Water conservancy internet of things multi-dimensional data acquisition and hierarchical fusion method

Abstract

The application provides a water conservancy Internet of Things multi-dimensional data acquisition and hierarchical fusion method, and relates to the field of intelligent water conservancy; a three-layer data fusion architecture of edge-fog-cloud is designed, and a numerical / video / knowledge graph fusion model is constructed; the edge layer collects hydrological and image data through a multi-protocol terminal and adds a four-dimensional space-time label, and adjusts the sampling frequency in combination with a dynamic entropy algorithm; the multi-dimensional water conservancy Internet of Things data collected by the sensor is hierarchically encrypted by adopting an elliptic curve encryption principle; a three-level verification mechanism is established; the fog computing layer aligns the data by adopting a sliding space-time window algorithm; the cloud computing layer realizes cross-domain fusion; and the method effectively supports real-time decision-making requirements such as flood warning and facility monitoring.

Description

Technical Field

[0001] This invention relates to the field of water conservancy Internet of Things (IoT) technology, specifically to a multi-dimensional data acquisition and hierarchical fusion method for water conservancy monitoring scenarios, applicable to fields such as hydrological monitoring, water conservancy facility operation and maintenance, and flood prevention early warning. I. The closest prior art to this patent:

[0002] 1. Brief description of existing technical solutions

[0003] CN117459557B provides a low-code method for fusing multi-dimensional data in the Internet of Things (IoT). This method involves first developing the IoT system and then adding hardware and communication protocol information based on user needs. While this solves the problem of high workload, long development cycle, and high cost in traditional IoT development where changes in devices, protocols, or business scenarios necessitate modifications to a series of algorithm modules based on the initial communication protocol, the method requires centralized cloud processing, resulting in heavy cloud load and poor real-time performance. This makes it unsuitable for scenarios with heterogeneous nodes and geographically dispersed locations in water conservancy IoT systems, and significantly limits its application to water conservancy IoT systems with high real-time requirements.

[0004] CN115550444B provides a water conservancy data acquisition method based on the Internet of Things for water conservancy. It establishes a listening service on the data receiving server and a message queue service on the real-time communication server. While this method overcomes the shortcomings of existing technologies, such as inconvenience in viewing acquired data and maintaining terminal equipment, inability to handle high-concurrency, large-volume reporting of monitoring data from water conservancy acquisition terminals, poor timeliness of data display, and low data parsing efficiency, it also addresses the issues of data encryption and data security during message transmission.

[0005] CN117591997A provides a multi-dimensional data fusion method, device, and medium based on agricultural IoT. This method belongs to the field of agricultural IoT technology. Although it can solve the technical problem that existing agricultural AI models have weak data processing capabilities for high-dimensional agricultural data, making it difficult to efficiently utilize the collected agricultural data and meet the accurate prediction function of agricultural AI models, this method cannot be transferred to the field of water conservancy IoT because the objects of water conservancy IoT and agricultural IoT are fundamentally different.

[0006] CN118736550B provides a system and method for acquiring and transmitting water level data based on the Internet of Things (IoT). While this method eliminates the influence of randomly searching for initial cluster centers and improves the accuracy of clustering results, enabling the prediction of water level data, water data is not limited to a single water level. Using only water level data for model prediction fails to consider the constraints of other hydrological data and spatiotemporal relationships on the model, resulting in a prediction model with poor practicality.

[0007] 2. Objective shortcomings of existing technologies

[0008] The data heterogeneity is prominent: water conservancy monitoring involves multimodal data such as water level, flow, water quality, and video images. Traditional acquisition systems have difficulty in achieving unified processing of structured and unstructured data. This invention collects hydrological, image, and video data through multi-protocol terminals and adds four-dimensional spatiotemporal tags.

[0009] Spatiotemporal reference differences: Different sensors have inconsistent sampling frequencies, including minute-level water level gauges and hour-level water quality monitoring. The discrete spatial distribution makes it difficult to align data in time and space. This invention combines a dynamic entropy algorithm to adjust the sampling frequency and establishes a three-level verification mechanism.

[0010] Poor data security: Sensor nodes in water conservancy scenarios are distributed across different administrative jurisdictions, and data from different regions needs to be centrally processed for training decision-making models. Therefore, there are high requirements for data security during data transmission. This invention proposes elliptic curve encryption technology to perform layered encryption on the collected multidimensional water conservancy IoT data.

[0011] Fog computing is lacking: Existing solutions mostly use centralized cloud processing, resulting in high latency (average >5s) and network bandwidth pressure (daily transmission volume >50GB). This invention designs a three-level data fusion architecture of edge-fog-cloud, with different fusion algorithms in each level to make full use of resources at each level and reduce latency. II. The technology of this patent

[0012] This invention provides a method for multi-dimensional data acquisition and hierarchical fusion in the field of smart water conservancy via the Internet of Things (IoT). It designs a three-level data fusion architecture (edge-fog-cloud) and constructs a numerical / video / knowledge graph fusion model. This effectively supports real-time decision-making needs such as flood warning and facility monitoring, and includes the following steps:

[0013] S1 achieves numerical-level fusion at the edge layer;

[0014] S2 achieves video-level fusion at the fog computing layer;

[0015] S3 achieves advanced convergence at the cloud computing layer;

[0016] The edge layer enables numerical-level fusion, primarily achieving multi-dimensional data acquisition, preprocessing, and numerical fusion processing of the terminal devices connected to this edge node, including the following steps:

[0017] S1.1 Employs elliptic curve encryption principles to perform layered encryption on the multidimensional water conservancy IoT data collected by sensors. Edge-layer intelligent sensing terminals are multifunctional, with sensors simultaneously monitoring multiple devices and parameters, collecting a large amount of diverse data, including sensitive information. Using elliptic curve encryption principles, layered encryption is applied to the multidimensional data collected by sensors. Based on the different characteristics and security requirements of the data, this method places data with consistent encryption rules within the same multidimensional data type. This method not only improves the organization and structuring capabilities of the data but also ensures data security during transmission and storage. Furthermore, layered encryption enables data access control and permission management, preventing unauthorized access and leakage. The elliptic curve encryption includes the following steps:

[0018] S1.11 Elliptic Curve Parameter Settings

[0019] The network host performs layered processing of the source parameters in the key template using a simplified scalar product approach. Let β represent the average value of the parameters of the n code source permutations, and let β represent the curve encryption threshold, guaranteeing... The inequality always holds true;

[0020] S1.12 Layered Encryption Service Implementation

[0021] The core idea of ​​the layered encryption mechanism is to divide data into different layers according to the characteristics and requirements of the data, and set unique encryption keys and algorithm parameters for each layer. By combining OpenSSL source code definition items, compilation sample classification items, and multidimensional data sampling vector physical quantities, a layered encryption expression based on elliptic curve cryptography is constructed to achieve classified collection and encryption processing of multidimensional data. The layered encryption expression is as follows:

[0022]

[0023] In the formula: δ represents the encryption parameters related to different levels, γ represents the multidimensional data sampling vector; γ represents the encrypted data sampling coefficient based on the elliptic curve mechanism. An elliptic encrypted vector representing multidimensional data in a network transmission environment;

[0024] S1.13 Multidimensional Data Deconstruction

[0025] The security of multidimensional data is improved after layered encryption to avoid the risk of unauthorized access or tampering during data transmission and storage. Multidimensional data deconstruction further ensures data integrity and authenticity. During deconstruction, signature initialization provides an authentication and verification mechanism, enabling the recipient to verify the trustworthiness of the data source and preventing malicious tampering from unauthorized senders. Network multidimensional data deconstruction requires the cooperation of the edge layer and the fog computing layer. The edge layer is responsible for defining the specific elliptic curve cryptography algorithm principles; the fog computing layer is responsible for executing the specific multidimensional data deconstruction scheme and defining a complete signature tag document for the sampled data. The deconstruction process verifies the data source through signature initialization, ensuring the recipient confirms the data comes from a trusted sender, preventing unauthorized tampering, and enhancing data security and trustworthiness. The multidimensional data deconstruction signature initialization expression is:

[0026]

[0027] In the formula: t represents the multidimensional data message value, and a represents the deconstruction coefficient of the multidimensional data in the network host after layered encryption. This represents the mean of samples that can be completely deconstructed per unit time. This represents the sample encoding features in the signed document;

[0028] S1.2 Collects hydrological and image data through multi-protocol terminals; ensures the secure transmission of critical data through a sensor node security scheduling mechanism; and solves the correlation problem between data through maximum likelihood estimation of data aggregation.

[0029] The multi-protocol terminal acquisition method deeply analyzes the structure, function, and data transmission mechanism of intelligent sensing terminals, customizes multi-protocol acquisition boxes, and adopts diverse transmission protocols to improve sampling accuracy. The secure scheduling mechanism describes the encryption specificity of multi-dimensional data samples in the sensing terminal set by calculating the weight vector of multi-dimensional heterogeneous data. This weight vector is the aggregation degree calculation value of information objects based on the network multi-dimensional data matrix. After construction and processing by the network host, only information objects of the same type can belong to the same matrix set, solving the problem that the deconstructed data samples cannot meet the network host's operational requirements during real-time transmission. The maximum likelihood estimation controls the value of the multi-dimensional data weight vector within a predetermined numerical range, enabling the network host to formulate a secure aggregation processing scheme.

[0030] S1.3 Additional Four-Dimensional Spacetime Tags

[0031] Each data unit is assigned four-dimensional coordinates: longitude, latitude, elevation, and UTC time.

[0032] S1.4 Adjusting the sampling frequency using the dynamic entropy algorithm

[0033] S1.5 establishes a three-level verification mechanism

[0034] A three-level verification mechanism is adopted, which includes threshold verification based on physical laws, statistical verification based on the improved 3σ criterion, and correlation verification combined with equipment operating conditions.

[0035] The statistical verification criterion based on the improved 3σ criterion is calculated using dynamic standard deviation, and is expressed as follows:

[0036] σ · =σ*(1+|ΔV / V_avg|)

[0037] A mechanism is adopted to adaptively adjust the window size according to the data volatility.

[0038] The fog computing layer relies on the AI ​​acceleration module of the edge gateway to achieve numerical data fusion of multiple edge nodes in the region, primary feature extraction of video, data alignment using a sliding spatiotemporal window algorithm, automatic association of video detection results with spatiotemporal labels of sensor data, and output of structured feature vectors to the cloud platform.

[0039] The cloud computing layer runs on a distributed system, enabling advanced integration, including cross-domain data fusion, integrating meteorological and geographic external system data; multimodal deep association, using GNN to achieve semantic connection between video features and business knowledge; and using RGCN for dynamic graph updates to achieve knowledge graph fusion, constructing a global water conservancy entity relationship network to support comprehensive decision analysis.

[0040] The beneficial effects of this patent are as follows: Compared with the prior art, the innovation of this invention lies in the layered data fusion: the main body is in the fog computing layer, which completes more than 60% of the fusion calculation, including regional data weighting and video target detection; the enhancement is in the cloud platform layer, which handles fusion tasks requiring a global perspective and large-scale computing, completing 35% of the fusion calculation, including knowledge graphs and cross-system associations; the edge layer is responsible for data collection and basic preprocessing, completing 5% of the fusion calculation; the layered fusion method enables the system to balance real-time performance (fog computing layer) and decision-making depth (cloud platform layer), meeting the triple requirements of "real-time edge response + local fog processing + in-depth cloud analysis" in water conservancy IoT.

[0041] This patent innovatively adopts the elliptic curve encryption principle to perform layered encryption on the multi-dimensional water conservancy IoT data collected by sensors, thus solving the problem of data security in heterogeneous networks during multi-protocol data collection. III. Are there any alternatives to the above technical solutions?

[0042] Yes, but research and development is needed. IV. Description of the attached drawings

[0043] Figure 1 System Architecture Diagram V. Specific Implementation Methods

[0044] (I) Hardware Deployment Plan

[0045] 1. Edge Awareness Layer Configuration

[0046] The radar level gauges (12 sets) use the Modbus RTU communication protocol; the gate opening sensors (6 sets) use the LoRaWAN communication protocol; the panoramic cameras (4 sets) use the RTSP over 5G protocol; and the 9-parameter water quality meters (3 sets) use the NB-IoT protocol.

[0047] 2. Fog Computing Layer Equipment

[0048] The edge gateway uses Huawei Atlas 500 smart cell: 8-core ARM v8.2 CPU; AI computing power of 8 TOPS; 16GB LPDDR4X memory;

[0049] The protocol converter uses a customized multi-protocol terminal: supporting Modbus / CoAP / LoRaWAN / NB-IoT protocols; the positioning module uses BeiDou-3 dual-frequency positioning.

[0050] (II) Implementation Steps

[0051] 1. Protocol Adaptive Processing

[0052] Define the core logic for protocol identification, and use principle branch statements to prioritize matching the Modbus RTU protocol, then the LoRaWAN protocol, and finally use a deep learning model to infer the protocol type of the information collected by the terminal device.

[0053] 2. Generation of four-dimensional spatiotemporal tags

[0054] Obtain the original BeiDou coordinates; define the coordinate offset compensation installation azimuth angle θ = 30°, and the equipment offset from the antenna by 2.5m; then convert it into longitude and latitude increments;

[0055] Elevation data were obtained using barometers and temperature compensation algorithms; time synchronization was achieved using BeiDou time service.

[0056] 3. Adjust the sampling frequency using a dynamic entropy algorithm.

[0057] Sliding window configuration: Window size 60 seconds, dynamically expandable to 180 seconds; calculation cycle 5 seconds;

[0058] The pseudocode for the core algorithm is as follows:

[0059] #Dynamic frequency tuning on edge gateways (accelerated by NumPy) defadaptive_sampling(data_stream): window = data_stream[59:61] # Retrieves data from the last 60 seconds entropy=calculate_shannon_entropy(window) std = np.std(window)

[0060] #Dynamic Adjustment Formula ifentropy < 1.5 and std < 0.2: new_rate = base_rate * 0.8 elifentropy>3.0 or std>0.5: new_rate=min(base_rate*1.5,max_rate) else: new_rate=base_rate*(entropy / 2.0)

[0061] #Send frequency control command send_to_sensor(sensor_id,new_rate)

[0062] 4. Alignment of sliding time and space windows

[0063] Temporal alignment uses a cubic spline interpolation enhancement algorithm; spatial alignment uses hybrid interpolation. The spatial interpolation selection logic is defined. Based on the value range of the node health rate parameter, Delaunay triangulation + IDW is selected when it is greater than 0.8, k-NN + RBF is selected when it is between 0.6 and 0.8, and spatiotemporal kriging algorithm is selected when it is less than 0.6.

[0064] 5. Multimodal hierarchical fusion

[0065] Edge layer numerical fusion: entropy weight-CRITIC integrated weight calculation; fog computing layer video fusion: improved YOLOv5 water conservancy target detection; cloud platform layer advanced fusion: knowledge graph generation.

[0066] (III) Implementation Results

[0067] Data alignment takes 0.7 seconds; video analysis latency is 320ms; anomaly detection accuracy is 93%; average daily data transmission volume is 15G; emergency scenario early warning lead time is 2 hours.

Claims

1. A method for multi-dimensional data acquisition and hierarchical fusion in water conservancy Internet of Things (IoT), characterized in that: Design a three-layer data fusion architecture of edge-fog-cloud, and construct a fusion model of numerical data, video, and knowledge graph. The edge layer collects hydrological and image data through multi-protocol terminals and adds four-dimensional spatiotemporal labels, and adjusts the sampling frequency by combining dynamic entropy algorithm. The elliptic curve encryption principle is adopted to perform layered encryption on the multi-dimensional water conservancy IoT data collected by sensors. A three-level verification mechanism is established. The fog computing layer uses a sliding spatiotemporal window algorithm to align the data. Cross-domain integration is achieved at the cloud computing layer, including the following steps: S1 achieves numerical-level fusion at the edge layer; S2 achieves video-level fusion at the fog computing layer; S3 achieves advanced convergence at the cloud computing layer.

2. The method for multi-dimensional data acquisition and hierarchical fusion in water conservancy IoT according to claim 1, characterized in that: The edge layer enables numerical-level fusion, primarily realizing multi-dimensional data acquisition, preprocessing, and numerical fusion processing of the terminal devices connected to this edge node.

3. The method for multi-dimensional data acquisition and hierarchical fusion in water conservancy IoT according to claim 1, characterized in that: The fog computing layer relies on the AI ​​acceleration module of the edge gateway to achieve numerical data fusion of multiple edge nodes in the region, primary feature extraction of video, data alignment using a sliding spatiotemporal window algorithm, automatic association of video detection results with spatiotemporal labels of sensor data, and output of structured feature vectors to the cloud platform.

4. The method for multi-dimensional data acquisition and hierarchical fusion of water conservancy Internet of Things according to claim 1, characterized in that: The cloud computing layer runs on a distributed system, enabling advanced integration, including cross-domain data integration and the integration of meteorological and geographic external system data. Multimodal deep association, using GNN to achieve semantic connection between video features and business knowledge; RGCN is used for dynamic updating of the knowledge graph, realizing knowledge graph fusion and constructing a global water conservancy entity relationship network to support comprehensive decision analysis.

5. A method for multi-dimensional data acquisition and hierarchical fusion in water conservancy IoT according to claim 2, characterized in that: The edge layer enables numerical-level fusion, primarily achieving multi-dimensional data acquisition, preprocessing, and numerical fusion processing of the terminal devices connected to this edge node, including the following steps: S2.1 adopts the elliptic curve encryption principle to perform layered encryption on the multi-dimensional water conservancy Internet of Things data collected by sensors; S2.2 collects hydrological and image data through multi-protocol terminals, ensures the secure transmission of critical data through a sensor node security scheduling mechanism, and solves the correlation problem between data through maximum likelihood estimation of data aggregation. S2.3 Add four-dimensional spatiotemporal labels, adding four-dimensional coordinates to each data unit, namely: longitude, latitude, elevation, and UTC time; S2.4 combines the dynamic entropy algorithm to adjust the sampling frequency; S2.5 establishes a three-level verification mechanism, employing threshold verification based on physical laws; statistical verification based on the improved 3σ criterion; and correlation verification combined with equipment operating conditions.

6. A method for multi-dimensional data acquisition and hierarchical fusion in water conservancy IoT according to claim 5, characterized in that: The elliptic curve encryption principle is adopted to perform layered encryption on the multidimensional water conservancy Internet of Things data collected by sensors. The elliptic curve encryption includes the following steps: S5.1 Elliptic Curve Parameter Settings The network host performs layered processing of the source parameters in the key template using a simplified scalar product approach. Let β represent the average value of the parameters of the n code source permutations, and let β represent the curve encryption threshold, guaranteeing... The inequality always holds true; S5.2 Layered Encryption Service Implementation The core idea of ​​the layered encryption mechanism is to divide data into different layers according to the characteristics and requirements of the data, and set unique encryption keys and algorithm parameters for each layer. By combining OpenSSL source code definition items, compilation sample classification items, and multidimensional data sampling vector physical quantities, a layered encryption expression based on elliptic curve cryptography is constructed to achieve classified collection and encryption processing of multidimensional data. The layered encryption expression is as follows: In the formula: δ represents the encryption parameters related to different levels, γ represents the multidimensional data sampling vector; γ represents the encrypted data sampling coefficient based on the elliptic curve mechanism. An elliptic encrypted vector representing multidimensional data in a network transmission environment; S5.3 Multidimensional Data Deconstruction Multidimensional data, after being encrypted in layers, enhances security to prevent unauthorized access or tampering during data transmission and storage. Multidimensional data deconstruction further ensures data integrity and authenticity. During deconstruction, signature initialization provides an authentication and verification mechanism, enabling the recipient to verify the data's source and prevent malicious tampering from unauthorized senders. Network multidimensional data deconstruction requires the cooperation of the edge layer and fog computing layer. The edge layer defines the specific elliptic curve cryptography algorithm principles, while the fog computing layer executes the specific multidimensional data deconstruction scheme and defines complete signature tags for the sampled data. The deconstruction process verifies the data source through signature initialization, ensuring the recipient confirms the data comes from a trusted sender, preventing unauthorized tampering, and enhancing data security and trustworthiness. The multidimensional data deconstruction signature initialization expression is: In the formula: t represents the multidimensional data message value, and a represents the deconstruction coefficient of the multidimensional data in the network host after layered encryption. This represents the mean of samples that can be completely deconstructed per unit time. This represents the sample encoding features in the signed document.

Images