Power equipment multi-modal data real-time acquisition and monitoring system based on star flash ad hoc network

Abstract

The invention relates to a power equipment multi-modal data real-time acquisition and monitoring system based on a star flash ad hoc network, and belongs to the technical field of power Internet of Things. Aiming at the problems of high communication delay, large multi-source data fusion error and poor complex environment adaptability in power equipment monitoring, the method comprises the following steps: 1) constructing a star-flash dynamic ad hoc network, and modulating by adopting FHSS frequency hopping and QPSK-polarization code (error rate lt; 10-6) the anti-interference performance is improved; 2) designing a multi-modal data space-time alignment algorithm (IEEE 1588v2 + DTW interpolation) and a dynamic weight fusion strategy (normal / fault working condition weight 0.6-0.2); and 3) three-dimensional visualization and edge AI fault positioning are realized based on digital twinning, and the accuracy rate is 98.5%. The system reduces the operation and maintenance cost by 67%, and is suitable for high-precision monitoring scenes such as a transformer substation.

Description

Technical Field

[0001] The present invention belongs to the technical field of the power Internet of Things, and relates to a real-time multi-modal data acquisition and monitoring system for power equipment based on SparkLink self-organizing network. Background Art

[0002] In the field of intelligent monitoring of power equipment, the existing technologies face the following key problems to be solved urgently:

[0003] (1) Bottlenecks in communication real-time performance and reliability

[0004] Traditional wireless sensor networks (such as ZigBee, LoRa) have significant defects in the monitoring of power equipment:

[0005] The end-to-end delay of existing solutions in complex power grid topologies generally exceeds 20 ms (for example, the measured delay of the LoRa solution disclosed in CN112954749A reaches 32 ms), making it difficult to meet the millisecond-level real-time requirements such as breaker status monitoring, resulting in a fault location lag rate of 12%-15%;

[0006] In the strong electromagnetic environment of substations (interference intensity of 30-50 dBm), the bit error rate of traditional FSK modulation schemes is as high as 10-3 (see CN113114433B), and the local discharge signal loss rate exceeds 18%;

[0007] The maximum number of nodes supported by existing self-organizing network protocols (such as CN112737786A) is less than 100, making it difficult to cover the monitoring requirements of the entire 500 kV substation area.

[0008] (2) For the multi-modal data involved in power equipment monitoring, such as electrical parameters (voltage / current), mechanical vibration, environmental temperature and humidity, etc., the existing technologies have:

[0009] The clock deviation of sensor nodes generally exceeds 5 μs (such as the scheme in CN113452501A), resulting in an error of 8.3% in the temperature-vibration correlation analysis;

[0010] The traditional weighted average method (CN112269366B) uses fixed weight allocation, and the misjudgment rate is as high as 22% under the condition of equipment overload;

[0011] The spatial registration error of the existing digital twin system (CN114064820A) > 0.5 m, and it is impossible to accurately locate minute defects such as insulator cracks.

[0012] (3) Limitations in extreme environment adaptability

[0013] The memory failure rate of existing edge computing nodes (such as CN113656722A) increases suddenly by 37% at -40°C low temperature;

[0014] The wireless sensor node (CN112737658A) has a battery life of less than 72 hours and a data disconnection rate of over 25% in the typhoon monitoring scenario;

[0015] When 30% of the nodes of the traditional TDMA protocol (CN113347029A) fail, the network reconstruction time is greater than 200 ms, seriously affecting the continuity of fault monitoring.

[0016] CN114024647A proposes to use 5G slicing technology to improve the transmission bandwidth, but does not solve the problem of deteriorated bit error rate under strong electromagnetic interference; CN113852928A uses the LSTM algorithm to improve data fusion, but the calculation delay increases by 40 ms, which cannot meet the real-time warning requirements. None of the above solutions effectively overcome the technical bottlenecks of spatio-temporal alignment of multi-modal data, reliable networking of high-density nodes, and collaborative optimization of adaptability to complex electromagnetic environments.

[0017] In view of the above defects, the present invention realizes millisecond-level low-latency transmission through the NearLink dynamic self-organizing network technology, innovatively integrates chaotic frequency hopping and polar code modulation to improve the anti-interference ability, and combines the spatio-temporal two-dimensional alignment algorithm and the edge-cloud collaborative inference mechanism to break through the technical ceiling of the existing power monitoring system. Summary of the Invention

[0018] In view of this, the purpose of the present invention is to provide a real-time multi-modal data acquisition and monitoring system for power equipment based on NearLink self-organizing network.

[0019] To achieve the above purpose, the present invention provides the following technical solutions:

[0020] A real-time multi-modal data acquisition and monitoring system for power equipment based on NearLink self-organizing network, comprising the following steps:

[0021] S1: Construct a multi-source perception network for power equipment, deploy NearLink terminal nodes on substations, transmission lines or utility poles, and integrate electrical parameter sensors, environmental monitoring sensors and mechanical state sensors to form a multi-modal data set;

[0022] S2: Standardize the collected data, label spatio-temporal tags and equipment status classification tags, and divide the data set into a training set, a validation set and a test set;

[0023] S3: Establish a NearLink self-organizing network fusion model, including:

[0024] S3.1: Use the NearLink dynamic self-organizing network protocol to construct a 256-node cascaded network, and design an anti-interference communication link based on frequency hopping spread spectrum (FHSS) and QPSK modulation;

[0025] S3.2: Deploy a lightweight Kalman filter algorithm on the edge computing gateway to achieve dynamic weighted fusion of multi-source data;

[0026] S3.3: Build a 3D model of the power grid based on the digital twin platform and develop a real-time mapping algorithm for the data flow and the virtual model;

[0027] S4: Deploy the optimized model to the edge computing gateway and achieve fault location within minutes in combination with the digital twin platform.

[0028] Furthermore, the SparkLink terminal node includes:

[0029] An electrical parameter module for collecting voltage, current, and partial discharge signals;

[0030] An environmental monitoring module for collecting temperature, humidity, wind speed, and ultraviolet imaging data;

[0031] A mechanical state module for monitoring the vibration spectrum and deformation of the equipment;

[0032] The electrical parameter module, the environmental monitoring module, and the mechanical state module achieve spatio-temporal synchronous transmission of multi-source data through the SparkLink network protocol and perform dynamic weighted fusion based on the edge computing node;

[0033] The data is transmitted back through the SparkLink-5G fusion, and the end-to-end delay is less than 50 ms.

[0034] Furthermore, in the above-mentioned S2, the standardization process includes:

[0035] Hardware-level synchronization, adopting the IEEE 1588v2 precise time protocol, with the clock deviation between nodes less than 0.5 μs;

[0036] Software-level synchronization, performing data interpolation based on the dynamic time warping (DTW) algorithm, with the maximum allowable time difference being 5 ms;

[0037] The device status classification labels include 8 types of device statuses and 6 types of environmental events.

[0038] The 8 types of device statuses specifically include:

[0039] Normal status, detection index: all parameters are within the rated range;

[0040] Overheating anomaly, detection index: the surface temperature of the device > 85 °C or the temperature of the wire joint > 90 °C;

[0041] Partial discharge, detection index: discharge amount > 50 pC or ultraviolet imaging photon count > 2000 / s;

[0042] Mechanical looseness, detection index: vibration main frequency deviation > 10% or displacement > 0.5 mm;

[0043] Insulation deterioration, detection index: dielectric loss factor > 0.02 or leakage current > 10 mA;

[0044] Abnormal contact resistance, detection index: joint resistance > 1.2 times the initial value;

[0045] Abnormal oil level / gas, detection index: transformer oil level < calibration lower limit or SF6 gas pressure < 0.4 MPa;

[0046] Cooling system failure, detection index: fan speed < 80% of the rated value or oil pump flow < 50 L / min;

[0047] The 6 types of environmental event bodies include:

[0048] Typhoon / strong wind, detection index: wind speed > 20 m / s or conductor dancing amplitude > 0.5 m;

[0049] Lightning strike, detection index: lightning counter action times > 3 times / hour or ground grid current > 20 kA;

[0050] Icing, detection index: conductor icing thickness > 10 mm or tension sensor overload > 15%;

[0051] Heavy rain / flood, detection index: precipitation > 50 mm / h or water level sensor triggered;

[0052] High temperature / drought, detection index: ambient temperature > 40 °C for 6 hours continuously or humidity < 20% RH;

[0053] Pollution flashover, detection index: equivalent salt deposit density > 0.1 mg / cm 2 Or corona intensity detected by ultraviolet imaging > 5000 photons / s.

[0054] Furthermore, in the S3.1, the FHSS frequency hopping sequence generation adopts a chaotic encryption algorithm, and its formula is:

[0055] f n = f min +(x n ×N)mod(f max -f min )

[0056] x n+1 = μx n (1 - x n ), μ ∈ [3.57, 4]

[0057] Frequency hopping once every 15 ms, QPSK modulation combined with polar code error correction, and the coding efficiency satisfies R = 1 - H(p e ); where H(·) is the binary entropy function; p e is the bit error rate and p e < 10 -6 .

[0058] Furthermore, in the above S3.2, the lightweight Kalman filtering includes:

[0059] Using the ARM NEON instruction set for parallel calculation of the inverse of a 4×4 matrix, the operation speed is increased by 8 times;

[0060] The residual detection threshold is set to Trigger the elimination of abnormal data;

[0061] The dynamic allocation of multi-source data weights is as follows: under normal operating conditions, the electrical parameters are 0.6, the environmental data is 0.3, and the mechanical state is 0.1; under fault conditions, it is adjusted to electrical parameters 0.4, environmental data 0.4, and mechanical state 0.2.

[0062] Furthermore, in the above S3.3, the real-time mapping algorithm includes:

[0063] Parsing the data frame according to the IEEE 1888 protocol, and extracting the device ID, timestamp, and sensor data;

[0064] Based on the Iterative Closest Point (ICP) algorithm to achieve the matching of physical coordinates and virtual models, the objective function is:

[0065]

[0066] The registration error is less than 0.1 m.

[0067] Furthermore, in the above S4, the deployment of the edge computing gateway includes:

[0068] Temperature control, using the PID algorithm to adjust the fan speed to maintain the SoC temperature below 85°C;

[0069] Memory management, using CMA continuous memory allocation to ensure that the AI inference latency is less than 30 ms;

[0070] Communication fault tolerance, the local cached data can be stored for up to 72 hours when the network is disconnected.

[0071] Furthermore, the PID parameters of the above temperature control are K p = 0.8, K i = 0.05, K d = 0.1.

[0072] Furthermore, the above fault location adopts three levels of accuracy:

[0073] The ranging error of the SparkLink TOA is ±1 m; the positioning error assisted by visual markers is ±0.5 m; the spatial reasoning error of the digital twin is ±0.2 m.

[0074] Furthermore, the digital twin platform is built based on Unreal Engine 5.2, the grid three-dimensional model mesh resolution is less than 0.01 m, and the real-time rendering device temperature field, vibration cloud map, and discharge hot spots are rendered.

[0075] The beneficial effects of the present invention are as follows:

[0076] (1) Through the SparkLink dynamic self-organizing network protocol and TDMA time slot optimization, the end-to-end delay is achieved to be < 1 ms, which is 98.6% lower than the traditional scheme, meeting the millisecond-level monitoring requirements such as the opening and closing states of circuit breakers.

[0077] (2) Chaotic frequency hopping (frequency hopping every 15 ms) is combined with QPSK-polarization code modulation, and the bit error rate is < 8.2×10 -7 under strong electromagnetic interference of 30 dBm, ensuring data integrity in the complex environment of substations.

[0078] (3) Support for cascaded networking of 256 nodes, with a single-node coverage radius of 500 meters, which can meet the full-domain monitoring requirements of 500 kV extra-high voltage substations.

[0079] (4) Hardware-level synchronization (IEEE1588v2) and software-level DTW interpolation cooperate, with the spatio-temporal registration error < 0.5 μs, and the multi-source data fusion error rate is reduced to 2.3%.

[0080] (5) Based on the working condition adaptive weight allocation strategy (the weight of electrical parameters in normal working conditions is 0.6, and the weight of mechanical states in fault working conditions is increased to 0.2), the diagnosis accuracy rate of overload working conditions is increased from 78% to 96.8%.

[0081] (6) Through the ICP algorithm, the physical-virtual space registration error < 0.1 m is achieved, supporting the AR visual positioning of 1 mm-level cracks in insulators.

[0082] (6) Triple protection of edge computing nodes (PID temperature control maintains the SoC temperature < 85 °C / stable operation at low temperature of -40 °C), and the device life is extended by 3.2 times.

[0083] (7) When 30% of the nodes fail, the network reconstruction time < 50 ms, and the system availability remains above 99.9%.

[0084] Other advantages, objectives, and features of the present invention will be described to some extent in the subsequent specification, and to some extent, will be obvious to those skilled in the art based on the study of the following text, or can be taught from the practice of the present invention. The objectives and other advantages of the present invention can be achieved and obtained through the following specification. Brief Description of the Drawings

[0085] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be described in detail with reference to the accompanying drawings, where:

[0086] Figure 1 It is the overall system framework diagram provided in the embodiment of the present invention;

[0087] Figure 2 It is the application technology diagram of the SparkLink self-organizing network in the embodiment system of the present invention;

[0088] Figure 3 It is the FHSS frequency hopping timing and QPSK modulation flow chart implemented in the embodiment of the present invention;

[0089] Figure 4 It is the bar chart of the Kalman filter residual detection threshold and dynamic weight allocation implemented in the embodiment of the present invention;

[0090] Figure 5 It is the point cloud matching effect diagram of the ICP registration algorithm implemented in the digital twin platform in the embodiment of the present invention. Specific Embodiments

[0091] The following uses specific specific examples to illustrate the implementation manners of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific implementation manners. Various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the drawings provided in the following embodiments only illustrate the basic concept of the present invention in a schematic manner. Without conflict, the following embodiments and the features in the embodiments can be combined with each other.

[0092] Among them, the drawings are only for illustrative purposes, showing only schematic diagrams, rather than physical diagrams, and should not be construed as limiting the present invention; in order to better illustrate the embodiments of the present invention, some components in the drawings will be omitted, enlarged, or reduced, which do not represent the dimensions of actual products; for those skilled in the art, it is understandable that some well-known structures and their descriptions in the drawings may be omitted.

[0093] In the accompanying drawings of the embodiments of the present invention, the same or similar reference numerals correspond to the same or similar components; in the description of the present invention, it should be understood that if there are terms such as "upper", "lower", "left", "right", "front", "rear", etc. indicating the orientation or positional relationship, they are based on the orientation or positional relationship shown in the accompanying drawings. This is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, the terms describing the positional relationship in the accompanying drawings are only for illustrative purposes and should not be construed as a limitation of the present invention. For those of ordinary skill in the art, the specific meanings of the above terms can be understood according to specific circumstances.

[0094] Please refer to Figure 1 , which is a real-time multi-modal data acquisition and monitoring system for power equipment based on SparkLink self-organizing network, and it includes the following steps:

[0095] Step 1: Precise deployment of multi-modal sensors and data acquisition

[0096] Step 1.1 Sensor selection and configuration

[0097] Deploy SparkLink terminal nodes (T nodes) on substations, transmission lines, and utility poles. Each node integrates a tri-modal heterogeneous sensor:

[0098] Electrical parameter module: Deploy Hall sensors (HX-10P) at key nodes of main transformers, circuit breakers, and transmission lines in substations to collect voltage (accuracy ±0.5%), current (accuracy ±0.2%), and partial discharge signals (sensitivity ≤5 pC) in real time;

[0099] Environmental monitoring module: Install SHT35 temperature and humidity sensors (measurement range -40°C to 125°C) and FS3000 ultrasonic anemometers (resolution 0.1 m / s) at the poles and wire suspension points, and synchronously integrate ultraviolet imagers to detect corona discharge;

[0100] Mechanical state module: Install ADXL345 three-axis vibration sensors (measurement range ±16 g) and LVDT-S2 displacement sensors (accuracy ±0.1 mm) at the transformer base and the connection of the pole tower to monitor the vibration spectrum and deformation of the equipment.

[0101] Step 1.2 Multi-source data synchronization mechanism

[0102] Hardware-level synchronization: Adopt the IEEE1588v2 Precision Time Protocol (PTP) to achieve a clock deviation of ≤0.5 μs for all network nodes through the SparkLink network master clock;

[0103] Software-level alignment: Apply the Dynamic Time Warping (DTW) algorithm to asynchronous data streams to interpolate and compensate for a maximum time difference of 5 ms. The formula is simplified as:

[0104] minΣ|t i -t j '| 2 ,s·t·|i - j| ≤ δ(δ=5ms)

[0105] Step 2: SparkLink Self - Organizing Network Dynamic Networking and Anti - interference Transmission

[0106] Step 2.1 Network Architecture Deployment

[0107] Node Deployment: SparkLink terminal nodes (T nodes) are deployed on substations, transmission lines and poles. The single - node coverage radius is 500 meters, supporting 256 - node cascaded networking. The topology structure adopts a dynamic self - healing Mesh network;

[0108] Communication Link Design: Anti - interference Frequency Hopping: Based on the chaotic sequence (Logistic mapping) to generate the frequency - hopping pattern, and the frequency - switching formula is as follows:

[0109] f n =f min +(x n ×N) mod (f max -f min )

[0110] Frequency hopping occurs every 15ms, covering the 2.4GHz - 2.4835GHz frequency band, and the anti - interference gain is increased by 40%.

[0111] High - reliability Modulation: Adopt QPSK + polar code error - correction technology, and the coding efficiency formula is as follows:

[0112] R=1 - H(p e )

[0113] where H(·) is the binary entropy function, p e <10 -6

[0114] Step 2.2 Low - latency Transmission Guarantee

[0115] End - to - end Optimization: Through the time - slot allocation algorithm (TDMA) and priority queue scheduling, ensure that the transmission delay of key data (such as partial discharge) < 1ms;

[0116] Edge Caching Mechanism: The node has a 72 - hour local storage built - in. When the network is disconnected, it automatically caches data, and after the network is restored, it synchronizes incrementally to the cloud.

[0117] Step 3: Cloud - Edge - Terminal Collaborative Data Fusion and Intelligent Applications

[0118] Step 3.1 Edge - side Lightweight Processing

[0119] Feature-level fusion: Deploy an improved Kalman filtering algorithm on the RK3588 edge gateway, with a dynamic weighting strategy:

[0120] Normal operating condition: Electrical parameter weight 0.6, environmental data 0.3, mechanical state 0.1;

[0121] Fault condition: Electrical parameter weight 0.4, environmental data 0.4, mechanical state 0.2.

[0122] Anomaly detection: Set the residual threshold Implement the elimination of abnormal data.

[0123] Step 3.2 Cloud digital twin mapping

[0124] 3D modeling: Build a high-precision substation model (mesh resolution <0.01m) based on UnrealEngine5.2, and achieve the matching of physical coordinates and virtual space through the ICP registration algorithm. The formula is as follows:

[0125]

[0126] where p i is the physical coordinate point, q i is the virtual simulation point; the registration error <0.1m

[0127] Visualization warning: Real-time render the device temperature field, vibration cloud map and discharge hot spots on the digital twin interface, and support the AR inspection terminal to retrieve the 3D fault location (accuracy ±0.2m).

[0128] Please refer to Figure 2 , which is the implementation of SparkLink self-organizing network dynamic networking in this system, and it includes the following steps:

[0129] Step 1: Dynamic network architecture deployment

[0130] Step 1.1 Node configuration and networking

[0131] Deploy SparkLink terminal nodes (T nodes) in substations, transmission lines and towers. The single-node coverage radius is 500 meters, and 256-node cascaded networking is supported. Adopt a dynamic Mesh topology structure to achieve network self-healing function, and the recovery time after node failure <50ms. Set the main coordinator node to be responsible for the time slot allocation, frequency scheduling and routing optimization of the whole network to ensure the transmission priority of key data (such as partial discharge signals). Step 1.2 Communication frequency band planning

[0132] The working frequency band covers the 2.4GHz - 2.4835GHz ISM band and supports dynamic spectrum allocation. Real-time detect the interference intensity through the channel sensing algorithm and automatically avoid congested frequency points.

[0133] Step 2: Implementation of anti-interference communication protocol

[0134] Step 2.1 Frequency Hopping Spread Spectrum (FHSS) technology

[0135] A chaotic frequency hopping sequence is designed based on the Logistic map to generate a pseudo-random frequency hopping sequence. The formula is as follows:

[0136] f n = f min +(x n ×N) mod (f max - f min )

[0137] x n+1 = μx n (1 - x n ), μ ∈ [3.57, 4]

[0138] Frequency hopping occurs every 15 ms, covering the frequency band of 2.4 GHz - 2.4835 GHz. The anti-jamming gain is increased by 40%, adapting to the strong electromagnetic environment in the substation.

[0139] Step 2.2 Modulation and Error Correction Technology

[0140] QPSK modulation is combined with PolarCode error correction. The coding efficiency is:

[0141] R = 1 - H(p e )

[0142] where H(·) is the binary entropy function, and p e <10 -6

[0143] In an electromagnetic interference environment of 30 dBm, the measured bit error rate < 8.2×10 -7 , meeting the high-reliability transmission requirements.

[0144] Step 3: Low-latency Transmission Optimization

[0145] Step 3.1 Time Slot Allocation and Scheduling

[0146] Based on the TDMA (Time Division Multiple Access) protocol, the communication cycle is divided into 1 ms time slots, and exclusive time slots are allocated for critical data. A dynamic priority queue is introduced to adjust the data transmission order in real time, ensuring that the end-to-end latency of alarm information such as partial discharge < 1 ms.

[0147] Step 3.2 Edge Caching and Fault Tolerance Mechanism

[0148] The node has built-in 72-hour local storage, automatically caches data when the network is disconnected, and uploads it to the cloud through an incremental synchronization strategy after the network is restored. A hybrid mechanism of forward error correction (FEC) and automatic repeat request (ARQ) is adopted to reduce the packet retransmission rate to <0.1%.

[0149] Step 4: Network performance verification and optimization

[0150] Step 4.1 Performance test metrics

[0151] Throughput: The peak rate of a single link > 10 Mbps, meeting the concurrent transmission requirements of multi-modal data; Latency consistency: Under full load of 256 nodes, the end-to-end latency jitter < 0.2 ms; Anti-interference ability: At an interference intensity of 30 dBm, the communication success rate > 99.2%.

[0152] Step 4.2 Dynamic parameter tuning

[0153] The hopping sequence and routing path are optimized in real time through a reinforcement learning algorithm to reduce the probability of network congestion. A network health monitoring module is deployed to periodically diagnose the node status and predictively replace faulty devices.

[0154] Please refer to Figure 3 , which is an implementation of an anti-interference communication protocol, and includes the following steps:

[0155] Step 1: Hopping sequence generation and modulation optimization

[0156] A chaotic hopping sequence is designed based on the Logistic map to generate a pseudo-random hopping sequence, and its formula is as follows:

[0157] f n = f min +(x n × N) mod (f max - f min )

[0158] x n+1 = μx n (1 - x n ), μ ∈ [3.57, 4]

[0159] Hopping frequency once every 15 ms, covering the frequency band of 2.4 GHz - 2.4835 GHz, and the anti-interference gain is increased by 40%.

[0160] Step 2: QPSK modulation and polar code error correction

[0161] The modulation symbol mapping formula is as follows:

[0162]

[0163] The polar code coding efficiency formula is as follows:

[0164] R = 1 - H(p e )

[0165] where H(·) is the binary entropy function, p e <10 -6

[0166] Step 3: Communication link verification

[0167] Step 3.1 Basic performance verification (laboratory environment)

[0168] The test objectives are: to verify the theoretical performance indicators of a single communication link, including bit error rate, throughput, and end-to-end delay. The test equipment includes: 2 SparkLink terminal nodes (T nodes), a signal generator (simulating data source), a bit error rate tester (Keysight N4960A), and a high-precision oscilloscope (measuring delay). First, establish a point-to-point link between the two T nodes and send a standard test sequence (PRBS15); then adjust the transmit power to the nominal value (20 dBm) and measure the bit error rate at the receiving end; finally, capture the packet sending and receiving timestamps through the oscilloscope and calculate the end-to-end delay. The final test results are: the measured bit error rate ≤ 10 -6 (after polar code error correction); the peak rate of the single-link throughput ≥ 10 Mbps; the end-to-end transmission delay < 1 ms.

[0169] Step 3.2 Anti-interference ability verification (simulating electromagnetic environment)

[0170] The test objectives are: to verify the communication stability of the FHSS / QPSK combined technology in a high electromagnetic interference environment. First, conduct interference simulation: use a radio frequency interference source (Rohde & Schwarz SMBV100B) to generate 30 dBm continuous wave interference, with the frequency band covering 2.4 GHz to 2.4835 GHz; then repeat the bit error rate and delay tests in Step 3.1 in the interference environment; then dynamically switch the frequency hopping sequence and record the channel switching success rate. The final test results are: the bit error rate is still ≤ 10 -6 ; the channel switching success probability ≥ 99.5%; the link recovery time < 10 ms after the interference is removed.

[0171] Step 3.3 Multi-node networking verification (actual deployment scenario)

[0172] The test objectives are: to verify the dynamic networking ability of 256 nodes and the network self-healing performance. First, network configuration is carried out. 256 T nodes are deployed inside the substation to build a dynamic Mesh network; the master coordinator node allocates time slots and routing paths; then, fault simulation is carried out. 10% of the nodes are randomly disconnected (simulating equipment failures), and the topology reconstruction time is recorded; high-priority alarm data (such as partial discharge signals) is injected to test the end-to-end delay consistency. The final test results are: support for 256-node cascading, network load rate ≤ 80%; network recovery time < 50 ms after node failure; delay standard deviation < 0.2 ms under full load.

[0173] Please refer to Figure 4 , multi-modal data fusion and anomaly detection, including the following steps:

[0174] Step 1: Lightweight Kalman filtering

[0175] Its state equation is as follows:

[0176] x k = Fx k-1 + ω k

[0177] Its observation equation is as follows:

[0178]

[0179] Use the ARM NEON instruction set to calculate the 4×4 covariance matrix P in parallel k , accelerate the matrix, and the inverse calculation speed is increased by 8 times.

[0180] Step 2: Residual detection and anomaly rejection

[0181] The trigger condition formula is as follows:

[0182]

[0183] When the residual value exceeds the threshold, the abnormal data is rejected.

[0184] Step 3: Dynamic weight allocation

[0185] Normal operating condition: electrical parameter weight 0.6, environmental data 0.3, mechanical state 0.1;

[0186] Fault condition: electrical parameter weight 0.4, environmental data 0.4, mechanical state 0.2.

[0187] Please refer to Figure 5 , the application of the ICP registration algorithm in the digital twin platform, including the following steps:

[0188] Step 1: Generate a high-precision 3D model of the substation based on Unreal Engine 5.2, with a mesh resolution < 0.01m.

[0189] Step 2: ICP registration algorithm

[0190] Objective function:

[0191]

[0192] where p i is the physical coordinate point, q i is the virtual simulation point; the registration error < 0.1m

[0193] Step 3: Three-level fault location

[0194] First, for the star flash TOA ranging, it is based on the time of arrival difference (TOA) for positioning, with an error of ±1m; then through AR tag recognition, with the assistance of visual markers, the error is ±0.5m; finally, combining the device topology relationship and historical data, the digital twin system conducts reasoning, with an error of ±0.2m.

[0195] The 8 types of device states include:

[0196] Normal state

[0197] Detection indicators: All parameters are within the rated range (such as voltage ±5%, temperature ≤ 75°C).

[0198] Overheating anomaly

[0199] Detection indicators: The surface temperature of the device > 85°C (transformer) or the temperature of the wire joint > 90°C.

[0200] Partial discharge

[0201] Detection indicators: Discharge amount > 50pC (transformer) or ultraviolet imaging photon count > 2000 / s (insulator).

[0202] Mechanical looseness

[0203] Detection indicators: The main vibration frequency deviation > 10% (such as from 1kHz to 1.1kHz) or the displacement > 0.5mm.

[0204] Insulation deterioration

[0205] Detection indicators: Dielectric loss factor (tanδ) > 0.02 or leakage current > 10mA.

[0206] Abnormal contact resistance

[0207] Detection indicators: The joint resistance > 1.2 times the initial value (such as rising from 50μΩ to 60μΩ).

[0208] Oil level / gas abnormality

[0209] Detection index: Transformer oil level < calibration lower limit or SF6 gas pressure < 0.4 MPa.

[0210] Cooling system failure

[0211] Detection index: Fan speed < 80% of the rated value or oil pump flow < 50 L / min.

[0212] Six types of environmental events include:

[0213] Typhoon / strong wind

[0214] Detection index: Wind speed > 20 m / s (tower tilt angle > 1°) or conductor galloping amplitude > 0.5 m.

[0215] Lightning strike

[0216] Detection index: Lightning counter operation times > 3 times / hour or ground grid current > 20 kA.

[0217] Icing

[0218] Detection index: Conductor ice thickness > 10 mm or tension sensor overload > 15%.

[0219] Heavy rain / flood

[0220] Detection index: Precipitation > 50 mm / h or water level sensor triggered (substation internal water accumulation > 0.3 m).

[0221] High temperature / drought

[0222] Detection index: Ambient temperature > 40 °C for 6 hours continuously or humidity < 20% RH.

[0223] Pollution flashover

[0224] Detection index: Equivalent salt deposit density (ESDD) > 0.1 mg / cm 2 Or corona intensity detected by ultraviolet imaging > 5000 photons / s.

[0225] The present invention realizes the efficient acquisition and intelligent fusion of multimodal data of power equipment through SparkLink self-organizing network technology, breaking through the technical bottlenecks of traditional monitoring systems in terms of communication real-time performance, environmental adaptability, and data fusion accuracy. Its core innovation lies in: Based on the SparkLink networking dynamic networking and anti-interference communication protocol, it solves the problems of high reliability and low latency of data transmission in complex electromagnetic environments; through the spatio-temporal alignment of multi-source data and the edge-cloud collaborative processing mechanism, it realizes the generalization and fusion analysis of non-ideally distributed data without relying on preset data distribution models or fixed scenario assumptions, significantly improving the universality of the system.

[0226] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the present technical solution, and they should all be covered within the scope of the claims of the present invention.

Claims

1. A real-time multi-modal data acquisition and monitoring system for power equipment based on Star Flash Ad Hoc Network, characterized by: The following steps are involved: S1: Build a multi-source sensing network for power equipment, deploy Star Flash terminal nodes on substations, transmission lines or poles, integrate electrical parameter sensors, environmental monitoring sensors and mechanical status sensors to form a multimodal data set; S2: Standardize the collected data, annotate the spatiotemporal labels and equipment status classification labels, and divide the data set into training set, validation set and test set; S3: Establish the Star Flash Self-organizing Network Fusion Model, including: S3.1: Use the Star Flash dynamic self-organizing network protocol to build a 256-node cascade network and design an anti-interference communication link based on Frequency-Hopping Spread Spectrum (FHSS) and Quadrature Phase Shift Keying (QPSK) modulation; S3.2: Deploy a lightweight Kalman filter algorithm on the edge computing gateway to achieve dynamic weighted fusion of multi-source data; S3.3: Build a three-dimensional model of the power grid based on the digital twin platform and develop a real-time mapping algorithm between data flow and virtual model; S4: Deploy the optimization model to the edge computing gateway and combine it with the digital twin platform to locate faults within minutes.

2. The real-time multi-modal data acquisition and monitoring system for electric power equipment based on Star Flash Ad Hoc Network according to claim 1 is characterized in that: The Star Flash terminal node includes: Electrical parameter module, used to collect voltage, current and partial discharge signals; Environmental monitoring module, used to collect temperature, humidity, wind speed and ultraviolet imaging data; Mechanical status module, used to monitor the vibration spectrum and deformation of equipment; The electrical parameter module, environmental monitoring module and mechanical status module realize the time-space synchronous transmission of multi-source data through the Star Flash network protocol, and perform dynamic weighted fusion based on edge computing nodes; The data is backhauled via StarFlash-5G fusion, with an end-to-end latency of less than 50ms.

3. The real-time multi-modal data acquisition and monitoring system for electric power equipment based on Star Flash Ad Hoc Network according to claim 1 or 2 is characterized in that: In S2, the standardization process includes: Hardware-level synchronization, using IEEE 1588v2 precision time protocol, the clock deviation between nodes is less than 0.5μs; Software-level synchronization, data interpolation based on the Dynamic Time Warping (DTW) algorithm, with a maximum allowable time difference of 5ms; The device status classification labels include 8 types of device status and 6 types of environmental events; The 8 types of device status include: Normal state, test indicators: all parameters are within the rated range; Overheating abnormality, detection indicators: equipment surface temperature>85℃ or wire connector temperature>90℃; Partial discharge, detection index: discharge amount>50pC or UV imaging photon number>2000 / s; Mechanical looseness, detection index: vibration main frequency deviation> 10% or displacement> 0.5mm; Insulation degradation, detection indicators: dielectric loss factor> 0.02 or leakage current> 10mA; Abnormal contact resistance, detection index: joint resistance>1.2 times the initial value; Oil level / gas abnormality, detection index: transformer oil level < calibration lower limit or SF6 gas pressure < 0.4MPa; Cooling system failure, detection indicators: fan speed < 80% of rated value or oil pump flow < 50L / min; The six types of environmental events include: Typhoon / strong wind, detection index: wind speed>20m / s or conductor dancing amplitude>0.5m; Lightning strike, detection index: lightning counter action times> 3 times / hour or ground grid current> 20kA; Ice coating, detection index: conductor ice coating thickness> 10mm or tension sensor overload> 15%; Heavy rain / flood, detection index: precipitation > 50mm / h or water level sensor trigger; High temperature / drought, detection indicators: ambient temperature>40℃ for 6 hours or humidity<20%RH; Pollution flashover, detection index: equivalent salt density> 0.1mg / cm 2 Or ultraviolet imaging detects corona intensity > 5000 photons / s.

4. The real-time multi-modal data acquisition and monitoring system for electric power equipment based on Star Flash Ad Hoc Network according to any one of claims 1 to 3, characterized in that: In S3.1, the FHSS frequency hopping sequence is generated using a chaotic encryption algorithm, and its formula is: f n =f min +(x n ×N)mod(f max -f min ) x n+1 =μx n (1-x n ),μ∈[3.57,4] Frequency hopping is performed every 15ms, QPSK modulation is combined with polarization code error correction, and the coding efficiency satisfies R = 1-H (p e ), where H(·) is the binary entropy function; p e is the bit error rate and p e <10 -6 .

5. The real-time multi-modal data acquisition and monitoring system for electric power equipment based on Star Flash Ad Hoc Network according to claim 4 is characterized by: In S3.2, the lightweight Kalman filter includes: The ARM NEON instruction set is used to parallelize the 4×4 matrix inversion, increasing the calculation speed by 8 times; The residual detection threshold is set to Trigger the elimination of abnormal data; The weights of multi-source data are dynamically allocated as follows: electrical parameters 0.6, environmental data 0.3, and mechanical status 0.1 under normal conditions; under fault conditions, they are adjusted to electrical parameters 0.4, environmental data 0.4, and mechanical status 0.

2.

6. The real-time multi-modal data acquisition and monitoring system for electric power equipment based on Star Flash Ad Hoc Network according to any one of claims 1 to 5, characterized in that: In S3.3, the real-time mapping algorithm includes: Parse data frames according to IEEE 1888 protocol to extract device ID, timestamp and sensor data; The physical coordinates are matched with the virtual model based on the iterative closest point (ICP) algorithm. The objective function is: The registration error is less than 0.1m.

7. The real-time multi-modal data acquisition and monitoring system for electric power equipment based on Star Flash Ad Hoc Network according to any one of claims 1 to 6, characterized in that: In S4, the edge computing gateway deployment includes: Temperature control, using the Proportional-Integral-Derivative (PID) algorithm to adjust the fan speed and maintain the SoC temperature below 85°C; Memory management uses Contiguous Memory Allocation (CMA) to ensure that AI inference latency is less than 30ms; Communication fault tolerance, local cache data for up to 72 hours when the network is disconnected.

8. The real-time multi-modal data acquisition and monitoring system for electric power equipment based on Star Flash Ad Hoc Network according to claim 7 is characterized in that: The PID parameter of the temperature control is K p =0.8, K i =0.05, K d =0.

1.

9. The real-time multi-modal data acquisition and monitoring system for electric power equipment based on Star Flash Ad Hoc Network according to any one of claims 1 to 8, characterized in that: The fault location adopts three levels of accuracy: Star flash TOA ranging error is ±1m; visual marker assisted positioning error is ±0.5m; digital twin space reasoning error is ±0.2m.

10. The real-time multi-modal data acquisition and monitoring system for electric power equipment based on Star Flash Ad Hoc Network according to any one of claims 1 to 9, characterized in that: The digital twin platform is built on Unreal Engine 5.

2. The grid resolution of the three-dimensional model of the power grid is less than 0.01m, and it can render the equipment temperature field, vibration cloud map and discharge hotspots in real time.

Images