An intelligent water conservancy internet-of-things method and system based on artificial intelligence

Abstract

The application discloses a kind of water conservancy intelligent internet-of-things method and system based on artificial intelligence, comprising: real-time acquisition water conservancy equipment state data by field sensor and low-power wireless transmission;Edge computing node obtains equipment preliminary abnormal diagnosis information based on hydrological model real-time analysis data;Data is encrypted by using hardware platform loaded with water conservancy Hong Meng OS and national encryption algorithm;Data is fused and analyzed by using end-edge-cloud three-layer collaborative fusion analysis mechanism, and equipment health intelligent diagnosis model is constructed to identify equipment abnormal state in real time;Video AI model analyzes field equipment operation video in real time, and obtains visual diagnosis result;Device abnormal alarm information is output by using intelligent management platform, and equipment operation and maintenance scheme is automatically formed.The application realizes the accurate perception and efficient fault disposal of water conservancy equipment operation state, and improves the intelligent level of water conservancy operation and maintenance.

Description

Technical Field

[0001] This invention relates to the field of smart water conservancy technology, and in particular to a smart water conservancy IoT method and system based on artificial intelligence. Background Technology

[0002] In recent years, with the continuous integration of artificial intelligence and Internet of Things (IoT) technologies in the water conservancy field, smart water conservancy IoT technology has gradually become an important development direction in water conservancy engineering and water resources management. Currently, the monitoring and maintenance of water conservancy equipment operation status mostly relies on a combination of manual on-site inspections and traditional monitoring equipment. Data is collected in real time through a single type of sensor, and the equipment operation status is then analyzed and judged manually on-site or remotely. This manual inspection and single-sensor technology approach cannot comprehensively and accurately identify equipment anomalies, and the response speed is slow, making it difficult to meet the needs of accurate monitoring and timely maintenance of water conservancy facilities operation status.

[0003] Existing technologies include some solutions that combine IoT devices and sensors for remote monitoring. These solutions primarily use conventional wireless communication technology to upload basic operational status data from the equipment to a remote monitoring platform, where it is then manually analyzed by back-end personnel. However, these solutions are limited to basic data collection and transmission, lacking effective data fusion processing and intelligent analysis methods. This makes it impossible to make rapid and accurate judgments about equipment anomalies, leading to delayed fault detection and low maintenance efficiency. Furthermore, most existing water conservancy monitoring equipment does not employ secure data transmission mechanisms, posing security risks to data during remote transmission. In addition, most current monitoring systems fail to effectively utilize video data, relying solely on numerical sensor data. This makes it difficult to accurately identify subtle changes in the equipment's appearance and operational status, reducing the accuracy and reliability of monitoring and hindering the further development of smart water conservancy IoT technology.

[0004] Therefore, how to provide a water conservancy intelligent IoT method and system based on artificial intelligence is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] One objective of this invention is to propose an AI-based intelligent IoT method and system for water conservancy. Addressing the issues of low accuracy in anomaly identification and slow response speed in existing technologies for diagnosing the operational status of water conservancy equipment, this invention proposes an integrated technical solution that combines real-time data acquisition by on-site sensors, hydrological model diagnosis via edge computing nodes, secure transmission via the HarmonyOS hardware platform, end-edge-cloud collaborative fusion analysis, real-time identification by a health intelligent diagnostic model, visual diagnosis via a video AI model, and automated operation and maintenance via a honeycomb intelligent management platform. This invention offers advantages such as accurate equipment status identification, rapid response to anomalies, and intelligent operation and maintenance decision-making.

[0006] An artificial intelligence-based intelligent IoT method for water conservancy according to an embodiment of the present invention includes:

[0007] Real-time data collection of water conservancy equipment operation status is used, and the data is transmitted using a LoRa gateway and LoRa-MCU to form a preliminary sensing data set.

[0008] By using the hydrological model analysis of edge computing nodes, preliminary sensing data sets are analyzed to obtain preliminary equipment anomaly diagnosis information;

[0009] The initial abnormality diagnosis information of the equipment is transmitted to the hardware platform equipped with the Water Conservancy HarmonyOS for data encryption and encapsulation processing, forming encrypted data transmission information;

[0010] A three-layer collaborative analysis mechanism involving the edge, cloud, and device is used to fuse and analyze the data transmission information to obtain a set of device status characteristic parameters.

[0011] A smart diagnostic model for equipment health is constructed based on a set of equipment status characteristic parameters. An abnormal equipment status is identified through multi-parameter fusion training, and the diagnostic results and risk warning information for the abnormal equipment status are obtained.

[0012] Based on the water conservancy video AI model, real-time video recognition and analysis of video data are performed to obtain visual diagnostic results and abnormal event warning information;

[0013] The abnormal equipment status diagnosis results, risk warning information, visual diagnosis results and abnormal event warning information are simultaneously input into the Honeycomb Intelligent Management Platform, and the abnormal equipment alarm information and comprehensive diagnostic analysis report are output.

[0014] An automatic classification mechanism for equipment fault levels is constructed based on equipment anomaly alarm information and comprehensive diagnostic analysis reports, and equipment fault handling plans are generated.

[0015] Equipment failure handling plans and feedback on their implementation effects will be continuously entered into the intelligent operation and maintenance knowledge base.

[0016] Optionally, the real-time acquisition of water conservancy equipment operating status data, and the transmission using a LoRa gateway and LoRa-MCU to form a preliminary sensing data set, specifically includes:

[0017] Water level data, video data, temperature data, humidity data, and vibration data are collected by preset water level sensors, video sensors, temperature and humidity sensors, and vibration sensors, respectively. The data is then encapsulated into structured data packets by the Lora-MCU according to a preset data protocol format and wirelessly transmitted through a wireless transmission channel built based on the Lora wireless communication protocol at a preset transmission power and transmission rate, forming a preliminary sensing data set.

[0018] Optionally, the step of using the hydrological model analysis of edge computing nodes to obtain preliminary sensing data set and obtain preliminary equipment anomaly diagnosis information specifically involves:

[0019] The water level data, video data, temperature data, humidity data, and vibration data in the initial sensing data set are synchronously input into the hydrological model of the edge computing node to form a time-series correlated data stream.

[0020] The time-series variation trend features of water level data, video data, temperature data, humidity data, and vibration data are extracted from the time-series correlated data stream using a hydrological model.

[0021] Based on the preset normal operation baseline curve in the hydrological model, the trend deviation of the time-series change trend characteristics of water level data, video data, temperature data, humidity data and vibration data are calculated to obtain the time-series trend anomaly characteristics of each sensor data.

[0022] When the time-series trend anomaly of any sensor data exceeds the preset trend deviation threshold, the edge computing node is automatically triggered to start the anomaly interaction verification mechanism, comparing the change direction and amplitude of the trend anomaly characteristics of other sensor data in the same time period to obtain the cross-validated anomaly state characteristics.

[0023] The cross-validated abnormal state features are weighted together by the duration and magnitude of the abnormality to form a joint feature vector of equipment state abnormalities.

[0024] The Euclidean distance between the joint weight value of the joint feature vector of equipment status anomalies and the pre-stored standard feature vector of equipment anomalies is calculated. The anomaly category is determined according to the principle of minimum distance, and preliminary equipment anomaly diagnosis information is generated.

[0025] Optionally, the step of transmitting the preliminary anomaly diagnosis information of the device to a hardware platform equipped with the HarmonyOS for water conservancy data encryption and encapsulation to form encrypted data transmission information specifically involves:

[0026] The preliminary anomaly diagnosis information of the device is transmitted to the hardware platform through the data transmission channel between the edge computing node and the hardware platform equipped with the Water Conservancy HarmonyOS.

[0027] Using Rockchip RK3576 or RK3588 chips mounted on the hardware platform, the preliminary anomaly diagnosis information of the equipment is analyzed to extract the equipment anomaly type, anomaly occurrence time, anomaly duration and anomaly severity.

[0028] A security chip with built-in national cryptographic SM3 algorithm is used to calculate message digests of device anomaly type, anomaly occurrence time, anomaly duration and anomaly severity to obtain device anomaly information digest data.

[0029] A security chip with a built-in national cryptographic SM4 algorithm is used to perform symmetric encryption on the device anomaly type, anomaly occurrence time, anomaly duration and anomaly severity using a pre-stored symmetric encryption key to obtain symmetric encrypted anomaly information data.

[0030] The security chip with built-in national cryptographic SM2 algorithm uses a pre-stored asymmetric key on the hardware platform to perform asymmetric encryption on the device anomaly information digest data and the symmetric encrypted anomaly information data to obtain ciphertext data;

[0031] The encrypted data is encapsulated using Rockchip RK3576 or RK3588 chips on a hardware platform equipped with the HarmonyOS for water conservancy, forming encrypted data transmission information.

[0032] Optionally, the step of employing a three-layer collaborative analysis mechanism (end-edge-cloud) to fuse and analyze the data transmission information to obtain a set of device status characteristic parameters is as follows:

[0033] The encrypted data is synchronously transmitted to end computing devices, edge computing nodes, and cloud computing servers.

[0034] The edge computing device uses the built-in national cryptographic SM2 private key to decrypt encrypted data transmission information, forming initial characteristic data of the edge device status;

[0035] Edge computing nodes use Rockchip RK3576 or RK3588 chips and built-in national cryptographic SM2 private keys to decrypt encrypted data transmission information and obtain edge device status association feature data.

[0036] The cloud computing server uses a pre-stored national cryptographic SM2 private key to decrypt encrypted data transmission information, forming comprehensive feature data of cloud device status;

[0037] A three-layer collaborative analysis mechanism of end-edge-cloud is built on the cloud computing server. Data consistency comparison rules are used to verify the consistency of the initial feature data of the end-side device status, the associated feature data of the edge-side device status, and the comprehensive feature data of the cloud device status, and to determine the fusion weight of the data at each layer.

[0038] By utilizing a three-layer collaborative fusion analysis mechanism of end-edge-cloud, the initial feature data of the end-side device status, the associated feature data of the edge-side device status, and the comprehensive feature data of the cloud-side device status with fusion weights higher than the confidence threshold are weighted and fused to obtain a set of device status feature parameters.

[0039] Optionally, the step of constructing an intelligent diagnostic model for equipment health based on a set of equipment status characteristic parameters, identifying abnormal equipment states through multi-parameter fusion training, and obtaining diagnostic results and risk warning information for abnormal equipment states specifically involves:

[0040] A training sample dataset for real-time classification of device health status is constructed based on a set of device status feature parameters.

[0041] The health intelligent diagnostic model is used to iteratively train the training sample dataset, and the model parameters are adjusted based on the difference between the abnormal state categories output by the model and the actual abnormal state categories.

[0042] The trained intelligent health diagnosis model is used to perform feature vectorization on the set of real-time input device status feature parameters to form a real-time device feature vector.

[0043] The health intelligent diagnostic model is used to match the feature vectors of real-time devices with the feature vectors of standard abnormal state categories one by one, and the weighted distance value is calculated to determine the degree of matching.

[0044] The abnormal state category of the real-time device feature vector is determined based on the standard abnormal state category feature vector with the highest matching degree, thus forming the real-time device abnormal state diagnosis result;

[0045] Based on the real-time equipment abnormality diagnosis results and the set of equipment status characteristic parameters, the risk level is determined according to the risk level assessment rules, and risk warning information is generated.

[0046] Optionally, the step of performing real-time video recognition and analysis on video data based on a water conservancy video AI model to obtain visual diagnostic results and abnormal event early warning information specifically includes:

[0047] The video data of the on-site water conservancy equipment operation is divided into a continuous video data sequence according to the preset video resolution and video frame rate;

[0048] A water conservancy video AI model is used to divide each frame of the video data sequence into multiple spatial regions, and video features are extracted for the appearance, structure and operating status of water conservancy equipment in each spatial region.

[0049] By using a water conservancy video AI model to analyze the changes in video features between consecutive frames in each spatial region, dynamic features of the operating status of water conservancy equipment in each spatial region can be obtained.

[0050] The dynamic features between spatial regions are correlated using a water conservancy video AI model, and the correlation of operational status change features is fused to form fused video dynamic features.

[0051] The dynamic features of the fused video are continuously analyzed, and the changes in the device's operating status are judged according to the preset abnormal event evolution trend rules to determine whether the changes in the device's operating status conform to the abnormal event trend and to determine the abnormal type of the device's operating status.

[0052] Visual diagnostic results and early warning information for abnormal events are generated based on the type of abnormal equipment operation status.

[0053] Optionally, the step of simultaneously inputting the equipment abnormality diagnosis results, risk warning information, visual diagnosis results, and abnormal event warning information into the Honeycomb Intelligent Management Platform, and outputting equipment abnormality alarm information and a comprehensive diagnostic analysis report, specifically includes:

[0054] The device abnormal status diagnosis results, risk warning information, visual diagnosis results and abnormal event warning information are synchronously input to the Honeycomb Intelligent Management Platform at fixed intervals through a preset interface.

[0055] Based on the diagnostic results of abnormal equipment status, determine the location of abnormal equipment and generate a network topology diagram of the equipment.

[0056] Based on the visual diagnostic results and abnormal event warning information, the system calls on on-site video data to generate remote monitoring images that identify the location, type, and evolution trend of the abnormality.

[0057] Based on the diagnostic results of abnormal equipment status and risk warning information, maintenance measures are automatically matched to generate an operation and maintenance plan that includes maintenance personnel arrangements, operation steps and maintenance cycles;

[0058] Based on the equipment abnormality diagnosis results and abnormal event early warning information, generate equipment abnormality alarm information;

[0059] A comprehensive diagnostic analysis report is generated based on the equipment abnormality diagnosis results, risk warning information, visual diagnosis results, and abnormal event warning information.

[0060] Optionally, the step of constructing an automatic equipment fault level classification mechanism based on equipment anomaly alarm information and comprehensive diagnostic analysis report, and generating equipment fault handling plan, specifically includes:

[0061] Establish equipment fault level classification rules based on equipment abnormal alarm information and comprehensive diagnostic analysis reports;

[0062] The equipment fault level is determined by comparing and analyzing the abnormal alarm information and comprehensive diagnostic analysis report according to the equipment fault level classification rules to determine whether the equipment fault level is minor or serious.

[0063] For equipment with minor malfunctions, determine the current location of the equipment based on the equipment abnormality alarm information, and retrieve the list of maintenance personnel adjacent to the location of the malfunctioning equipment;

[0064] Send a fault notification message to each maintenance personnel in the maintenance personnel list;

[0065] For equipment with serious malfunctions, the intelligent operation and maintenance knowledge base is invoked according to the alarm event type and equipment anomaly type to extract historical handling measures and applicable conditions;

[0066] Based on the abnormality type of the severely faulty equipment and the applicable conditions of the historical handling measures, the corresponding technical experts are selected from the list of experts, and the fault information and historical handling measures are sent in real time through the remote collaborative communication system.

[0067] Based on the analysis results and suggested measures returned by experts, equipment failure handling plans are formulated in real time.

[0068] Optionally, an artificial intelligence-based intelligent water conservancy IoT system includes:

[0069] The field sensing unit is equipped with a water level sensor, video sensor, temperature and humidity sensor and vibration sensor, and is also equipped with a Lora-MCU and Lora gateway to collect real-time operating status data of water conservancy equipment and form a preliminary sensing data set.

[0070] The edge computing unit, equipped with a hydrological model and Rockchip RK3576 or RK3588 chip, is used to receive the initial sensing data set, perform trend feature extraction and cross-validation analysis of abnormal states in real time, and form preliminary abnormal diagnosis information of the equipment.

[0071] The data security transmission unit is equipped with the HarmonyOS OS and integrates Rockchip RK3576 or RK3588 chips and security chips with built-in national cryptographic algorithms SM2, SM3 and SM4. It is used to parse the preliminary abnormal diagnosis information of the equipment, calculate message digests, and perform symmetric and asymmetric encryption to generate encrypted data transmission information.

[0072] The collaborative fusion analysis unit includes end computing devices, edge computing nodes, and cloud computing servers. It synchronously receives encrypted data transmission information through a secure communication protocol channel, decrypts and extracts feature data respectively, and performs feature fusion based on trust level and fusion weight to form a set of device status feature parameters.

[0073] The equipment health diagnosis unit has a built-in intelligent health diagnosis model, which is used to receive the set of equipment status feature parameters in real time, perform feature vectorization processing and feature similarity matching, determine the real-time equipment abnormal status category and risk level, and output the equipment abnormal status diagnosis results and risk warning information in real time.

[0074] The video analysis and diagnosis unit has a built-in water conservancy video AI model. It receives video data collected by video sensors in real time, performs spatial partitioning, video feature extraction, regional feature trajectory tracking and fusion, and performs temporal continuity analysis of video dynamic features based on the rule library of abnormal event evolution trends to form visual diagnostic results and abnormal event early warning information.

[0075] The Honeycomb Intelligent Management Platform receives real-time diagnostic results of abnormal equipment status, risk warning information, visual diagnostic results, and abnormal event warning information through data interface protocols. It generates equipment network topology diagrams, remote monitoring screens, and automated operation and maintenance plans. It outputs real-time equipment abnormal alarm information and comprehensive diagnostic analysis reports. Based on the alarm information and diagnostic analysis reports, it automatically classifies equipment fault levels and calls upon the intelligent operation and maintenance knowledge base to form equipment fault handling solutions.

[0076] The beneficial effects of this invention are:

[0077] (1) This invention achieves accurate diagnosis and rapid response of the operating status of water conservancy equipment by real-time multi-parameter data acquisition by the field perception unit, extraction of time-series trend features and cross-verification of anomalies by the edge computing unit, secure encryption based on national cryptographic algorithms by the data security transmission unit, end-edge-cloud three-layer collaborative fusion analysis mechanism, real-time identification by the health intelligent diagnosis model and visual diagnosis by the water conservancy video AI model. This effectively improves the accuracy of equipment anomaly identification and response efficiency, and enhances the intelligent operation and maintenance decision-making capability.

[0078] (2) This invention significantly improves the efficiency and accuracy of handling abnormal equipment faults through topology node mapping of the honeycomb intelligent management platform, remote monitoring screen generation, automated operation and maintenance plan formulation, automatic fault level classification and remote collaborative analysis of intelligent operation and maintenance knowledge base, and shows better adaptability and stability in the remote intelligent operation and maintenance management of water conservancy facilities.

[0079] (3) In terms of comprehensive monitoring and maintenance of the operating status of water conservancy equipment, this invention effectively solves the shortcomings of inaccurate equipment anomaly identification and slow maintenance response in the prior art through multi-dimensional data fusion analysis, rapid diagnosis of abnormal status and intelligent decision-making mechanism. It breaks through the bottleneck of single data collection, manual analysis and simple alarm in the prior art, and realizes a significant improvement in rapid and accurate diagnosis of abnormal faults of water conservancy equipment and intelligent operation and maintenance decision-making, effectively improving the automation level of smart water conservancy facility operation and maintenance. Attached Figure Description

[0080] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0081] Figure 1 This is a flowchart of an artificial intelligence-based intelligent IoT method for water conservancy proposed in this invention;

[0082] Figure 2 This is a structural diagram of an artificial intelligence-based intelligent IoT system for water conservancy proposed in this invention. Detailed Implementation

[0083] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0084] refer to Figures 1-2 A water conservancy intelligent IoT method based on artificial intelligence includes:

[0085] The system collects real-time data on the operation of water conservancy equipment by using on-site water level sensors, video sensors, temperature and humidity sensors, and vibration sensors. It also uses a LoRa gateway and LoRa-MCU for low-power wireless data transmission to form a preliminary sensing data set.

[0086] The preliminary sensing data set is input into the edge computing node equipped with the hydrological model. The hydrological model is used to analyze the preliminary sensing data set in real time, extract features that characterize the abnormal state of water conservancy equipment, and obtain preliminary abnormality diagnosis information of the equipment.

[0087] The preliminary abnormal diagnosis information of the device is transmitted to the hardware platform equipped with the Water Conservancy HarmonyOS. The data is encrypted and encapsulated by the Rockchip RK3576 or RK3588 chip and the security chip with built-in national cryptographic SM2, SM3 and SM4 algorithms to form encrypted data transmission information.

[0088] The encrypted data transmission information is fused and analyzed using a three-layer collaborative analysis mechanism of end-edge-cloud to obtain a set of device status characteristic parameters for constructing a health intelligent diagnosis model.

[0089] A health intelligent diagnosis model is constructed based on the set of equipment status characteristic parameters. Through multi-parameter fusion training, abnormal equipment status is identified in real time, and the diagnosis results and risk warning information of abnormal equipment status are obtained.

[0090] Video sensors are used to collect video data of the operation of on-site equipment. Based on the water conservancy video AI model, the video data is analyzed in real time to obtain visual diagnostic results of the on-site equipment operation status and early warning information of abnormal events.

[0091] The abnormal status diagnosis results, risk warning information, visual diagnosis results and abnormal event warning information of the equipment are synchronously input into the Honeycomb Intelligent Management Platform to generate the equipment network topology diagram, remote monitoring screen and automated operation and maintenance plan, and output the equipment abnormal alarm information and comprehensive diagnostic analysis report in real time.

[0092] Based on the equipment anomaly alarm information and comprehensive diagnostic analysis report, an automatic equipment fault level classification mechanism is constructed. For faults classified as minor, nearby maintenance personnel are automatically notified to handle them. For faults classified as serious, the intelligent maintenance knowledge base is automatically invoked, and experts are organized to conduct remote collaborative analysis to form equipment fault handling plans in real time.

[0093] The equipment fault handling plan and its implementation effect feedback will be continuously entered into the intelligent operation and maintenance knowledge base for automatic recommendation and decision support in subsequent equipment diagnosis and maintenance processes.

[0094] In this embodiment, the real-time acquisition of operational status data of the hydraulic equipment is achieved through on-site water level sensors, video sensors, temperature and humidity sensors, and vibration sensors. Low-power wireless data transmission is then performed using a LoRa gateway and a LoRa-MCU to form a preliminary sensing data set. Specifically:

[0095] The water level sensor installed on site measures the water level height at the location of the water conservancy equipment in real time, and outputs the water level data periodically according to the preset sampling frequency.

[0096] The video sensors installed on site capture real-time video data of the appearance, structure and operating status of the water conservancy equipment, and output the video data periodically at a preset video resolution and frame rate.

[0097] The temperature and humidity values ​​of the environment where the water conservancy equipment is located are measured in real time by on-site temperature and humidity sensors, and the temperature and humidity data are output periodically with preset measurement accuracy and sampling frequency.

[0098] The vibration frequency and amplitude data of the hydraulic equipment during operation are measured in real time by the vibration sensor installed on site, and the vibration data is output within the preset measurement range and sampling period.

[0099] The Lora-MCU is used to receive the water level data, video data, temperature data, humidity data and vibration data, and encapsulates the received data into structured data packets according to a preset data protocol format;

[0100] A wireless transmission channel is established between the field and the edge computing node using a LoRa gateway based on the LoRa wireless communication protocol. The structured data packets are then wirelessly transmitted through the wireless transmission channel at a preset transmission power and transmission rate to form the preliminary sensing data set.

[0101] In this embodiment, the step of inputting the preliminary sensing data set to an edge computing node equipped with a hydrological model, and using the hydrological model to analyze the preliminary sensing data set in real time, extracting features characterizing the abnormal state of the water conservancy equipment, and obtaining preliminary abnormality diagnosis information of the equipment, specifically involves:

[0102] The water level data, video data, temperature data, humidity data, and vibration data in the preliminary sensing data set are input into the hydrological model of the edge computing node according to the synchronous temporal relationship to form a temporal correlation data stream. The hydrological model is a multi-sensor data correlation analysis model that is pre-trained using historical water conservancy equipment operation data. It has built-in normal operation benchmark curves for water level height, video status features, temperature values, humidity values, and vibration amplitude, and preset temporal trend feature extraction rules, trend deviation calculation rules, anomaly interaction verification rules, and joint weight calculation rules. It is used to perform synchronous temporal correlation analysis and cross-validation on the operation status data of various water conservancy equipment to generate preliminary equipment anomaly diagnosis information.

[0103] A multi-sensor data cross-validation mechanism is established within the edge computing node. The hydrological model is used to extract the temporal variation trend features of each sensor data from the time-series correlated data stream. The multi-sensor data cross-validation mechanism synchronously processes the time-series correlated data stream composed of water level data, video data, temperature data, humidity data, and vibration data through the hydrological model. After extracting the corresponding temporal variation trend features from each type of sensor data, the extracted trend features are then cross-analyzed pairwise to determine whether there is a correlation between the trend changes of different sensor data, thereby realizing the validity verification of the temporal variation trend features of each sensor data.

[0104] Based on the baseline curve of the time-series change trend characteristics of the normal operation state preset in the hydrological model, the trend deviation of the time-series change trend characteristics of water level data, video data, temperature data, humidity data and vibration data is calculated to obtain the time-series trend anomaly characteristics of each sensor data.

[0105] The anomaly verification mechanism is triggered by the edge computing node based on the anomaly characteristics of the time series trend. The consistency of the anomaly characteristics of the time series trend among water level data, video data, temperature data, humidity data and vibration data is verified through cross-validation to obtain the anomaly state characteristics that have been cross-validated. The anomaly verification mechanism is automatically triggered when the anomaly characteristics of the time series trend of any one of the water level data, video data, temperature data, humidity data and vibration data exceed the preset trend deviation threshold. The mechanism compares the change direction and amplitude of the anomaly characteristics of the trend of other types of data in the same anomaly period. If the anomaly characteristics of the trend of multiple sensor data occur simultaneously in the same direction and the amplitude exceeds their respective preset thresholds, the anomaly state is determined to exist and the anomaly state characteristics that have been cross-validated are obtained.

[0106] The cross-validated abnormal state features are weighted jointly by the duration and magnitude of the abnormality, and a joint feature vector of equipment state abnormalities is formed.

[0107] Based on the joint weight values ​​of various sensor data anomalies in the joint feature vector of the equipment status anomalies, the Euclidean distance between the joint feature vector and the pre-stored standard feature vector of the equipment anomalies is calculated. Then, the anomaly category is determined from the pre-stored anomaly status types corresponding to the standard feature vector according to the principle of minimum distance, thereby generating preliminary equipment anomaly diagnosis information that includes equipment anomaly type, anomaly occurrence time, anomaly duration, and anomaly degree.

[0108] In this embodiment, the preliminary anomaly diagnosis information of the device is transmitted to a hardware platform equipped with the HarmonyOS for water conservancy. Data encryption and encapsulation are performed using a Rockchip RK3576 or RK3588 chip and a security chip with built-in national cryptographic algorithms SM2, SM3, and SM4 to form encrypted data transmission information. Specifically:

[0109] Through the pre-defined data transmission channel between the edge computing node and the hardware platform equipped with the Water Conservancy HarmonyOS, the preliminary abnormal diagnosis information of the device is transmitted to the hardware platform in a predetermined data format.

[0110] The Rockchip RK3576 or RK3588 chip, which is mounted on the hardware platform, is used to perform compliance analysis on the data format of the preliminary abnormality diagnosis information of the equipment, and to extract the equipment abnormality type, abnormality occurrence time, abnormality duration and abnormality degree from the data.

[0111] A security chip with built-in national cryptographic SM3 algorithm is used to calculate message digests of the extracted device anomaly type, anomaly occurrence time, anomaly duration and anomaly severity to obtain device anomaly information digest data.

[0112] A security chip with built-in national cryptographic SM4 algorithm is used to perform symmetric encryption on the device anomaly type, anomaly occurrence time, anomaly duration and anomaly severity using a pre-stored symmetric encryption key, to obtain symmetrically encrypted anomaly information data;

[0113] The security chip with built-in national cryptographic SM2 algorithm uses the asymmetric key pre-stored in the hardware platform to perform asymmetric encryption on the device anomaly information digest data and the symmetrically encrypted anomaly information data to obtain ciphertext data;

[0114] The Rockchip RK3576 or RK3588 chip, which is equipped with the HarmonyOS hardware platform, encapsulates the ciphertext data to form encrypted data transmission information containing ciphertext data, encryption algorithm type identifier, and key identifier.

[0115] In this embodiment, the encrypted data transmission information is fused and analyzed using a three-layer collaborative analysis mechanism (end-edge-cloud) to obtain a set of device status characteristic parameters for constructing a health intelligent diagnostic model. Specifically:

[0116] Encrypted data transmission information is synchronously transmitted to end computing devices, edge computing nodes, and cloud computing servers through a pre-established secure communication protocol channel at the end-edge-cloud three-layer layer.

[0117] The edge computing device uses the built-in national cryptographic SM2 private key to decrypt the received encrypted data transmission information, and extracts the device anomaly type, anomaly occurrence time, anomaly duration and anomaly degree based on the data identification information to form the initial feature data of the edge device status.

[0118] Edge computing nodes decrypt received encrypted data transmission information based on Rockchip RK3576 or RK3588 chips and built-in national cryptographic SM2 private key, and perform matching calculations based on pre-stored historical state feature datasets to obtain edge device state association feature data.

[0119] The cloud computing server uses a pre-stored national cryptographic SM2 private key to decrypt the received encrypted data transmission information. Combined with a pre-stored historical water conservancy equipment abnormal status feature database, it performs comprehensive analysis and calculation on the equipment abnormality type, abnormality occurrence time, abnormality duration and abnormality degree to form comprehensive feature data of cloud equipment status.

[0120] A three-layer collaborative analysis mechanism of end-edge-cloud is constructed in the cloud computing server. Based on the preset data synchronization channel between end computing devices, edge computing nodes and cloud computing servers, the consistency of the initial feature data of the end device status, the associated feature data of the edge device status and the comprehensive feature data of the cloud device status is verified one by one using data consistency comparison rules. Based on the consistency verification results, the trust level and fusion weight of each layer of data of end side, edge side and cloud are dynamically allocated.

[0121] The edge-cloud three-layer collaborative fusion analysis mechanism is used to perform weighted feature fusion calculation on the initial feature data of the end-side device status, the associated feature data of the edge-side device status, and the comprehensive feature data of the cloud device status, which have a trust level and fusion weight higher than the preset trust threshold, to obtain the fused set of device status feature parameters.

[0122] The aforementioned edge-cloud three-layer collaborative fusion analysis mechanism takes initial device status characteristic data, associated device status characteristic data, and comprehensive device status characteristic data as inputs:

[0123] A timestamp synchronization calibration method is used to achieve time-series alignment of data among the three layers of terminal computing devices, edge computing nodes, and cloud computing servers.

[0124] The consistency of the three-layer device status feature data is verified item by item by using the preset data consistency comparison rules to check the anomaly type, anomaly occurrence time, anomaly duration and anomaly degree.

[0125] Based on the consistency verification results, the data output by the end computing device, edge computing node and cloud computing server are assigned dynamically adjusted trust levels and fusion weights.

[0126] The multi-source data credibility level and fusion weight calculation method is used to perform weighted fusion calculation on the three-layer equipment status feature data whose credibility level and fusion weight are higher than the preset credibility threshold, so as to form a unified set of equipment status feature parameters.

[0127] The data consistency comparison rules are as follows:

[0128] Calculate the feature differences in the initial feature data of the terminal device status, the associated feature data of the edge device status, and the comprehensive feature data of the cloud device status, respectively, for the corresponding anomaly type, anomaly occurrence time, anomaly duration, and anomaly severity.

[0129] A predetermined consistency threshold is set for each feature difference value;

[0130] When the feature difference value of a certain feature in the three-layer data is less than or equal to the corresponding predetermined consistency threshold, the feature data is determined to pass the consistency check.

[0131] When the feature difference value is greater than the corresponding predetermined consistency threshold, the feature data is determined to have failed the consistency check.

[0132] The trust level and fusion weight of the three layers of data—end computing devices, edge computing nodes, and cloud computing servers—are dynamically adjusted based on the number of feature data that pass the consistency verification.

[0133] In this embodiment, the step of constructing a health intelligent diagnosis model based on the set of equipment status characteristic parameters, and identifying abnormal equipment states in real time through multi-parameter fusion training to obtain equipment abnormal state diagnosis results and risk warning information, specifically includes:

[0134] Based on the numerical combination of equipment anomaly type, anomaly occurrence time, anomaly duration, and anomaly severity in the set of equipment status feature parameters, a training sample dataset for real-time classification of equipment health status is constructed.

[0135] The health intelligent diagnosis model is used to iteratively train the training sample dataset. The health intelligent diagnosis model has a predefined training loss function. The training loss function is based on the difference between the abnormal state diagnosis category output by the model and the actual abnormal state category in the training sample dataset. The model parameters are adjusted through multiple rounds of training to reduce the value of the loss function.

[0136] The trained intelligent health diagnosis model is used to perform real-time feature vectorization on the equipment anomaly type, anomaly occurrence time, anomaly duration and anomaly degree in the real-time input set of equipment status feature parameters, forming a real-time equipment feature vector.

[0137] The health intelligent diagnosis model is used to perform similarity matching between the real-time device feature vector and the pre-stored standard abnormal state category feature vector one by one. The weighted distance value of the difference of each feature parameter between the two is calculated, and the degree of matching between the real-time device feature vector and the standard abnormal state category feature vector is judged by the magnitude of the weighted distance value.

[0138] Based on the abnormal state category corresponding to the feature vector of the standard abnormal state category with the highest matching degree, the abnormal state category of the real-time device feature vector is determined, and the abnormal state diagnosis result of the real-time device is formed.

[0139] Based on the real-time equipment abnormality diagnosis results and the abnormality duration and abnormality severity values ​​in the equipment status characteristic parameter set, the corresponding risk level is determined according to the preset risk level assessment rules, and risk warning information is generated.

[0140] The intelligent health diagnosis model includes a feature vector input module, a state feature difference calculation module, an abnormal state matching module, a dynamic risk assessment module, and an abnormal state output module.

[0141] The feature vector input module receives the set of device status feature parameters and extracts feature values ​​of device anomaly type, anomaly occurrence time, anomaly duration and anomaly degree in a preset order, and maps them into a standardized real-time device feature vector.

[0142] The state feature difference calculation module compares the real-time device feature vector with the pre-stored historical abnormal state standard feature vector feature by feature, and calculates the difference values ​​between the real-time device feature vector and each historical abnormal state standard feature vector in each dimension of device abnormality type, abnormality occurrence time, abnormality duration and abnormality degree, and generates a multi-dimensional state difference vector.

[0143] The abnormal state matching module assigns a difference weight to the difference values ​​of each dimension in the multidimensional state difference vector according to a preset difference threshold. Based on the difference weight values, the real-time device feature vector is weighted and accumulated with each historical abnormal state standard feature vector in a dimension-wise manner to obtain a comprehensive difference accumulation value for each historical abnormal state standard feature vector. The abnormal state category corresponding to the historical abnormal state standard feature vector with the smallest comprehensive difference accumulation value is selected as the abnormal state category for matching the real-time device feature vector.

[0144] The dynamic risk assessment module determines the urgency of the current abnormal event based on the duration and severity of the abnormality in the real-time device feature vector, and dynamically searches for and determines the risk level based on the abnormal state category and the preset risk assessment rule matrix.

[0145] The abnormal status output module constructs structured real-time equipment abnormal status diagnosis results and risk warning information based on the matched abnormal status category, the abnormal occurrence time, the abnormal duration, the abnormality degree in the real-time equipment feature vector, and the risk level determined by the dynamic risk assessment module.

[0146] In this embodiment, the step of using video sensors to collect video data of the operation of on-site equipment, and performing real-time video recognition and analysis on the video data based on a water conservancy video AI model to obtain visual diagnostic results of the on-site equipment's operating status and early warning information of abnormal events, specifically involves:

[0147] The video data of the operation of the on-site water conservancy equipment is divided into a continuous video data sequence according to the preset video resolution and video frame rate;

[0148] A water conservancy video AI model is used to spatially partition video data sequences, dividing each frame of the video data sequence into multiple non-overlapping spatial regions. Independent video features are extracted for the appearance, structure, and operational status of water conservancy equipment within each spatial region. The water conservancy video AI model incorporates a water conservancy equipment appearance and structure status feature library, an equipment operational status dynamic feature library, and an abnormal event evolution trend rule library. The water conservancy equipment appearance and structure status feature library is used to extract and identify features of the appearance and structure of water conservancy equipment in the video data sequence. The equipment operational status dynamic feature library is used to analyze the dynamic feature change trajectory during equipment operation and determine the correlation between dynamic features in different regions. The abnormal event evolution trend rule library is used to determine whether the temporal continuity of dynamic features conforms to the abnormal event evolution trend, thereby determining the abnormal event type, occurrence spatial region, and duration.

[0149] Based on video features within a spatial region, a water conservancy video AI model is used to track and analyze the video feature change trajectory between consecutive frames in the video data sequence for each spatial region according to the time-series evolution rules, thereby obtaining the dynamic characteristics of the changes in the operating status of water conservancy equipment in each spatial region.

[0150] By using a water conservancy video AI model, dynamic feature correlation analysis is performed on the video feature change trajectory of each spatial region to determine whether there is a logical correlation between the changes in equipment operating status between spatial regions. The operating status change features between the correlated regions are then fused to form the dynamic video features after spatial region fusion.

[0151] Temporal continuity analysis is performed on the dynamic features of the video after spatial region fusion. Based on the preset abnormal event evolution trend rule base, it is determined whether the continuity of the equipment operation status change features conforms to the abnormal event evolution trend, and the abnormal type of the on-site equipment operation status is determined.

[0152] Based on the abnormal operating status of on-site equipment, visual diagnostic results and early warning information for abnormal events are generated, including the spatial area where the abnormality occurred, the type of abnormal event, the event evolution trend, and the duration of the abnormal event.

[0153] In this embodiment, the step of simultaneously inputting the device abnormal status diagnosis results, risk warning information, visual diagnosis results, and abnormal event warning information into the honeycomb intelligent management platform to generate a device network topology diagram, remote monitoring screen, and automated operation and maintenance plan, and outputting device abnormal alarm information and comprehensive diagnostic analysis report in real time, specifically includes:

[0154] Through a preset data interface protocol, the device abnormal status diagnosis results, risk warning information, visual diagnosis results and abnormal event warning information are synchronously input to the Honeycomb Intelligent Management Platform at fixed time intervals.

[0155] Based on the received abnormal device status diagnosis results, the Honeycomb Intelligent Management Platform uses topology node mapping rules to determine the location of abnormal device nodes, and uses pre-stored water conservancy equipment location information and connection relationship information to generate a device network topology map containing device location, device identifier and connection path;

[0156] Based on the received visual diagnostic results and abnormal event warning information, the Honeycomb Intelligent Management Platform calls on the video data of the on-site water conservancy equipment to generate a remote monitoring screen of the on-site equipment corresponding to the equipment network topology. The monitoring screen identifies the location of the abnormal event, the type of abnormal event, and the evolution trend of the abnormal event.

[0157] Based on the diagnostic results of abnormal equipment status and risk warning information, the Honeycomb Intelligent Management Platform automatically matches maintenance measures corresponding to the abnormal status type and risk level according to the preset equipment maintenance strategy library, and generates an automated operation and maintenance plan that includes maintenance personnel arrangements, maintenance operation steps and equipment maintenance cycles.

[0158] Based on the received equipment abnormality status diagnosis results and abnormal event warning information, the Honeycomb Intelligent Management Platform generates equipment abnormality alarm information in real time, including alarm level, alarm device identifier, alarm event type and alarm timestamp, according to the pre-set alarm triggering rules.

[0159] Based on the diagnostic results of abnormal equipment status, risk warning information, visual diagnostic results, and abnormal event warning information, the Honeycomb Intelligent Management Platform automatically generates a comprehensive diagnostic analysis report that includes the type of equipment abnormality, duration of abnormality, risk level, on-site video screenshots, and recommended maintenance measures according to the preset diagnostic report template.

[0160] The honeycomb intelligent management platform is an intelligent management software platform deployed on a cloud computing server and developed based on the HarmonyOS OS. It has built-in modules for receiving equipment status diagnosis results, visual diagnosis analysis, generating equipment network topology, remote monitoring and display, generating automated operation and maintenance plans, generating equipment anomaly alarms, and generating comprehensive diagnostic analysis reports. It receives equipment anomaly status diagnosis results, risk warning information, visual diagnosis results, and anomaly event warning information in real time through data interface protocols. Based on equipment location information and connection relationship information, it constructs equipment network topology diagrams and remote monitoring screens. It automatically matches operation and maintenance plans according to the equipment maintenance strategy library and generates anomaly alarm information. At the same time, it automatically fills and outputs comprehensive diagnostic analysis reports based on preset diagnostic report templates.

[0161] In this embodiment, the automatic equipment fault level classification mechanism is constructed based on the equipment abnormal alarm information and comprehensive diagnostic analysis report. For faults classified as minor, nearby maintenance personnel are automatically notified for handling. For faults classified as serious, the intelligent maintenance knowledge base is automatically invoked, and experts are organized for remote collaborative analysis to generate equipment fault handling plans in real time. Specifically:

[0162] Based on the alarm level, alarm event type, and alarm device identifier in the equipment anomaly alarm information, as well as the equipment anomaly type, risk level, and anomaly duration in the comprehensive diagnostic analysis report, establish equipment fault level classification rules.

[0163] According to the equipment fault level classification rules, the received equipment abnormality alarm information and comprehensive diagnostic analysis report are automatically compared and analyzed item by item to determine whether the equipment fault level is a minor fault or a serious fault.

[0164] For equipment with a minor fault level, the current location of the equipment is determined based on the alarm equipment identifier contained in the equipment abnormality alarm information, and the list of maintenance personnel adjacent to the location of the faulty equipment is automatically retrieved based on the pre-stored maintenance personnel location information.

[0165] Send a fault notification message containing the device identifier, anomaly type, fault location, and anomaly duration to each maintenance personnel in the maintenance personnel list;

[0166] For equipment with a serious fault level, the system automatically matches and calls the corresponding historical equipment fault cases and maintenance experience information in the intelligent operation and maintenance knowledge base based on the alarm event type contained in the equipment abnormal alarm information and the equipment abnormality type in the comprehensive diagnostic analysis report, and extracts the historical handling measures and applicable conditions of the relevant fault cases.

[0167] Based on the abnormal type of the severely faulty equipment and the applicable conditions of the extracted historical handling measures, the system automatically selects technical experts corresponding to the fault type from the pre-stored list of experts, and sends the faulty equipment identifier, fault event type, fault duration and historical handling measures to the selected experts in real time through the preset remote collaborative communication system.

[0168] Based on the analysis results and suggested handling measures returned by experts through the remote collaborative communication system, a fault handling plan is generated in real time, which includes faulty equipment identification, expert-suggested handling steps, required maintenance resources, and maintenance time plan.

[0169] In this embodiment, a water conservancy intelligent IoT system based on artificial intelligence includes:

[0170] The field sensing unit is equipped with a water level sensor, video sensor, temperature and humidity sensor and vibration sensor to collect real-time operating status data of water conservancy equipment. It is also equipped with a Lora-MCU and Lora gateway to achieve low-power wireless data transmission and form a preliminary sensing data set.

[0171] The edge computing unit, equipped with a hydrological model and a Rockchip RK3576 or RK3588 chip, is used to receive the preliminary sensing data set formed by the field sensing unit, and to perform real-time extraction of time-series trend features and cross-validation analysis of abnormal states based on the built-in hydrological model to form preliminary abnormal diagnosis information of the equipment.

[0172] The data security transmission unit is equipped with the HarmonyOS OS and integrates a Rockchip RK3576 or RK3588 chip and a security chip with built-in national cryptographic algorithms SM2, SM3 and SM4. It is used to perform data format parsing, message digest calculation, symmetric encryption and asymmetric encryption processing on the preliminary abnormal diagnosis information of the device to generate encrypted data transmission information.

[0173] The collaborative fusion analysis unit adopts a three-layer collaborative fusion analysis mechanism of end-edge-cloud, including end computing devices, edge computing nodes and cloud computing servers. The three achieve data synchronization through a secure communication protocol channel, respectively decrypt the encrypted data transmission information and extract feature data, perform data consistency verification and perform feature fusion calculation based on trust level and fusion weight to form a set of device status feature parameters.

[0174] The equipment health diagnosis unit has a built-in intelligent health diagnosis model. It receives the set of equipment status feature parameters in real time and completes real-time feature vectorization processing. It performs feature vector similarity matching calculation and determines the category of real-time abnormal equipment status. At the same time, it determines the risk level according to the preset risk level assessment rules and outputs the equipment abnormal status diagnosis results and risk warning information in real time.

[0175] The video analysis and diagnosis unit has a built-in water conservancy video AI model. It receives video data collected by the video sensor of the field perception unit in real time, performs spatial partitioning of video data sequence, video feature extraction, regional feature trajectory tracking and fusion, and performs temporal continuity analysis of video dynamic features based on a preset abnormal event evolution trend rule library to form visual diagnosis results and abnormal event early warning information.

[0176] The Honeycomb Intelligent Management Platform receives real-time diagnostic results of abnormal device status, risk warning information, visual diagnostic results, and abnormal event warning information through data interface protocols. It performs device network topology node mapping, remote monitoring screen generation, and automated operation and maintenance plan formulation. It generates device abnormal alarm information and comprehensive diagnostic analysis reports in real time, and automatically classifies device fault levels and calls the intelligent operation and maintenance knowledge base based on alarm information and diagnostic analysis reports to form and output device fault handling solutions.

[0177] Example 1

[0178] To verify the feasibility of this invention in practice, it was applied to the intelligent equipment operation status monitoring and intelligent fault handling tasks of water conservancy facilities in a certain region, in order to achieve accurate perception and timely maintenance of abnormal equipment conditions. In the actual scenario, the region includes multiple scattered water conservancy facilities, such as river gates, pumping stations, and hydrological monitoring stations. Each facility is characterized by complex equipment operating conditions, a large amount of real-time monitoring data, and strict requirements for equipment maintenance response.

[0179] In traditional technical solutions, equipment operating status monitoring typically relies on periodic manual inspections and the collection of localized data from single-type sensors to determine the status. Inspection reports are then uploaded to a back-end management center for manual fault type analysis and repair plan development. This approach is not only slow in response and has long inspection cycles, but manual analysis is also prone to misjudgments, especially regarding the effective detection of latent or subtle faults, significantly impacting the safe and reliable operation of the equipment.

[0180] During implementation, field sensing units, including water level sensors, video sensors, temperature and humidity sensors, and vibration sensors, were first installed and deployed on each hydraulic equipment on site. These units collected real-time data on the water level, equipment appearance and operation video, ambient temperature and humidity, and equipment vibration. The data was then packaged into structured data packets by an integrated Lora-MCU and transmitted wirelessly to the edge computing node via a Lora gateway using low power.

[0181] The edge computing node is equipped with a hydrological model pre-trained using historical water conservancy equipment operation data. It receives the initial sensing data set transmitted in real time, and uses built-in time-series trend feature extraction rules, trend deviation calculation rules, and anomaly interaction verification rules to synchronously analyze the data from different sensors. After cross-validation, it generates preliminary anomaly diagnosis information for the equipment.

[0182] The initial abnormality diagnosis information of the equipment is then transmitted to the hardware platform equipped with the HarmonyOS for water conservancy. This platform integrates the Rockchip RK3576 chip and a security chip with built-in national cryptographic algorithms SM2, SM3 and SM4 to realize message digest calculation and encryption processing of data, and generate encrypted data transmission information containing algorithm identifier and key identifier, so as to ensure the security and reliability of data transmission.

[0183] The encrypted data transmission information is then synchronously sent to the edge computing device, edge computing node, and cloud computing server, forming a three-layer collaborative analysis mechanism of edge-cloud. Each layer of devices decrypts and extracts device status feature data based on its own historical data storage and analysis capabilities, and performs cross-layer data consistency verification. Based on the verification results, data fusion weights are dynamically allocated, and a set of device status feature parameters is formed through weighted feature fusion.

[0184] Next, the set of equipment status characteristic parameters is fed into the equipment health intelligent diagnosis model for real-time feature vectorization processing. By matching similarity with pre-stored standard abnormal state feature vectors, the model determines and outputs the equipment abnormal state diagnosis results and risk warning information in real time. Simultaneously, the video analysis and diagnosis unit utilizes a pre-trained hydraulic video AI model to perform spatial region segmentation and feature trajectory analysis on the equipment video data sequence. Combined with abnormal event evolution trend rules, it outputs visual diagnosis results of on-site equipment operation and abnormal event warning information.

[0185] The aforementioned diagnostic and early warning information is then uniformly input into the Honeycomb Intelligent Management Platform in real time. Based on the device location, operating status information, and video data, the platform generates a device network topology diagram, remote monitoring screen, and automated operation and maintenance plan, and outputs device anomaly alarm information containing anomaly level and event type, as well as a detailed comprehensive diagnostic analysis report in real time.

[0186] Based on abnormal alarm information and diagnostic analysis reports, the system automatically classifies equipment faults into minor or severe categories according to pre-established equipment fault level classification rules. For minor faults, the system automatically identifies nearby maintenance personnel and pushes a notification message containing the specific location of the fault, the type of abnormality, and the duration. For severe faults, the system automatically accesses the intelligent maintenance knowledge base, matches relevant historical cases and maintenance experience, selects experts in the corresponding technical field for real-time remote collaborative analysis, and quickly generates a detailed equipment fault handling plan recommended by experts.

[0187] After actual deployment, the technical solution of this invention significantly improved the accuracy of equipment operation status monitoring and fault diagnosis efficiency. Table 1 below shows the effect data of equipment operation status monitoring and fault identification in multiple water conservancy facilities after system deployment. It compares the implementation effects of five indicators: anomaly type identification accuracy, alarm response time, fault handling efficiency, false alarm rate, and fault recurrence rate, and lists the specific comparative data of traditional technology and the technical solution of this invention.

[0188] Table 1 Comparison of Equipment Monitoring and Maintenance Effects Before and After System Implementation

[0189]

[0190] As shown in Table 1, the implementation results of this invention demonstrate significant advantages over traditional manual inspections and single data analysis methods in all key performance indicators. Specifically, the accuracy rate of anomaly type identification is increased to over 95%, significantly reducing the risk of misjudgment; the average alarm response time is shortened from 120 minutes to approximately 15 minutes, effectively ensuring equipment operational safety; the fault handling efficiency is reduced from the traditional 24 hours to approximately 3 hours, greatly improving maintenance efficiency; and the false alarm rate and fault recurrence rate are significantly reduced, enhancing the overall stability and reliability of water conservancy equipment operation and maintenance.

[0191] Based on the detailed analysis and real data of the above embodiments, it can be seen that the present invention, through multi-dimensional data perception, real-time edge intelligent analysis, end-edge-cloud collaborative fusion, and video-assisted diagnosis, realizes accurate monitoring of the operating status of water conservancy facilities, rapid response to anomalies, and intelligent operation and maintenance management. It effectively solves the shortcomings of traditional technologies in equipment status monitoring and maintenance, achieves the expected technical goals and practical effects, and has broad practical application value.

[0192] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A water conservancy intelligent IoT method based on artificial intelligence, characterized in that, include: Real-time data collection of water conservancy equipment operation status is used, and the data is transmitted using a LoRa gateway and LoRa-MCU to form a preliminary sensing data set. The preliminary sensing data set is analyzed using the hydrological model of the edge computing node to obtain preliminary anomaly diagnosis information for the equipment. The analysis includes: inputting water level data, video data, temperature data, humidity data, and vibration data from the preliminary sensing data set into the hydrological model of the edge computing node in a synchronous time sequence to form a time-series correlated data stream; extracting the time-series trend characteristics of water level data, video data, temperature data, humidity data, and vibration data from the time-series correlated data stream using the hydrological model; and calculating the trend deviation of the time-series trend characteristics of water level data, video data, temperature data, humidity data, and vibration data based on the preset normal operation baseline curve within the hydrological model, thereby obtaining the results for each sensor. When the time-series trend anomaly of any sensor data exceeds the preset trend deviation threshold, the edge computing node is automatically triggered to start the anomaly interaction verification mechanism. It compares the change direction and amplitude of the trend anomaly features of other sensor data in the same time period to obtain cross-validated anomaly state features. The cross-validated anomaly state features are weighted by the duration and amplitude of the anomaly to form a joint feature vector of equipment anomaly. The Euclidean distance between the joint weight value of the joint feature vector of equipment anomaly and the pre-stored standard feature vector of equipment anomaly state is calculated. The anomaly category is determined according to the principle of minimum distance, and preliminary equipment anomaly diagnosis information is generated. The initial abnormality diagnosis information of the equipment is transmitted to a hardware platform equipped with the HarmonyOS for water conservancy, where the data is encrypted and encapsulated to form encrypted data transmission information. A three-layer collaborative analysis mechanism (edge-cloud) is used to fuse and analyze this data transmission information to obtain a set of equipment status characteristic parameters. Based on this set of parameters, an intelligent health diagnosis model for the equipment is constructed. Through multi-parameter fusion training, abnormal equipment states are identified, resulting in diagnostic results and risk warning information. Finally, a water conservancy video AI model is used to perform real-time video recognition and analysis of the video data, yielding visual diagnostic results and abnormal event warning information. The system synchronously inputs the equipment abnormal status diagnosis results, risk warning information, visual diagnosis results, and abnormal event warning information into the Honeycomb Intelligent Management Platform, outputting equipment abnormal alarm information and comprehensive diagnostic analysis reports; it constructs an automatic equipment fault level classification mechanism based on the equipment abnormal alarm information and comprehensive diagnostic analysis reports, and generates equipment fault handling plans; it continuously records the equipment fault handling plans and implementation effect feedback into the intelligent operation and maintenance knowledge base.

2. The AI-based intelligent IoT method for water conservancy as described in claim 1, characterized in that, The real-time acquisition of operational status data from water conservancy equipment, and its transmission via a LoRa gateway and LoRa-MCU, forms a preliminary sensing data set, specifically as follows: Water level data, video data, temperature data, humidity data, and vibration data are collected by preset water level sensors, video sensors, temperature and humidity sensors, and vibration sensors, respectively. The data is then encapsulated into structured data packets by the Lora-MCU according to a preset data protocol format and wirelessly transmitted through a wireless transmission channel built based on the Lora wireless communication protocol at a preset transmission power and transmission rate, forming a preliminary sensing data set.

3. The AI-based intelligent IoT method for water conservancy as described in claim 1, characterized in that, The process of transmitting preliminary equipment anomaly diagnosis information to a hardware platform running the HarmonyOS for water conservancy data encryption and encapsulation to form encrypted data transmission information specifically involves: The preliminary anomaly diagnosis information of the device is transmitted to the hardware platform through the data transmission channel between the edge computing node and the hardware platform equipped with the Water Conservancy HarmonyOS. Using Rockchip RK3576 or RK3588 chips mounted on the hardware platform, the preliminary anomaly diagnosis information of the equipment is analyzed to extract the equipment anomaly type, anomaly occurrence time, anomaly duration and anomaly severity. A security chip with built-in national cryptographic SM3 algorithm is used to calculate message digests of device anomaly type, anomaly occurrence time, anomaly duration and anomaly severity to obtain device anomaly information digest data. A security chip with a built-in national cryptographic SM4 algorithm is used to perform symmetric encryption on the device anomaly type, anomaly occurrence time, anomaly duration and anomaly severity using a pre-stored symmetric encryption key to obtain symmetric encrypted anomaly information data. The security chip with built-in national cryptographic SM2 algorithm uses a pre-stored asymmetric key on the hardware platform to perform asymmetric encryption on the device anomaly information digest data and the symmetric encrypted anomaly information data to obtain ciphertext data; The encrypted data is encapsulated using Rockchip RK3576 or RK3588 chips on a hardware platform equipped with the HarmonyOS for water conservancy, forming encrypted data transmission information.

4. The AI-based intelligent IoT method for water conservancy as described in claim 1, characterized in that, The process involves constructing an intelligent diagnostic model for equipment health based on a set of equipment status characteristic parameters, identifying abnormal equipment states through multi-parameter fusion training, and obtaining diagnostic results and risk warning information for abnormal equipment states. Specifically: A training sample dataset for real-time classification of device health status is constructed based on a set of device status feature parameters. The health intelligent diagnostic model is used to iteratively train the training sample dataset, and the model parameters are adjusted based on the difference between the abnormal state categories output by the model and the actual abnormal state categories. The trained intelligent health diagnosis model is used to perform feature vectorization on the set of real-time input device status feature parameters to form a real-time device feature vector. The health intelligent diagnostic model is used to match the feature vectors of real-time devices with the feature vectors of standard abnormal state categories one by one, and the weighted distance value is calculated to determine the degree of matching. The abnormal state category of the real-time device feature vector is determined based on the standard abnormal state category feature vector with the highest matching degree, thus forming the real-time device abnormal state diagnosis result; Based on the real-time equipment abnormality diagnosis results and the set of equipment status characteristic parameters, the risk level is determined according to the risk level assessment rules, and risk warning information is generated.

5. The AI-based intelligent IoT method for water conservancy as described in claim 1, characterized in that, The method of performing real-time video recognition and analysis on video data based on a water conservancy video AI model to obtain visual diagnostic results and abnormal event early warning information is as follows: The video data of the on-site water conservancy equipment operation is divided into a continuous video data sequence according to the preset video resolution and video frame rate; A water conservancy video AI model is used to divide each frame of the video data sequence into multiple spatial regions, and video features are extracted for the appearance, structure and operating status of water conservancy equipment in each spatial region. By using a water conservancy video AI model to analyze the changes in video features between consecutive frames in each spatial region, dynamic features of the operating status of water conservancy equipment in each spatial region can be obtained. The dynamic features between spatial regions are correlated using a water conservancy video AI model, and the correlation of operational status change features is fused to form fused video dynamic features. The dynamic features of the fused video are continuously analyzed, and the changes in the device's operating status are judged according to the preset abnormal event evolution trend rules to determine whether the changes in the device's operating status conform to the abnormal event trend and to determine the abnormal type of the device's operating status. Visual diagnostic results and early warning information for abnormal events are generated based on the type of abnormal equipment operation status.

6. The AI-based intelligent IoT method for water conservancy as described in claim 1, characterized in that, The process of simultaneously inputting equipment abnormality diagnosis results, risk warning information, visual diagnosis results, and abnormal event warning information into the Honeycomb Intelligent Management Platform, and outputting equipment abnormality alarm information and a comprehensive diagnostic analysis report, specifically includes: The device abnormal status diagnosis results, risk warning information, visual diagnosis results and abnormal event warning information are synchronously input to the Honeycomb Intelligent Management Platform at fixed intervals through a preset interface. Based on the diagnostic results of abnormal equipment status, determine the location of abnormal equipment and generate a network topology diagram of the equipment. Based on the visual diagnostic results and abnormal event warning information, the system calls on on-site video data to generate remote monitoring images that identify the location, type, and evolution trend of the abnormality. Based on the diagnostic results of abnormal equipment status and risk warning information, maintenance measures are automatically matched to generate an operation and maintenance plan that includes maintenance personnel arrangements, operation steps and maintenance cycles; Based on the equipment abnormality diagnosis results and abnormal event early warning information, generate equipment abnormality alarm information; A comprehensive diagnostic analysis report is generated based on the equipment abnormality diagnosis results, risk warning information, visual diagnosis results, and abnormal event warning information.

7. The AI-based intelligent IoT method for water conservancy as described in claim 1, characterized in that, The mechanism for automatically classifying equipment fault levels based on equipment anomaly alarm information and comprehensive diagnostic analysis reports, and generating equipment fault handling plans, specifically includes: Establish equipment fault level classification rules based on equipment abnormal alarm information and comprehensive diagnostic analysis reports; The equipment fault level is determined by comparing and analyzing the abnormal alarm information and comprehensive diagnostic analysis report according to the equipment fault level classification rules to determine whether the equipment fault level is minor or serious. For equipment with minor malfunctions, determine the current location of the equipment based on the equipment abnormality alarm information, and retrieve the list of maintenance personnel adjacent to the location of the malfunctioning equipment; Send a fault notification message to each maintenance personnel in the maintenance personnel list; For equipment with serious malfunctions, the intelligent operation and maintenance knowledge base is invoked according to the alarm event type and equipment anomaly type to extract historical handling measures and applicable conditions; Based on the abnormality type of the severely faulty equipment and the applicable conditions of the historical handling measures, the corresponding technical experts are selected from the list of experts, and the fault information and historical handling measures are sent in real time through the remote collaborative communication system. Based on the analysis results and suggested measures returned by experts, equipment failure handling plans are formulated in real time.

8. The AI-based intelligent IoT method for water conservancy as described in claim 1, characterized in that, The fusion analysis includes: synchronously transmitting encrypted data transmission information to edge computing devices, edge computing nodes, and cloud computing servers; the edge computing devices decrypting the encrypted data transmission information using a built-in national cryptographic SM2 private key to form initial feature data of the edge device status; the edge computing nodes decrypting the encrypted data transmission information using Rockchip RK3576 or RK3588 chips and a built-in national cryptographic SM2 private key to obtain edge device status association feature data; and the cloud computing server decrypting the encrypted data transmission information using a pre-stored national cryptographic SM2 private key to form comprehensive feature data of the cloud device status. A three-layer collaborative fusion analysis mechanism is constructed on the cloud computing server, and consistency verification is performed on the initial feature data of the edge device status, the associated feature data of the edge device status, and the comprehensive feature data of the cloud device status using data consistency comparison rules to determine the fusion weight of each layer of data. The three-layer collaborative fusion analysis mechanism is then used to perform weighted fusion calculations on the initial feature data of the edge device status, the associated feature data of the edge device status, and the comprehensive feature data of the cloud device status with fusion weights higher than a trusted threshold to obtain a set of device status feature parameters.

9. A water conservancy intelligent IoT system based on artificial intelligence, executing the water conservancy intelligent IoT method based on artificial intelligence as described in any one of claims 1 to 8, characterized in that, include: The field sensing unit is equipped with a water level sensor, video sensor, temperature and humidity sensor and vibration sensor, and is also equipped with a Lora-MCU and Lora gateway to collect real-time operating status data of water conservancy equipment and form a preliminary sensing data set. The edge computing unit, equipped with a hydrological model and Rockchip RK3576 or RK3588 chip, is used to receive the initial sensing data set, perform trend feature extraction and cross-validation analysis of abnormal states in real time, and form preliminary abnormal diagnosis information of the equipment. The data security transmission unit is equipped with the HarmonyOS OS and integrates Rockchip RK3576 or RK3588 chips and security chips with built-in national cryptographic algorithms SM2, SM3 and SM4. It is used to parse the preliminary abnormal diagnosis information of the equipment, calculate message digests, and perform symmetric and asymmetric encryption to generate encrypted data transmission information. The collaborative fusion analysis unit includes end computing devices, edge computing nodes, and cloud computing servers. It synchronously receives encrypted data transmission information through a secure communication protocol channel, decrypts and extracts feature data respectively, and performs feature fusion based on trust level and fusion weight to form a set of device status feature parameters. The equipment health diagnosis unit has a built-in intelligent health diagnosis model, which is used to receive the set of equipment status feature parameters in real time, perform feature vectorization processing and feature similarity matching, determine the real-time equipment abnormal status category and risk level, and output the equipment abnormal status diagnosis results and risk warning information in real time. The video analysis and diagnosis unit has a built-in water conservancy video AI model. It receives video data collected by video sensors in real time, performs spatial partitioning, video feature extraction, regional feature trajectory tracking and fusion, and performs temporal continuity analysis of video dynamic features based on the rule library of abnormal event evolution trends to form visual diagnostic results and abnormal event early warning information. The Honeycomb Intelligent Management Platform receives real-time diagnostic results of abnormal equipment status, risk warning information, visual diagnostic results, and abnormal event warning information through data interface protocols. It generates equipment network topology diagrams, remote monitoring screens, and automated operation and maintenance plans. It outputs real-time equipment abnormal alarm information and comprehensive diagnostic analysis reports. Based on the alarm information and diagnostic analysis reports, it automatically classifies equipment fault levels and calls upon the intelligent operation and maintenance knowledge base to form equipment fault handling solutions.

Images