity.

[0095] Reflux ratio control unit: integrates an electromagnetic flow sensor and a variable frequency reflux pump. The flow sensor collects the reflux flow of nitrification liquid and sludge in real time, and the reflux pump adjusts the pump speed to achieve dynamic control of the reflux ratio.

[0096] Sludge discharge control unit: The sludge discharge action is triggered by real-time data from the sludge concentration sensor. When the sludge concentration exceeds the set threshold, the opening degree is increased; when it is below the threshold, the opening degree is decreased.

[0097] Specifically, the aeration control unit, based on the analyzed instructions, adjusts the aeration blower speed using frequency conversion control technology, collects real-time feedback data from the dissolved oxygen sensor in the aerobic tank, and employs a closed-loop adjustment based on the difference between the target value and the actual value. When the actual dissolved oxygen value is lower than the target value by 0.5 mg / L, the blower speed is gradually increased by 5% per cycle; when it is higher than the target value by 0.5 mg / L, the speed is decreased by 3% per cycle, ensuring a dissolved oxygen control accuracy of ±0.2 mg / L. The mixing control unit regulates the mixing uniformity by adjusting the mixer's operating power. When the mixing uniformity in the anaerobic / anoxic tank is detected to be lower than 90%, the mixing power is increased by 10%-20%, and the sludge suspension state in the tank is continuously monitored until the mixing uniformity meets the standard. The system maintains the current power. The return ratio control unit is linked to the return pump flow controller, which adjusts the return flow rate by regulating the pump speed. The nitrification liquid return ratio is dynamically adjusted according to the ammonia nitrogen concentration in the aerobic tank. When the ammonia nitrogen concentration is higher than 8 mg / L, the return ratio is increased to 180%-200%, and when it is lower than 2 mg / L, it is reduced to 100%-120%. The sludge return ratio is adjusted in conjunction with the sludge concentration in the secondary sedimentation tank to ensure that the sludge concentration in the tank is stable at 2000-4000 mg / L. The sludge discharge control unit controls the opening of the sludge discharge valve based on the sludge age target value and real-time sludge concentration data. The opening adjustment range is 0-100%. When the sludge age is too high, the opening is increased to increase the sludge discharge volume, and when it is too low, the opening is decreased, while avoiding excessive sludge discharge that could cause a sudden drop in sludge concentration.

[0098] The formula for adjusting the sludge concentration discharge valve opening is:

[0099]

[0100] in, The adjusted mud discharge valve opening, This represents the current opening degree of the sludge discharge valve. The opening adjustment coefficient is fixed at 20, determined based on process debugging and optimization. This represents the real-time sludge concentration in the secondary sedimentation tank. This is the preset target value for sludge concentration.

[0101] See Figure 7 As shown, the early warning push module specifically includes:

[0102] Multi-level early warning judgment unit: Based on the diagnosis results of the fault diagnosis module, the degree of parameter exceedance of the edge computing data processing module, and combined with the operating threshold benchmarks of each unit of the multi-level AO process, the first, second and third level early warning levels are determined;

[0103] Multi-channel push execution unit: Construct a multi-channel redundant push architecture that integrates GSM SMS, 4G communication, audible and visual alarms and industrial Ethernet, and customize push content and push frequency based on the identity of the recipient;

[0104] Early warning ledger and feedback unit: Records the early warning time, level, triggering reason, push target, reception status, and processing result in real time to form a traceable early warning ledger. Operation and maintenance personnel can provide feedback on the early warning processing progress through the APP.

[0105] Specifically, the system receives diagnostic results from the intelligent fault diagnosis module and abnormal data alerts from the edge computing data processing module in real time. It performs timestamp alignment and correlation verification on both types of data, eliminating false alert trigger signals caused by data transmission delays. The received data is then graded: if a single parameter anomaly triggers an alert, the validity of the alert must be confirmed in conjunction with other related parameters within the same unit; if the fault diagnosis module directly outputs the fault level, the corresponding alert level is directly matched without repeated judgment. After judgment, the alert level, triggering cause, and associated process / equipment unit are marked.

[0106] Based on the warning level and push rule base, standardized warning information is generated, including the warning level, trigger time, abnormal location, core abnormal parameters, associated fault type, and preliminary handling suggestions. A multi-channel push mechanism is activated: SMS push is sent to the preset mobile phone number through the industrial-grade SMS module; APP push is connected to the operation and maintenance management APP through the cloud API interface, and the push information includes a link to the real-time monitoring data page; the plant area audible and visual alarms are activated for the corresponding area, with a level 1 warning consisting of a yellow light and a low-frequency buzzer, and a level 3 warning consisting of a red light and a high-frequency buzzer.

[0107] See Figure 8 As shown, the cloud data management module specifically includes:

[0108] Distributed data storage unit: It adopts a hybrid architecture of "distributed file storage + relational database + time series database". The distributed file storage adopts a layered design of SSD and mechanical hard disk. The relational database stores equipment information and structured data of operation and maintenance ledger. The time series database stores water quality and process parameter time series data.

[0109] Data processing and retrieval unit: Data cleaning optimizes data quality by removing outliers, standardizing data formats, and deduplicating duplicate data; data retrieval supports multi-dimensional combined retrieval based on time, process unit, data type, and fault type.

[0110] Model optimization unit: By analyzing historical operating data and fault handling records, it optimizes the parameter thresholds of the fault diagnosis module and the control strategy of the control execution module, and generates an optimized parameter package to push to the edge.

[0111] Specifically, a redundant architecture is constructed using three or more storage nodes, interconnected via high-speed fiber optic links. Data receiving ports and MQTT protocol subscription services are configured, preset data classification rules and storage strategies are loaded, data security parameters are configured, AES-256 encryption storage mechanism is enabled, sensitive data is encrypted, and three levels of access permissions are set, with different permissions corresponding to different data viewing and operation permissions. After configuration is completed, a system self-check is initiated to verify the connectivity of storage nodes, data encryption function, and the effectiveness of access control.

[0112] Based on preset classification rules, valid data is classified and archived in three levels according to "module type - data type - time period", such as "multi-level AO process monitoring module - water quality parameters - October 1, 2024". A data fragmentation storage strategy is adopted, and the classified data packets are split into 100MB / fragment and distributed to various nodes. At the same time, a data index table is generated to record the data storage location, fragmentation information and access path. Monitoring data with high real-time requirements is stored in the high-speed cache area. Non-real-time data such as historical operation data and diagnostic reports are stored in the large-capacity archive area.

[0113] See Figure 9 As shown, the remote operation and maintenance module specifically includes:

[0114] Command interaction unit: Built-in hierarchical permission management module, which divides permissions into three levels: administrator, maintenance personnel and operator. Only the administrator can issue maintenance commands. After the command is issued, the execution status is fed back in real time. If the execution fails, a retry mechanism and anomaly alarm are triggered.

[0115] Equipment maintenance management unit: Pre-stores model parameters, maintenance cycle and spare parts information of aeration blowers and return pumps, and generates periodic maintenance plans based on equipment runtime, failure frequency and load rate data;

[0116] Remote upgrade and maintenance support unit: It adopts differential upgrade technology to remotely update the firmware of core modules, only transmitting the difference data of the upgrade package, and has a built-in local maintenance interface for on-site debugging and emergency fault handling.

[0117] Specifically, managers can log in to the platform via the web or mobile app and initiate a real-time data acquisition request. The platform retrieves operational status data, fault diagnosis reports, early warning logs, and control execution records from the cloud data management module through the cloud API interface. Data synchronization adopts a "real-time incremental synchronization + scheduled full synchronization" mechanism.

[0118] Based on data such as device uptime, fault records, and parameter decay trends stored in the cloud, the module automatically generates personalized maintenance plans, clearly defining the maintenance objects, maintenance content, planned time, responsible personnel, and required tools and consumables. The maintenance plan is pushed to the corresponding maintenance personnel's mobile APP, and the maintenance personnel respond with "confirm execution" or "request adjustment" after receiving it. After the maintenance is completed, a maintenance acceptance report is generated.

[0119] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0120] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0121] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A real-time monitoring and fault diagnosis system for multi-level AO wastewater treatment based on the Internet of Things, characterized in that, include: Multi-stage AO process monitoring module: Deployed in the anaerobic tank, anoxic tank, aerobic tank, and secondary sedimentation tank of the multi-stage AO wastewater treatment process, including multi-dimensional sensing components and adaptive sampling units; The IoT data transmission module adopts a three-level transmission architecture of "edge node-gateway-cloud", which gathers and pre-filters data from various sensors nearby, encrypts the transmitted data using encryption algorithms, and converts the private data protocol into the MQTT standard protocol. Edge computing data processing module: Removes random noise from sensor data, intelligently completes missing data based on the material balance principle of multi-level AO process, and generates preliminary control instructions; Fault diagnosis module: It extracts features from preprocessed data and historical operation data using deep convolutional neural network algorithm, combines gradient boosting tree algorithm to accurately classify fault types, builds a correlation model between faults and causes, and generates diagnostic results; Control and execution module: Connects to the execution equipment of multi-stage AO process, and performs aeration control, stirring control, reflux ratio control and sludge discharge control based on preliminary control commands and diagnostic results; Early warning push module: Based on the diagnostic results of the fault diagnosis module, multi-level early warnings are set, with each level corresponding to different early warning thresholds and processing priorities. Early warning information is pushed through various push methods. Cloud-based data management module: It adopts a distributed storage architecture to securely store all data, performs multi-dimensional retrieval based on time, process, and fault type, builds a historical data mining model, and optimizes the parameters and control strategy thresholds of the fault diagnosis model by analyzing historical operation data and fault handling data. Remote operation and maintenance module: View the running status, fault diagnosis results and early warning information of each module in real time through the web and APP, remotely issue operation and maintenance instructions, and generate equipment maintenance plans based on equipment data.

2. The IoT-based multi-level AO wastewater treatment real-time monitoring and fault diagnosis system according to claim 1, characterized in that, The multi-level AO process monitoring module specifically includes: Multi-dimensional sensing component unit: includes dissolved oxygen sensor, COD sensor, ammonia nitrogen sensor, total nitrogen sensor, sludge concentration sensor, pH sensor and tank level sensor, each sensor is equipped with PTFE composite anti-fouling membrane; Adaptive sampling unit: integrates a flow sensor and an electric lifting sampling mechanism. It sets a fluctuation threshold based on real-time flow data. When the flow fluctuation exceeds ±10% or the water quality parameter fluctuation exceeds ±5%, it automatically increases the sampling frequency. The electric lifting sampling rod mechanism is driven by a stepper motor to locate the sampling depth.

3. The IoT-based multi-level AO wastewater treatment real-time monitoring and fault diagnosis system according to claim 1, characterized in that, The IoT data transmission module specifically includes: Edge node communication unit: It adopts a dual-mode integrated design of LoRa and NB-IoT to gather and initially filter the data from various sensors nearby, and automatically switches to NB-IoT signal when LoRa signal is interrupted; Gateway processing unit: performs hardware encryption on data uploaded by edge nodes, with the key dynamically generated through the device's unique identifier, and converts the sensor's private protocol into the MQTT standard protocol in real time; Cloud communication unit: It adopts dual-mode redundant transmission of 4G / 5G and optical fiber, and performs integrity verification of transmitted data based on CRC-32 algorithm. If the verification fails, a retransmission mechanism is triggered.

4. The IoT-based multi-level AO wastewater treatment real-time monitoring and fault diagnosis system according to claim 1, characterized in that, The edge computing data processing module specifically includes: Data preprocessing unit: A sliding window filtering algorithm is used to remove random noise from sensor data. The window length is dynamically adjusted according to data fluctuations. System error correction is performed by comparing the standard reference values ​​of the calibration curve parameters of each sensor model in real time. Data completion is performed based on the parameters of adjacent processing units. Edge decision computing unit: The edge decision model is built based on a lightweight neural network algorithm. The model structure is optimized for scenarios with limited computing power at the edge, and redundant network layers are removed. Control command generation unit: It has a built-in fault symptom recognition rule library. It inputs preprocessed real-time data into the edge decision model to obtain simple fault symptoms, and generates preliminary control commands by matching the fault symptom recognition rule library.

5. The IoT-based multi-level AO wastewater treatment real-time monitoring and fault diagnosis system according to claim 1, characterized in that, The fault diagnosis module specifically includes: Feature extraction unit: Extracts features from the preprocessed data transmitted by the edge computing data processing module, including water quality parameter fluctuation coefficient, equipment operating parameter deviation value, and process unit response delay. It removes interference features through signal smoothing processing and retains key features strongly correlated with faults based on the mutual information entropy principle. Fault Classification and Localization Unit: A fusion model of deep convolutional neural network and gradient boosting tree is used to classify faults. A fault-cause association network is constructed through knowledge graph to locate the source of the fault and generate diagnostic results containing fault type, scope of impact and handling steps. Diagnostic model iteration and storage unit: caches fault data and diagnostic results from the past month, receives model parameters optimized from the cloud based on historical data in real time, and performs incremental updates to the model.

6. The IoT-based multi-level AO wastewater treatment real-time monitoring and fault diagnosis system according to claim 1, characterized in that, The control execution module specifically includes: Aeration control unit: It adopts an industrial-grade vector frequency converter and a high-efficiency aeration head array. It receives control commands through a 4-20mA analog signal. Each aerobic tank is divided into zones with independent aeration branches. The aeration volume of each zone is controlled by the branch solenoid valve. The stirring control unit is a submersible low-speed impeller mixer with a propeller-type impeller. The impeller speed is adjusted to control the impeller intensity. Reflux ratio control unit: integrates an electromagnetic flow sensor and a variable frequency reflux pump. The flow sensor collects the reflux flow of nitrification liquid and sludge in real time, and the reflux pump adjusts the pump speed to achieve dynamic control of the reflux ratio. Sludge discharge control unit: The sludge discharge action is triggered by real-time data from the sludge concentration sensor. When the sludge concentration exceeds the set threshold, the opening degree is increased; when it is below the threshold, the opening degree is decreased.

7. The IoT-based multi-level AO wastewater treatment real-time monitoring and fault diagnosis system according to claim 1, characterized in that, The early warning push module specifically includes: Multi-level early warning judgment unit: Based on the diagnosis results of the fault diagnosis module, the degree of parameter exceedance of the edge computing data processing module, and combined with the operating threshold benchmarks of each unit of the multi-level AO process, the first, second and third level early warning levels are determined; Multi-channel push execution unit: Construct a multi-channel redundant push architecture that integrates GSM SMS, 4G communication, audible and visual alarms and industrial Ethernet, and customize push content and push frequency based on the identity of the recipient; Early warning ledger and feedback unit: Records the early warning time, level, triggering reason, push target, reception status, and processing result in real time to form a traceable early warning ledger. Operation and maintenance personnel can provide feedback on the early warning processing progress through the APP.

8. The IoT-based multi-level AO wastewater treatment real-time monitoring and fault diagnosis system according to claim 1, characterized in that, The cloud-based data management module specifically includes: Distributed data storage unit: It adopts a hybrid architecture of "distributed file storage + relational database + time series database". The distributed file storage adopts a layered design of SSD and mechanical hard disk. The relational database stores equipment information and structured data of operation and maintenance ledger. The time series database stores water quality and process parameter time series data. Data processing and retrieval unit: Data cleaning optimizes data quality by removing outliers, standardizing data formats, and deduplicating duplicate data; data retrieval supports multi-dimensional combined retrieval based on time, process unit, data type, and fault type. Model optimization unit: By analyzing historical operating data and fault handling records, it optimizes the parameter thresholds of the fault diagnosis module and the control strategy of the control execution module, and generates an optimized parameter package to push to the edge.

9. The IoT-based multi-level AO wastewater treatment real-time monitoring and fault diagnosis system according to claim 1, characterized in that, The remote operation and maintenance module specifically includes: Command interaction unit: Built-in hierarchical permission management module, which divides permissions into three levels: administrator, maintenance personnel and operator. Only the administrator can issue maintenance commands. After the command is issued, the execution status is fed back in real time. If the execution fails, a retry mechanism and anomaly alarm are triggered. Equipment maintenance management unit: Pre-stores model parameters, maintenance cycle and spare parts information of aeration blowers and return pumps, and generates periodic maintenance plans based on equipment runtime, failure frequency and load rate data; Remote upgrade and maintenance support unit: It adopts differential upgrade technology to remotely update the firmware of core modules, only transmitting the difference data of the upgrade package, and has a built-in local maintenance interface for on-site debugging and emergency fault handling.

Images