A method and system for corrosion sensing data transmission and communication optimization for marine environment power equipment

Abstract

This invention discloses a method and system for optimizing corrosion sensing data transmission and communication for power equipment in marine environments. It employs a four-layer heterogeneous architecture, with Chirp spread spectrum modulation-based communication between sensing nodes and the aggregation node. The spreading factor, bandwidth, and coding rate are adaptively adjusted by calculating the channel quality index. The aggregation node runs a hybrid time-slot allocation MAC protocol, combining frequency division multiplexing and time division multiplexing to support high-concurrency access for thousands of nodes. Multi-source data undergoes three-level compression: dead-zone compression, differential coding, and wavelet transform. A random forest model is deployed for edge-side corrosion risk prediction. A closed-loop power control and bidirectional time synchronization mechanism optimize energy efficiency and communication quality. This invention achieves long-distance, high-capacity, highly reliable, and low-power transmission for corrosion monitoring in harsh marine environments, with a data delivery success rate ≥95%, single-hop distance ≥2km, node battery life ≥2 years, and warning delay ≤10 seconds.

Description

Technical Field

[0001] This invention relates to the field of power equipment condition monitoring and Internet of Things communication technology, and in particular to a method and system for corrosion sensing data transmission and communication optimization for power equipment in marine environments. Background Technology

[0002] The marine environment is globally recognized as one of the most severely corrosive natural environments for materials. According to NACE International (National Association of Corrosion Engineers), global economic losses due to corrosion reach $2.5 trillion annually, with marine corrosion accounting for over 30%. For power systems, critical infrastructure in marine environments, such as transmission towers, wind turbine foundations, substation grounding grids, and cable trays, are exposed to high concentrations of chloride ions, high humidity, periodic salt spray, acidic gases (SO2, H2S), and biofouling, resulting in corrosion rates 5 to 10 times higher than inland areas. For example, the annual corrosion depth of carbon steel in the marine atmosphere can reach 0.2 to 0.5 mm, while the zinc coating of galvanized steel may be completely consumed within two years. The resulting structural strength reduction, increased contact resistance, and deterioration of grounding performance directly threaten the safe and stable operation of the power grid.

[0003] Currently, the main corrosion monitoring methods used in the power industry include:

[0004] The hanging plate weight loss method involves suspending a standard test piece at the monitoring point and periodically weighing it to calculate the corrosion rate. This method offers high accuracy but lacks real-time performance and cannot reflect the dynamic changes in corrosion.

[0005] Resistance probe method: This method estimates the amount of thinning caused by corrosion by measuring the change in resistance of a metal film during corrosion. It allows for continuous monitoring, but the probe has a limited lifespan, and data transmission requires a wired connection.

[0006] Electrochemical noise method: It has high sensitivity for monitoring potential / current fluctuations during corrosion, but it requires a high signal acquisition frequency (≥1Hz) and a large amount of data.

[0007] Quartz crystal microbalance (QCM): It can measure mass changes at the nanogram level with extremely high accuracy, but it is expensive and sensitive to ambient temperature.

[0008] The data generated by the aforementioned sensors ranges from a few bytes to hundreds of bytes, but they all face a common challenge: how to reliably, in real time, and at low cost transmit the collected corrosion data from widely distributed monitoring points to the data center.

[0009] Currently, the main wireless communication technologies used in the Industrial Internet of Things (IIoT) include:

[0010]

[0011] However, directly applying these technologies presents the following specific problems:

[0012] Although LoRa / Wi-SUN has a long range and low power consumption, its standard MAC protocol (such as LoRaWAN) uses pure ALOHA or simple CSMA. In high-concurrency (≥500 nodes / gateway) and high-frequency (≤1 minute) scenarios, the probability of collision increases sharply, resulting in an actual data delivery rate of less than 70%.

[0013] NB-IoT / 4G relies on operator base stations, and offshore wind farms are often in signal coverage blind spots, and the monthly rental cost is high (tens of yuan per node per month), making large-scale deployment uneconomical.

[0014] ZigBee / BLE has weak anti-interference capabilities. Frequency converters (switching frequency 2~8kHz) in offshore wind farms, electromagnetic pulses generated by thunderstorms, high-power radars, etc. can all cause serious co-channel interference, with measured packet loss rates reaching 30%~50%.

[0015] Existing systems lack environmental adaptability: fixed spreading factors, transmit power, and data rates cannot adapt to the drastic fluctuations in channel quality in marine environments (such as rain attenuation and multipath variations caused by tides).

[0016] Lack of edge intelligence: Uploading large amounts of raw data directly not only wastes bandwidth but also puts enormous pressure on the platform's processing, making it impossible to achieve second-level early warning.

[0017] In summary, existing technologies cannot simultaneously meet the six requirements of long-distance operation, high capacity, high frequency, anti-interference, low power consumption, and edge intelligence in marine environments. This invention aims to address these bottlenecks through systematic innovation at the physical layer, MAC layer, network layer, and application layer. Summary of the Invention

[0018] In view of this, the purpose of this invention is to provide a corrosion sensing data transmission and communication optimization method and system for marine power equipment, which realizes long-distance, large-capacity, highly reliable, and low-power transmission for corrosion monitoring in harsh marine environments. The data delivery success rate is ≥95%, the single-hop distance is ≥2km, the node battery life is ≥2 years, and the early warning delay is ≤10 seconds. It can be widely used in offshore wind farms, coastal substations, cross-sea transmission towers and other scenarios.

[0019] To achieve the above objectives, the present invention adopts the following technical solution: a method for optimizing corrosion sensing data transmission and communication for marine environmental power equipment, comprising the following steps:

[0020] S1: Construct a four-layer architecture including a sensing layer, a convergence layer, a gateway layer, and a platform layer;

[0021] S2: The sensing nodes of the sensing layer and the aggregation nodes of the aggregation layer communicate using Chirp spread spectrum modulation, and adaptively adjust the spreading factor, bandwidth and coding rate based on real-time measured received signal strength, signal-to-noise ratio and historical packet loss rate.

[0022] S3: The aggregation node adopts a hybrid time slot allocation MAC protocol, aligns the superframe period with the default reporting frequency of the sensor node, and supports at least 1,000 sensor nodes to access concurrently through a combination of frequency division multiplexing and time division multiplexing.

[0023] S4: The sensing node reports corrosion sensing data to the aggregation node at a default frequency of at least once per minute, and switches to event-driven reporting mode when abnormal corrosion parameters are detected.

[0024] S5: After receiving data reported by multiple sensor nodes, the aggregation node sequentially performs three-level data compression: node-level dead zone compression, differential coding, and wavelet transform, and calls a pre-deployed random forest model to predict edge-side erosion risk.

[0025] S6: The aggregation node uses a closed-loop transmit power control algorithm to adjust the transmit power of the sensor node, and uses a two-way time exchange mechanism to achieve time synchronization with the sensor node.

[0026] S7: The aggregation node uploads the compressed data and prediction results to the platform layer.

[0027] This invention also provides a corrosion sensing data transmission and communication optimization system for marine environmental power equipment, comprising the following steps:

[0028] Multiple sensing nodes, each integrating a microcontroller, a LoRa RF chip, and at least one corrosion or environmental sensor, are used to periodically acquire data and communicate with the aggregation node based on Chirp spread spectrum modulation.

[0029] At least one edge aggregation node, comprising an industrial-grade processor, a multi-channel LoRa gateway chip, and a synchronous positioning module, wherein the aggregation node is used for: executing adaptive modulation and coding based on the channel quality index; running a hybrid time slot allocation MAC protocol and sending beacon frames for time slot scheduling; performing three-level compression of received multi-source data, including dead-zone compression, differential coding, and wavelet transform; deploying and running a random forest model for edge-side erosion risk prediction; and managing the link through closed-loop power control and bidirectional time synchronization.

[0030] An optional gateway layer is used to aggregate and upload data from multiple aggregation nodes;

[0031] A platform layer is used for data storage, model training and updates, visualization, and early warning push notifications.

[0032] In a preferred embodiment, the adaptive adjustment of the spreading factor, bandwidth, and coding rate specifically involves calculating the Channel Quality Index (CQI), using the following formula: The minimum received signal strength Maximum received signal strength minimum signal-to-noise ratio maximum signal-to-noise ratio Weighting coefficient , , Based on the CQI range, the spreading factor SF = 7~12, bandwidth BW = 62.5~250kHz, and coding rate CR = 4 / 5~4 / 8 are selected from a table. PER is the packet error rate.

[0033] In a preferred embodiment, the superframe of the hybrid time-slot allocation MAC protocol includes beacon time slots, dedicated time slots, dynamic time slots, and contention-based access periods; node priority. Calculate and use the following formula for dedicated time slot allocation: in, The rate of change of corrosion rate, For corrosion rate, As a comprehensive indicator of environmental factors, As the comprehensive index of the largest environmental factors, Historical packet loss rate For the remaining energy, For nodes The initial total energy and the weighting coefficient of the rate of change of corrosion rate. Weighting coefficients of comprehensive environmental factor indicators Reliability weighting coefficient Weighting coefficient of the remaining energy proportion .

[0034] In a preferred embodiment, the three-level data compression includes: a first level performing dead-zone compression at the node end, where the difference is not reported when it is within a threshold; and a second level performing differential encoding at the aggregation node end, transmitting the difference between adjacent time points. The third stage performs Haar wavelet transform on the corrosion rate sequence, retaining the first 20% of the large coefficients, achieving a compression ratio of over 10:1.

[0035] The transformation formula is: .

[0036] Where aj[k] represents the approximation coefficients (low-frequency components) of the j-th level wavelet decomposition, reflecting the smoothing trend of the signal; dj[k] represents the approximation coefficients (low-frequency components) of the j-th level wavelet decomposition. The detail coefficients (high-frequency components) of the layer wavelet decomposition reflect the local changes in the signal; cj−1[⋅] is the coefficient sequence of the next layer (finer scale); when hour, The original discrete signal sequence; k is the coefficient index. ,in The decomposition layer number is 2k, and 2k+1 are adjacent pairing indices, reflecting the summation and difference of two adjacent samples by the Haar wavelet.

[0037] In a preferred embodiment, the random forest model contains 30 decision trees with a maximum depth of 8. The input features include 12 dimensions: temperature, humidity, SO2, H2S, salt spray, corrosion potential, electrochemical noise, cumulative loss, corrosion rate slope, wind speed, light intensity, and rainfall. The output is categorized into four types: normal, attention, warning, and alarm. The warning accuracy rate is no less than 85%.

[0038] In a preferred embodiment, the closed-loop transmit power control employs a proportional-derivative algorithm: Where Ptx(k+1) is the transmit power value after the next adjustment, and Ptx(k) is the current (k+1) transmit power value. The transmit power value at the time of adjustment (k-1); Ptx(k-1) is the transmit power value at the time of the previous adjustment (k-1). (Next) The transmit power value before adjustment; α is the proportional coefficient, which controls the response speed to the deviation of the received signal strength, and its value is 0.5; β is the differential coefficient, which is used to suppress the oscillation of the transmit power (damping of the rate of power change), and its value is 0.2; The target received signal strength is the desired signal strength that the sink node expects to achieve, with a value of 90 dBm; RSSI measured This represents the actual received signal strength measured at the aggregation node; the transmit power adjustment range is 5~20dBm, with a step size of 1dB.

[0039] In a preferred embodiment, the bidirectional time exchange mechanism is as follows: the sink node transmits its local time T1 in the beacon, the nodes record the reception time T2 and transmit T1 and T2 back in the uplink frame, the sink node records the reception time T3, and the round-trip delay is calculated. and clock offset The nodes correct their clocks accordingly, with a synchronization error of less than 1ms.

[0040] In a preferred embodiment, the sensing node employs an adaptive duty cycle energy-saving strategy: when the absolute value of the corrosion rate change rate is less than 0.01 μm / year / hour for one consecutive hour and all environmental factors are within the safety threshold, the reporting frequency is reduced to once every 5 minutes; otherwise, it is restored to once every 1 minute. The node and the aggregation node negotiate the session key using elliptic curve Diffie-Hellman, encrypt the data payload using SM4 or AES-128, and switch between four orthogonal channels using a pseudo-random frequency hopping mechanism to combat interference.

[0041] In a preferred embodiment, when the aggregation node outputs a warning or alarm level in the random forest model, it triggers an event-driven report within 10 seconds. The report includes raw data fragments from each second within the 5 minutes prior to the warning trigger. The system is applied to online monitoring of metal structure corrosion in offshore wind farms, coastal substations, or cross-sea transmission towers.

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] 1. Longer communication distance and wider coverage.

[0044] This invention, employing Chirp spread spectrum modulation and adaptive spread spectrum factor adjustment technology, achieves a single-hop communication distance of over 2.5 kilometers under line-of-sight conditions, with a maximum measured distance of 3.1 kilometers. In contrast, existing LoRaWAN standard solutions typically achieve a communication distance of only 1 to 2 kilometers under the same transmit power and environment. This invention effectively solves the signal coverage problem for large-scale distributed monitoring points (such as wind turbine clusters and transoceanic towers) in marine environments.

[0045] 2. Large node access capacity, supporting large-scale deployment.

[0046] This invention utilizes a superframe structure combining frequency division multiplexing and time division multiplexing, along with a dynamic priority time slot allocation algorithm, enabling a single aggregation node to stably support at least 1000 sensor nodes concurrently reporting data at a high frequency of 1 report per minute. In contrast, existing LoRaWAN standard protocols typically support only about 300 nodes at the same reporting frequency, with the collision rate increasing sharply beyond that. This invention increases node capacity by more than three times, meeting the large-scale monitoring needs of scenarios such as offshore wind farms and large substations.

[0047] 3. High data transmission reliability and strong anti-interference capability.

[0048] This invention employs adaptive modulation and coding, forward error correction (convolutional code + interleaving), and frequency hopping mechanisms. Even in environments with strong electromagnetic interference (such as inverter noise and lightning electromagnetic pulses) at sea, where the interference signal strength is 10 dB higher than the useful signal, the bit error rate can still be maintained below 10^-5. In a scenario with 1000 concurrent nodes, the data packet delivery success rate reaches 96% to 98%, while the standard LoRaWAN solution only achieves a delivery rate of 65% to 75% under the same conditions. This invention reduces the packet loss rate by approximately 30 percentage points, significantly improving data integrity and system availability.

[0049] 4. High real-time early warning capability and strong edge intelligence capabilities.

[0050] This invention deploys a random forest edge inference model at the aggregation node, enabling real-time assessment of corrosion risk levels (normal, caution, warning, alarm) based on 12-dimensional environmental and corrosion characteristics. When a warning or alarm is detected, the system triggers an event-driven report within 10 seconds, achieving second-level warnings. In contrast, most existing solutions upload all raw data to the cloud for analysis, resulting in warning delays typically in the minutes or even longer. This invention's edge intelligent architecture not only significantly shortens warning time but also reduces cloud computing pressure and network bandwidth usage.

[0051] 5. High data compression ratio, saving storage and bandwidth resources.

[0052] This invention employs a three-stage compression method: node-level dead-zone compression, convergence node-level differential coding, and optional wavelet transform. This achieves a compression ratio of 8:1 to 15:1 for slowly varying signals such as corrosion rates. Existing technologies typically do not perform data compression or only use simple dead-zone compression, with compression ratios not exceeding 3:1. This invention significantly reduces uplink data volume, extends the effective utilization of wireless channel resources, and also reduces platform-side storage costs.

[0053] 6. Low node power consumption and long battery life.

[0054] This invention achieves an average power consumption of approximately 43.6 mJ / cycle for a node at a 1-minute reporting frequency through adaptive duty cycle (normally 1 minute / cycle, automatically decreasing to 5 minutes / cycle during idle periods), dynamic power control (adaptive adjustment of transmit power from 5 to 20 dBm), and a refined time-slot wake-up mechanism. Using two 18650 batteries (total energy approximately 45 Wh), the theoretical battery life is up to 7 years, with a conservative estimate of at least 2 years in actual use. In contrast, existing LoRaWAN nodes typically only last 1 to 1.5 years at the same reporting frequency. This invention extends battery life by at least 33%.

[0055] 7. High communication resource utilization, supporting high-concurrency reporting.

[0056] This invention designs a hybrid time-slot allocation MAC protocol that ensures the real-time performance of high-priority nodes (such as eroded or abnormal nodes) through dedicated time slots, while accommodating low-priority nodes and bursty data through dynamic time slots and contention periods. Furthermore, it employs multi-channel frequency division multiplexing (4 channels, a total of 640 time slots per superframe), resulting in channel resource utilization far exceeding that of the pure ALOHA method in standard LoRaWAN. Real-world testing shows that under pressure from 1000 nodes and a reporting frequency of 1 minute, this invention still maintains a delivery rate of over 96%, while the standard solution fails to function properly due to severe conflicts.

[0057] 8. High safety, meeting the requirements of the power industry.

[0058] This invention employs pre-shared key two-way authentication when nodes join the network, uses SM4 / AES-128 encryption (CTR mode) for data transmission, HMAC-SHA256 for integrity protection, distributes session keys through ECDH negotiation, and uses dynamic anonymous identifiers for node IDs. These measures effectively prevent data eavesdropping, tampering, replay attacks, and node spoofing, meeting the high standards of information security required by the power industry. Most existing corrosion monitoring systems only use simple fixed keys or plaintext transmission, which is clearly insufficient in terms of security.

[0059] 9. Highly expandable and adaptable to various sensor types.

[0060] The data frames defined in this invention adopt a Type-Length-Value (TLV) encoding format, which can be flexibly extended to new sensor types (currently supporting at least 15 environmental and corrosion-related factors) and can adapt to changes in future monitoring needs without modifying the underlying communication protocol. In contrast, existing solutions often fix the frame structure for specific sensors, resulting in poor scalability.

[0061] 10. Significant economic and social benefits

[0062] The distributed corrosion data resource integration system for power systems constructed in this invention can create an overall demonstration effect, improve the intelligence and digitalization level of corrosion safety for marine materials, and promote the upgrading of corrosion protection technologies. Applying the results of this invention can enhance the safety and reliability of marine power equipment, reduce unplanned outages and maintenance costs caused by corrosion, and provide technical support for the construction, system design, and intelligent enhancement of new power system architectures. The project results have been demonstrated and verified in scenarios such as offshore wind farms and coastal substations, and have the potential for widespread application to power companies in various provinces, demonstrating clear prospects for technology transfer and market value. Attached Figure Description

[0063] Figure 1 This is a four-layer architecture diagram of the system in a preferred embodiment of the present invention. It illustrates the connection relationships and data flows between the sensor nodes, aggregation nodes, gateways, and cloud platform.

[0064] Figure 2 This is a flowchart of the adaptive modulation and coding process in a preferred embodiment of the present invention. It includes steps such as RSSI / SNR measurement, CQI calculation, SF / BW lookup table, power adjustment, and PER feedback.

[0065] Figure 3 This is a superframe structure diagram in a preferred embodiment of the present invention. It illustrates the time allocation and interrelationships of beacon time slots, dedicated time slots, dynamic time slots, and contention time slots.

[0066] Figure 4 This is a diagram of the data frame format in a preferred embodiment of the present invention. The length and meaning of each field are clearly labeled.

[0067] Figure 5 This is a flowchart of the random forest edge inference process in a preferred embodiment of the present invention. It includes feature extraction, model loading, voting decision-making, and early warning triggering.

[0068] Figure 6 This is a timing diagram of node network access and key negotiation in a preferred embodiment of the present invention. It illustrates the interaction processes between nodes and the aggregation node, including authentication, key exchange, and time slot application.

[0069] Figure 7 This is a power control convergence curve diagram in a preferred embodiment of the present invention. It simulates the change in transmit power as a node approaches the sinking node from a distance.

[0070] Figure 8 This is a comparison chart of data delivery success rates under different node densities in a preferred embodiment of the present invention (the present invention vs. LoRaWAN). Detailed Implementation

[0071] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0072] It should be noted that the following detailed descriptions are illustrative and intended to provide further explanation of this application. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0073] It should be noted that the terminology used herein is for the purpose of describing particular implementations only and is not intended to limit the exemplary implementations according to this application; as used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise; furthermore, it should be understood that when the terms “comprising” and / or “including” are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or combinations thereof.

[0074] refer to Figure 1-8 This invention adopts a four-layer heterogeneous architecture, with each layer communicating with the others through standardized interfaces, as described below:

[0075] Sensing Layer: Composed of multiple sensor nodes, each node integrates a microcontroller (MCU), a LoRa RF chip, and various sensors (via I²C / SPI / ADC interfaces). Nodes are responsible for periodically acquiring data, timestamping, initial filtering (median filtering to remove spike noise), and data buffering (storing up to the most recent 100 records). Nodes run a lightweight real-time operating system or a bare-metal event-driven system.

[0076] Aggregation Layer: Edge aggregation nodes employ more powerful processors (such as industrial-grade ARM Cortex-A7), run Linux, and are equipped with LoRa gateway chips (supporting 8-channel concurrent reception), GPS / BeiDou modules, 4G / 5G modules, and Ethernet interfaces. Their main functions include: receiving data frames from all nodes, performing time slot scheduling and resource allocation, running random forest models for edge inference, data compression and encryption, and communicating with the platform layer.

[0077] Gateway Layer: Optional component. When the number of aggregation nodes is large (e.g., a wind farm has 30 aggregation nodes), a primary gateway can be set up to aggregate data from multiple aggregation nodes and upload it to the platform via a fiber optic ring network or 5G. In a single aggregation node scenario, the gateway layer can be merged with the aggregation node.

[0078] Platform Layer: Deployed on the enterprise intranet or public cloud, adopting a microservice architecture. It includes data receiving services (MQTT / HTTP), time-series databases, big data analytics engines, visualization dashboards, and alert push services (SMS / APP). The platform layer also handles tasks such as model training (updating the random forest model and distributing it to the aggregation node), user management, and device management.

[0079] Modulation method and parameter selection of the sensing layer:

[0080] This invention employs Chirp Spread Spectrum (CSS) modulation, a linear frequency modulation spread spectrum technique. The main characteristics of CSS include:

[0081] The demodulation sensitivity is proportional to the spreading factor SF. When SF=12 and bandwidth BW=62.5kHz, the sensitivity can reach -148dBm.

[0082] It is not sensitive to Doppler frequency shift, making it suitable for slight node movement caused by wind turbine blade rotation and ocean waves.

[0083] It has strong anti-narrowband interference capability and a spreading gain of up to 36dB.

[0084] The default parameter configurations for this invention are shown in the table below:

[0085]

[0086] Adaptive Modulation and Coding (AMC) Algorithm

[0087] Objective: To maximize data rate and energy efficiency while ensuring communication reliability.

[0088] Input variables:

[0089] Received signal strength measured at the aggregation node (Unit: dBm)

[0090] Signal-to-noise ratio (Unit: dB)

[0091] Packet loss rate (Statistics within a sliding window)

[0092] Battery voltage (Used for power back-off)

[0093] Decision-making method: First, define the channel quality index. The calculation formula is as follows:

[0094] in, , ; , Weighting coefficient , , .

[0095] The optimal combination of (SF, BW, CR) is selected based on the CQI value, and the mapping relationship is as follows:

[0096] When CQI is between 0.8 and 1.0, with SF=7, BW=250kHz, and CR=4 / 5, the effective data rate is approximately 5470bps.

[0097] When CQI is between 0.6 and 0.8, with SF=8, BW=125kHz, and CR=4 / 6 selected, the effective data rate is approximately 2500bps.

[0098] With CQI between 0.4 and 0.6, selecting SF=9, BW=125kHz, and CR=4 / 7 results in an effective data rate of approximately 1760bps.

[0099] With CQI between 0.2 and 0.4, selecting SF=10, BW=62.5kHz, and CR=4 / 8, the effective data rate is approximately 440bps.

[0100] When CQI is between 0.0 and 0.2, with SF=11, BW=62.5kHz, and CR=4 / 8, the effective data rate is approximately 220bps.

[0101] When CQI is below 0, select SF=12, BW=62.5kHz, CR=4 / 8, and the effective data rate is approximately 110bps.

[0102] Closed-loop adjustment process:

[0103] The aggregation node periodically (e.g., every 10 minutes) broadcasts beacon frames containing the current CQI threshold table.

[0104] After receiving the beacon, each node measures its local RSSI and SNR, calculates its own CQI, and looks up the corresponding SF / BW / CR.

[0105] The node sends data under the new parameters and indicates the parameters such as SF used in the frame header (to facilitate demodulation by the aggregation node).

[0106] The aggregation node calculates the PER of each node. If the PER exceeds 10% for three consecutive times, the node will be forced to reduce its data rate by one level (increase SF) in the next beacon.

[0107] Adaptive power control: The nodes use a closed-loop power control algorithm based on the RSSI fed back from the sink node.

[0108]

[0109] in, (Ensure sufficient margin) , The power adjustment step size is limited to ±1 dB / cycle, ranging from 5 to 20 dBm. The algorithm incorporates a proportional-derivative term to reduce overshoot and oscillation.

[0110] Convolutional codes (constrained length 7, code rate 1 / 2) are used in conjunction with interleaving. The generator polynomial for the convolutional code is (133, 171) (octal). The encoded data is processed by a block interleaver, with a block size of 4×4 (i.e., 16 bits) to resist sudden errors at sea (such as lightning pulses). After interleaving, the data is transmitted in frames.

[0111] Detailed Design of MAC Layer Protocol

[0112] 1) Superframe structure

[0113] Set superframe period Seconds, aligned with the default reporting frequency of 1 time per minute. A superframe consists of the following time slots:

[0114] Beacon slot: 10ms in length, used to send synchronization information, slot allocation table, and system configuration.

[0115] Dedicated time slots: each time slot is 500ms, and the number is fixed at 120 (30 per sub-channel, for a total of 4 sub-channels).

[0116] Dynamic time slots: each time slot is 500ms, and the number is fixed at 40 (10 per sub-channel).

[0117] Competition access period: 10ms in duration, used for node network access or emergency data.

[0118] The total superframe period is 10ms + (120+40)×500ms + 10ms = 60.02s, which is slightly longer than 60 seconds. It can be precisely aligned by fine-tuning the beacon interval.

[0119] 2) Multi-channel frequency division multiplexing

[0120] To improve capacity, the 470-510MHz frequency band is divided into four orthogonal channels with center frequencies of 470.3MHz, 475.5MHz, 480.7MHz, and 485.9MHz. Each channel has a bandwidth of 125kHz and a guard interval of 200kHz. The gateway chip at the aggregation node can simultaneously monitor eight channels (this solution uses only four), and each channel operates the aforementioned superframe structure independently. Therefore, the total number of time slots = (120+40)×4 = 640 per superframe. Theoretically, this can support 640 nodes reporting at a rate of 1 minute per report. If the number of nodes exceeds 640, the superframe period can be increased to 120 seconds (reducing the reporting frequency to 2 minutes per report) or the number of channels can be increased (if permitted by regulations).

[0121] 3) Priority calculation and dynamic time slot allocation

[0122] Priority of each node Calculate using the following formula:

[0123]

[0124] Meaning of each variable:

[0125] Corrosion rate change rate (slope of linear regression of the most recent 3 sampling points), in μm / year / minute.

[0126] Comprehensive index of environmental factors Normalized to [0,1].

[0127] Historical packet loss rate (sliding window of 10 packets).

[0128] Node remaining energy (estimated based on voltage).

[0129] Weighting coefficient , , , .

[0130] Dynamic time slot allocation process:

[0131] At the start of each superframe, the aggregation node collects the latest priorities of all nodes (nodes carry their remaining energy and packet loss rate statistics in the uplink data frames).

[0132] Sort all nodes in descending order of priority.

[0133] Take before Nodes ( Dedicated time slots are allocated to them. During allocation, nodes are mapped to time slots in the dedicated time slot list in the order of node sorting (the time slot order is fixed, but the nodes can be rotated).

[0134] The remaining nodes enter the dynamic time slot pool. Dynamic time slots are allocated on demand: nodes receive a dynamic time slot request window broadcast by the aggregation node in the beacon time slot and send a request during the contention access period; the aggregation node broadcasts the allocation result in the beacon of the next superframe.

[0135] If a node fails to obtain any time slots for three consecutive superframes, its priority is automatically raised to the highest level. (to ensure they don't starve to death.)

[0136] 4) Conflict detection and retransmission mechanism

[0137] During the contention-based access period (10ms), time-slot CSMA / CA is used: the 10ms is divided into 10 micro-slots (1ms / slot). Before transmitting, a node randomly selects a micro-slot and checks the channel energy. If the channel is idle, it transmits immediately; if busy, it waits for the next micro-slot and decrements the backoff counter by 1.

[0138] Maximum retransmission count: 3 times. If the packet fails after 3 attempts, the node marks it as "unacknowledged," stores it in the cache, and attempts to retransmit it during the contention period of the next superframe. Once the cache is full (100 packets), the oldest data is discarded.

[0139] ACK Confirmation: After receiving data, the aggregation node broadcasts an ACK bitmap (1 bit per node) in the next beacon slot. Nodes check the ACK bitmap; if no ACK is received, a retransmission is triggered.

[0140] 5) Precise Time Synchronization Algorithm

[0141] Employs a two-way time exchange mechanism (similar to a simplified version of NTP):

[0142] The aggregation node sends a synchronization frame, which includes the local time, in the beacon time slot. .

[0143] The node receives the synchronization frame and records the local reception time. .

[0144] The node carries in the next uplink data frame and .

[0145] After receiving the data, the aggregation node records the reception time. And calculate the round-trip delay:

[0146]

[0147] in The sending time of the aggregation node recorded on the node side (actual and...) (Same). Due to the symmetry between uplink and downlink, the one-way delay is... Node clock offset:

[0148]

[0149] The nodes adjust their local real-time clocks accordingly. The synchronization error can be controlled within ±1ms, meeting the minute-level data alignment requirements.

[0150] Data compression and edge computing

[0151] Two-level compression architecture

[0152] Node-level compression: After data acquisition, nodes first perform dead-zone compression. A dead-zone threshold is defined for each sensor (e.g., temperature ±0.2℃, humidity ±1%, corrosion rate ±0.01μm / year). If the difference between the current value and the previously reported value is within the dead zone, the sensor value is not reported; instead, a "no change" flag is set. Nodes also perform lightweight run-length encoding: when consecutive values ​​are identical, only one value and the number of repetitions are recorded.

[0153] Aggregator-level compression: After collecting data from multiple nodes, the aggregation node performs differential coding and wavelet compression. Mathematical expression of differential coding: Let the original time series... ,but , for Because the corrosion process is gradual, They are usually very small and can be represented with fewer bits (e.g., using exponential Golomb coding).

[0154] Wavelet compression: Applying Haar wavelet transform to the erosion rate sequence. Let the sequence length be... The transformation formula is:

[0155]

[0156] The first 20% of the large coefficients (approximate coefficients + some detail coefficients) are retained, and the rest are set to zero. The signal is reconstructed after inverse transformation, and the distortion is controllable (root mean square error <0.01μm / year). The compression ratio can reach over 10:1.

[0157] Random forest edge inference model:

[0158] Parameters of the random forest model deployed on the aggregation node:

[0159] Number of trees:

[0160] Maximum depth:

[0161] The number of features considered during splitting: take the square root of the total number of features and round down. Therefore, consider Features

[0162] Node splitting criterion: Gini impurity

[0163] Input feature vector ,include:

[0164] Temperature (°C)

[0165] Relative humidity (%)

[0166] SO2 concentration (ppb)

[0167] H2S concentration (ppb)

[0168] Salt spray deposition rate (mg / (m²·d))

[0169] Corrosion potential (mV vs. SCE)

[0170] Root mean square (μA) of electrochemical noise current

[0171] Cumulative metal loss (μm)

[0172] The corrosion rate slope over the past hour (μm / year / hour)

[0173] Wind speed (m / s)

[0174] Light intensity (lux)

[0175] Rainfall (mm / h)

[0176] The output consists of four categories: Normal (0), Attention (1), Warning (2), and Alarm (3). Decision rule: Each tree outputs one category, and the majority vote determines the final category. If the vote is tied, the higher risk level is chosen.

[0177] Model training: The model is trained on the platform using historical data (at least 3 months) and updated quarterly. New model files (JSON format) are distributed to the aggregation node via 4G.

[0178] Warning trigger conditions: When the inference results are "warning" or "alarm" twice in a row, the aggregation node immediately generates a warning message, which includes the node ID, timestamp, original feature value, and inference result, and sends it to the platform through a high-priority queue (using a contention access period or a dedicated time slot).

[0179] Security mechanisms:

[0180] Device authentication: When a node joins the network, it uses a pre-shared key to perform two-way authentication with the aggregation node. A challenge-response mechanism is adopted: the aggregation node sends a random number R, the node calculates a response using HMAC-SHA256, and the aggregation node verifies the response.

[0181] Data encryption: Employs SM4 (Chinese national cryptographic algorithm) or AES-128, operating in CTR mode (counter mode), requiring no padding, facilitating the encryption of variable-length data. Session keys are generated by the aggregation node and distributed through ECDH (Elliptic Curve Diffie-Hellman) negotiation after authentication.

[0182] Integrity protection: Each data frame is appended with a CRC-32C checksum, verified by the aggregation node, and erroneous frames are discarded. Additionally, critical frames (such as configuration commands) are protected using a message authentication code (HMAC-SHA256).

[0183] Privacy protection: The node ID uses a dynamic anonymous identifier that changes with each session to prevent location tracking.

[0184] Energy-saving strategies (including power consumption estimation):

[0185] Node power consumption model:

[0186] Sleep mode: Current Voltage 3.3V, Power

[0187] Sensor data acquisition (20ms): Current ,power

[0188] MCU processing (10ms): Current ,power

[0189] RF transmission (100ms, SF9, 14dBm): Current ,power

[0190] RF reception (50ms, listening beacon): Current ,power

[0191] Typical work cycle (reporting once every 1 minute):

[0192] Sleep duration: 59.82 seconds; Energy consumption: 0.598 mJ

[0193] Energy consumption for data collection: 0.66 mJ

[0194] Energy consumption for processing: 0.264 mJ

[0195] Transmission energy consumption: 39.6 mJ

[0196] The energy consumption of the receiving beacon is 2.475 mJ.

[0197] The total energy consumption per cycle is approximately 43.6 mJ.

[0198] Using two 18650 batteries (capacity 3400mAh, voltage 3.7V, total energy approximately 45Wh = 162000J), the theoretical battery life is:

[0199]

[0200] Considering factors such as battery self-discharge, low-temperature capacity decay, and fluctuations in actual transmission power, the estimated range is ≥2 years, which meets engineering requirements.

[0201] Adaptive duty cycle: When a node detects that the absolute value of the corrosion rate change rate is less than 0.01 μm / year / hour for 1 consecutive hour, and all environmental factors are stable within a safe range, the reporting frequency is automatically reduced to once every 5 minutes; when the change rate exceeds the threshold, it is restored to once every 1 minute.

[0202] Specific Implementation Example 1: Large-scale offshore wind farm (30 wind turbines, 600 nodes)

[0203] Deployment Details: This offshore wind farm in Fujian province is located 15km from the shore, with turbine spacing ranging from 1.8 to 2.5km. Each turbine is equipped with 20 corrosion sensing nodes, located on the outer wall of the tower (6 at different heights), the foundation flange (4), bolt connections (4), the grounding down conductor (3), and the bottom of the nacelle (3). The nodes are fitted with IP67 protective shells and secured magnetically or adhesively.

[0204] Aggregation Node: One edge aggregation node is deployed inside each wind turbine tower, using the SX1302 gateway chip and an external 6dBi fiberglass antenna (placed on top of the nacelle). The aggregation node is connected to the offshore booster station via an optical fiber ring network (OPGW cable), and then transmitted to the onshore control center via microwave relay.

[0205] Parameter configuration: Default SF=9, BW=125kHz, transmit power 14dBm. Since the wind turbine spacing is approximately 2km, and the measured RSSI is between -105dBm and -90dBm, SF=9 or SF=10 is selected. Adaptive modulation is enabled, adjusting every 10 minutes. Dynamic time slot allocation is enabled, with priority weights set according to contract requirements.

[0206] Operational Results: After 6 months of continuous operation, data from 600 nodes were collected: the average delivery success rate was 98.7%, the maximum communication distance was 3.1km (with a node partially obstructed by an adjacent wind turbine tower), and 3 abnormal corrosion rate warnings were successfully issued (2 of which were actual bolt corrosion, and 1 was a false alarm due to sensor malfunction, with an accuracy rate of 66.7%, which was improved to 85% after subsequent model optimization). Battery voltage monitoring showed that the lowest voltage was still 3.2V (initially 3.7V), and the estimated battery life is over 2.5 years.

[0207] Specific Implementation Example 2: Large Coastal Substation (450 nodes)

[0208] Deployment Details: A 220kV substation in Zhejiang Province, covering approximately 200 acres, has strong electromagnetic interference sources (circuit breaker operation, main transformer vibration). 450 nodes are deployed, covering the transformer casing (50 nodes), disconnector support brackets (120 nodes), busbar fittings (80 nodes), and grounding grid outgoing points (200 nodes). Three convergence nodes are set up within the substation, located on the roof of the main control building, the roof of the power distribution equipment building, and near the capacitor bank, respectively.

[0209] Challenges and Solutions: Strong transient electromagnetic interference exists within the station (pulse field strengths up to 10kV / m are generated during disconnection switch operation), leading to frequent packet loss in ordinary wireless communication. This invention employs the following measures:

[0210] Enabling convolutional coding with interleaving (depth 8) enhances error correction capabilities.

[0211] Frequency hopping is used between nodes and the sink node: each superframe switches between four channels in a pseudo-random sequence (the seed is broadcast by the sink node), so that even if one channel is interfered with, the other channels can still work.

[0212] Increase the number of retransmissions to 3 and enable ARQ (Automatic Repeat Request).

[0213] Results: In three disconnector operation tests, the packet loss rate was as high as 35% before optimization, but decreased to 6.2% after adopting this invention. The packet loss rate during daily operation is less than 3%.

[0214] Specific Implementation Example 3: Cross-sea transmission tower (80 nodes per tower)

[0215] Deployment Details: A 500kV cross-sea transmission tower, 120m high, is located in a tidal flat area. Corrosion sensors are deployed at 80 nodes on the tower legs (underwater, tidal, and atmospheric zones), crossarms, bolt connections, and ladders. The convergence node is installed at a height of 40m on the tower body, using a directional antenna (120° sector) to provide downward coverage.

[0216] Special characteristics: The steel structure of the tower reflects and blocks signals, and the large current generates a strong magnetic field (50Hz power frequency and its harmonics). Solution:

[0217] The node antenna uses a flexible PCB antenna, which is closely attached to the surface of the steel structure, using the structure as a reflector to enhance the signal.

[0218] The aggregation node uses dual-antenna diversity reception (two vertically polarized antennas) and selects the channel with a high signal-to-noise ratio.

[0219] Data compression uses differential + dead zone, with a compression ratio of 12:1, reducing the number of transmissions.

[0220] Results: The furthest communication distance was between the node on the tower leg (0m height) and the convergence node (40m), with a path of approximately 50m, but requiring penetration through multiple layers of steel structure. The actual equivalent distance was approximately 300m. SF=12 was used to ensure reliability. The data delivery rate was 99.5%. The corrosion rate in the tidal zone of the tower leg was successfully monitored to be 5 times higher than that in the atmospheric zone, providing guidance for anti-corrosion coating maintenance.

Claims

1. A method for corrosion sensing data transmission and communication optimization for marine environment power equipment, characterized in that, Includes the following steps: S1: Construct a four-layer architecture including a sensing layer, a convergence layer, a gateway layer, and a platform layer; S2: The sensing nodes of the sensing layer and the aggregation nodes of the aggregation layer communicate using Chirp spread spectrum modulation, and adaptively adjust the spreading factor, bandwidth and coding rate based on real-time measured received signal strength, signal-to-noise ratio and historical packet loss rate. S3: The aggregation node adopts a hybrid time slot allocation MAC protocol, aligns the superframe period with the default reporting frequency of the sensor node, and supports at least 1,000 sensor nodes to access concurrently through a combination of frequency division multiplexing and time division multiplexing. S4: The sensing node reports corrosion sensing data to the aggregation node at a default frequency of at least once per minute, and switches to event-driven reporting mode when abnormal corrosion parameters are detected. S5: After receiving data reported by multiple sensor nodes, the aggregation node sequentially performs three-level data compression: node-level dead zone compression, differential coding, and wavelet transform, and calls a pre-deployed random forest model to predict edge-side erosion risk. S6: The aggregation node uses a closed-loop transmit power control algorithm to adjust the transmit power of the sensor node, and uses a two-way time exchange mechanism to achieve time synchronization with the sensor node. S7: The aggregation node uploads the compressed data and prediction results to the platform layer.

2. A corrosion sensing data transmission and communication optimization system for marine environmental power equipment, characterized in that, The method for optimizing corrosion sensing data transmission and communication for marine environmental power equipment as described in claim 1 includes: Multiple sensing nodes, each integrating a microcontroller, a LoRa RF chip, and at least one corrosion or environmental sensor, are used to periodically acquire data and communicate with the aggregation node based on Chirp spread spectrum modulation. At least one edge aggregation node, comprising an industrial-grade processor, a multi-channel LoRa gateway chip, and a synchronous positioning module, wherein the aggregation node is used for: executing adaptive modulation and coding based on the channel quality index; running a hybrid time slot allocation MAC protocol and sending beacon frames for time slot scheduling; performing three-level compression of received multi-source data, including dead-zone compression, differential coding, and wavelet transform; deploying and running a random forest model for edge-side erosion risk prediction; and managing the link through closed-loop power control and bidirectional time synchronization. An optional gateway layer is used to aggregate and upload data from multiple aggregation nodes; A platform layer is used for data storage, model training and updates, visualization, and early warning push notifications.

3. The corrosion sensing data transmission and communication optimization system for marine environmental power equipment according to claim 2, characterized in that, The adaptive adjustment of the spreading factor, bandwidth, and coding rate specifically involves calculating the Channel Quality Index (CQI), with the following formula: in, The minimum received signal strength, The maximum received signal strength. This represents the minimum signal-to-noise ratio. This represents the maximum signal-to-noise ratio. Weighting coefficients.

4. The corrosion sensing data transmission and communication optimization system for marine environmental power equipment according to claim 2, characterized in that, The superframe of the hybrid time-slot allocation MAC protocol includes beacon time slots, dedicated time slots, dynamic time slots, and contention-based access periods; node priority. Calculate and use the following formula for dedicated time slot allocation: in, The rate of change of corrosion rate, For corrosion rate, As a comprehensive indicator of environmental factors, As the comprehensive index of the largest environmental factors, Historical packet loss rate For the remaining energy, For nodes The initial total energy and the weighting coefficient of the rate of change of corrosion rate. Weighting coefficients of comprehensive environmental factor indicators Reliability weighting coefficient Weighting coefficient of the remaining energy proportion .

5. A corrosion sensing data transmission and communication optimization system for marine environmental power equipment according to claim 2, characterized in that, The three-level data compression includes: the first level performs dead-zone compression at the node end, and does not report when the difference is within a threshold; the second level performs differential encoding at the aggregation node end, transmitting the difference between adjacent time points. The third stage performs Haar wavelet transform on the corrosion rate sequence, retaining the first 20% of the large coefficients, achieving a compression ratio of over 10:

1. The transformation formula is: ; Where aj[k] are the approximation coefficients of the j-th level wavelet decomposition, reflecting the smoothing trend of the signal; dj[k] are the approximation coefficients of the j-th level wavelet decomposition. The detail coefficients of the layer wavelet decomposition reflect the local changes in the signal; cj−1[⋅] is the coefficient sequence of the previous layer; when hour, The original discrete signal sequence; k is the coefficient index. ,in The decomposition layer number is 2k, and 2k+1 are adjacent pairing indices, reflecting the summation and difference of two adjacent samples by the Haar wavelet.

6. A corrosion sensing data transmission and communication optimization system for marine environmental power equipment according to claim 2, characterized in that, The random forest model contains 30 decision trees with a maximum depth of 8. The input features include 12 dimensions: temperature, humidity, SO2, H2S, salt spray, corrosion potential, electrochemical noise, cumulative loss, corrosion rate slope, wind speed, light intensity, and rainfall. The output is categorized into four types: normal, attention, warning, and alarm. The warning accuracy rate is no less than 85%.

7. A corrosion sensing data transmission and communication optimization system for marine environmental power equipment according to claim 2, characterized in that, The closed-loop transmit power control uses a proportional-derivative algorithm: Where Ptx(k+1) is the transmit power value after the next adjustment, and Ptx(k) is the power value of the k-th generation. The transmit power value at the time of the adjustment; Ptx(k−1) is the power value at the time of the adjustment. The transmit power value before the adjustment; α is the proportional coefficient, which controls the response speed to the deviation of the received signal strength, and its value is 0.5; β is the differential coefficient, which is used to suppress the oscillation of the transmit power, and its value is 0.2; The target received signal strength is the desired signal strength that the sink node expects to achieve, with a value of 90 dBm; RSSI measured This represents the actual received signal strength measured at the aggregation node; the transmit power adjustment range is 5~20dBm, with a step size of 1dB.

8. A corrosion sensing data transmission and communication optimization system for marine environmental power equipment according to claim 2, characterized in that, The bidirectional time exchange mechanism is as follows: the sink node sends its local time T1 in the beacon, the node records the received time T2 and sends back T1 and T2 in the uplink frame, the sink node records the received time T3, and calculates the round-trip delay. and clock offset The nodes correct their clocks accordingly, with a synchronization error of less than 1ms.

9. A corrosion sensing data transmission and communication optimization system for marine environmental power equipment according to claim 2, characterized in that, The sensing node adopts an adaptive duty cycle energy-saving strategy: when the absolute value of the corrosion rate change rate is less than 0.01 μm / year / hour for 1 consecutive hour and all environmental factors are within the safety threshold, the reporting frequency is reduced to 5 minutes / time; otherwise, it is restored to 1 minute / time. The node and the aggregation node use elliptic curve Diffie-Hellman negotiation session key, use SM4 or AES-128 to encrypt the data payload, and use a pseudo-random frequency hopping mechanism to switch between 4 orthogonal channels to combat interference.

10. A corrosion sensing data transmission and communication optimization system for marine environmental power equipment according to claim 2, characterized in that, When the aggregation node outputs a warning or alarm level in the random forest model, it triggers an event-driven report within 10 seconds. The report includes raw data fragments from each second within the 5 minutes prior to the warning trigger. The system is applied to online monitoring of metal structure corrosion in offshore wind farms, coastal substations, or cross-sea transmission towers.

Images