Intelligent water level monitoring system and data transmission method thereof

Abstract

The invention discloses an intelligent water level monitoring system and a data transmission method thereof, and belongs to the technical field of Internet of Things sensing. The system comprises a distributed water level sensing node, an edge computing gateway and a cloud management platform, wherein the sensing node is integrated with a piezoresistive water level transmitter and an adaptive Kalman filtering module; the data transmission method comprises a dynamic threshold triggering mechanism and a compression algorithm based on improved Huffman coding, and low-power-consumption and high-reliability transmission is realized by optimizing transmission frequency and a data packet structure. The edge computing gateway is internally provided with an LSTM water level prediction model, historical data are analyzed through an LSTM neural network, and an early warning signal is generated in real time. And the cloud management platform stores the key data by adopting a block chain technology. Through multi-sensor data fusion and layered encryption transmission, the water level monitoring error rate is reduced to + / -0.5 cm, the data transmission energy consumption is reduced by 40%, and the monitoring efficiency and the data safety are remarkably improved in flood early warning and reservoir management scenes.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of Internet of Things sensing, and in particular to an intelligent water level monitoring system and a data transmission method thereof. BACKGROUND

[0002] The intelligent water level monitoring system is a typical application of Internet of Things sensing technology in the field of water conservancy monitoring. Its core consists of a water level sensing device, a data transmission module, and a data processing platform. The existing technology generally uses a float type or ultrasonic sensor to collect water level data, transmits the data back through a mobile communication network such as GPRS / 4G, and performs simple early warning based on a threshold alarm mechanism. However, due to the limitations of sensing accuracy, communication efficiency, and algorithm model, the existing system faces problems such as high data distortion rate, excessive energy consumption, and prediction lag in complex environments, making it difficult to meet the comprehensive requirements of real-time, reliability, and energy efficiency ratio in smart water conservancy construction.

[0003] Chinese patent CN114754834A discloses a "water level monitoring device and method based on NB-IoT". This scheme transmits sensor data through Narrow Band Internet of Things (NB-IoT) and uses cloud storage to achieve remote monitoring. Although this patent reduces the standby power consumption of the communication module through a low-power wide-area network, its technical solution still has significant defects:

[0004] Data redundancy transmission: using a fixed sampling frequency (e.g., once every minute) to continuously upload raw data results in a daily transmission volume exceeding 2MB, and the data transmission trigger mechanism is not optimized for water level change characteristics;

[0005] Lack of noise suppression: sensor data is directly uploaded without deploying filtering algorithms on the edge, and actual measurement data shows that the data false alarm rate under heavy rain conditions is as high as 12%;

[0006] Weak security architecture: relying on a centralized cloud server to store data poses a single point of failure risk, and no data integrity verification mechanism is introduced.

[0007] Based on the existing technology of CN114754834A, the current intelligent water level monitoring system has the following technical problems that need to be solved urgently:

[0008] Insufficient dynamic noise suppression: the sensor lacks a filtering algorithm that adapts to environmental interference, resulting in a raw data signal-to-noise ratio (SNR) of less than 20dB;

[0009] Low data transmission energy efficiency ratio: the fixed period transmission mechanism does not match the actual water level change pattern, causing waste of communication resources (invalid data ratio > 40%);

[0010] Prediction model unification: Existing early warning algorithms only rely on historical water level data, without integrating rainfall, basin terrain and other multi-source information, resulting in an average absolute error (MAE) of more than 0.3 meters;

[0011] Data credibility is missing: Centralized storage architecture is vulnerable to malicious tampering, and lacks traceable evidence preservation mechanism, making it difficult to meet the audit compliance requirements of water conservancy supervision. SUMMARY

[0012] In view of the above existing problems, the present application is proposed.

[0013] Therefore, the present application provides an intelligent water level monitoring system and its data transmission method, which solves the problems of insufficient dynamic noise suppression, low data transmission energy efficiency, prediction model unification, and lack of data credibility.

[0014] To solve the above technical problems, the present application provides the following technical solutions:

[0015] In a first aspect, the present application provides an intelligent water level monitoring system, comprising:

[0016] A distributed water level sensing node, which is built-in with a piezoresistive water level transmitter, an adaptive Kalman filter module and a LoRa communication unit; the adaptive Kalman filter module is used for dynamic filtering processing, automatically adjusting noise parameters and eliminating measurement noise; the LoRa communication unit is combined with a dynamic threshold trigger mechanism and only transmits when the data changes exceed the threshold or is periodically updated;

[0017] An edge computing gateway, which is deployed with an LSTM water level prediction model and a dynamic threshold optimizer, and is connected with a cloud management platform through an MQTT protocol; the edge computing gateway runs an LSTM neural network model, analyzes historical and real-time water level data, predicts short-term water level trends (flood risk), and generates early warning signals in advance; an improved Huffman encoding compression algorithm is applied to optimize the data packet structure and transmission frequency; layered encryption (lightweight encryption protocol) is performed;

[0018] A cloud management platform, which is integrated with a blockchain evidence preservation module and a visual early warning interface; the cloud management platform stores key data (early warning records, calibration parameters) using distributed ledger technology; the LSTM water level prediction model is optimized based on global data and is issued to the edge computing gateway.

[0019] As a preferred scheme of the intelligent water level monitoring system, the piezoresistive water level transmitter is used to sense the real-time water level change and provide raw data input; the temperature compensation algorithm of the piezoresistive water level transmitter is:

[0020]

[0021] And limit the change rate after compensation ;

[0022] Wherein The water level height after the temperature compensation algorithm correction, The original water level height;

[0023] The noise covariance matrix Q of the adaptive Kalman filtering module is dynamically adjusted formula:

[0024]

[0025] In the formula, h is the water level, The water level change rate;

[0026] The LoRa communication unit adopts low power wide area network (LPWAN) technology, and the filtered water level data is transmitted to the edge computing gateway in a long distance and low energy consumption mode;

[0027] The LoRa communication unit adopts improved Huffman encoding to compress data, and the encoding tree weight The calculation method is:

[0028]

[0029] Wherein, The water level value Frequency, The adjacent sampling difference value, The highest water level value, The lowest water level value, The water level value Total frequency, The water level coupling coefficient.

[0030] As a preferred scheme of the intelligent water level monitoring system, wherein: the hidden layer state updating formula of the LSTM water level prediction model contains a meteorological factor correction term:

[0031]

[0032] Wherein, The rainfall sensor input, =0.15 is the meteorological coupling coefficient, Is the weight matrix used for linear transformation in cell state updating process, The meteorological factor correction term, The Water level height at time t, The hidden layer State at time t, The updated state of the hidden layer.

[0033] As a preferred scheme of the intelligent water level monitoring system, the condition for triggering data transmission by the dynamic threshold optimizer is met:

[0034]

[0035] wherein, , is the mean of the previous N water levels, and Δt is the sampling interval, is the dynamic threshold.

[0036] As a preferred scheme of the intelligent water level monitoring system, the blockchain storage module adopts a lightweight Merkle tree structure, and the leaf node hash value generation formula is:

[0037]

[0038] wherein, is the denoised water level value obtained after adaptive Kalman filtering, is the data packet generated after improved Huffman encoding compression, is the warning level, and Location is the geographic location identifier of the water level sensing node, is the timestamp of the data record, accurately marking the time when the water level data is generated, is a 256-bit cryptographic hash algorithm, represents data splicing, constructing a lightweight Merkle tree and calculating the root hash value;

[0039] The root hash value is synchronized to the public chain and the regulatory node every 10 minutes.

[0040] As a preferred scheme of the data transmission method of the intelligent water level monitoring system, it includes the following steps:

[0041] S1. The distributed water level sensing node collects water level physical signals (water pressure) in real time through a piezoresistive water level transmitter, converts them into electrical signals and quantizes the water level value, dynamically adjusts the noise parameters by the adaptive Kalman filter module, and obtains the denoised water level value after adaptive Kalman filtering ; S2. Calculate the difference between the current water level and the historical data, when , i.e. exceeding the preset dynamic threshold (±0.5 cm), activate data transmission, trigger improved Huffman encoding compression, and generate data packet D p ={Header, , CRC16}; otherwise, enter hibernation, reduce energy consumption by 40%; send the compressed data packet to the edge computing gateway through the LoRa communication unit, and use the time division multiple access (TDMA) protocol to avoid channel conflict;

[0042] S3. The edge computing gateway receives multi-node data and performs the following operations:

[0043] Check the CRC and eliminate abnormal values, meet

[0044]

[0045] where is the mean value of the nodes in the same area, is the variance of the nodes in the same area;

[0046] Fusion of multi-source data through weighted average or Kalman filtering algorithm;

[0047] Input the effective data into the LSTM water level prediction model to predict the future short-term (within the next hour) water level change trend and generate warning levels

[0048] ,

[0049] where is the rectified linear unit, an activation function for deep learning, is the weight matrix, is the bias vector, is the water level change value predicted by the LSTM water level prediction model;

[0050] Apply hierarchical encryption (such as AES+RSA combination) to the fused data and use improved Huffman encoding for further compression;

[0051] S4. After the cloud management platform receives the data, it constructs a lightweight Merkle tree and writes as a leaf node hash value into the blockchain; integrate multi-region water level data, provide real-time water level, warning map and device status through time series analysis and spatial heat map, support remote monitoring and emergency response; optimize LSTM model parameters based on global data and issue them to the edge computing gateway to realize dynamic upgrading of the prediction algorithm;

[0052] S5. According to the analysis results of the cloud management platform, issue instructions to the distributed water level sensing nodes in reverse, dynamically adjust their sampling frequency and dynamic trigger threshold (increase sensitivity during flood season); optimize TDMA time slot allocation and compression algorithm parameters to balance transmission reliability and energy consumption; periodically update the encryption key between the edge computing gateway and the cloud management platform, and verify the integrity of the data in the blockchain storage module.

[0053] As a preferred solution of the data transmission method of the intelligent water level monitoring system described in the present invention, the compression rate CR of the improved Huffman coding in step S2 is calculated as:

[0054]

[0055] in, The water level value after encoding The bit length is 16, which is the number of original ADC sampling bits.

[0056] As a preferred solution of the data transmission method of the intelligent water level monitoring system of the present invention, wherein: the warning level L in step S3 w The grading rules are:

[0057]

[0058] The conditions for triggering blockchain evidence storage are: ≥1 or single Δh ≥0.3 m.

[0059] As a preferred solution of the data transmission method of the intelligent water level monitoring system of the present invention, the step S4 includes the following steps:

[0060] The cloud management platform receives the encrypted compressed data packet from the edge computing gateway and parses it to obtain the denoised water level value. , warning level , timestamp and geographic location identifier (Location);

[0061] The blockchain evidence storage module generates a leaf node hash value for each water level data according to the following rules:

[0062] ;

[0063] The leaf node hash value is written into the blockchain distributed ledger, and the timestamp and geographic location metadata are stored in association to ensure the immutability and traceability of the water level data;

[0064] Conduct aggregate analysis of multi-region water level data to generate real-time water level heat maps and warning level distribution maps, and dynamically display them through the management platform interface;

[0065] Based on historical and real-time data training, the LSTM neural network parameters are optimized, and the updated LSTM water level prediction model is sent to the edge computing gateway to achieve dynamic upgrades of the edge prediction algorithm.

[0066] As a preferred scheme of the data transmission method of the intelligent water level monitoring system, the step S5 comprises the following steps,

[0067] The cloud management platform generates optimization instructions and sends them to the distributed water level sensing nodes according to historical water level trends, early warning records and energy consumption statistics, dynamically adjusts the sampling frequency, dynamic trigger threshold and sleep cycle, and adapts to environmental changes (increases the sampling frequency to once per minute during the flood season).

[0068] The edge computing gateway balances data transmission reliability and node energy consumption by statistically analyzing the energy consumption and network load data of each sensing node, optimizing the TDMA time slot allocation strategy of the LoRa communication unit and improving the compression rate parameter of the Huffman encoding.

[0069] The cloud management platform regularly updates the layered encryption protocol keys between the edge computing gateway and the distributed water level sensing nodes, and verifies the hash consistency of historical data through the block chain evidence module to detect and repair potential tampering or abnormal nodes.

[0070] The edge computing gateway periodically performs health status detection (battery level, signal strength) of the distributed water level sensing nodes, and uploads abnormal node information to the cloud management platform in real time to trigger device maintenance alarms and dynamically adjust the monitoring network topology.

[0071] In a second aspect, the application provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and wherein the computer program, when executed by the processor, implements any step of the data transmission method of the intelligent water level monitoring system according to the first aspect of the application.

[0072] In a third aspect, the application provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements any step of the data transmission method of the intelligent water level monitoring system according to the first aspect of the application.

[0073] The application has the following advantages:

[0074] 1. High-precision water level monitoring

[0075] The piezoresistive sensor and adaptive Kalman filter are used to dynamically compensate for temperature drift (compensation coefficient γ=0.0002 / ℃), reducing the measurement error from ±3cm in the traditional scheme to ±0.5cm, and improving the signal-to-noise ratio (SNR) to 35dB.

[0076] Through abnormal value clustering verification, turbulent interference data is effectively eliminated, and the false alarm rate is reduced from 12% to 0.8%.

[0077] 2. Ultra-low power data transmission

[0078] The dynamic threshold trigger mechanism intelligently adjusts the sampling frequency based on the rate of water level change, reducing invalid transmissions by more than 60%. Combined with improved Huffman coding (compression ratio CR=0.42), the average daily data volume is compressed from 2MB to 0.84MB.

[0079] The LoRa communication unit optimizes its sleep strategy, reducing the overall system power consumption to 40% of traditional GPRS solutions, and extending the battery life of a 2000mAh battery from 3 months to 18 months.

[0080] 3. Accurate water level forecasting and early warning

[0081] The meteorological fusion LSTM water level prediction model reduces the mean absolute error (MAE) of water level prediction from 0.3m to 0.185m by coupling rainfall data, and the accuracy of trend prediction for the next two hours is >92%;

[0082] It supports multi-level early warnings, shortening the trigger response time from 15 minutes in traditional solutions to 3 seconds, making it particularly suitable for flash flooding scenarios.

[0083] 4. Trusted Data Storage System

[0084] An edge cloud collaborative evidence storage network is built based on a lightweight consortium chain, with a latency of less than 5 minutes for key data to be uploaded to the chain, and storage costs are only 22% of the full-chain solution.

[0085] Through Merkle tree hash anchoring (SHA256) and regulatory penetration audits, data tampering is ensured to be traceable, meeting the compliance requirements of the Ministry of Water Resources' "Hydrological Monitoring Data Management Measures."

[0086] 5. Strong adaptability to complex environments

[0087] The LoRa communication unit has a packet loss rate of ≤0.3% within a range of 10km (traditional solutions ≥8.7%) and supports a wide operating temperature range of -40℃~85℃;

[0088] The system uses FPGA acceleration to achieve LSTM water level prediction model inference time of less than 10ms, and can support large-scale network monitoring of 1000+ nodes.

[0089] Through multi-level technological innovation, this invention achieves significant improvements in sensing accuracy, communication energy efficiency, predictive capabilities, and data security. Experiments with the Henan Smart Water Conservancy Project have shown that the false alarm rate in flood warning scenarios is reduced by 90%, emergency response efficiency is increased by 70%, and annual operation and maintenance costs are reduced by 45%, providing reliable technical support for the intelligent management of water conservancy facilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0090] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without any creative effort based on these drawings.

[0091] Figure 1 A schematic diagram of an intelligent water level monitoring system in Embodiment 1;

[0092] Figure 2 A flow chart of a data transmission method of an intelligent water level monitoring system in Embodiment 2. DETAILED DESCRIPTION

[0093] In order to make the above objectives, features and advantages of the present application more apparent and comprehensible, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0094] In the following description, a large number of specific details are set forth in order to facilitate a thorough understanding of the present application, but the present application can also be implemented in other ways different from those described herein, and those skilled in the art can make similar generalizations without departing from the connotation of the present application, therefore the present application is not limited to the specific embodiments disclosed below.

[0095] Secondly, the "one embodiment" or "embodiment" referred to herein means that the specific features, structures or characteristics can be included in at least one implementation of the present application. The "in one embodiment" appearing in different places in the specification does not mean the same embodiment, nor is it an independent or alternative embodiment that excludes other embodiments.

[0096] Embodiment 1, refer to Figure 1 , the first embodiment of the present application, the embodiment provides an intelligent water level monitoring system, comprising:

[0097] The distributed water level sensing node 101 is deployed at a monitoring point and is used for water level data collection, local processing and low-power consumption transmission. The distributed water level sensing node 101 is internally provided with a piezoresistive water level transmitter 101a, an adaptive Kalman filtering module 101b and a LoRa communication unit 101c. The piezoresistive water level transmitter 101a converts a water level physical signal (water pressure) into an electrical signal through a piezoresistive principle, realizes high-precision water level measurement (error rate ±0.5 cm), is used for real-time sensing of water level changes and provides original data input. The adaptive Kalman filtering module 101b is used for dynamic filtering processing of sensor original data, automatically adjusts noise parameters to adapt to environmental interference (water flow fluctuation, temperature drift) and eliminates measurement noise. The LoRa communication unit 101c adopts a low-power wide-area network (LPWAN) technology, transmits filtered water level data to an edge computing gateway 102 in a long-distance and low-energy consumption manner.

[0098] The edge computing gateway 102 is deployed with an LSTM water level prediction model 102a and a dynamic threshold optimizer 102b and is connected with a cloud management platform 103 through an MQTT protocol. The edge computing gateway 102 is used for realizing data aggregation, real-time analysis and early warning. The edge computing gateway 102 receives LoRa data of multiple sensing nodes, performs multi-sensor data fusion, runs an LSTM neural network model, analyzes historical and real-time water level data, predicts a short-term water level trend (flood risk), generates an early warning signal in advance, applies an improved Huffman encoding compression algorithm, optimizes data packet structure and transmission frequency, performs layered encryption (a lightweight encryption protocol) and guarantees transmission safety of data at the edge layer.

[0099] The cloud management platform 103 is integrated with a block chain storage module 103a and a visual early warning interface 103b. The cloud management platform 103 is used for global data management, safe storage and high-level application services. Distributed ledger technology is used to store key data (early warning records, calibration parameters) and ensure that data cannot be tampered with. Multi-region data is integrated to perform long-term trend analysis and reservoir scheduling optimization. A management interface is provided to display real-time water level, early warning maps and equipment status, support remote monitoring and emergency response, train and optimize the LSTM water level prediction model 102a based on global data and issue the model to the edge computing gateway 102 to realize algorithm iteration.

[0100] The temperature compensation algorithm of the piezoresistive water level transmitter 101a is as follows:

[0101]

[0102] and limits the change rate after compensation ;

[0103] wherein is the water level height after the temperature compensation algorithm correction, is the original water level height;

[0104] The noise covariance matrix Q of the adaptive Kalman filter module 101b is dynamically adjusted according to the formula:

[0105]

[0106] wherein h is the water level height, is the water level change rate;

[0107] The LoRa communication unit 101c adopts improved Huffman encoding to compress data, and the encoding tree weight value is The calculation method is:

[0108]

[0109] wherein, is the water level value occurrence frequency, is the adjacent sampling difference value, is the highest water level value, is the lowest water level value, is the water level value total frequency of occurrence, is the water level coupling coefficient.

[0110] The hidden layer state update formula of the LSTM water level prediction model 102a contains a meteorological factor correction term:

[0111]

[0112] wherein, is the rainfall sensor input, =0.15 is the meteorological coupling coefficient, is the weight matrix used for linear transformation in the cell state update process, is the meteorological factor correction term, is the water level height at time t, is the hidden layer state at time t, is the updated state of the hidden layer.

[0113] The condition for triggering data transmission of the dynamic threshold optimizer 102b is satisfied:

[0114]

[0115] wherein, , is the average of the previous N water levels, and Δt is the sampling interval, The dynamic threshold value is determined.

[0116] The blockchain storage module 103a adopts a lightweight Merkle tree structure, and a leaf node hash value generation formula is as follows:

[0117]

[0118] The root hash value is synchronized to the public chain 103a1 and the regulatory node 103a2 every 10 minutes.

[0119] Embodiment 2, with reference to Figure 2 The second embodiment of the present application provides a data transmission method of an intelligent water level monitoring system, comprising the following steps:

[0120] S1. The distributed water level sensing node 101 collects water level physical signals (water pressure) in real time through the piezoresistive water level transmitter 101a, converts them into electrical signals, quantifies the water level value, dynamically adjusts the noise parameters through the adaptive Kalman filter module 101b, and obtains the denoised water level value after adaptive Kalman filtering ;

[0121] This step realizes high-precision water level data acquisition through the cooperative design of the sensing node hardware architecture and the adaptive filtering algorithm. The specific implementation process is as follows:

[0122] 1. Hardware acquisition layer

[0123] Sensor selection and installation:

[0124] MPM4899 type piezoresistive water level transmitter (range 0-10m, accuracy ±0.1%FS) is used, which is installed on the vertical support of the monitoring point revetment, and the probe is immersed in water with a depth of ≥20cm to avoid water surface fluctuation interference.

[0125] Signal conditioning circuit: built-in 24-bit ADC (ADS1248), sampling rate 1Hz, output raw data

[0126]

[0127] Wherein, k=0.5m / V, is the reference sea level elevation.

[0128] Temperature compensation module:

[0129] DS18B20 temperature sensor is integrated to collect water temperature T in real time, and the following formula is used for correction:

[0130]

[0131] And limit the change rate after correction to prevent sudden noise.

[0132] 2. Adaptive Kalman filter algorithm

[0133] State space model:

[0134] Define system state (water level h, rate of change v), observation equation

[0135]

[0136] where

[0137]

[0138] Process noise covariance Dynamic adjustment, measurement noise covariance R is fixed.

[0139] Filter iteration process:

[0140] (1) Prediction stage:

[0141]

[0142]

[0143] (2) Update stage:

[0144]

[0145]

[0146]

[0147] Output denoised water level value .

[0148] Adaptive strategy:

[0149] Every 60 seconds, calculate the water level change rate according to the historical 10 data points , according to the formula

[0150]

[0151] Adjust , realize the dynamic adaptation of noise model.

[0152] 3. Embedded software implementation

[0153] Microcontroller configuration: STM32L476RG (Cortex-M4F, 80MHz) is used, resource allocation: timer TIM2 triggers 1Hz sampling interrupt; DMA channel transmits ADC data to memory; FPU unit accelerates matrix operation.

[0154] Code optimization:

[0155] Matrix operations use ARM CMSIS-DSP library functions (arm_mat_mult_f32, etc.), Kalman gain K k The calculation uses Cholesky decomposition to avoid matrix singularity, and the state variables are stored in single-precision floating point, with a memory usage of ≤2KB.

[0156] Exception handling mechanism:

[0157] If 3 times in a row , triggering the sensor self-test procedure; when the ADC reference voltage drift is detected to be greater than 1%, automatic calibration is started.

[0158] 4. Effect verification data Indicator Before filtering After filtering Lifting amplitude Signal-to-noise ratio (SNR) 18.2 dB 35.6 dB +95% Root mean square error (RMSE) 2.7 cm 0.48 cm -82% Instantaneous mutation suppression rate 23% 89% +287%

[0159] Example (Flood Monitoring Scenario)

[0160] 1. Sensor nodes are deployed at river sections to collect raw data

[0161]

[0162] 2. After temperature compensation

[0163] 3. Adaptive Kalman filter output , rate of change v=0.02m / s

[0164] 4. Dynamic threshold judgment , because |3.81-3.80|=0.01<0.075, transmission is not triggered for the time being.

[0165] It should be noted that through Q k Dynamic adjustment to adapt to different working conditions of still water and turbulent flow, filter stability is improved by 60%, the algorithm only occupies 15% of CPU resources on STM32, meeting real-time requirements, temperature compensation and anomaly detection mechanism extend the sensor life to 8 years (MTBF ≥ 50,000 hours). S2. Calculate the difference between the current water level and historical data. When the preset dynamic threshold (±0.5cm) is exceeded, data transmission is activated, the improved Huffman coding compression is triggered, and data packet D is generated. p ={Header, ,CRC16}; otherwise, it enters sleep mode, reducing energy consumption by 40%; the compressed data packet is sent to the edge computing gateway 102 through the LoRa communication unit 101c, and the time division multiple access (TDMA) protocol is used to avoid channel conflicts;

[0166] The compression rate CR of the improved Huffman coding in step S2 is calculated as:

[0167]

[0168] wherein, is the bit length of the water level value after encoding, and 16 is the original ADC sampling bit number.

[0169] This step realizes efficient event-driven data transmission through the synergistic optimization of the dynamic threshold triggering mechanism and the improved Huffman coding algorithm. The specific implementation process is as follows:

[0170] 1. Dynamic threshold determination and triggering logic

[0171] Threshold calculation:

[0172] Based on the formula , the dynamic threshold is updated every 10 seconds:

[0173]

[0174] wherein , the calculation process uses a sliding window method with a window length of 10 (N=10).

[0175] Triggering conditions:

[0176] When any of the following conditions is met, data transmission is activated:

[0177] (1) Instantaneous mutation:

[0178] (2) Periodic keep-alive: When no transmission is triggered for 30 consecutive minutes, the latest data is forced to be sent (to prevent communication link from dying).

[0179] Hardware implementation:

[0180] In the STM32 microcontroller, a low-power comparator (LPCOMP) is configured to monitor in real time. When , an external interrupt is triggered to wake up the LoRa module.

[0181] 2. Improved Huffman coding compression

[0182] Statistical feature analysis:

[0183] 10,000 sets of historical water level data are pre-stored to statistically analyze the occurrence frequency and the adjacent difference to build a dynamic encoding table.

[0184] Weight optimization formula: ​

[0185] The symbol weight is calculated by the formula

[0186]

[0187] where = 10.0 m, = 0.0 m, and the weight update period is 24 hours.

[0188] Binary tree construction:

[0189] According to , the water level values (resolution 0.01 m, range 0.00-10.00 m) are sorted to generate a priority queue;

[0190] The minimum weight node is merged using a greedy algorithm to construct a binary tree, with the left branch being 0 and the right branch being 1.

[0191] Limit the encoding length ≤ 12 bits to avoid long tail effects.

[0192] Compression packet generation:

[0193] Header (2 bytes):

[0194] Bit 0-3: Packet type (0001 indicates water level data)

[0195] Bit 4-7: Sensor ID (supports 16 nodes)

[0196] Bit 8-15: Timestamp low 8 bits (precision 1 minute)

[0197] Payload (variable length):

[0198] Encoded water level value , for example, "3.85 m" corresponds to the encoding "01011011" (length 8 bits)

[0199] CRC16 (2 bytes): Calculate the check code using the CCITT polynomial x 16 +x 12 +x 5 +1, covering the Header and Payload.

[0200] 3. Low-power transmission control

[0201] LoRa parameter configuration:

[0202] Operating frequency band: 470 MHz (China's water conservancy special frequency band)

[0203] ​Spreading factor SF = 10, bandwidth BW = 125 kHz, coding rate CR = 4 / 5

[0204] Transmit power: 14 dBm (communication distance 10 km)

[0205] Timing management:

[0206] (1) After triggering transmission, the LoRa module completes power-on initialization within 50 ms;

[0207] (2) Time consumption of sending data packet ;

[0208] (3) Immediately enter deep sleep mode (current <1 μA) after sending is completed.

[0209] Conflict avoidance:

[0210] Random backoff algorithm (CSMA / CA) is adopted, retry sending after 10-100 ms random delay, maximum retry number 3 times.

[0211] 4. Effect verification data Indicator Traditional scheme (fixed period) Invention scheme (dynamic trigger) Lifting amplitude Daily average transmission times 1440 times (1 Hz) 53 times (average) -96.3% Single data packet size 16 bytes (original) 4.2 bytes (after compression) -73.8% Average power consumption 120 mW 28 mW -76.7% Effective data ratio 41% 92% +124%

[0212] Example (sudden water level rise scenario)

[0213] (1) Initial state: water level stable at 3.80 m, dynamic threshold = 0.07 m;

[0214] (2) Mutation trigger: heavy rain causes water level to rise to 4.15 m within 10 minutes, triggering |4.15-3.80|=0.35 m 0.07 m;

[0215] (3) Encoding compression: encode 4.15 m as "10110101" (weight w = 0.12 + 0.3 x 0.35 / 10 = 0.1225);

[0216] (4) Data packet sending: generate D p ={Header: 0x1A3C, Payload: 0xB5, CRC16: 0x7E2F}, time consumption 412 ms to complete transmission;

[0217] (5) Recovery sleep: module sleep current drops to 1 μA until next trigger.

[0218] It should be noted that by dynamic threshold, invalid transmission is reduced by 96%, prolonging the service life of the device; improved Huffman algorithm increases compression rate to 58% (traditional solution ≤45%); CRC16 check + random backoff mechanism, packet loss rate from 8.7% to 0.3%; encoding table only occupies 512B Flash, suitable for low-end MCU.

[0219] S3. The edge computing gateway 102 receives multi-node data and performs the following operations:

[0220] Check CRC and eliminate outliers, meet

[0221]

[0222] Where is the mean value of nodes in the same area, is the variance of nodes in the same area;

[0223] Fusion of multi-source data by weighted average or Kalman filtering algorithm;

[0224] Input the effective data into the LSTM water level prediction model 102a to predict the future short-term (within the next hour) water level change trend, and generate warning level

[0225] ,

[0226] Where is the rectified linear unit, an activation function of deep learning, is the weight matrix, is the bias vector, is the water level change value after the LSTM water level prediction model 102a prediction;

[0227] Apply hierarchical encryption (such as AES+RSA combination) to the fused data, and use improved Huffman encoding for further compression;

[0228] The grading rules of the warning level L w in step S3 are as follows:

[0229]

[0230] The conditions for triggering the blockchain notarization at the same time are ≥1 or single Δh≥0.3 m.

[0231] This step realizes efficient and reliable water level monitoring through multi-dimensional data processing and intelligent early warning decision of the edge computing gateway 102. The specific implementation process is as follows:

[0232] 1. Data reception and verification

[0233] Protocol analysis:

[0234] Edge computing gateway 102 receives sensor node data packet D through LoRaWAN gateway (such as Semtech SX1302) p , parses the sensor ID (4 bits) and timestamp (8 bits) in the Header, and distributes it to the corresponding processing thread according to the MQTT topic.

[0235] CRC check and exception filtering:

[0236] (1) Calculate the check code using CCITT-16 polynomial:

[0237]

[0238] (2) If , discard the data packet and record the error log;

[0239] (3) The data that passes the check enters the clustering check:

[0240]

[0241] Where M is the number of adjacent nodes in the same basin, σ cluster is the cluster standard deviation.

[0242] 2. LSTM water level prediction model inference

[0243] Model input features:

[0244] Time window features: past 1 hour water level sequence (sampling interval 1 minute)

[0245] External features: real-time rainfall P rain (obtained through weather API, resolution 0.1 mm / h)

[0246] Geographical features: basin area A (preset parameter, unit km²)

[0247] LSTM network structure:

[0248] Weather coupling calculation: inject rainfall influence (θ=0.15) in LSTM cell state update

[0249] Real-time inference: trigger prediction once for each new data received, inference time <10ms (RK3396 processor)

[0250] 3. Warning level generation and decision making

[0251] Classification logic:

[0252] According to the following formula to divide the warning level:

[0253]

[0254] (Normal): Green indicator light, data is only stored but not uploaded to the chain

[0255] (Alert): Yellow warning, triggering SMS notification + local sound and light alarm

[0256] (Critical): Red alert, start emergency broadcast + automatically report to flood control headquarters

[0257] Blockchain evidence trigger:

[0258] When the following conditions are met, the blockchain evidence storage module 103a is called:

[0259]

[0260] Evidence data package format:

[0261] BlockData={Timestamp,SensorID, , ,Location(GPS)}

[0262] 4. Edge-Cloud Co-processing

[0263] Data distribution strategy:

[0264] Low-level data ( ): Batch upload to the cloud every 15 minutes (saving bandwidth)

[0265] High-level data ( / ): Real-time synchronization to the cloud and supervision nodes

[0266] Model online update:

[0267] The cloud management platform 103 aggregates the data of each edge computing gateway 102 every day to generate an incremental training set ΔD;

[0268] Update the weights of LSTM water level prediction model 102a through the federated learning framework ;

[0269] The model differential file (<100KB) is digitally signed and pushed to the edge computing gateway 102.

[0270] 5. Hardware Implementation Example

[0271] Edge computing gateway 102 configuration:

[0272] Processor: Rockchip RK3399 (Dual-core Cortex-A72 + Quad-core Cortex-A53)

[0273] Memory: 4GB LPDDR4 + 32GB eMMC

[0274] Communication Interface: LoRaWAN Concentrator + 4G Module (Yiyuan EC20)

[0275] Example (Flash Flood Warning Scenario)

[0276] Data Reception: The edge computing gateway 102 receives the data packet of sensor ID = 5 , and the analysis result is = 2.85 m;

[0277] Cluster Verification: The average of the other three nodes in the same river basin is μ cluster = 2.83 m, and the standard deviation is σ cluster = 0.05 m, and the verification is passed;

[0278] Model Prediction: Input the water level data in the past 60 minutes and the rainfall P rain = 15 mm / h, and output Δh pred = 0.63 m;

[0279] Warning Decision: Trigger = 2, start red warning and chain storage;

[0280] Blockchain Record: Generate hash , synchronized to the supervision node within 5 minutes.

[0281] Technical Effect Verification

[0282] It should be noted that the false positive rate is reduced by 86% through CRC + cluster analysis double filtering; the inference speed of the LSTM water level prediction model 102a is <10 ms, supporting hundred-node concurrency; the dynamic warning threshold is intelligently linked with the blockchain storage; the federal learning incrementally updates the model, saving 94% of traffic.

[0283] S4. After receiving the data, the blockchain storage module 103a of the cloud management platform 103 constructs a lightweight Merkle tree, writes as a leaf node hash value to the blockchain; integrates multi-region water level data, provides a management interface to display real-time water level, warning map and device status through time series analysis and spatial heat map, supports remote monitoring and emergency response; optimizes the LSTM model parameters based on global data, and issues them to the edge computing gateway 102 to realize dynamic upgrade of the prediction algorithm;

[0284] The cloud management platform 103 receives the encrypted compressed data packets from the edge computing gateway 102, and parses the denoised water level value , early warning level , timestamp (Timestamp) and location identification (Location);

[0285] The blockchain storage module 103a generates a leaf node hash value for each water level data according to the following rules:

[0286] ;

[0287] Write the leaf node hash value into the blockchain distributed ledger, and store the timestamp and geographic location metadata to ensure the non-tamperability and traceability of the water level data;

[0288] Aggregate and analyze multi-region water level data to generate real-time water level heat maps and early warning level distribution maps, and dynamically display them through the management platform interface;

[0289] Optimize the LSTM neural network parameters based on historical and real-time data, and update the LSTM water level prediction model 102a to the edge computing gateway 102 to realize dynamic upgrading of the edge prediction algorithm.

[0290] S5. According to the analysis results of the cloud management platform 103, the instructions are sent back to the distributed water level sensing nodes 101 to dynamically adjust their sampling frequency, dynamic trigger threshold (increase sensitivity during flood season); statistical analysis of the energy consumption and network load of each node, optimization of TDMA time slot allocation and compression algorithm parameters, balance of transmission reliability and energy consumption; Regularly update the encryption key between the edge computing gateway 102 and the cloud management platform 103, and verify the integrity of the data of the blockchain storage module 103a.

[0291] The cloud management platform 103 generates optimization instructions based on historical water level trends, warning records and energy consumption statistics, and sends them to the distributed water level sensing nodes 101 to dynamically adjust their sampling frequency, dynamic trigger threshold and sleep cycle, and adapt to environmental changes (increase the sampling frequency to once per minute during the flood season);

[0292] The edge computing gateway 102 collects the energy consumption and network load data of each sensing node, optimizes the TDMA time slot allocation strategy of the LoRa communication unit, and improves the compression rate parameters of the improved Huffman encoding, balancing data transmission reliability and node energy consumption;

[0293] The cloud management platform 103 regularly updates the layered encryption protocol key between the edge computing gateway 102 and the distributed water level sensing nodes 101, and verifies the hash consistency of historical data through the blockchain storage module 103a, detects and repairs potential tampering or abnormal nodes;

[0294] The edge computing gateway 102 periodically performs distributed water level sensing node 101 health state detection (battery level, signal strength), and uploads abnormal node information to the cloud management platform 103 in real time, triggers device maintenance alarm and dynamically adjusts the monitoring network topology.

[0295] The embodiment also provides a computer device suitable for the case of the data transmission method of the intelligent water level monitoring system, including a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions to realize the data transmission method of the intelligent water level monitoring system proposed in the above embodiment.

[0296] The computer device can be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The communication interface of the computer device is used to communicate with external terminals in a wired or wireless manner. The wireless manner can be achieved through WIFI, an operator network, NFC (Near Field Communication) or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the computer device, or an external keyboard, touchpad or mouse, etc.

[0297] The embodiment also provides a storage medium having a computer program stored thereon, the program being executed by a processor to realize the data transmission method of the intelligent water level monitoring system proposed in the above embodiment. The storage medium can be realized by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic memory, a flash memory, a magnetic disk or an optical disk.

[0298] In summary, the application adopts a piezoresistive sensor and an adaptive Kalman filter to dynamically compensate for temperature drift (compensation coefficient γ = 0.0002 / ℃), reducing the measurement error from ±3cm in the traditional scheme to ±0.5cm, and improving the signal-to-noise ratio (SNR) to 35dB; through outlier clustering verification, effectively eliminating turbulent interference data, and reducing the false positive rate from 12% to 0.8%. The dynamic threshold triggering mechanism intelligently adjusts the sampling frequency according to the water level change rate, reducing the number of invalid transmissions by more than 60%, combined with the improved Huffman encoding (compression ratio CR = 0.42), which compresses the daily data volume from 2MB to 0.84MB; the LoRa communication unit optimizes the sleep strategy, and the overall power consumption of the system is reduced to 40% of the traditional GPRS scheme, and the 2000mAh battery endurance is extended from 3 months to 18 months. The weather fusion LSTM water level prediction model 102a reduces the average absolute error (MAE) of water level prediction from 0.3m to 0.185m by coupling rainfall data, and the future 2-hour trend prediction accuracy is >92%; supports multi-level early warning, and the trigger response time is shortened from 15 minutes in the traditional scheme to 3 seconds, especially suitable for flash flood scenarios. Based on the lightweight consortium chain, an edge cloud collaborative storage network is constructed, the key data on-chain delay is <5 minutes, and the storage cost is only 22% of the full-chain on-chain scheme; through Merkle tree hash anchoring (SHA256) and regulatory penetration audit, it ensures that data tampering is traceable, and meets the compliance requirements of the Ministry of Water Resources "Hydrological Monitoring Data Management Regulations". The LoRa communication unit has a packet loss rate ≤0.3% within 10km (traditional scheme ≥8.7%), supporting a wide temperature range of -40℃~85℃; the system realizes LSTM water level prediction model 102a inference time consumption <10ms through FPGA acceleration, and can support large-scale networking monitoring of 1000+ nodes. Through multi-level technical innovation, the application significantly improves sensing accuracy, communication energy efficiency, prediction capability, and data security, etc. The application is tested in the smart water conservancy project in Henan Province, and the actual measurement shows that in the flood warning scenario, the false positive rate is reduced by 90%, the emergency response efficiency is improved by 70%, and the annual operation cost is reduced by 45%, providing reliable technical support for intelligent management of water conservancy facilities.

[0299] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and not limit the present application. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present application, which should be covered by the scope of the claims of the present application.

Claims

1. An intelligent water level monitoring system, characterized in that: include: A distributed water level sensing node (101) has a built-in piezoresistive water level transmitter (101a), an adaptive Kalman filter module (101b), and a LoRa communication unit (101c); the adaptive Kalman filter module (101b) is used for dynamic filtering processing, automatically adjusting noise parameters, and eliminating measurement noise; the LoRa communication unit (101c), combined with a dynamic threshold trigger mechanism, transmits data only when the data change exceeds a threshold or is periodically updated; An edge computing gateway (102) is deployed with an LSTM water level prediction model (102a) and a dynamic threshold optimizer (102b), and is connected to a cloud management platform (103) via an MQTT protocol; the edge computing gateway (102) runs an LSTM neural network model to analyze historical and real-time water level data, predict short-term water level trends, and generate early warning signals; applies an improved Huffman coding compression algorithm to optimize data packet structure and transmission frequency; and performs layered encryption; A cloud management platform (103) is integrated with a blockchain evidence storage module (103a) and a visual warning interface (103b); the cloud management platform (103) uses distributed ledger technology to store key data; and an LSTM water level prediction model (102a) is optimized based on global data training and sent to an edge computing gateway (102).

2. The intelligent water level monitoring system according to claim 1, characterized in that: The piezoresistive water level transmitter (101a) is used to sense water level changes in real time and provide raw data input; the temperature compensation algorithm of the piezoresistive water level transmitter (101a) is: ; And limit the rate of change after compensation ; in is the water level height corrected by the temperature compensation algorithm, is the original water level; The dynamic adjustment formula of the noise covariance matrix Q of the adaptive Kalman filter module (101b) is: ； Where h is the water level height, is the water level change rate; The LoRa communication unit (101c) uses low-power wide area network technology to transmit the filtered water level data to the edge computing gateway (102) in a long-distance and low-energy manner; The LoRa communication unit (101c) uses improved Huffman coding to compress data, and its coding tree weight The calculation method is: ； in, Water level value Frequency of occurrence, is the difference between adjacent samples, is the highest water level value, is the lowest water level value, Water level value The total frequency of occurrence, is the water level coupling coefficient.

3. The intelligent water level monitoring system according to claim 1, characterized in that: The hidden layer state update formula of the LSTM water level prediction model (102a) includes a meteorological factor correction term: ； in, For rainfall sensor input, is the meteorological coupling coefficient, is the weight matrix used for linear transformation in the cell state update process, is the meteorological factor correction term, for The water level at the moment, For the hidden layer The state of the moment, is the updated state of the hidden layer.

4. The intelligent water level monitoring system according to claim 1, characterized in that: The dynamic threshold optimizer (102b) triggers data transmission when: ； in, , For the front The mean of the secondary water level, Δt is the sampling interval, is the dynamic threshold.

5. The intelligent water level monitoring system according to claim 1, characterized in that: The blockchain evidence storage module (103a) adopts a lightweight Merkle tree structure, and the leaf node hash value generation formula is: ； in, is the denoised water level value obtained after adaptive Kalman filtering, This is the data packet generated after the improved Huffman coding compression. is the warning level, Location is the geographical location identifier of the water level sensor node, Timestamp for data recording, accurately marking the moment when water level data is generated, It is a 256-bit cryptographic hash algorithm. Represents data concatenation, constructs a lightweight Merkle tree and calculates the root hash value; The root hash value is synchronized to the public chain (103a1) and the supervisory node (103a2) every 10 minutes.

6. A data transmission method for an intelligent water level monitoring system, which is implemented based on the intelligent water level monitoring system according to any one of claims 1 to 5, characterized in that: The following steps are involved: S1. The distributed water level sensor node (101) collects the water level physical signal in real time through the piezoresistive water level transmitter (101a), converts it into an electrical signal and quantifies the water level value. The adaptive Kalman filter module (101b) dynamically adjusts the noise parameters, and the denoised water level value is obtained after the adaptive Kalman filter. ; S2. Calculate the difference between the current water level and the historical data. When the preset dynamic threshold is exceeded, data transmission is activated, the improved Huffman coding compression is triggered, and data packet D is generated. p ={Header, CRC16}; otherwise, it enters sleep mode, reducing energy consumption by 40%; the compressed data packet is sent to the edge computing gateway (102) via the LoRa communication unit (101c), and a time-division multiple access protocol is used to avoid channel conflicts; S3. The edge computing gateway (102) receives multi-node data and performs the following operations: Check CRC and remove abnormal values ​​to meet ； in is the mean value of nodes in the same region, is the variance of nodes in the same region; Fusion of multi-source data through weighted averaging or Kalman filtering algorithms; Input valid data into the LSTM water level prediction model (102a) to predict the future short-term water level change trend and generate the warning level ； in It is a rectified linear unit, an activation function for deep learning. is the weight matrix, is the bias vector, is the water level change value predicted by the LSTM water level prediction model (102a); Apply layered encryption to the fused data and further compress it using modified Huffman coding; S4. After receiving the data, the blockchain evidence module (103a) of the cloud management platform (103) constructs a lightweight Merkle tree. Write the hash value of the leaf node into the blockchain; integrate water level data from multiple regions, and provide a management interface to display real-time water level, warning map and equipment status through time series analysis and spatial heat map, supporting remote monitoring and emergency response; optimize LSTM model parameters based on global data and send them to the edge computing gateway (102) to achieve dynamic upgrade of the prediction algorithm; S5. Based on the analysis results of the cloud management platform (103), reverse instructions are sent to the distributed water level sensor nodes (101) to dynamically adjust their sampling frequency and dynamic trigger threshold; the energy consumption and network load of each node are counted, the TDMA time slot allocation and compression algorithm parameters are optimized, and the transmission reliability and energy consumption are balanced; the encryption key between the edge computing gateway (102) and the cloud management platform (103) is regularly updated, and the integrity of the blockchain evidence module (103a) data is verified.

7. The data transmission method of the intelligent water level monitoring system according to claim 6, characterized in that: The compression ratio CR of the improved Huffman coding in step S2 is calculated as: ； in, The coded water level value The bit length is 16, which is the number of original ADC sampling bits.

8. The data transmission method of the intelligent water level monitoring system according to claim 6, characterized in that: The warning level in step S3 The grading rules are: ； The conditions for triggering blockchain evidence storage are: ≥1 or single Δh ≥0.3 m.

9. The data transmission method of the intelligent water level monitoring system according to claim 6, characterized in that: The step S4 includes the following steps: The cloud management platform (103) receives the encrypted compressed data packet from the edge computing gateway (102) and parses it to obtain the denoised water level value. , warning level , timestamp and geographic location identification; The blockchain evidence storage module (103a) generates a leaf node hash value for each water level data according to the following rules: ； The leaf node hash value is written into the blockchain distributed ledger, and the timestamp and geographic location metadata are stored in association to ensure the immutability and traceability of the water level data; Aggregate and analyze water level data from multiple regions to generate real-time water level heat maps and warning level distribution maps, which are dynamically displayed through the management platform interface; The parameters of the LSTM neural network are optimized based on historical and real-time data training, and the updated LSTM water level prediction model (102a) is sent to the edge computing gateway (102) to achieve dynamic upgrade of the edge prediction algorithm.

10. The data transmission method of the intelligent water level monitoring system according to claim 6, characterized in that: The step S5 includes the following steps: The cloud management platform (103) generates optimization instructions based on historical water level trends, warning records, and energy consumption statistics, and sends them to the distributed water level sensor nodes (101), dynamically adjusting their sampling frequency, dynamic trigger threshold, and sleep cycle to adapt to environmental changes; The edge computing gateway (102) collects statistics on the energy consumption and network load data of each sensor node, optimizes the TDMA time slot allocation strategy of the LoRa communication unit and the compression rate parameters of the improved Huffman coding, and balances the data transmission reliability and node energy consumption; The cloud management platform (103) regularly updates the layered encryption protocol keys between the edge computing gateway (102) and the distributed water level sensor nodes (101), and verifies the hash consistency of historical data through the blockchain evidence module (103a), and detects and repairs potential tampering or abnormal nodes; The edge computing gateway (102) periodically performs health status detection of the distributed water level sensor nodes (101), and uploads abnormal node information to the cloud management platform (103) in real time, triggering equipment maintenance alarms and dynamically adjusting the monitoring network topology.

Images