Power distribution data security compression transmission system based on internet of things

Abstract

The application belongs to the technical field of power internet of things and information security, and specifically discloses a power distribution data security compression transmission system based on an internet of things. The system comprises a power distribution terminal data acquisition unit, a compression sensing measurement unit, a secure transmission unit, a gateway receiving unit and a data reconstruction unit. Signal compression and semantic confusion are simultaneously realized through random linear measurement of compression sensing, and a measurement matrix is dynamically generated by a physically unclonable function as a natural key. The secure transmission unit omits an independent encryption module and relies on endogenous security in the compression process. The data reconstruction unit adopts a pre-trained generative adversarial network, combines statistical characteristics of the measurement matrix, and restores the original waveform with high precision. Through the above scheme, the transmission security and the reconstruction robustness of non-sparse fault signals are improved while reducing terminal power consumption and delay, and the application is suitable for large-scale power distribution internet of things deployment scenarios with limited resources.

Description

Technical Field

[0001] This invention belongs to the field of power Internet of Things and information security technology, specifically relating to an Internet of Things-based power distribution data secure compression and transmission system. Background Technology

[0002] With the widespread application of IoT technology in smart distribution networks, the digital transformation of distribution systems has become crucial for achieving efficient operation and refined management of power systems. Distribution terminals, as the core unit for data acquisition, are responsible for real-time monitoring of grid operation status, capturing transient fault waveforms, and uploading them to the management platform. In complex power operation environments, the distribution IoT formed by massive sensor nodes places extremely high demands on the real-time performance, reliability, and security of data transmission. This makes efficient data processing and security defense mechanisms essential technical support for ensuring stable grid operation.

[0003] For high-frequency sampled waveform data in power distribution networks, existing data processing solutions typically employ a serial mechanism of compression followed by encryption, aiming to balance the conflict between transmission bandwidth limitations and sensitive data protection. This process usually involves discretizing the original signal to achieve dimensionality reduction, supplemented by symmetric encryption algorithms for secure encapsulation. Through this step-by-step data processing logic, the system attempts to ensure the privacy and integrity of power production data in public communication channels while reducing communication overhead.

[0004] Traditional transform-domain-based compression algorithms are highly dependent on the computing resources of edge terminals, leading to a surge in power consumption and significant computational latency in resource-constrained IoT nodes. Because data sampling, compression, and encryption operations are performed in separate pipelines, computational resources are easily blocked in high-frequency sampling scenarios, making it difficult to meet the instantaneous response requirements of power distribution fault monitoring. Furthermore, traditional encryption mechanisms are disconnected from the sampling process, resulting in a system that cannot balance computational efficiency and semantic security. It also struggles to achieve efficient feature extraction and high-precision reconstruction of multi-source heterogeneous data from the power distribution network, thus limiting the application effectiveness of power distribution IoT in large-scale, complex scenarios.

[0005] Therefore, a secure compressed transmission system for power distribution data based on the Internet of Things is desired. Summary of the Invention

[0006] The purpose of this invention is to provide a secure compressed transmission system for power distribution data based on the Internet of Things, which can solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0008] The IoT-based secure compressed transmission system for power distribution data includes a power distribution terminal data acquisition unit, a compressed sensing measurement unit, a secure transmission unit, a gateway receiving unit, and a data reconstruction unit, wherein:

[0009] The power distribution terminal data acquisition unit is configured to perform high-frequency sampling of voltage, current and fault transient waveforms in the power distribution network to obtain the original power distribution sensing signals;

[0010] The compressed sensing measurement unit is connected to the power distribution terminal data acquisition unit and is configured to perform random linear measurement operations on the original power distribution sensing signal to generate a low-dimensional measurement vector. This measurement process simultaneously achieves signal compression and semantic obfuscation.

[0011] The secure transmission unit is connected to the compressed sensing measurement unit and is configured to encapsulate the low-dimensional measurement vector into a lightweight data frame and send it to the gateway receiving unit via a wireless communication link.

[0012] The gateway receiving unit is configured to receive security data frames from multiple power distribution terminals and forward them to the data reconstruction unit.

[0013] The data reconstruction unit is configured to perform high-precision signal reconstruction on the received low-dimensional measurement vector based on a pre-trained generative adversarial network model, thereby recovering the time-domain waveform characteristics of the original power distribution sensing signal.

[0014] Preferably, the random linear measurement matrix used by the compressed sensing measurement unit is dynamically generated by a physically non-clonable function, serving as the system's natural key to ensure that different terminals output different measurement results under the same original signal, thereby achieving semantic-level security protection.

[0015] Furthermore, the compressed sensing measurement unit does not perform highly complex compression algorithms such as discrete wavelet transform or tensor decomposition, but only relies on linear projection operations under sparse representation theory, thereby reducing the computational load and energy consumption of the power distribution terminal.

[0016] Furthermore, the secure transmission unit omits the independent encryption module found in traditional serial processing flows. Its security is inherent in the incoherent measurement mechanism of compressed sensing, avoiding pipeline blockage and transmission delay caused by the step-by-step execution of encryption and compression.

[0017] Preferably, the data reconstruction unit incorporates an adversarial training architecture consisting of a generator and a discriminator. The generator takes a low-dimensional measurement vector as input and outputs a reconstructed signal, while the discriminator evaluates the consistency between the reconstructed signal and the distribution of the real power distribution waveform. The two achieve robust recovery of sparse or non-sparse power distribution signals through game optimization.

[0018] Furthermore, the generative adversarial network model is trained end-to-end using historical power distribution fault waveform datasets during the offline phase. The statistical properties of the measurement matrix are introduced as a conditional constraint during the training process, enabling the reconstruction process to tolerate channel noise and measurement errors.

[0019] Furthermore, the power distribution terminal data acquisition unit, compressed sensing measurement unit, and secure transmission unit are integrated on the same edge hardware platform, with overall power consumption controlled at an extremely low level, making it suitable for long-term deployment at power distribution line monitoring points without external power supply conditions.

[0020] Preferably, the gateway receiving unit is configured with a data caching and priority scheduling mechanism, which can differentiate the data frames from different terminals according to the urgency of the power distribution event, and ensure the priority reconstruction and reporting of critical fault information.

[0021] Furthermore, the low-dimensional measurement vector generated by the compressed sensing measurement unit has a dimension much lower than the number of original signal sampling points, and its compression ratio is adaptively adjusted according to the dynamic sparsity characteristics of the power distribution scenario, ensuring that bandwidth utilization is maximized while meeting reconstruction accuracy.

[0022] Furthermore, the overall system architecture abandons the traditional three-stage processing paradigm of "acquisition-compression-encryption" and achieves deep integration of sensing, compression and encryption functions at the physical layer, so that the data has security attributes from the source without the need for additional encryption and decryption overhead.

[0023] Compared with the prior art, the present invention has the following beneficial effects:

[0024] The IoT-based power distribution data secure compression transmission system provided by this invention organically integrates compressed sensing theory, semantic security mechanisms and generative adversarial network reconstruction technology, achieving a balance between low power consumption, low latency and high security on the power distribution terminal side.

[0025] The system utilizes the random linear measurement process of compressed sensing to simultaneously complete data dimensionality reduction and information obfuscation, making the measurement matrix itself a natural key to resist known-plaintext attacks and replay attacks. The receiver adopts an intelligent reconstruction algorithm based on generative adversarial networks to improve the recovery accuracy of non-stationary signals such as fault transient waveforms, overcoming the performance problem of traditional compression methods in scenarios with insufficient sparsity.

[0026] The integrated physical layer design completely eliminates computational blocking in traditional serial processing architectures, reduces the computing power requirements and energy consumption of edge terminals, and is particularly suitable for large-scale deployment scenarios of distribution IoT with limited resources, providing reliable technical support for real-time monitoring and secure communication of smart distribution networks. Attached Figure Description

[0027] Figure 1 This is a schematic diagram of the overall technical solution architecture of the present invention;

[0028] Figure 2This is a schematic diagram of the core principle framework of the deep integration of physical layer sensing, compression and encryption based on compressed sensing in this invention.

[0029] Figure 3 This is a logical flowchart of the original signal acquisition and lightweight random linear measurement on the power distribution terminal side in this invention;

[0030] Figure 4 This is a schematic diagram of the internal logic architecture and adversarial training principle of the data reconstruction unit based on generative adversarial networks in this invention;

[0031] Figure 5 This is a schematic diagram of the multi-level interaction relationship and data flow involving power distribution terminals, gateways and reconfiguration centers in this invention. Detailed Implementation

[0032] Example 1: To make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to specific embodiments.

[0033] The IoT-based power distribution data secure compression transmission system includes a power distribution terminal data acquisition unit, a compression sensing measurement unit, a secure transmission unit, a gateway receiving unit, and a data reconstruction unit.

[0034] The power distribution terminal data acquisition unit is configured to be deployed at key nodes of the power distribution network. It includes multiple high-precision sensing channels for high-frequency sampling of instantaneous voltage values, instantaneous current values, and transient waveforms caused by various faults during the operation of the power distribution network at the millisecond or even microsecond level, so as to obtain raw power distribution sensing signals containing rich power operation characteristics.

[0035] The compressed sensing measurement unit, whose input terminal is electrically connected to the signal output terminal of the power distribution terminal data acquisition unit, is configured to perform a non-adaptive random linear measurement operation on the original power distribution sensing signal. By projecting the high-dimensional signal into a low-dimensional space, a low-dimensional measurement vector containing the core feature information of the signal is generated. In this process, the non-correlation characteristic of the random linear measurement enables the operation to simultaneously achieve semantic obfuscation processing of the original signal while realizing data dimensionality reduction and compression, so that the output measurement vector has natural encryption properties at the physical layer.

[0036] The secure transmission unit, whose input is connected to the compressed sensing measurement unit, is configured to encapsulate the low-dimensional measurement vector through a hardware abstraction layer to form a lightweight data frame, and asynchronously send the lightweight data frame to a remote gateway receiving unit using a low-power wireless communication link.

[0037] The gateway receiving unit is configured as a regional communication aggregation node to receive security data frames from multiple power distribution terminals deployed in different geographical locations, and to verify the integrity of the data using built-in verification logic. Then, the low-dimensional measurement vector that has passed the verification is forwarded to the data reconstruction unit on the central side through the high-speed backbone network.

[0038] The data reconstruction unit is configured to be deployed in the computing server cluster of the scheduling center. Its core logic is based on a preset generative adversarial network model. The generative adversarial network model performs nonlinear mapping and feature recovery on the received low-dimensional measurement vector through the game mechanism between the generator and the discriminator, and recovers the time-domain waveform features of the original power distribution sensing signal under high precision requirements.

[0039] The power distribution terminal data acquisition unit further includes a current transformer sensing subunit, a signal conditioning subunit, a high-speed analog-to-digital converter (ADC) subunit, and a clock synchronization subunit. The current transformer sensing subunit includes a current transformer and a voltage transformer, configured to proportionally convert high-voltage, high-current signals on the power distribution line into measurable low-voltage signals. The signal conditioning subunit, connected to the current transformer sensing subunit, includes a low-pass filter and a gain-adjustable amplifier to filter out high-frequency electromagnetic interference and adjust the signal amplitude to the operating range of the ADC. The high-speed ADC subunit is configured to employ multi-channel parallel sampling technology, with its sampling frequency set to be more than twice the highest harmonic frequency of the power distribution signal to ensure compliance with the physical basis of the Nyquist sampling theorem and to provide an oversampled data source for subsequent compressed sensing processing. The clock synchronization subunit, connected to the high-speed ADC subunit, uses satellite timing signals to provide a unified time reference, ensuring high consistency of signals acquired by different power distribution terminals on the time axis and supporting collaborative analysis of wide-area power distribution networks.

[0040] The compressed sensing measurement unit further includes a measurement matrix dynamic generation module, a sparse representation conversion module, and a linear projection calculation module. The measurement matrix dynamic generation module is configured to use a random entropy source generated by a physically unclonable function as an initial seed, combined with a preset matrix construction algorithm to dynamically generate the measurement matrix. The physically unclonable function utilizes microscopic process deviations in semiconductor device manufacturing to provide a unique hardware fingerprint for each power distribution terminal, ensuring that different terminals use different measurement matrices when processing the same original power distribution signal. This measurement matrix serves as the system's natural key, preventing attackers from mathematically reconstructing the original signal even if they intercept the transmitted measurement vector, without the corresponding hardware fingerprint-generated measurement matrix, thus achieving physical-level semantic security. The sparse representation conversion module is configured to utilize the sparsity characteristics of the power distribution signal in the wavelet domain or discrete cosine transform domain to convert the original signal into a sparse coefficient vector. Although this module does not directly perform complex transformations in actual computation, it provides a priori constraints on sparsity for the design of the measurement matrix. The linear projection calculation module is configured to perform a multiplication operation between the measurement matrix and the original power distribution sensing signal. This multiplication operation at the edge involves only basic multiply-accumulate logic, avoiding complex floating-point operations and iterative processing, and reducing the processor power consumption of the power distribution terminal.

[0041] The secure transmission unit further includes a lightweight encapsulation module and a link state awareness module. The lightweight encapsulation module is configured to split the low-dimensional measurement vector into fixed-length data blocks without increasing redundant encryption overhead, and add frame header information containing terminal identifier, sequence number, and cyclic redundancy check (CRC) code to construct a lightweight frame structure adapted to IoT communication protocols. Since the compressed sensing process has already completed information concealment, the encapsulation process omits the secondary computation of the payload data by symmetric or asymmetric encryption algorithms, effectively eliminating the computational blockage caused by complex encryption in traditional architectures. The link state awareness module is configured to monitor the signal-to-noise ratio and packet loss rate of the wireless channel in real time, and dynamically adjust the compression ratio of the compressed sensing measurement unit based on channel quality feedback, increasing the compression factor to improve transmission success rate when the channel environment deteriorates.

[0042] The gateway receiving unit further includes a concurrent processing module, a data caching module, and a priority scheduling module. The concurrent processing module is configured to simultaneously monitor multiple communication channels using a multi-threaded concurrency mechanism, processing access requests from distribution terminals at different geographical coordinates. The data caching module is configured to use a first-in, first-out (FIFO) queue structure to temporarily store the massive influx of measurement data, buffering instantaneous traffic surges caused by network jitter. The priority scheduling module is configured to perform hierarchical processing of different types of data frames based on preset power service importance labels; for example, when the original distribution sensor signal contains fault transient waveform characteristics, the priority scheduling module marks such data frames as the highest priority, prioritizing their use of bandwidth resources and allowing the data reconstruction unit to perform real-time recovery.

[0043] The data reconstruction unit further includes a model storage module, an adversarial generative reconstruction module, and a precision evaluation module. The model storage module stores parameters of the generative adversarial network trained on a large-scale historical power distribution dataset, covering signal feature distributions under various typical fault conditions. The adversarial generative reconstruction module consists of a generator hardware acceleration unit and a discriminator-assisted verification unit. The generator hardware acceleration unit takes a low-dimensional measurement vector as input and performs dimensionality-increasing operations through a multi-layer deep residual convolutional network structure, attempting to reconstruct an output signal that most closely approximates the original signal distribution. The discriminator-assisted verification unit guides the generator to learn the statistical laws of real power distribution waveforms during the training phase, and assists in evaluating the confidence level of the reconstructed signal during the online reconstruction phase. The precision evaluation module is configured to calculate the mean square error and structural similarity between the features of the reconstructed signal and the original signal. If the error exceeds a preset threshold, an online model fine-tuning mechanism is triggered.

[0044] Furthermore, the ratio of the number of rows to the number of columns in the random linear measurement matrix used by the compressed sensing measurement unit is defined as the compression ratio. The compressed sensing measurement unit is configured to dynamically adjust the compression ratio based on the signal complexity identified by the distribution terminal data acquisition unit. When the system detects that the distribution network is in a steady-state operation mode, the signal sparsity is high, and the compressed sensing measurement unit is configured to increase the compression ratio, i.e., reduce the dimension of the measurement vector, to maximize bandwidth savings. When the system detects transient events such as short-circuit faults or harmonic pollution in the distribution network, the signal sparsity decreases and contains high-frequency non-stationary components, and the compressed sensing measurement unit is configured to decrease the compression ratio and increase the measurement sampling density to ensure that the reconstructed fault waveform has sufficient resolution. This adaptive compression mechanism achieves optimal energy efficiency throughout the entire lifecycle of the distribution IoT while ensuring monitoring accuracy.

[0045] Furthermore, the generative adversarial network model of the data reconstruction unit adopts an end-to-end learning strategy during the offline training phase. During training, the system uses historical power distribution waveforms as raw samples, simulates the compressed sensing process to generate corresponding low-dimensional measurement vectors, and inputs these vectors into the generator. Simultaneously, the statistical features of the measurement matrix are used as conditional constraint vectors and injected into the generator's intermediate hidden layer. By minimizing the probability distribution difference between the raw and reconstructed samples, and by introducing a reconstruction consistency loss function related to the measurement matrix, the generator acquires robustness in recovering weak fault features under noisy interference environments. This deep learning-based reconstruction method eliminates the complex iterative optimization process in traditional compressed sensing reconstruction algorithms, shortening the delay from data reception to waveform display.

[0046] Example 2: Based on Example 1, Example 2 provides a distribution data security compression and transmission system based on the Internet of Things with a distributed edge computing architecture. This system aims to solve the problem of excessive computing load on the central server in large-scale distribution networks.

[0047] The IoT-based power distribution data secure compression and transmission system includes a power distribution terminal cluster, edge computing nodes, a core gateway, and a cloud management platform.

[0048] The distribution terminal cluster comprises several distribution terminals deployed in different transformer substations. Each distribution terminal integrates a data acquisition subunit and a physical layer compression and encryption subunit. The data acquisition subunit is configured to acquire voltage and current data from the low-voltage side of the distribution transformer using high-frequency synchronous sampling technology. The physical layer compression and encryption subunit is configured to map the original signal to a low-dimensional linear space. This mapping process uses a random measurement matrix bound to the terminal hardware serial number to ensure that the encrypted data stream output by each terminal is logically orthogonal and unpredictable.

[0049] The edge computing nodes are deployed at the edge of the distribution substation area and are connected to the distribution terminal cluster via a local area wireless network. Each edge computing node includes a preliminary reconstruction module and a data fusion module. The preliminary reconstruction module is configured to use a pre-deployed lightweight neural network model to quickly reconstruct the received measurement vectors. Since the computing power of the edge computing node is between that of the terminal and the central server, its reconstruction accuracy is set to meet the level required for initial fault diagnosis, aiming to identify the anomaly type of the line as quickly as possible. The data fusion module is configured to perform spatial correlation analysis on the preliminary reconstruction results of different nodes within the same substation area, eliminating false feature data caused by sensor faults, and achieving data noise reduction and purification at the edge.

[0050] The core gateway is configured to connect multiple edge computing nodes and includes a protocol conversion bridging module and an intelligent routing module. The protocol conversion bridging module is responsible for converting the edge-side proprietary communication protocols into standardized power IoT uplink protocols. The intelligent routing module dynamically selects the optimal transmission path based on the congestion level of the backbone network, ensuring that critical reconfiguration data can reach the cloud with minimal latency.

[0051] The cloud management platform includes a deep reconstruction server cluster and a security key management server. The deep reconstruction server cluster is configured to run a fully parameterized generative adversarial network to perform secondary fine-grained reconstruction of the feature vectors reported from the edge side, generating high-fidelity waveforms for incident tracing and power quality analysis. The security key management server is configured to store and manage the physical non-clonable function mapping table for all terminals, providing the cloud server with an accurate decompression "reference frame" during the reconstruction process.

[0052] In this embodiment, each terminal in the power distribution terminal cluster has a dynamic sparsity assessment function. The processor integrated inside the terminal is configured to perform a first-order differential operation on the original signal before executing compressed sensing operations to calculate the signal's fluctuation entropy value. If the fluctuation entropy value is lower than a preset steady-state threshold, the signal is determined to be in a highly sparse state, and the system automatically switches to an ultra-high compression mode, at which point the length of the measurement vector is reduced to less than 1 / 10 of the original signal length. If the fluctuation entropy value increases instantaneously due to load changes, the system automatically reduces the compression factor and increases the sampling dimension. This distributed sparsity adaptive strategy pushes complex analysis logic down to the terminal, further reducing the pressure on subsequent transmission and reconstruction modules.

[0053] Furthermore, this embodiment employs a multi-path parallel transmission mechanism for secure transmission. The secure transmission unit is configured to split a complete cycle of measurement vector into several sub-components and transmit them to the gateway via two different wireless standards, NB-IoT and LoRa. This physical layer diversity design ensures that even if an attacker completely cracks one of the wireless protocols, they will be unable to recover any meaningful power operation information because they cannot obtain the complete measurement components, thus providing additional security enhancement at the transmission layer.

[0054] Example 3: Based on the above examples, Example 3 describes an IoT-based power distribution data secure compression transmission system with self-healing capabilities, suitable for extreme electromagnetic interference environments.

[0055] The IoT-based power distribution data secure compression and transmission system includes an anti-interference acquisition unit, a fusion compression processing unit, a redundant transmission gateway, and an enhanced anti-reconfiguration server.

[0056] The anti-interference acquisition unit employs a fully differential symmetrical circuit design at the hardware level, supplemented by a metal shielding cavity with a shielding effectiveness of no less than 60 dB. It is configured to acquire high signal-to-noise ratio raw power distribution waveform signals even in environments with extremely strong electromagnetic transient interference, such as UHV substations. The anti-interference acquisition unit also includes an adaptive zero-adjustment module to compensate for zero-point drift caused by drastic fluctuations in ambient temperature, ensuring the high fidelity of the data input to the subsequent compression module.

[0057] The fusion compression processing unit is configured to employ a compressed sensing algorithm based on structured sparsity characteristics. Unlike ordinary random measurement matrices, the fusion compression processing unit internally stores a structured measurement dictionary specifically designed for the characteristic harmonics of the distribution network. This structured measurement dictionary uses common harmonic components of the distribution network as basis vectors, enabling the linear projection operation to not only complete the compression task but also simultaneously extract the basic features of the signal. This means that in the generated low-dimensional measurement vector, elements at different positions directly correspond to the energy distribution of different frequency bands, allowing for faster capture of the frequency characteristics of faults at the receiving end.

[0058] The redundant transmission gateway is configured as a gateway device with hot backup switching capability. It simultaneously maintains connections to both the terrestrial mobile communication public network and the low-Earth orbit satellite communication system. The monitoring logic within the redundant transmission gateway continuously detects the bit error rate of the current link. When it detects that the terrestrial base station is damaged or maliciously interfered with, causing the data frame retransmission rate to exceed a preset 30%, the gateway is configured to automatically switch the compressed, secure data frames to the satellite link for transmission without interrupting the data flow. To adapt to the bandwidth-limited characteristics of the satellite link, the gateway sends a feedback command to the front-end fusion compression processing unit, forcing a higher proportion of feature compression.

[0059] The enhanced adversarial reconstruction server incorporates a transfer learning mechanism in its software architecture. When running a generative adversarial network (GAN) model, the server is configured to automatically load corresponding pre-trained weight branches based on the load characteristics of the current power distribution area (e.g., a higher proportion of industrial load or a higher proportion of residential load). This model specialization for specific load scenarios makes the reconstructed waveforms more consistent with actual physical processes in terms of detail and texture. For example, for nonlinear industrial loads, the reconstruction model can more accurately reproduce the spike pulses and gaps in the current waveform.

[0060] Furthermore, the enhanced adversarial reconstruction server includes a security assessment submodule. This submodule is configured to simulate various known plaintext attacks and brute-force methods to perform security stress tests on the received measurement vectors. If the security assessment submodule determines that the security of the current measurement matrix is ​​at risk of exposure due to a lack of updates over a long period, it will trigger a new random seed through a secure channel to regenerate the physical non-cloning function on the terminal side. This completes the dynamic rolling update of the system's "natural key," ensuring the long-term security of the system.

[0061] Furthermore, the system in this embodiment also possesses a closed-loop feedback function for reconstruction quality. After completing one round of waveform reconstruction, the enhanced adversarial reconstruction server regenerates the reconstructed waveform using a virtual "simulated compressor" to generate a simulated measurement vector, and then performs a difference calculation between this vector and the actually received measurement vector. The magnitude of this difference serves as a direct indicator for evaluating the reconstruction quality. If the difference continues to increase, the system determines that there is a mismatch between the current terminal's compression matrix and the server's decompression matrix, and automatically initiates a synchronization calibration process.

[0062] Example 4: This example describes an IoT-based power distribution data secure compression transmission system applied in distributed photovoltaic access scenarios, which is specifically optimized for the signal complexity caused by bidirectional power flow.

[0063] The IoT-based power distribution data secure compression transmission system includes a bidirectional energy flow monitoring unit, a joint compression sensing unit, a heterogeneous secure communication unit, and a collaborative reconfiguration center.

[0064] The bidirectional energy flow monitoring unit is configured to be installed at the distributed photovoltaic grid connection point and includes a sensor array capable of distinguishing the direction of active power and the distribution of reactive power. Because the harmonics generated by the photovoltaic inverter and the inherent harmonics of the distribution network are nonlinearly superimposed, the bidirectional energy flow monitoring unit is configured to capture these minute waveform distortions with extremely high sampling depth.

[0065] The joint compressed sensing unit is configured to not only project the time-domain waveform but also perform feature masking in the frequency domain. Its internal logic module is configured to dynamically construct a measurement matrix with a specific frequency-domain window function based on the fluctuations in photovoltaic output. This matrix interweaves the unique texture features of photovoltaic power generation with the background signal of the power distribution network, making it impossible for unauthorized third parties to distinguish whether the data originates from the photovoltaic side or the grid side after interception, thus protecting the operational privacy of distributed power sources.

[0066] The heterogeneous secure communication unit is configured to support complementary transmission of power line carrier communication and wireless private network communication. In urban power distribution networks, the heterogeneous secure communication unit preferentially uses power line carrier communication to transmit the compressed measurement vector along the power lines to the concentrator at the distribution transformer; when the carrier communication is interfered with by the start-up and shutdown of high-power electrical equipment, the internal network 5G slicing technology is automatically activated for compensation. This redundancy at the physical medium level ensures that the compressed core feature data can arrive at the reconfiguration node on time in complex power distribution environments.

[0067] The collaborative reconstruction center is configured to employ a knowledge graph-guided generative adversarial network. Internally, the center constructs a knowledge graph of the distribution network topology. The reconstruction process relies not only on measurement vectors uploaded by individual terminals but also on voltage and current constraints with neighboring nodes. A special "physical constraint term" is incorporated into the generator's loss function. This constraint term is configured to verify the reconstructed waveform based on Kirchhoff's current law and voltage law. If the reconstruction result violates basic electrical engineering principles, the discriminator will assign a very low score, forcing the generator to readjust its output to ensure that the reconstructed data is mathematically sound and physically accurate.

[0068] In this embodiment, to further improve the real-time performance of the reconstruction, the collaborative reconstruction center employs a hardware-based neural network inference engine. This neural network inference engine, based on an application-specific integrated circuit (ASIC) architecture, transforms complex matrix convolution operations into simple lookup table operations and fixed-point shift operations. Compared to general-purpose processors, this hardware inference engine reduces the processing time for a single measurement vector by orders of magnitude, meeting the ultra-real-time response requirements of distribution network relay protection.

[0069] Example 5: To address the maintenance difficulties caused by large-scale sensor deployment, this example provides an IoT-based power distribution data secure compression transmission system with self-diagnostic and adaptive characteristics.

[0070] The IoT-based power distribution data secure compression and transmission system includes intelligent sensing terminals, adaptive transmission gateways, and self-evolving reconfiguration servers.

[0071] The intelligent sensing terminal integrates a sensor health monitoring module. This module is configured to monitor the hardware status parameters of the data acquisition unit in real time, including fluctuations in the power supply voltage, the operating temperature of the core chip, and the leakage current of the sample-and-hold circuit. When the monitoring module detects that hardware aging has led to a decrease in the signal-to-noise ratio of the original signal, it automatically notifies the backend compressed sensing measurement unit to adjust the structure of the measurement matrix and add redundant measurement dimensions to compensate for the accuracy loss caused by the attenuation of the front-end hardware.

[0072] The adaptive transmission gateway is configured as an intelligent routing device with data importance assessment capabilities. It integrates a resource scheduling engine based on a Markov decision process. This engine is configured to analyze the current network bandwidth availability and the latency sensitivity of the data to be transmitted in real time. For low-dimensional measurement vectors identified as short-circuit faults, the gateway is configured to perform preemptive transmission using the highest power level and the most reliable encoding method; while for routine power quality monitoring data, it is configured to perform asynchronous transmission during network idle periods and use a higher compression ratio to save energy.

[0073] The self-evolving reconstruction server is configured as an intelligent platform with continuous learning capabilities. It not only runs the current generative adversarial network but also continuously performs "online incremental training" on the model in the background using newly collected, manually verified real fault waveforms. As the system's runtime increases, the server accumulates a wider range of samples, and the reconstruction accuracy for rare faults (such as arcing grounding and lightning overvoltage) improves accordingly. Furthermore, the server is configured to automatically switch model parameters based on changes in the statistical characteristics of power distribution signals under different seasons and climatic conditions, ensuring that the system maintains optimal compression and transmission performance throughout the year.

[0074] In terms of security, the system in this embodiment introduces a blockchain-based measurement matrix auditing mechanism. Each time the compressed sensing measurement unit generates a new measurement matrix seed, it uploads its hash value to the distributed power blockchain ledger. When the data reconstruction unit performs signal recovery, it first checks whether the hash value in the ledger matches the currently used matrix. This mechanism prevents unauthorized nodes from maliciously replacing the measurement matrix to launch man-in-the-middle attacks, achieving end-to-end security and trust without relying on a centralized certification authority.

[0075] Furthermore, the self-evolving reconfiguration server also includes a fault early warning module. This module is configured to perform deep pattern recognition on the reconfigured high-fidelity waveform. Once a weak partial discharge pulse matching the characteristics of an insulation defect is detected in the waveform, even if the operating parameters of the power system are still within the normal range, the system will automatically issue an early warning command. This command is sent to the corresponding distribution terminal via an adaptive transmission gateway, instructing the terminal to temporarily stop compressed sensing operations and instead perform full-rate, full-length original waveform capture to provide the most detailed raw data support for subsequent accurate diagnosis.

[0076] Example 6: This example describes a system configuration for a miniaturized monitoring node with extremely limited resources, which minimizes hardware cost and power consumption while maintaining basic security.

[0077] The IoT-based power distribution data secure compression transmission system includes miniaturized sensing components, a simplified compression logic unit, a low-pulse transmission module, and a lightweight cloud reconfiguration engine.

[0078] The miniaturized sensing component employs an integrated microelectromechanical system (MEMS) current sensor, which can be directly mounted on the sheath of a power distribution cable in a very small size. This component is configured to extract energy using ambient magnetic fields, requiring no external power supply; therefore, the processor's power consumption requirement is at the nanowatt level.

[0079] The simplified compression logic unit is configured for random measurements implemented entirely based on bitwise operations. It does not involve any multiplier circuitry, but instead employs a sparse binary measurement matrix. In this configuration, the random linear measurement operation is simplified to the selective accumulation or ignoring of the original signal sample values. While the reconstruction performance of the binary measurement matrix is ​​mathematically slightly inferior to that of the Gaussian measurement matrix, it reduces the size of the logic gate array, enabling even the most basic microcontrollers to perform compression and physical layer encryption in real time.

[0080] The low-pulse transmission module is configured to use ultra-wideband incoherent communication technology. Instead of transmitting continuous carrier signals, it carries compressed data bits by sending extremely short time pulses. This transmission method has an extremely low power spectral density, making it difficult to detect by eavesdropping devices and effectively resisting strong narrowband interference within the power plant.

[0081] The lightweight cloud reconstruction engine is configured to be deployed within a cloud container and employs a pruned generative adversarial network (GAN) model. This model reduces computational cost by removing redundant convolutional channels, sacrificing only a small amount of accuracy. This allows the system to support tens of thousands of micro-monitoring nodes at extremely low cloud rental costs. Furthermore, the cloud reconstruction engine features multi-node collaborative completion capabilities. When a micro-node experiences data loss due to occlusion, the engine can intelligently fill in the missing parts using reconstruction information from neighboring nodes, combined with a spatiotemporal correlation model, ensuring the continuity of the monitoring footage.

[0082] In this embodiment, the low-dimensional vector generated by the simplified compression logic unit has a fixed bit length, which facilitates underlying memory management. Meanwhile, to prevent unstable output of physically non-clonable functions under extremely low power consumption, the system also introduces lightweight soft-decision error correction logic to ensure that the generated measurement matrix sequence maintains high consistency even under voltage fluctuations, guaranteeing the correctness of backend reconstruction.

[0083] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the invention. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention. Several variations and improvements can be made without departing from the concept of the present invention, and these all fall within the scope of protection of the present invention. Contents not described in detail in this specification are prior art known to those skilled in the art.

Claims

1. A secure compressed transmission system for power distribution data based on the Internet of Things, characterized in that, It includes a power distribution terminal data acquisition unit, a compressed sensing measurement unit, a secure transmission unit, a gateway receiving unit, and a data reconstruction unit; The power distribution terminal data acquisition unit is configured to sample voltage, current and fault transient waveforms in the power distribution network to obtain the original power distribution sensing signals. The compressed sensing measurement unit is connected to the power distribution terminal data acquisition unit and is configured to perform random linear measurement operations on the original power distribution sensing signal to generate a low-dimensional measurement vector. The secure transmission unit is connected to the compressed sensing measurement unit and is configured to encapsulate the low-dimensional measurement vector into a lightweight data frame and transmit it through a wireless communication link. The gateway receiving unit is configured to receive data frames from multiple power distribution terminals and forward the low-dimensional measurement vector carried in the data frames. The data reconstruction unit is configured to reconstruct the signal from the low-dimensional measurement vector based on a pre-trained generative adversarial network model, thereby recovering the time-domain waveform characteristics of the original power distribution sensing signal.

2. The IoT-based power distribution data secure compression and transmission system according to claim 1, characterized in that, The power distribution terminal data acquisition unit includes: The current transformer sensing subunit is configured to convert high voltage and high current signals on the power distribution line into low voltage signals according to a preset ratio. The signal conditioning subunit is connected to the current transformer sensing subunit and integrates a low-pass filter and a gain-adjustable amplifier. It is configured to filter out high-frequency electromagnetic interference and adjust the signal amplitude to the operating range of the analog-to-digital converter. The high-speed analog-to-digital conversion subunit is configured to use multi-channel parallel sampling technology, with its sampling frequency set to a preset multiple higher than the highest harmonic frequency of the power distribution signal to generate an oversampled data source. The clock synchronization subunit is connected to the high-speed analog-to-digital conversion subunit and is configured to receive external timing signals and provide a unified time reference for the sampling process, ensuring that power distribution terminals in different geographical locations remain consistent on the time axis.

3. The IoT-based secure compressed transmission system for power distribution data as described in claim 1, characterized in that, The compressed sensing measurement unit includes: The measurement matrix dynamic generation module is configured to use a random entropy source generated by a physically non-clonable function as an initial seed, and dynamically generate the measurement matrix in combination with matrix construction logic. The physical unclonable function utilizes the micro-process deviations in the semiconductor device manufacturing process to provide a unique hardware fingerprint for each power distribution terminal, so that the measurement matrices corresponding to different terminals are orthogonal. A sparse representation conversion module is configured to provide sparsity prior constraints for the design of the measurement matrix based on the sparsity characteristics of the power distribution signal in the transform domain. The linear projection calculation module is configured to perform the multiplication operation between the measurement matrix and the original power distribution sensing signal. By projecting the high-dimensional signal into a low-dimensional space, it achieves data dimensionality reduction and compression while simultaneously realizing semantic obfuscation of the original signal.

4. The IoT-based secure compressed transmission system for power distribution data as described in claim 1, characterized in that, The compression sensing measurement unit has an adaptive compression function and is configured to dynamically adjust the compression ratio according to the signal complexity identified by the power distribution terminal data acquisition unit. When the system detects that the distribution network is in a steady-state operation mode and the signal fluctuation entropy value is lower than the preset steady-state threshold, the compressed sensing measurement unit increases the compression ratio, so that the length of the low-dimensional measurement vector is less than 1 / 10 of the total number of original signal sampling points; When the system detects a short-circuit fault or harmonic pollution in the distribution network that causes a sudden increase in the signal fluctuation entropy value, the compressed sensing measurement unit reduces the compression ratio and increases the dimension of measurement sampling to improve the resolution of the fault transient waveform.

5. The IoT-based secure compressed transmission system for power distribution data as described in claim 1, characterized in that, The secure transmission unit includes: A lightweight encapsulation module is configured to split the low-dimensional measurement vector into fixed-length data blocks and add frame header information containing terminal identifier, serial number and cyclic redundancy check code to construct a lightweight frame structure adapted to the Internet of Things communication protocol. The link status awareness module is configured to monitor the signal-to-noise ratio and packet loss rate of the wireless channel in real time, and adjust the compression parameters according to the compression sensing measurement unit as instructed by the channel quality feedback command. The secure transmission unit omits independent symmetric or asymmetric encryption modules in the processing flow, and its security is inherently provided by the incoherent measurement mechanism in the compressed sensing process.

6. The IoT-based power distribution data secure compression and transmission system according to claim 1, characterized in that, The gateway receiving unit includes: The concurrent processing module is configured to use a multi-threaded concurrency mechanism to listen to multiple communication channels simultaneously and process access requests from different power distribution terminals. The data caching module is configured to temporarily store the received measurement data using a first-in-first-out queue structure to buffer the instantaneous traffic surges caused by network jitter. The priority scheduling module is configured to classify data frames according to the importance of power services. When the data frame contains fault transient waveform characteristics, the priority scheduling module marks such data frames as the highest priority, prioritizes their use of bandwidth resources, and forwards them to the data reconstruction unit.

7. The IoT-based secure compressed transmission system for power distribution data as described in claim 1, characterized in that, The data reconstruction unit includes: The model storage module is configured to store the parameters of a generative adversarial network trained on a historical power distribution fault waveform dataset. The adversarial generation and reconstruction module includes a generator hardware acceleration unit and a discriminator-assisted verification unit; The generator hardware acceleration unit is configured to perform dimensionality upscaling operations on the low-dimensional measurement vector through a multi-layer deep residual convolutional network structure. The discriminator-assisted verification unit is configured to evaluate the consistency of the reconstructed signal with the actual power distribution waveform during the training phase, and to assist in evaluating the confidence level of the reconstructed signal during the online reconstruction phase. The accuracy evaluation module is configured to calculate the mean square error between the features of the reconstructed signal and the original signal, and to trigger an online model fine-tuning mechanism when the mean square error exceeds a preset threshold.

8. The IoT-based secure compressed transmission system for power distribution data according to claim 1, characterized in that, The system also includes an edge computing node deployed at the edge of the distribution area. The edge computing node is configured between the distribution terminal and the gateway receiving unit. It includes a preliminary reconstruction module configured to use a lightweight neural network model to perform preliminary recovery of the received low-dimensional measurement vector in order to identify the anomaly type of the line. The data fusion module is configured to perform spatial correlation analysis on the preliminary reconstruction results of different power distribution terminals within the same distribution area, and to achieve data noise reduction and purification on the edge side by eliminating pseudo-feature data.

9. The IoT-based secure compressed transmission system for power distribution data according to claim 1, characterized in that, The secure transmission unit is configured with redundant transmission paths, including a first wireless communication link and a second wireless communication link. The secure transmission unit is configured to split the low-dimensional measurement vector of the complete cycle into several sub-components, and transmit them in parallel through the first wireless communication link and the second wireless communication link, respectively. When the retransmission rate of the first wireless communication link is detected to exceed a preset ratio, the secure transmission unit automatically switches the data stream to the second wireless communication link and instructs the compressed sensing measurement unit to perform a higher ratio of feature compression to adapt to the bandwidth limitations of the second wireless communication link.

10. The IoT-based secure compressed transmission system for power distribution data according to any one of claims 1 to 9, characterized in that, The system also integrates the following security and optimization function modules: The blockchain auditing subsystem is configured to upload the hash value of the compressed sensing measurement unit to the distributed power blockchain ledger each time a new measurement matrix seed is generated, so that the data reconstruction unit can verify the legality of the measurement matrix before the signal is restored. A hardware-based neural network inference engine is deployed in the data reconstruction unit and configured to convert the matrix convolution operation of the generative adversarial network model into lookup table operations and fixed-point shift operations to reduce reconstruction latency. The physical constraint verification module is configured to evaluate the physical consistency of the reconstructed waveform based on Kirchhoff's current law and voltage law, and to force the reconstruction result to conform to the basic principles of electrical engineering by correcting the loss function term of the generative adversarial network model.

Images