Distributed intelligent monitoring and early warning system for sewage treatment facilities based on Internet of Things

Abstract

The invention relates to the technical field of sewage treatment, and particularly discloses a sewage treatment facility distributed intelligent monitoring and early warning system based on the Internet of Things, which comprises distributed intelligent monitoring nodes, a region edge computing gateway and a central cloud platform. The monitoring node is integrated with a state self-sensing module, and is used for quantitatively evaluating self corrosion, power supply and environment health degree in real time and locally warning; the edge gateway carries out spatial correlation diagnosis on the multi-node data, and distinguishes node fault false anomalies and real water quality events; and the cloud platform performs macroscopic situation analysis and early warning decision scheduling. According to the invention, dual reliable monitoring of water quality parameters and node health states is realized, and the credibility of monitoring data and the reliability of the system are improved.

Description

Technical Field

[0001] This invention belongs to the field of wastewater treatment technology, specifically relating to a distributed intelligent monitoring and early warning system for wastewater treatment facilities based on the Internet of Things. Background Technology

[0002] As a core component of the next generation of information technology, the Internet of Things (IoT) connects various physical devices and sensors to a network, enabling comprehensive perception, reliable transmission, and intelligent processing of the physical world. It is widely applied in key areas such as industrial manufacturing, smart cities, and environmental monitoring. Wastewater treatment facility management in the environmental monitoring field is an important application of IoT technology. Its core objective is to achieve real-time, continuous, and accurate monitoring of key water quality parameters during wastewater treatment by deploying a distributed sensor network.

[0003] Existing technologies typically employ the deployment of fixed or mobile sensor nodes to collect water quality data, which is then transmitted to a central server for processing and analysis via a wireless network. However, this approach faces a series of integrated challenges in practice: sensor nodes are exposed to complex and corrosive wastewater environments for extended periods, and the corrosion rate of their core sensing units (such as pH electrodes and dissolved oxygen probes) exhibits significant non-uniformity due to fluctuations in water quality. Traditional nodes lack the ability to diagnose and assess their own physical state in real time. This leads to the unnoticed degradation or even failure of node performance, while the central system can only receive abnormal or missing data, unable to effectively distinguish between sudden changes in water quality and node malfunctions. Consequently, unpredictable blind spots and data distortion risks arise in the monitoring network. Summary of the Invention

[0004] The purpose of this invention is to provide a distributed intelligent monitoring and early warning system for wastewater treatment facilities based on the Internet of Things, so as to solve the problem in the prior art that the monitoring data is distorted and blind spots are formed because the sensor nodes cannot be perceived and diagnosed in real time due to the performance degradation or failure of the sensor nodes themselves.

[0005] The technical solution of this invention is a distributed intelligent monitoring and early warning system for wastewater treatment facilities based on the Internet of Things. This system includes distributed intelligent monitoring nodes deployed in key process units of the wastewater treatment facility, regional edge computing gateways, and a central cloud platform. By constructing a three-layer architecture integrating physical state self-sensing, multi-source data fusion diagnosis, and hierarchical collaborative early warning decision-making, the system achieves dual, synchronous, and reliable monitoring of water quality parameters in the wastewater treatment process and the health status of the monitoring nodes themselves.

[0006] The distributed intelligent monitoring node is a composite device integrating a water quality sensing module, a node status self-sensing module, an embedded microprocessor, and a first wireless communication module. The water quality sensing module collects water quality parameters at the target location, including but not limited to pH value, dissolved oxygen concentration, chemical oxygen demand, ammonia nitrogen concentration, and turbidity. The node status self-sensing module monitors the node's own physical state parameters in real time; this module further includes a sensing unit corrosion monitoring submodule, a power supply status monitoring submodule, and an environmental stress monitoring submodule.

[0007] The corrosion monitoring submodule of the sensing unit measures the rate of change of the equivalent resistance of the sensitive membrane surface in real time through a micro-resistance probe array integrated on the surface of the water quality sensor. This rate of change is negatively correlated with the corrosion rate. The power supply status monitoring submodule continuously collects the voltage, current, and internal resistance parameters of the internal batteries of the node. The environmental stress monitoring submodule collects data on the fluid impact force, internal temperature and humidity, and external pressure experienced by the node's outer shell.

[0008] The embedded microprocessor incorporates a node health fusion calculation engine. This engine receives raw water quality data from the water quality sensing module and all physical state parameters from the node status self-sensing module. The engine first preprocesses the raw water quality data, including noise filtering and outlier removal. Then, the engine performs a quantitative calculation of node health, as follows: Based on the equivalent resistance change rate output by the sensor unit corrosion monitoring submodule, combined with a preset baseline corrosion curve for the sensor model, a sensor unit corrosion health index is calculated. This index is a value between 0 and 1, where 1 represents no corrosion and 0 represents complete failure. Based on the voltage, current, and internal resistance data output by the power supply status monitoring submodule, a power supply stability index is calculated by querying the battery discharge characteristic curve and load model. Based on the impact force, temperature, humidity, and pressure data output by the environmental stress monitoring submodule, an environmental adaptability index is calculated by comparing it with preset tolerance thresholds for the node's mechanical structure and electronic components. Finally, the embedded microprocessor uses a preset weighted fusion algorithm to fuse the sensor unit corrosion health index, power supply stability index, and environmental adaptability index to generate a comprehensive node health score, which is also normalized to the range of 0 to 1.

[0009] The distributed intelligent monitoring nodes are configured with a local early warning strategy. The embedded microprocessor compares the real-time calculated node health score with multiple preset early warning thresholds. When the health score is lower than the first threshold but higher than the second threshold, the node sends a status attention signal to the regional edge computing gateway through the first wireless communication module. This signal includes the node identifier, health score, and main attenuation parameters. When the health score is lower than the second threshold, the node determines that it is in a faulty or imminent failure state. At this time, in addition to sending a fault alarm signal, the node automatically initiates a data credibility marking process, adding a low credibility flag to the water quality data subsequently collected by the water quality sensing module.

[0010] The regional edge computing gateway is deployed in various functional areas of the wastewater treatment facility, establishing communication connections with all distributed intelligent monitoring nodes within its jurisdiction via a second wireless communication module. The regional edge computing gateway has a built-in edge collaborative diagnostic engine. This engine receives water quality data, node health scores, and early warning signals from each node. The core function of the edge collaborative diagnostic engine is to perform spatial correlation diagnosis, the process of which is as follows: The engine acquires water quality data sequences and health scores from multiple monitoring nodes adjacent to the treatment processes within the same region. When a sudden change in water quality data reported by a single node is detected, and the node's health score is below a second threshold or its data has been marked as low confidence, while the water quality data of other adjacent nodes within the same region remain stable, the edge collaborative diagnostic engine diagnoses the water quality change as a false anomaly caused by the node's own failure, rather than a true deterioration in water quality. Conversely, when multiple adjacent nodes within the same region simultaneously monitor consistent trends in water quality parameters, and the health scores of these nodes are all within the normal range, the engine diagnoses it as a genuine water quality anomaly event.

[0011] The central cloud platform is connected to all regional edge computing gateways via wired or wireless wide area networks. The central cloud platform includes a data warehouse, a macro-situation analysis engine, and an early warning decision-making and dispatch center. The data warehouse is used for long-term storage of historical water quality data, historical health data, and all diagnostic event records for all nodes. Based on the long-term data in the data warehouse, the macro-situation analysis engine performs two analytical tasks: The first task is facility-level corrosion trend analysis. The engine aggregates historical corrosion health index data of all similar sensor nodes across the entire plant, uses time series analysis methods to predict the overall remaining lifespan of sensor clusters in different process units, and generates preventative maintenance recommendations. The second task is water quality-node health correlation rule mining. The engine uses machine learning algorithms to analyze the potential correlation between historical water quality fluctuation patterns and node health decay patterns, continuously optimizing the diagnostic rules in the edge collaborative diagnostic engine.

[0012] The early warning decision-making and dispatch center is the final decision-making layer of the system. It receives collaborative diagnostic results from the regional edge computing gateway and predictive information from the macro-situation analysis engine. The early warning decision-making and dispatch center implements a hierarchical early warning response mechanism: for pseudo-anomalies determined by the edge collaborative diagnostic engine to be caused by node failures, the dispatch center records the event and generates a maintenance work order for the specific node, while automatically ignoring low-reliability water quality data reported by the node during the failure period to avoid triggering false water quality alarms.

[0013] For water quality anomalies identified as genuine, the dispatch center activates the corresponding level of water quality safety warning based on the type, magnitude, and spread of the abnormal parameters, and coordinates with the central control system of the wastewater treatment facilities to provide adjustment suggestions. For preventative maintenance suggestions issued by the macro-situation analysis engine, the dispatch center incorporates them into the planned maintenance schedule.

[0014] Furthermore, the weighted fusion algorithm in the node health fusion calculation engine does not have fixed weight coefficients, but is dynamically adjusted through a self-learning mechanism. The self-learning mechanism is implemented as follows: the embedded microprocessor periodically uploads the node's health score and various sub-indices to the central cloud platform; the macro-situation analysis engine statistically analyzes a large amount of data from similar nodes before their eventual failure, and uses regression analysis to infer the contribution of each sub-indice to the eventual failure, i.e., the actual weight; the central cloud platform periodically distributes the updated weight coefficient model to each distributed intelligent monitoring node, completing the online optimization of the weighted fusion algorithm.

[0015] Furthermore, the edge collaborative diagnostic engine of the regional edge computing gateway introduces temporal consistency verification during the diagnostic process. Specifically, when faced with suspected false anomalies, the engine not only checks spatial correlations but also retrieves historical water quality data of the faulty node during the decline in its health score. It analyzes whether the data fluctuation pattern shows abnormal noise increase or trend drift before the reported abrupt change point, using this as an auxiliary judgment basis to further improve diagnostic accuracy.

[0016] Furthermore, the first wireless communication module of the distributed intelligent monitoring node adopts a dual-mode chip supporting both low-power wide-area network (LPWAN) and ad hoc network (ANR) protocols. Under normal monitoring conditions, the node uses LPWAN to communicate with the regional edge computing gateway via a single hop to conserve energy. When a partial network outage or gateway failure occurs, the node can automatically switch to ANR to form a multi-hop relay network with neighboring nodes, relaying critical status alarm data to other reachable gateways or directly uploading it to the central cloud platform, ensuring that early warning information is not lost.

[0017] Compared with existing technologies, the beneficial effects of this invention are as follows: 1. By integrating a node status self-sensing module and a health fusion calculation engine into each distributed intelligent monitoring node, this invention endows sensor nodes with real-time self-diagnosis capabilities. Nodes can quantitatively assess their performance degradation caused by corrosion, power supply issues, and environmental stress, and issue early warnings. This enables the monitoring system to clearly distinguish whether data anomalies originate from water quality changes or node failures, effectively eliminating monitoring blind spots and data distortion risks caused by latent node failures, and greatly improving the data reliability and trustworthiness of the entire monitoring network.

[0018] 2. This invention constructs a three-layer intelligent analysis architecture: local node early warning, edge collaborative diagnosis, and cloud-based macro-level decision-making. The regional edge computing gateway, through spatial correlation diagnosis and temporal consistency verification, can quickly and accurately distinguish between false anomalies and real events at the network edge, reducing the processing burden on the central cloud platform and improving response speed. The central cloud platform focuses on macro-level situation analysis and model optimization, achieving strategic management from micro-node health to macro-level facility maintenance. This layered collaborative processing mechanism balances real-time performance, accuracy, and global optimization, forming a complete closed loop of monitoring, diagnosis, early warning, and maintenance decision-making.

[0019] 3. This invention introduces dynamic self-learning and network self-healing mechanisms. The node health assessment model can be optimized online through cloud-based big data analysis, allowing the system assessment standards to continuously approach real failure patterns with accumulated experience, thus possessing continuous evolution capabilities. Simultaneously, the dual-mode design of the communication network ensures robust information transmission under partial failure conditions. These designs enable the system not only to cope with known node decay patterns but also to adapt to complex and changing field environments, exhibiting high adaptability and reliability, providing a solid technical guarantee for the long-term, stable, and intelligent operation of wastewater treatment facilities. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the overall technical architecture of the distributed intelligent monitoring and early warning system for wastewater treatment facilities based on the Internet of Things proposed in this invention.

[0021] Figure 2 This is a schematic diagram of the core principle framework of the node health fusion calculation engine in this invention.

[0022] Figure 3 This is a logical flow diagram of the edge collaborative diagnostic engine performing spatial and temporal correlation diagnostics in this invention.

[0023] Figure 4 This is a schematic diagram of the three-level interaction relationship and data flow of node local early warning, edge collaborative diagnosis and cloud macro decision-making in this invention.

[0024] Figure 5 This is a schematic diagram illustrating the principle framework of the system self-learning optimization and network self-healing mechanism in this invention. Detailed Implementation

[0025] This invention provides a distributed intelligent monitoring and early warning system for wastewater treatment facilities based on the Internet of Things (IoT). Please refer to the appendix. Figures 1 to 5 The system's overall architecture consists of distributed intelligent monitoring nodes deployed in key process units of the wastewater treatment facility, regional edge computing gateways deployed in different functional areas within the facility, and a central cloud platform located in the remote monitoring center. These three layers are tightly connected via wireless and wired networks, forming a complete closed loop integrating data acquisition, edge intelligent diagnostics, and cloud-based macro-level decision-making. The system's core innovation lies in achieving dual, synchronous, and reliable monitoring and diagnostics of both water quality parameters during the wastewater treatment process and the health status of the monitoring nodes themselves. This effectively distinguishes between genuine water quality anomalies and false anomalies caused by sensor node performance degradation or malfunction, ensuring the authenticity and reliability of the monitoring data.

[0026] Please continue to refer to the appendix. Figure 1 The distributed intelligent monitoring node is the front-end data sensing and primary intelligent processing unit of the system. Physically, this node is a sealed enclosure with a certain level of protection, integrating a water quality sensing module, a node status self-sensing module, an embedded microprocessor, and a primary wireless communication module. The node is mechanically mounted at specific monitoring points in the wastewater treatment facility, such as the inlet, biochemical reaction tank, sedimentation tank outlet, and disinfection tank—key process units. Its sensing probe is either submerged in the water being measured or in contact with the main fluid through a flow tank.

[0027] The water quality sensing module is the part of the node that performs its core monitoring functions. This module integrates various water quality parameter sensors to collect real-time physicochemical indicators of the water body at the target location. Typical sensor types include, but are not limited to, pH sensors, dissolved oxygen sensors, chemical oxygen demand (COD) sensors, ammonia nitrogen sensors, and turbidity sensors.

[0028] Each sensor is calibrated, and its analog or digital signal output is directly connected to the dedicated data acquisition interface of the embedded microprocessor. Taking the pH sensor as an example, it uses the glass electrode method. The millivolt-level potential signal generated by the electrode is converted into a digital quantity by a high-precision analog-to-digital converter inside the node. This digital quantity is then converted into a specific pH value reading according to a preset Nernst equation calibration curve. Data acquisition from all water quality sensors follows a unified timing schedule, triggered by a timer within the embedded microprocessor. Data is acquired synchronously or sequentially at fixed sampling intervals, such as every 5 minutes or every 10 minutes. The acquired raw data packets contain timestamps, node identifiers, and the raw measured values ​​of each parameter.

[0029] The node status self-sensing module is a key component of this invention, enabling monitoring nodes to perform self-diagnostic functions. This module is not an independent peripheral device, but rather consists of a series of miniature sensors and detection circuits integrated into key locations within the node, specifically designed for real-time monitoring of the node's physical state and health parameters. The node status self-sensing module is further subdivided into three functional sub-modules: a sensor unit corrosion monitoring sub-module, a power supply status monitoring sub-module, and an environmental stress monitoring sub-module.

[0030] The core task of the corrosion monitoring submodule of the sensing unit is to quantitatively assess the degree of corrosion or fouling of the sensitive elements of the water quality sensor. In specific implementations, a set of micron-sized resistive probe arrays is integrated onto the surfaces of key sensitive membranes, such as the oxygen-permeable membrane of the dissolved oxygen electrode or the glass membrane of the pH electrode. This array is fabricated as interdigitated electrodes in or near the non-functional region of the sensor's sensitive membrane, and its electrode material is completely consistent with the working environment of the sensitive membrane. During node operation, the embedded microprocessor periodically applies a constant microcurrent to the resistive probe array, for example, once per hour, and measures the voltage drop across it, thereby calculating the equivalent resistance of the probe array.

[0031] As the sensitive membrane gradually wears away due to chemical corrosion, biofouling, or mechanical abrasion, the conductive paths of the micro-resistance probes covering it also change, leading to a change in the equivalent resistance value. The embedded microprocessor records the resistance value measured each time and calculates its rate of change relative to an initial reference resistance value. This rate of change in resistance is negatively correlated with the actual corrosion or wear rate of the sensitive membrane; that is, an increased rate of change in resistance usually indicates more severe corrosion. The output of this submodule is a continuous time-series data of the equivalent resistance change rate.

[0032] The power supply status monitoring submodule is responsible for comprehensively monitoring the energy supply status of the node. The input of this module is directly connected to the node's internal power supply system, typically a combination of rechargeable lithium-ion battery packs or solar panels and energy storage batteries. The power supply status monitoring submodule includes voltage sampling circuits, current sampling circuits, and internal resistance measurement circuits. The voltage sampling circuit measures the total output voltage of the battery pack in real time using a high-impedance voltage divider network, with an accuracy down to the millivolt level.

[0033] The current sampling circuit employs a precision sampling resistor or Hall effect current sensor to monitor the total load current of the node in real time during both operating and sleep states. The internal resistance measurement circuit estimates the battery's internal equivalent resistance by injecting a brief high-frequency AC test signal and measuring the voltage response at the battery terminals. The embedded microprocessor periodically acquires complete power supply parameters—voltage, current, and estimated internal resistance—for example, every 30 minutes, and records the trends of these parameters over time.

[0034] The environmental stress monitoring submodule is used to sense the stress impact of the external physical environment on the node. This submodule integrates multiple miniature sensors, including a miniature triaxial accelerometer or force sensor to monitor the instantaneous impact force and continuous vibration caused by fluid impact on the node shell; a temperature and humidity sensor installed inside the node shell to monitor the temperature and relative humidity of the environment in which the electronic components inside the node are located; and a pressure sensor installed on the outside of the node shell or connected to the external water body through a pressure guide hole to monitor the water pressure or atmospheric pressure at the node installation depth. These sensors continuously or periodically send analog signals to an analog-to-digital converter, which reads and converts them into digital quantities by an embedded microprocessor. For example, the triaxial accelerometer data is used to calculate the amplitude and frequency of the fluid impact force; the temperature and humidity data is used to determine whether condensation or overheating occurs inside the node; and the pressure data can be used to assist in verifying whether the installation depth has changed or to monitor water level fluctuations.

[0035] Embedded microprocessors are the computing and control core of distributed intelligent monitoring nodes, typically employing low-power microcontroller units. Internally, they integrate the node operating system, communication protocol stack, and, most critically, the node health fusion computing engine. Please refer to the appendix. Figure 2 This engine is the core algorithm for performing node self-health assessment. The node health fusion calculation engine runs continuously, and its input data stream includes the raw water quality data stream from the water quality sensing module and the complete physical state parameter stream output from the three sub-modules of the node state self-sensing module.

[0036] The engine first preprocesses the input raw water quality data. The preprocessing stage includes two main steps: noise filtering and outlier removal. Noise filtering uses a moving average filter or a Kalman filter to smooth the continuous sampling sequence of each water quality parameter, such as pH, to suppress short-term high-frequency fluctuations caused by water turbulence or electronic noise. Outlier removal uses statistical methods, such as calculating the mean and standard deviation of the data within the current sampling window. Data points deviating from the mean by more than three times the standard deviation are identified as outliers and removed, replaced by the previous valid value or an interpolated value. The preprocessed water quality data is marked as valid data, awaiting subsequent transmission or local use.

[0037] Subsequently, the engine initiates the node health metric calculation process. This process calculates three sub-health indices in parallel and then performs a weighted fusion.

[0038] The first item is the calculation of the sensor unit corrosion health index. The engine reads the latest equivalent resistance change rate data reported by the sensor unit corrosion monitoring submodule and records it as... The engine internally stores a preset baseline corrosion curve for the specific sensor model configured for that node. This curve was obtained through accelerated aging tests in the laboratory, establishing a mapping relationship from the rate of change of equivalent resistance to the percentage of sensor performance degradation or remaining lifetime. The calculation process involves the engine using the current... Substituting the values ​​into the inverse function of the baseline corrosion curve or a lookup table yields the raw score characterizing the health status of the sensor's sensitive membrane. Its value ranges from 0 to 1. A value of 1 indicates that the sensor membrane is brand new or undamaged, with no performance degradation; a value of 0 indicates that the sensor membrane is completely ineffective and cannot provide valid measurements. To smooth out short-term fluctuations, the engine may... After a period of moving average processing, the final output is the corrosion health index of the sensing unit. .

[0039] The second item is the calculation of the power supply stability index. The engine reads the voltage V, current I, and internal resistance reported by the power supply status monitoring submodule. The engine internally stores the discharge characteristic curve model of the battery used by the node and the typical load power consumption model of the node. The calculation process is divided into several sub-steps. First, based on the current voltage V, the discharge characteristic curve is queried to assess the remaining percentage of battery charge. Second, combined with the current load current I, the estimated continuous power supply time of the battery is calculated. Third, based on the internal resistance... The ratio of the remaining capacity to the initial internal resistance is used to assess the degree of battery aging. Finally, the three sub-indicators—remaining capacity, sustainable power supply time, and degree of aging—are combined using a preset scoring function to generate a power supply stability index. Its value range is also normalized to 0 to 1. A value of 1 represents sufficient and stable power supply and good battery condition; a value of 0 represents power interruption or battery depletion.

[0040] The third item is the calculation of the environmental adaptability index. The engine reads the impact force reported by the environmental stress monitoring submodule. Internal temperature Internal humidity And external pressure P data. The engine's internal mechanical structures and electronic components have preset tolerance thresholds, including the maximum permissible impact force. Operating temperature range to Maximum permissible humidity and pressure sensor range to The calculation process involves independently evaluating each environmental parameter and taking the worst-case value. For example, the temperature adaptation score is calculated as follows: ,in To determine the optimal operating temperature; calculate the impact force adaptor fraction. The environmental adaptability index is generated by taking the minimum value of all sub-scores or by performing a weighted average. The value ranges from 0 to 1. A value of 1 indicates that the environmental stress is completely within the ideal range; a value of 0 indicates that at least one environmental parameter has exceeded the safety threshold, and the node faces the risk of immediate damage.

[0041] After obtaining the three sub-indices , , Then, the node health fusion calculation engine performs the final weighted fusion. The fusion algorithm uses a linear weighted sum model, and its calculation formula is as follows: Where H represents the final comprehensive node health score, the value of which is normalized and truncated within a closed interval of 0 to 1. , , These represent the weighting coefficients for the corrosion health index, power supply stability index, and environmental adaptability index of the sensing unit, respectively, and satisfy the following conditions: + + =1. The weighting coefficient is not fixed; its dynamic adjustment mechanism will be explained in detail later in conjunction with the functions of the central cloud platform. In the initial state, the weighting coefficient can be set based on expert experience, for example... =0.5, =0.3, =0.2, emphasizing the critical impact of the sensor's corrosion status on data quality. Each time the embedded microprocessor completes the calculation of the health score H, for example, once per hour, it stores it along with the various sub-indices in the node's local cache.

[0042] The distributed intelligent monitoring nodes are configured with a local early warning strategy, which is the first-level proactive early warning mechanism of the system. Two early warning thresholds are preset within the embedded microprocessor: a first threshold TH1 and a second threshold TH2, where TH1 > TH2. For example, TH1 can be set to 0.7 and TH2 to 0.4. Each time a new node health score H is calculated, the engine immediately compares it with these two thresholds. If H is lower than TH1 but higher than TH2, it indicates a slight decline in the node's health status, but it can still maintain basic functionality. At this time, the node sends a "status attention" signal to the regional edge computing gateway of its area via its first wireless communication module. This signal data packet contains the node's unique identifier ID, the current timestamp, the health score H value, and the main attenuation parameters that caused the score decline, such as a significantly reduced [specific parameters]. value or This signal is intended to alert the upper-level system to the state trend of this node, but does not require immediate intervention.

[0043] If the health score H drops below the second threshold TH2, the node determines that it is in a critical state of serious failure or impending failure. At this time, the node immediately performs two key operations. First, it sends a high-priority "fault alarm" signal to the regional edge computing gateway through the first wireless communication module. This signal contains more detailed diagnostic information. Second, and more importantly, the node automatically initiates a data credibility marking process. The embedded microprocessor sets a flag bit on its internal data flow processing path. Before packaging and sending all water quality data subsequently collected and preprocessed by the water quality sensing module, it automatically attaches a "low credibility" flag. This flag is stored and transmitted as metadata along with the water quality measurement value, clearly informing the data recipient that the node's own state is abnormal at this stage, the reliability of the water quality data it provides is questionable, and it should be used with caution or further verification is required.

[0044] The first wireless communication module is crucial for the connection between the node and the upper-layer network. This module employs a dual-mode communication chip that supports both low-power wide-area network (LPWAN) protocols and ad hoc network protocols. Under most normal monitoring conditions, the node uses LPWAN protocols for direct single-hop communication with the regional edge computing gateway. This mode consumes extremely low power, making it suitable for long-term operation of battery-powered sensor nodes, and can stably transmit periodic water quality data, health reports, and warning signals to the gateway. Communication data packets are encapsulated using a lightweight application-layer protocol, including a frame header, node ID, data length, payload, and cyclic redundancy check (CRC) code to ensure transmission integrity.

[0045] Please continue to refer to the appendix. Figure 1 The regional edge computing gateway constitutes the second layer of the system, namely the edge intelligent collaborative diagnostic layer. The gateway hardware is deployed in a relatively central location within the wastewater treatment plant, such as the power distribution room, control room, or a dedicated edge computing box. Each gateway is responsible for covering a specific process area, such as the pretreatment area, biochemical reaction area, or sludge-water separation area. The regional edge computing gateway establishes and maintains a stable star network connection with all distributed intelligent monitoring nodes deployed within its wireless signal coverage area through its second wireless communication module. The second wireless communication module is compatible with low-power wide-area network protocols and is used to receive uplink data from each node, while also issuing configuration commands or model updates. The gateway typically has a wired Ethernet interface or a cellular network module for uplink connection with the central cloud platform.

[0046] The core of a regional edge computing gateway is its internally running edge collaborative diagnostic engine. This engine is a software application running on the gateway's edge computing unit, and its core task is to quickly identify the authenticity of water quality data anomalies by leveraging the spatial correlation of data from multiple monitoring nodes within the region. Please refer to the appendix. Figure 3 With appendix Figure 4The edge collaborative diagnostic engine continuously monitors and receives data streams from all nodes within its jurisdiction, including periodic water quality data packets, node health score reports, and various early warning signals.

[0047] The core function of the edge collaborative diagnostic engine is to perform spatial correlation diagnosis. Its diagnostic logic flow is as follows: The engine maintains a regional data view in memory, which is updated in real time and includes the latest water quality parameter values, node health score H, and data reliability flags for each node within the region. When the engine detects a sudden change in water quality data reported by a specific node—for example, if the ammonia nitrogen concentration of a node spikes by more than 50% of its historical average over two consecutive sampling periods—the diagnostic analysis process is triggered. The engine first checks the status information of the node with the change. If it finds that the node's health score H is below the second threshold TH2, or its reported water quality data has been marked as "low reliability," it initially suspects that the water quality change may originate from a node malfunction. To verify this suspicion, the engine immediately retrieves data from other monitoring nodes within the same process area that are adjacent or close to each other in the wastewater treatment process. For example, multiple monitoring nodes are typically deployed within the same aerobic biological treatment tank. The engine checks the monitoring data of these adjacent nodes for the same water quality parameter, ammonia nitrogen concentration, within the same time period.

[0048] The diagnostic conclusion is generated based on the following rules: If the mutation node is abnormal, while the ammonia nitrogen concentration data of multiple adjacent nodes in the same area remain stable and do not show a consistent upward trend, the edge collaborative diagnostic engine has sufficient reason to diagnose this water quality mutation as a "pseudo-anomaly," that is, a false alarm signal caused by the failure of the specific node itself. Conversely, if multiple adjacent nodes in the area almost simultaneously detect a consistent and significant upward trend in ammonia nitrogen concentration, and the health scores of these nodes are all within the normal range (e.g., higher than TH1), and the data has no low confidence indicators, the engine diagnoses it as a "real water quality anomaly event," indicating that there is indeed a problem in the wastewater treatment process.

[0049] Furthermore, to improve diagnostic accuracy, the edge collaborative diagnostic engine introduces temporal consistency verification as an auxiliary judgment method. Please refer to the appendix. Figure 3When faced with suspected false anomalies, the engine not only performs the aforementioned spatial correlation checks but also proactively retrieves historical water quality data sequences from the central cloud platform's data warehouse or its own cached historical data during the period of declining health scores for the faulty node. The engine analyzes the data fluctuation patterns of the node in the period preceding the reported water quality mutation, such as within 24 hours. If the analysis reveals that the node's water quality data showed signs of increased abnormal noise, irregular drift in measurements unrelated to the process, or gradually weakening correlation with adjacent node data long before the health score dropped to the threshold, this provides strong temporal evidence for the judgment that "gradual degradation of node performance leads to inaccurate final data," further supporting the diagnosis of false anomalies. The combination of spatial correlation and temporal consistency significantly improves the accuracy of edge diagnosis.

[0050] After completing the diagnosis, the edge collaborative diagnostic engine generates a structured diagnostic event report. This report includes a unique event identifier, triggering node ID, a list of associated node IDs, the type of abnormal water quality parameter, the magnitude of the change, the diagnostic conclusion, the diagnostic confidence level, and a timestamp. For events diagnosed as false anomalies, the report clearly indicates the suspected cause of the fault, such as sensor corrosion or insufficient power supply. This diagnostic report is immediately uploaded to the central cloud platform via the gateway's uplink network interface.

[0051] Please continue to refer to the appendix. Figure 1 With appendix Figure 4 The central cloud platform is the third layer of the system, serving as the macro-decision-making and optimization center. Deployed on cloud servers or in a private enterprise data center, the platform establishes a secure and stable network connection with regional edge computing gateways within all wastewater treatment facilities via wired fiber optic or wireless wide area networks. Logically, the central cloud platform comprises three core components: a data warehouse, a macro-situation analysis engine, and an early warning decision-making and dispatch center.

[0052] A data warehouse is a large-scale, time-series database system used for long-term persistent storage of all data aggregated from edge computing gateways across all regions. The stored data types include raw and pre-processed historical water quality data from all distributed intelligent monitoring nodes, continuous historical node health scores and their sub-indices, all local early warning signal records from nodes, and diagnostic event reports from all edge collaborative diagnostic engines. The data warehouse employs partitioning and indexing techniques to support efficient querying and analysis by time range, facility region, node type, and parameter type, providing data support for upper-layer analysis engines.

[0053] The macro-level situation analysis engine is the core of the platform's data intelligence analysis. Based on the massive historical data accumulated in the data warehouse, it performs macro-level analysis that goes beyond a single node or event. The engine primarily performs two strategic analysis tasks.

[0054] The first task is facility-level corrosion trend analysis and predictive maintenance planning. The engine periodically, for example weekly, extracts corrosion health indices from the data warehouse for all similar sensor nodes across the entire plant, such as all dissolved oxygen sensors. Historical time-series data is used. The engine employs time-series analysis methods, such as autoregressive integral moving average models or long short-term memory networks, to aggregate and analyze this cluster data. The goal of the analysis is to predict the overall performance degradation trend of the entire sensor cluster over a future period, such as the next three months, and to estimate its mean remaining life. Based on the prediction results, the engine can automatically generate preventative maintenance recommendation reports, such as "The dissolved oxygen sensor cluster in the aerobic tank of area A is expected to drop below the attention threshold in 60 days; batch calibration or replacement is recommended after 45 days." This shift from microscopic corrosion data to macroscopic maintenance strategies realizes a transformation from passively responding to faults to proactively preventing them.

[0055] The second task is water quality-node health association rule mining and model optimization. The engine uses machine learning algorithms, such as association rule mining or gradient boosting decision trees, to analyze historical data, aiming to discover potential, undefined associations between water quality fluctuation patterns and node health degradation patterns. For example, the analysis might reveal that when the influent pH value fluctuates frequently and drastically, the corrosion health index of the pH sensor node... The rate of decline will accelerate significantly; or when the ambient temperature remains above a certain threshold, the power supply stability index of the node will decrease. Battery degradation exhibits a specific correlation with the increase in battery internal resistance. These newly discovered rules and knowledge are used to continuously optimize and update the diagnostic rule base in the edge collaborative diagnostic engine, and to adjust the parameters of the node local health assessment model, enabling the system to learn from historical experience and improve itself. Specifically, this engine is responsible for implementing the dynamic self-learning optimization of the weight coefficients in the node health fusion calculation model. The specific process is attached. Figure 5 As shown: The embedded microprocessor of each node periodically sends the node's health score H and its various sub-indices. , , Uploaded to the central cloud platform.

Claims

1. A distributed intelligent monitoring and early warning system for wastewater treatment facilities based on the Internet of Things, characterized in that, include: Distributed intelligent monitoring nodes deployed in key process units of wastewater treatment facilities; Regional edge computing gateways deployed in various functional areas of wastewater treatment facilities; and the central cloud platform; The regional edge computing gateway establishes a communication connection and has a built-in edge collaborative diagnostic engine; The central cloud platform is connected to all regional edge computing gateways via wired or wireless wide area networks.

2. The distributed intelligent monitoring and early warning system for wastewater treatment facilities based on the Internet of Things as described in claim 1, characterized in that, The distributed intelligent monitoring node includes a water quality sensing module, a node status self-sensing module, an embedded microprocessor, and a first wireless communication module. The water quality sensing module is used to collect water quality parameters at the target location; The node status self-sensing module is used to monitor the physical status parameters of the node itself in real time. The node status self-sensing module includes a sensing unit corrosion monitoring submodule, a power supply status monitoring submodule, and an environmental stress monitoring submodule. The corrosion monitoring submodule of the sensing unit is used to measure the rate of change of equivalent resistance on the surface of the sensitive membrane in real time through a micro-resistance probe array integrated on the surface of the sensitive membrane of the water quality sensor. The power supply status monitoring submodule is used to continuously collect the voltage, current and internal resistance parameters of the battery inside the node; The environmental stress monitoring submodule is used to collect data on the fluid impact force, internal temperature and humidity, and external pressure borne by the node shell.

3. The distributed intelligent monitoring and early warning system for wastewater treatment facilities based on the Internet of Things as described in claim 2, characterized in that, The embedded microprocessor contains a node health fusion calculation engine; The node health fusion calculation engine is used to receive raw water quality data from the water quality sensing module and all physical state parameters from the node state self-sensing module, preprocess the raw water quality data, and then perform node health quantification calculation. The node health quantification calculation process is as follows: Based on the equivalent resistance change rate output by the sensor unit corrosion monitoring submodule, combined with a preset benchmark corrosion curve, the sensor unit corrosion health index is calculated; based on the voltage, current, and internal resistance data output by the power supply status monitoring submodule, the power supply stability index is calculated by querying the battery discharge characteristic curve and load model; based on the impact force, temperature, humidity, and pressure data output by the environmental stress monitoring submodule, the environmental adaptability index is calculated by comparing it with the preset tolerance thresholds of the node mechanical structure and electronic components; finally, the sensor unit corrosion health index, power supply stability index, and environmental adaptability index are fused using a preset weighted fusion algorithm to generate a comprehensive node health score. The embedded microprocessor is also used to execute a local early warning strategy, comparing the real-time calculated node health score with a preset first threshold and a second threshold; when the health score is lower than the first threshold but higher than the second threshold, a status attention signal is sent to the regional edge computing gateway through the first wireless communication module; when the health score is lower than the second threshold, a fault alarm signal is sent to the regional edge computing gateway through the first wireless communication module, and a low confidence flag is added to the water quality data subsequently collected by the water quality sensing module.

4. The distributed intelligent monitoring and early warning system for wastewater treatment facilities based on the Internet of Things as described in claim 3, characterized in that, The edge collaborative diagnostic engine is used to receive water quality data, node health scores and early warning signals from each node, and to perform spatial correlation diagnosis. The spatial correlation diagnosis process is as follows: Water quality data sequences and health scores of multiple monitoring nodes within the same region and adjacent to the treatment process are acquired; when a sudden change in water quality data reported by a single node is detected, and the node's health score is below a second threshold or its data has been marked as low confidence, while the water quality data of other adjacent nodes within the same region remain stable, the water quality change is diagnosed as a pseudo-anomaly caused by the node's own failure; when multiple adjacent nodes within the same region simultaneously monitor consistent water quality parameter change trends, and the health scores of these nodes are all within the normal range, it is diagnosed as a genuine water quality anomaly event.

5. The distributed intelligent monitoring and early warning system for wastewater treatment facilities based on the Internet of Things as described in claim 4, characterized in that, The central cloud platform includes a data warehouse, a macro-situation analysis engine, and an early warning decision-making and dispatch center. The data warehouse is used for long-term storage of historical water quality data, historical health data, and all diagnostic event records for all nodes; The macro-situation analysis engine is used to perform facility-level corrosion trend analysis and water quality-node health association rule mining based on long-term data in the data warehouse. The early warning decision-making and scheduling center is used to receive collaborative diagnostic results from the regional edge computing gateway and predictive information from the macro-situation analysis engine, and to implement a hierarchical early warning response mechanism.

6. The distributed intelligent monitoring and early warning system for wastewater treatment facilities based on the Internet of Things as described in claim 5, characterized in that, The weighted fusion algorithm in the node health fusion calculation engine has its weight coefficients dynamically adjusted through a self-learning mechanism. The self-learning mechanism is implemented as follows: the embedded microprocessor periodically uploads the health score and various sub-indices of its node to the central cloud platform; the macro-situation analysis engine statistically analyzes a large amount of data on similar nodes before their eventual failure in the background, and uses regression analysis to infer the contribution of each sub-indice to the eventual failure, i.e., the actual weight; the central cloud platform periodically distributes the updated weight coefficient model to each distributed intelligent monitoring node to complete the online optimization of the weighted fusion algorithm.

7. The distributed intelligent monitoring and early warning system for wastewater treatment facilities based on the Internet of Things as described in claim 6, characterized in that, The edge collaborative diagnostic engine introduces time-domain consistency verification during the diagnostic process; The temporal consistency verification process is as follows: When faced with a suspected false anomaly, retrieve the historical water quality data of the fault node during the decline of its health score, and analyze whether the data fluctuation pattern shows abnormal noise increase or trend drift earlier than the reported mutation point, so as to serve as an auxiliary judgment basis.

8. The distributed intelligent monitoring and early warning system for wastewater treatment facilities based on the Internet of Things as described in claim 7, characterized in that, The first wireless communication module adopts a dual-mode chip that supports low-power wide-area network protocols and self-organizing network protocols; Under normal monitoring conditions, the node uses a low-power wide-area network protocol to communicate with the regional edge computing gateway via a single hop. When a partial network outage or gateway failure occurs, the node automatically switches to the self-organizing network protocol and forms a multi-hop relay network with neighboring nodes to relay critical status alarm data to other reachable gateways or directly upload it to the central cloud platform.

9. The distributed intelligent monitoring and early warning system for wastewater treatment facilities based on the Internet of Things as described in claim 8, characterized in that, The corrosion health index, power supply stability index, and environmental adaptability index of the sensing unit are all values ​​between 0 and 1, where 1 represents the best condition and 0 represents complete failure; the node health score is also normalized to the range of 0 to 1.

10. The distributed intelligent monitoring and early warning system for wastewater treatment facilities based on the Internet of Things as described in claim 9, characterized in that, The status attention signal includes node identifier, health score, and main attenuation parameters.

Images