Time series based sensor data compression transmission method and system

By compressing and projecting sensor data and extracting features, real-time monitoring and efficient transmission of abnormal events under data compression are achieved, solving the problems of high computational overhead and information loss in existing technologies, and ensuring transmission efficiency and complete analysis of abnormal events.

CN122513484BActive Publication Date: 2026-11-10GUANGDONG OPEN UNIV (GUANGDONG POLYTECHNIC VOCATIONAL COLLEGE)
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202611009915.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2026-07-08
Publication Date
2026-11-10
Estimated Expiration
2046-07-08

AI Technical Summary

Technical Problem

Existing sensor data compression and transmission methods incur high computational overhead during anomaly detection and are difficult to implement in real time. Furthermore, conventional methods lack the ability to cache and backtrack pre-event data, resulting in the loss of contextual information during the anomaly development process.

Method used

The current signal of the power distribution network line is collected by the sensor, compressed sensing projection is performed, compressed domain features are extracted, transient disturbance events are identified, and only the compressed domain features are transmitted when no event is triggered. When the event is triggered, detailed data layers before and after the event are transmitted. The receiving end decompresses and reconstructs the data for analysis.

Benefits of technology

It enables real-time monitoring of abnormal events under data compression, reducing computational overhead and transmission latency, as well as communication energy consumption and bandwidth usage, while ensuring complete information transmission and accurate analysis of abnormal events.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122513484B_ABST
    Figure CN122513484B_ABST
Patent Text Reader

Abstract

The application provides a time series-based sensor data compression transmission method and system, and belongs to the field of data compression transmission; through compression sensing projection of time series data in each time window, detail data layers retaining the sparse structure of the signal are obtained, statistical feature extraction is performed on the detail data layers to obtain compression domain features in each time window; the compression domain features are read to determine the condition of triggering a transient disturbance event; if the transient disturbance event is not triggered, only the compression domain features of the current time window are sent to the receiving end, otherwise, the detail data layers and the corresponding compression domain features of a predetermined time length before and after the event point are sent with high priority; the receiving end monitors the received compression domain features. By using the scheme of the application, real-time monitoring of abnormal events can be directly performed in the data compression state, mixed compression transmission of complete information of abnormal events is realized, and transmission efficiency is ensured.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This application relates to the field of data compression and transmission technology, and more specifically, to a method and system for time-series-based sensor data compression and transmission. Background Technology

[0002] Sensors are the core devices of the sensing layer of the Industrial Internet of Things (IIoT) and are widely deployed in scenarios such as power equipment condition monitoring, rotating machinery fault diagnosis, and structural health monitoring. With the continuous improvement of sensor sampling frequency and accuracy, the amount of time-series data generated by a single sensor node has increased dramatically. The storage and transmission of massive amounts of data pose a severe challenge to communication bandwidth and sensor node energy consumption.

[0003] Existing sensor data compression and transmission methods, such as dead-zone compression, rotating door algorithms, or predictive residual coding, require complete decompression and reconstruction of the compressed data before basic analyses such as anomaly detection can be performed. This incurs high computational costs and makes it difficult to perform real-time monitoring directly in the compressed domain. Furthermore, conventional event-driven compression methods only switch to high-fidelity transmission after an anomaly is detected, lacking the ability to cache and backtrack key precursory data before the event occurs, leading to the loss of contextual information during the anomaly's development. Therefore, how to perform real-time monitoring of anomaly events directly in a compressed state, achieving hybrid compressed transmission of complete anomaly event information while ensuring transmission efficiency, has become a challenge for the industry. Summary of the Invention

[0004] This application provides a time-series-based sensor data compression and transmission method and system, which can directly perform real-time monitoring of abnormal events under data compression, so as to realize the mixed compression and transmission of complete information of abnormal events and ensure transmission efficiency.

[0005] In a first aspect, this application provides a time-series-based sensor data compression and transmission method, comprising the following steps:

[0006] By collecting current signals from power distribution network lines using sensors, continuous time-series data of transient current waveforms can be obtained.

[0007] Compressed sensing projection is performed on the time series data within each time window to obtain a detailed data layer that preserves the sparse structure of the signal. Statistical feature extraction is performed on the detailed data layer to obtain the compressed domain features of the current state within each time window.

[0008] The sensor reads the compressed domain features to determine the conditions that trigger the transient disturbance event.

[0009] If no transient disturbance event is triggered, only the compressed domain features of the current time window are sent to the receiving end; otherwise, the context transmission mode is triggered, and the detailed data layer and the corresponding compressed domain features for a predetermined time length before and after the event point are sent with high priority.

[0010] The receiving end monitors the characteristics of the received compressed domain. When it receives a complete data packet triggered by a transient disturbance event, it decompresses and reconstructs the relevant detailed data layer to backtrack and analyze the entire event process.

[0011] In some embodiments, acquiring current signals from power distribution network lines using sensors to obtain continuous time-series data of transient current waveforms specifically includes:

[0012] The current signal of the power distribution network line is continuously sampled by a sensor and then converted from analog to digital to obtain a current sequence.

[0013] The current sequence is continuously divided according to a preset time window length to form time-series data of the transient current waveform in continuous time.

[0014] In some embodiments, performing compressed sensing projection on the time-series data within each time window to obtain a detailed data layer that preserves the sparse structure of the signal specifically includes:

[0015] A random measurement matrix is ​​generated using a random seed shared with the receiver;

[0016] A sparse transformation is applied to the time series data within each time window to obtain multiple sparse vectors;

[0017] The measurement matrix is ​​used to compress and project all sparse vectors to obtain low-dimensional measurement value vectors for each time window.

[0018] All measured value vectors are used as detailed data layers that preserve the sparse structure of the signal within the corresponding time window, and stored in the sensor's sliding buffer in chronological order.

[0019] In some embodiments, performing statistical feature extraction on the detailed data layer to obtain compressed domain features of the current state within each time window specifically includes:

[0020] For the measurement value vector in the detailed data layer, without reconstructing the original signal, its element-squared sum is directly calculated as the energy feature of the current waveform within the current window;

[0021] Determine the sparsity estimate of the measurement vector in the detailed data layer;

[0022] The energy characteristics and the sparsity estimates are used together as the compressed domain characteristics of the current state within each time window.

[0023] In some embodiments, the sensor reads the compressed domain features to determine the conditions that trigger the transient disturbance event, specifically including:

[0024] Obtain the baseline energy range and baseline sparsity range of the current waveform established within the historical time window;

[0025] The sensor reads the compressed domain features and compares the energy features in the compressed domain features of the current window with the baseline energy range;

[0026] Compare the sparsity estimate in the compressed domain features of the current window with the baseline sparsity interval;

[0027] When the energy feature in the compressed domain of the current window exceeds the baseline energy range, and the sparsity estimate exceeds the baseline sparsity range, it is determined that the current time window has triggered a transient disturbance event.

[0028] In some embodiments, if no transient disturbance event is triggered, sending only the compressed domain features of the current time window to the receiving end specifically includes:

[0029] Extract the compressed domain features of the current time window and the random seed shared with the receiver;

[0030] The compressed domain features of the current time window are encapsulated into a lightweight state packet, and the random seed is placed in a fixed field in the packet header;

[0031] Only the lightweight status packet is sent to the receiving end.

[0032] In some embodiments, triggering a context transmission mode to send detailed data layers and corresponding compressed domain features for a predetermined time period before and after the event point with high priority specifically includes:

[0033] Extract all detailed data layers within a predetermined time period prior to the event from the sensor's sliding buffer.

[0034] Get the detailed data layer of the current time window and concatenate it with all the detailed data layers within the predetermined time length in chronological order;

[0035] The random seed, the spliced ​​detail data layer, and the corresponding compressed domain features are encapsulated into a high-priority data frame, which is then transmitted to the receiving end via an independent control channel or a priority transmission queue.

[0036] Secondly, this application provides a time-series-based sensor data compression and transmission system, comprising:

[0037] The acquisition module is used to acquire current signals from power distribution network lines through sensors to obtain continuous time-series data of current transient waveforms.

[0038] The compression module is used to perform compressed sensing projection on the time series data within each time window to obtain a detailed data layer that preserves the sparse structure of the signal, and to extract statistical features from the detailed data layer to obtain the compressed domain features of the current state within each time window.

[0039] The judgment module is used by the sensor to read the compressed domain features in order to determine the conditions that trigger the transient disturbance event.

[0040] The transmission module is also used to send only the compressed domain features of the current time window to the receiving end if no transient disturbance event is triggered; otherwise, it triggers the context transmission mode and sends the detailed data layer and corresponding compressed domain features of a predetermined time length before and after the event point with high priority.

[0041] The decompression module is used by the receiving end to monitor the characteristics of the received compressed domain. When a complete data packet triggered by a transient disturbance event is received, it decompresses and reconstructs the relevant detailed data layer to backtrack and analyze the entire event process.

[0042] Thirdly, this application provides a computer device, the computer device including a memory and a processor, the memory storing code, the processor being configured to acquire the code and execute the above-described time-series-based sensor data compression and transmission method.

[0043] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described time-series-based sensor data compression and transmission method.

[0044] The technical solutions provided by the embodiments disclosed in this application have the following beneficial effects:

[0045] The time-series-based sensor data compression and transmission method and system provided in this application first acquires current signals from power distribution network lines using sensors to obtain continuous time-series data of transient current waveforms. The time-series data is then compressed and projected within each time window to obtain a detailed data layer that preserves the sparse structure of the signal. Statistical feature extraction is performed on the detailed data layer to obtain compressed domain features of the current state within each time window. The sensor reads the compressed domain features to determine the conditions for triggering a transient disturbance event. If no transient disturbance event is triggered, only the compressed domain features of the current time window are sent to the receiving end; otherwise, a context transmission mode is triggered, sending the detailed data layers and corresponding compressed domain features for predetermined time lengths before and after the event point with high priority. The receiving end monitors the received compressed domain features. When a complete data packet triggered by a transient disturbance event is received, the relevant detailed data layers are decompressed and reconstructed to retrospectively analyze the entire event process.

[0046] Therefore, this application's time-series-based sensor data compression and transmission method firstly extracts compressed domain features synchronously during the compressed sensing projection process, enabling the receiving end to directly complete current state monitoring and transient disturbance judgment without decompression, significantly reducing computational overhead and transmission latency under normal monitoring conditions. Secondly, when no disturbance is triggered, only lightweight compressed domain features are transmitted, significantly reducing the communication power consumption and bandwidth occupation of sensor nodes. Simultaneously, when a disturbance is triggered, the detailed data layer of the complete time length before and after the event is dynamically traced back, ensuring the reconstructability and accurate analysis of the entire abnormal process. Thus, it meets the dual requirements of long-term low-power lightweight transmission and accurate event tracing. The above scheme can directly perform real-time monitoring of abnormal events under data compression conditions, so as to realize the hybrid compressed transmission of complete abnormal event information and ensure transmission efficiency. Attached Figure Description

[0047] Figure 1 This is an exemplary flowchart of a time-series-based sensor data compression and transmission method according to some embodiments of this application;

[0048] Figure 2 This is a flowchart illustrating the transmission control of dual-branch data according to some embodiments of this application;

[0049] Figure 3 This is a schematic diagram of a time-series-based sensor data compression and transmission system according to some embodiments of this application, wherein 301 is a data acquisition module, 302 is a compression module, 303 is a judgment module, 304 is a transmission module, and 305 is a decompression module;

[0050] Figure 4 This is a schematic diagram of the structure of a computer device implementing a time-series-based sensor data compression and transmission method according to some embodiments of this application, wherein 401 is a processor, 402 is a communication bus, 403 is a memory, and 404 is a communication interface. Detailed Implementation

[0051] To better understand the technical solution of this application, the technical solution of this application will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0052] refer to Figure 1 The figure is an exemplary flowchart of a time-series-based sensor data compression and transmission method according to some embodiments of this application. The figure mainly includes the following steps:

[0053] In step S101, the current signal of the power distribution network line is collected by the sensor to obtain continuous time-series data of the transient current waveform.

[0054] In some embodiments, the acquisition of continuous time-series data of the transient waveform of the current in the power distribution network by means of a sensor can be achieved by the following steps:

[0055] The current signal of the power distribution network line is continuously sampled by a sensor and then converted from analog to digital to obtain a current sequence.

[0056] The current sequence is continuously divided according to a preset time window length to form time-series data of the transient current waveform in continuous time.

[0057] In specific implementation, firstly, the sensor continuously samples the current signal of the distribution network line at a fixed sampling frequency. The sampling frequency is set according to the highest frequency component of the transient signal of the distribution network to ensure that the Nyquist sampling theorem is satisfied. Secondly, the sampled analog current signal is converted into a digital quantity by an analog-to-digital converter, and each sampling moment corresponds to a digitized current amplitude. The current amplitudes of all sampling moments are arranged in chronological order to form a current sequence. Then, using a preset fixed time window length as the dividing unit, starting from the beginning of the current sequence, all sampling points in the current sequence that fall within the same time window are grouped into a data group. This data group contains the current amplitudes of all sampling moments within the time window. The continuous time windows are connected end to end on the time axis and do not overlap. Each time window corresponds to a data group, and all data groups are arranged in chronological order to form the time-series data of the transient current waveform in continuous time. Other embodiments may also use other methods to implement this, which are not limited here.

[0058] It should be noted that the fixed time window length preset in this application is determined based on the duration of the transient signal of the distribution network. When a ground fault or short circuit occurs in the distribution network, the duration of the current transient signal is usually in the range of several milliseconds to tens of milliseconds, of which the transient process of a stable ground fault is about 2 to 3 milliseconds. The time window length is preferably set to 10 to 30 milliseconds, corresponding to 0.5 to 1 cycle at a power frequency of 50 Hz. This range can not only fully cover most transient disturbance processes, but also ensure that each window contains enough sampling points to meet the accuracy requirements of compressed sensing reconstruction. Secondly, the current sequence is a current sequence that represents the discrete amplitude sequence of the current signal after digitization and arranged in the order of sampling time. The time series data represents the time series data of the continuous time window data group obtained by dividing the current sequence into time windows.

[0059] In step S102, compressed sensing projection is performed on the time series data within each time window to obtain a detailed data layer that preserves the sparse structure of the signal. Statistical feature extraction is performed on the detailed data layer to obtain the compressed domain features of the current state within each time window.

[0060] In some embodiments, performing compressed sensing projection on the time-series data within each time window to obtain a detailed data layer that preserves the sparse structure of the signal can be achieved using the following steps:

[0061] A random measurement matrix is ​​generated using a random seed shared with the receiver;

[0062] A sparse transformation is applied to the time series data within each time window to obtain multiple sparse vectors;

[0063] The measurement matrix is ​​used to compress and project all sparse vectors to obtain low-dimensional measurement value vectors for each time window.

[0064] All measured value vectors are used as detailed data layers that preserve the sparse structure of the signal within the corresponding time window, and stored in the sensor's sliding buffer in chronological order.

[0065] In practice, generating a random measurement matrix using a shared random seed can be achieved as follows: the sensor and the receiver negotiate and store the same random seed in advance during the communication establishment phase; the sensor uses this random seed as the initial value of the pseudo-random number generator to generate a set of random number sequences with independent and identically distributed characteristics, and arranges the random number sequence into a matrix by rows, where the number of rows of the matrix is ​​the compressed target dimension and the number of columns is the number of sampling points in the original time window; then, the columns of the matrix are orthogonalized to satisfy the constraint isometry condition, and finally the measurement matrix is ​​obtained.

[0066] In practice, applying a sparse transform to the time-series data within each time window to obtain multiple sparse vectors can be achieved in the following way: Select a Fourier transform basis or a wavelet transform basis as the sparse transform basis, and then perform matrix multiplication on the time-series data within each time window with the sparse transform basis to transform the current waveform signal in the time domain to the frequency domain or wavelet domain, thereby obtaining a sparse vector containing frequency domain coefficients or wavelet coefficients corresponding to each time window; where, since the current waveform under normal operating conditions of the distribution network is close to a standard sine wave, its energy in the frequency domain is concentrated near the fundamental frequency component, the absolute values ​​of most coefficients in this sparse vector are close to zero, and only a few coefficients have absolute values ​​significantly greater than zero, thus satisfying the sparsity requirement.

[0067] In specific implementation, the measurement matrix is ​​used to compress and project all sparse vectors to obtain low-dimensional measurement value vectors for each time window. This can be achieved by performing matrix multiplication on the measurement matrix and the sparse vectors within each time window, i.e., projecting the high-dimensional sparse vectors onto the low-dimensional measurement space to obtain low-dimensional measurement value vectors for each time window. The dimension of these measurement value vectors is much smaller than the number of sampling points in the original time window, thus achieving data compression. The compressed sensing theory guarantees that when the measurement matrix satisfies the constraint isometry condition and the sparse vectors satisfy the sparsity requirement, the measurement value vectors contain all the information required to reconstruct the original signal.

[0068] In specific implementation, all measurement value vectors are treated as detail data layers that preserve the sparse structure of the signal within the corresponding time window and stored in the sensor's sliding buffer in chronological order. This can be achieved in the following way: the low-dimensional measurement value vector corresponding to each time window is marked as a detail data layer and stored sequentially in the sensor's built-in sliding buffer in chronological order of the time window. The sliding buffer adopts a first-in-first-out queue structure, and the size of the buffer is preset to store a predetermined number of detail data layers for consecutive time windows. When the buffer is full, the earliest stored detail data layer is removed from the buffer, and the newly arrived detail data layer is stored at the end of the buffer, ensuring that the buffer always stores all detail data layers within the most recent time period. Other embodiments may also use other methods to achieve this, which are not limited here.

[0069] It should be noted that the measurement matrix in this application is a measurement matrix that projects a high-dimensional sparse vector onto a low-dimensional measurement space to achieve data compression; the sparse vector is a sparse vector representing the energy of the original current waveform concentrated in a few components in the frequency domain or wavelet domain; the measurement value vector is the measurement value vector representing the representation vector of the original signal in the low-dimensional measurement space after compressed sensing projection; compressed sensing projection is the process by which the sensor uses the measurement matrix to project high-dimensional original time-series data into a low-dimensional measurement space in the sparse transform domain, which is used to convert the high-dimensional original signal into a low-dimensional measurement value vector to achieve lossless data compression, so that the sensor can significantly reduce the amount of data storage and transmission, and reduce communication energy consumption and bandwidth occupation while maintaining reconstruction capability; the detailed data layer is a detailed data layer that retains all the signal information required to reconstruct the original current waveform in the compressed domain; it is used to retain all the signal information required to reconstruct the original current waveform in the compressed domain, so that the sensor can cache historical data with low storage overhead when no disturbance is triggered, and send the complete data before and after the event to the receiving end at one time after the event is triggered to support full-process backtracking analysis.

[0070] It should be noted that the current signal of the distribution network under normal operating conditions is highly periodic, and its waveform is close to a standard sine wave. In the frequency domain, almost all the energy is concentrated on the fundamental frequency component, and the amplitude of the high-frequency harmonic components is extremely small, which naturally meets the requirement of signal sparsity for compressed sensing. When a transient disturbance occurs, such as a sudden current change caused by a short-circuit fault, a transient component will be superimposed on the current waveform, which manifests as the simultaneous change of multiple frequency domain coefficients in the sparse transform domain. Compressed sensing theory utilizes this sparsity characteristic of the signal to project the high-dimensional signal into a low-dimensional space through a random measurement matrix, which significantly reduces the data dimensionality while ensuring the reconstruction quality. Thus, the low-dimensional measurement value vector is stored as a detailed data layer in the sliding buffer, so that the sensor only needs to store lightweight measurement values ​​instead of the original high-dimensional waveform data when no disturbance is triggered, reducing storage overhead. At the same time, the time-sequential storage structure of the sliding buffer ensures that historical data from multiple windows before the disturbance can be quickly retrieved after an event is triggered, providing complete contextual information for fault diagnosis.

[0071] In some embodiments, the statistical feature extraction of the detailed data layer to obtain the compressed domain features of the current state within each time window can be achieved by the following steps:

[0072] For the measurement value vector in the detailed data layer, without reconstructing the original signal, its element-squared sum is directly calculated as the energy feature of the current waveform within the current window;

[0073] Determine the sparsity estimate of the measurement vector in the detailed data layer;

[0074] The energy characteristics and the sparsity estimates are used together as the compressed domain characteristics of the current state within each time window.

[0075] In specific implementation, the sum of squares of the elements in the measurement value vector in the detail data layer can be directly calculated as the energy feature of the current waveform in the current window without reconstructing the original signal. This can be achieved in the following way: read the detail data layer of the current time window and obtain the low-dimensional measurement value vector contained in the detail data layer; perform a square operation on each element in the measurement value vector to obtain the square value of each element, and then sum the square values ​​of all elements. The summation result is used as the energy feature of the current waveform in the current time window. According to the constrained isometry in compressed sensing theory, when the measurement matrix satisfies the constrained isometry condition, this energy feature maintains an approximately proportional relationship with the energy of the original current waveform in the sparse transform domain, and can reliably characterize the true energy intensity of the original current waveform.

[0076] In specific implementation, the sparsity estimate of the measurement value vector in the detail data layer can be determined as follows: Read the measurement value vector in the detail data layer of the current time window and obtain the absolute value of each element in the vector; count the number of elements whose absolute value is less than a preset sparsity threshold, divide this number by the total number of elements in the measurement value vector, and use the resulting ratio as the sparsity estimate of the current time window; wherein, the sparsity estimate characterizes the proportion of insignificant elements in the measurement value vector, reflecting the degree of aggregation of coefficients of the original signal in the sparse transform domain: the larger the value, the more concentrated the energy of the signal in the transform domain, and the smaller the value, the more dispersed the energy; wherein, the preset sparsity threshold is usually set according to the statistical characteristics of the elements in the measurement value vector, for example, taking 1% to 5% of the maximum absolute value of all elements or setting it to 2 to 3 times the noise standard deviation according to the noise level, and the specific value needs to be determined by experiments or adaptive algorithms in combination with the sparsity of the actual signal and the noise level.

[0077] In specific implementation, the energy feature and the sparsity estimate can be used together as the compressed domain feature of the current state within each time window in the following way: the energy feature and the sparsity estimate are combined into a two-dimensional vector, which is the compressed domain feature of the current state within the current time window; wherein, the compressed domain feature only contains two dimensions, the energy feature and the sparsity estimate, with extremely low dimensionality, which can complete the transmission between the sensor and the receiver with a very small amount of data, and the receiver can directly determine the current state based on the compressed domain feature without performing any decompression operation; other embodiments may also use other methods, which are not limited here.

[0078] It should be noted that the energy characteristics in this application are energy characteristics that characterize the energy intensity of the current waveform within the current time window without reconstructing the original signal; the sparsity estimate is a sparsity estimate that characterizes the degree of energy concentration of the current signal in the sparse transform domain within the current time window; and the compressed domain characteristics are compressed domain characteristics that characterize the current state in the compressed domain to support direct monitoring without decompression. They are used to characterize the state of the current waveform within the current time window without reconstructing the original signal, so that the receiver can complete the real-time monitoring of the current state of the distribution network and the preliminary judgment of transient disturbance events under normal conditions without decompression.

[0079] It should be noted that in compressed sensing theory, when the measurement matrix satisfies the constrained isometry, the L2 norm of the measured value vector approximately preserves the L2 norm of the original signal in the sparse transform domain. For current waveforms, their energy in the wavelet or Fourier transform domain is positively correlated with the signal's effective value. Therefore, this step directly utilizes the energy characteristics calculated by the detail data layer, which can reliably characterize the intensity of the current waveform within the current window without reconstructing the signal, thereby achieving disturbance event perception in the compressed domain. In traditional methods, both calculating the current's effective value and analyzing spectral characteristics require processing on the reconstructed original signal, resulting in high computational cost and delay. This scheme, however, utilizes the constrained isometry of compressed sensing to directly establish the signal energy characterization on the measured value vector, obtaining the signal energy without waiting for the reconstruction process. The energy characteristics of the original signal are approximately proportional to its energy. Simultaneously, by statistically measuring the proportion of elements in the vector whose absolute value is below a threshold, the sparsity of the signal can be estimated in the compressed domain. This estimate is highly sensitive to changes in signal structure: under normal operating conditions, the current waveform spectrum is highly concentrated, and the sparsity estimate is close to 1; when transient disturbances occur, the waveform deviates from a sinusoidal shape, the spectral energy is dispersed, and the sparsity estimate decreases significantly. Thus, under constrained equidistant conditions, these two characteristics have a definite approximate proportional relationship with the signal's energy and sparse structure. Furthermore, combining the energy characteristics and the sparsity estimate forms compressed domain characteristics, enabling the receiver to directly perceive and judge transient disturbance events using these two indicators without relying on the reconstruction results, reducing the required computation and latency.

[0080] In step S103, the sensor reads the compressed domain features to determine the conditions that trigger the transient disturbance event.

[0081] In some embodiments, the sensor reads the compressed domain features to determine the conditions that currently trigger a transient disturbance event, which can be achieved through the following steps:

[0082] Obtain the baseline energy range and baseline sparsity range of the current waveform established within the historical time window;

[0083] The sensor reads the compressed domain features and compares the energy features in the compressed domain features of the current window with the baseline energy range;

[0084] Compare the sparsity estimate in the compressed domain features of the current window with the baseline sparsity interval;

[0085] When the energy feature in the compressed domain of the current window exceeds the baseline energy range, and the sparsity estimate exceeds the baseline sparsity range, it is determined that the current time window has triggered a transient disturbance event.

[0086] In specific implementation, the baseline energy range and baseline sparsity range of the current waveform established by the historical time window can be obtained in the following way: During the training phase, the sensor collects current waveform data of multiple time windows under normal operation of the distribution network, and calculates the energy characteristics and sparsity estimates of each time window; the minimum and maximum values ​​of the energy characteristics under all normal conditions are taken, and the minimum and maximum values ​​are used as endpoints to form the baseline energy range; the minimum and maximum values ​​of the sparsity estimates under all normal conditions are taken, and the minimum and maximum values ​​are used as endpoints to form the baseline sparsity range; the baseline energy range and baseline sparsity range are stored in the sensor's local memory.

[0087] In specific implementation, the sensor reads the compressed domain features and compares the energy features in the compressed domain features of the current window with the baseline energy range. This can be achieved in the following way: the sensor reads the lower limit and upper limit of the baseline sparsity range from the local memory; the sensor reads the compressed domain features of the current time window and extracts the sparsity estimate from these features; the extracted sparsity estimate is compared with the lower limit and upper limit of the baseline sparsity range respectively: if the sparsity estimate is less than the lower limit or greater than the upper limit, it is determined that the sparsity estimate exceeds the baseline sparsity range; if the sparsity estimate is greater than or equal to the lower limit and less than or equal to the upper limit, it is determined that the sparsity estimate does not exceed the baseline sparsity range; thus, when both conditions are met—the energy feature exceeding the baseline energy range and the sparsity estimate exceeding the baseline sparsity range—the sensor determines that the current time window triggers a transient disturbance event; when either condition is not met, the sensor determines that the current time window does not trigger a transient disturbance event. Other embodiments may also use other methods, which are not limited here.

[0088] It should be noted that the baseline energy range represents the normal fluctuation range of the current waveform energy characteristics under normal operating conditions of the distribution network; the baseline sparsity range represents the normal fluctuation range of the estimated sparsity value of the current waveform under normal operating conditions of the distribution network; and the transient disturbance event represents a transient disturbance event that occurs in the distribution network line, causing the current waveform to deviate from the normal operating conditions. Among these, the interval comparison method is used, and the baseline energy range and the baseline sparsity range are used together to characterize the allowable range of the normal operating conditions. A disturbance event is only determined to be triggered when the energy characteristics and the estimated sparsity value of the current window both exceed their respective baseline ranges, which effectively reduces false alarms caused by normal load fluctuations.

[0089] In step S104, if no transient disturbance event is triggered, only the compressed domain features of the current time window are sent to the receiving end; otherwise, the context transmission mode is triggered, and the detailed data layer and the corresponding compressed domain features for a predetermined time length before and after the event point are sent with high priority.

[0090] It should be noted that the reference Figure 3 As shown, this figure is a flowchart of the transmission control of dual-branch data in some embodiments of this application; the figure shows the dual-mode data interaction process between the acquisition terminal and the edge server: the left side of the figure is a sensor, which determines whether a transient disturbance event is triggered. If no transient disturbance event is triggered, only the compressed domain feature is transmitted to the receiving end; if a transient disturbance event is triggered, the detailed data layer before and after the event point for a predetermined period of time is sent to the receiving end together with the corresponding compressed domain feature.

[0091] In some embodiments, if no transient disturbance event is triggered, sending only the compressed domain features of the current time window to the receiving end can be achieved using the following steps:

[0092] Extract the compressed domain features of the current time window and the random seed shared with the receiver;

[0093] The compressed domain features of the current time window are encapsulated into a lightweight state packet, and the random seed is placed in a fixed field in the packet header;

[0094] Only the lightweight status packet is sent to the receiving end.

[0095] In specific implementation, the extraction of compressed domain features of the current time window and the random seed shared with the receiver can be achieved in the following way: After determining that no transient disturbance event has been triggered in the current time window, the sensor reads the compressed domain features corresponding to the current time window from the local cache. The compressed domain features contain two dimensions: energy features and sparsity estimates. At the same time, the sensor reads the shared random seed, which has been pre-negotiated and stored with the receiver, from the local memory. Next, according to the data packet format pre-agreed between the sensor and the receiver, the compressed domain features of the current time window are written into the payload area of ​​the data packet, and the shared random seed is written into the fixed field position of the data packet header to form a lightweight state packet. The total length of the lightweight state packet is determined by adding the fixed length of the packet header to the length of the compressed domain features, where the compressed domain features contain only two values. Finally, the sensor sends the encapsulated lightweight state packet to the receiver through the wireless communication interface. The lightweight state packet contains only the compressed domain features and the random seed, and does not carry any detailed data layer data of the time window. After the transmission is completed, the sensor continues to enter the data acquisition and processing flow of the next time window. Other embodiments can also be implemented in other ways, which are not limited here.

[0096] It should be noted that the lightweight status packet is used to report the current status to the receiver with minimal data volume during normal operation; the random seed is used to enable the receiver to independently recover the same measurement matrix as the sensor when needed. If the sensor transmits the complete detailed data layer in every time window, it will consume a large amount of communication bandwidth and power. This solution, however, transmits only compressed domain features—two values ​​plus a random seed—when no disturbance is triggered. The data volume is typically only tens of bytes, significantly reducing data transmission volume compared to transmitting the complete window of detailed data, thus greatly reducing the sensor's communication power consumption and bandwidth usage. Although the receiver cannot obtain the complete current waveform after receiving the lightweight status packet, it can still maintain continuous awareness of the distribution network's current status through the energy and sparsity information in the compressed domain features. The purpose of sending the random seed with the packet is that the receiver continuously updates the random seed during normal reception of the lightweight status packet, ensuring that the random seed of the sensor and the receiver remains consistent when an event is triggered, without requiring an additional seed synchronization process. This minimizes the data transmission volume under normal conditions, transmitting complete data only when necessary.

[0097] In some embodiments, triggering a context transmission mode to send detailed data layers and corresponding compressed domain features for a predetermined time period before and after the event point with high priority can be achieved through the following steps:

[0098] Extract all detailed data layers within a predetermined time period prior to the event from the sensor's sliding buffer.

[0099] Get the detailed data layer of the current time window and concatenate it with all the detailed data layers within the predetermined time length in chronological order;

[0100] The random seed, the spliced ​​detail data layer, and the corresponding compressed domain features are encapsulated into a high-priority data frame, which is then transmitted to the receiving end via an independent control channel or a priority transmission queue.

[0101] In practice, firstly, after determining that a transient disturbance event has been triggered in the current time window, the sensor records the index of the current time window as the event trigger time. Based on a preset backtracking time length, the sensor calculates the number of time windows that need to be extracted before the event trigger time. The sensor then reads the detailed data layers of these time windows from the sliding buffer. These detailed data layers are arranged chronologically from earliest to latest, covering the complete historical data within the predetermined time length before the event trigger. Secondly, the sensor acquires the detailed data layers of the currently triggered time window and concatenates all the extracted historical detailed data layers with the detailed data layers of the current time window in chronological order into a continuous detailed data sequence. The starting position of this concatenated sequence is the event trigger time. The start time of the predetermined time before triggering and the end position of the event triggering time completely cover the historical data window before the transient disturbance event and the current data window when the event occurs. Finally, the sensor reads the shared random seed stored locally, writes the random seed into the fixed field of the frame header according to the high-priority data frame format agreed upon by the sensor and the receiver, writes the spliced ​​detailed data sequence into the payload area of ​​the frame, and writes the compressed domain features corresponding to each time window as additional information of the frame. The data frame is marked as high priority and sent to the receiver with priority over other unsent regular data packets through an independent control channel or a priority-based transmission queue. Other embodiments may also use other methods to implement this, which are not limited here.

[0102] It should be noted that the context transmission mode in this application is a mode that sends complete data before and after an event to the receiving end at once after the event is triggered to support full-process retrospective analysis. This mode sends detailed data layers and corresponding compressed domain features of a predetermined time length before and after the event to the receiving end, facilitating the receiving end to obtain complete waveform data before and after the transient disturbance event to support full-process retrospective analysis and fault diagnosis. High-priority data frames are high-priority data frames that ensure critical data arrives at the receiving end first when communication resources are limited. The analysis and diagnosis of transient disturbance events in the distribution network not only relies on the current waveform at the moment of the disturbance but also requires waveforms from several or even dozens of cycles prior to the disturbance as a basis for analysis. The reference and waveform evolution process after the disturbance are used to determine the fault type and development trend. Traditional event-driven transmission schemes only transmit the current data after the event is triggered, losing key historical information before the event, which makes it impossible for the receiver to judge the change trajectory of the current state. In contrast, this scheme uses a sliding buffer to store the detailed data layer of all windows within the most recent predetermined time length. After the event is triggered, it can send the historical data of the predetermined time length before the event together with the current data, providing the receiver with complete spatiotemporal context information. By adopting an independent control channel or a priority transmission queue, it ensures that the complete data frame containing key data is not delayed or dropped when the channel is congested, ensuring that the key data can reach the receiver in a timely manner.

[0103] In step S105, the receiving end monitors the characteristics of the received compressed domain. When a complete data packet triggered by a transient disturbance event is received, it decompresses and reconstructs the relevant detailed data layer to backtrack and analyze the entire event process.

[0104] In some embodiments, when a complete data packet triggered by a transient disturbance event is received, decompressing and reconstructing the relevant detailed data layer to retrospectively analyze the entire event process can be achieved using the following steps:

[0105] Extract detailed data layers and random seeds for each time window from the complete data packet;

[0106] The random measurement matrix corresponding to each time window is recovered using the random seed, and the original current waveform of each event window is reconstructed from the measurement value vector of the detail data layer using a sparse reconstruction algorithm.

[0107] The reconstructed original current waveforms are stitched together in chronological order to obtain continuous high-fidelity transient current waveforms before and after the event point, which are used for fault diagnosis and backtracking analysis.

[0108] In practice, firstly, after receiving a high-priority data frame, the receiver reads a random seed from a fixed field in the frame header according to the pre-agreed data frame format between the sensor and the receiver, and then reads the spliced ​​detail data sequence from the payload area of ​​the frame. Based on the order of each time window in the spliced ​​sequence and the data length corresponding to each window, the receiver divides the spliced ​​sequence into independent detail data layers for each time window. Each detail data layer contains a low-dimensional measurement value vector for that window. Secondly, using the extracted random seed as the initial value for a pseudo-random number generator, the receiver regenerates the data according to the same generation rules as the sensor. The receiver obtains a random number sequence that is exactly the same as the one used at the sensor end, and arranges this random number sequence into a measurement matrix. This measurement matrix is ​​completely consistent with the measurement matrix used at the sensor end. The receiver obtains the sparse transform basis agreed upon with the sensor, uses this sparse transform basis as the reconstruction dictionary, and uses a sparse reconstruction algorithm, such as the basis pursuit algorithm or the orthogonal matching pursuit algorithm, to iteratively solve for the sparsest frequency domain or wavelet domain coefficient vector from the measurement value vector contained in the detail data layer of each time window. Then, the coefficient vector is restored to the current waveform sampling sequence within the original time window through inverse sparse transformation. Thus, the high-dimensional original current waveform acquired by the sensor end is completely restored. For a data packet containing multiple time windows, the receiving end recovers the original current waveform of each time window one by one. Finally, the receiving end splices the recovered original current waveforms of each time window in the order corresponding to the time segmentation of each window to form a continuous current waveform sequence. The start time of this continuous sequence is the start time of a predetermined time length before the event is triggered, and the end time is the end time of the event triggering window, which completely covers all waveform data of each window before, during and after the transient disturbance event. Thus, the receiving end uses this continuous waveform sequence as a complete data record of the entire event process, uses this record to extract parameters such as the amplitude, duration and attenuation characteristics of the transient disturbance, and compares them with a typical fault waveform library to complete the diagnosis of the fault type and the retrospective analysis of the event. Other embodiments may also use other methods to achieve this, which are not limited here.

[0109] Furthermore, in another aspect of this application, in some embodiments, this application provides a time-series-based sensor data compression and transmission system, referencing... Figure 3 The figure is a schematic diagram of a time-series-based sensor data compression and transmission system according to some embodiments of this application. The time-series-based sensor data compression and transmission system includes: an acquisition module 301, a compression module 302, a judgment module 303, a transmission module 304, and a decompression module 305, which are described below:

[0110] The acquisition module 301 in this application is mainly used to acquire the current signal of the power distribution network line through the sensor to obtain the continuous time-series data of the current transient waveform;

[0111] Compression module 302, in this application, is mainly used to perform compressed sensing projection on the time series data within each time window to obtain a detailed data layer that retains the sparse structure of the signal, and to extract statistical features from the detailed data layer to obtain the compressed domain features of the current state within each time window.

[0112] The judgment module 303 in this application is mainly used by the sensor to read the compressed domain features in order to determine the conditions that trigger the transient disturbance event.

[0113] The transmission module 304 in this application is mainly used to send the compressed domain features of the current time window to the receiving end if no transient disturbance event is triggered; otherwise, the context transmission mode is triggered to send the detailed data layer and the corresponding compressed domain features of a predetermined time length before and after the event point with high priority.

[0114] The decompression module 305 in this application is mainly used by the receiving end to monitor the characteristics of the received compressed domain. When a complete data packet triggered by a transient disturbance event is received, it decompresses and reconstructs the relevant detailed data layer to backtrack and analyze the entire event process.

[0115] In addition, this application also provides a computer device, the computer device including a memory and a processor, the memory storing code, the processor being configured to acquire the code and execute the above-described time-series-based sensor data compression and transmission method.

[0116] In some embodiments, reference Figure 4 The figure is a schematic diagram of the structure of a computer device implementing a time-series-based sensor data compression and transmission method according to some embodiments of this application. The time-series-based sensor data compression and transmission method in the above embodiments can... Figure 4 The computer device shown is used to implement this, and the computer device includes at least one processor 401, a communication bus 402, a memory 403, and at least one communication interface 404.

[0117] The processor 401 can be a general-purpose central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more devices used to control the execution of the time-series-based sensor data compression and transmission method in this application.

[0118] The communication bus 402 can be used to transmit information between the aforementioned components.

[0119] The memory 403 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disks or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. The memory 403 may exist independently and be connected to the processor 401 via the communication bus 402. The memory 403 may also be integrated with the processor 401.

[0120] The memory 403 stores program code for executing the scheme of this application, and its execution is controlled by the processor 401. The processor 401 executes the program code stored in the memory 403. The program code may include one or more software modules. In the above embodiments, the determination of the time-series-based sensor data compression and transmission method can be implemented by the processor 401 and one or more software modules in the program code in the memory 403.

[0121] Communication interface 404 uses any transceiver-like device for communicating with other devices or communication networks, such as Ethernet, radio access network (RAN), wireless local area networks (WLAN), etc.

[0122] In a specific implementation, as one example, a computer device may include multiple processors, each of which may be a single-core (single-CPU) processor or a multi-core (multi-CPU) processor. Here, a processor may refer to one or more devices, circuits, and / or processing cores used to process data (e.g., computer program instructions).

[0123] The aforementioned computer device can be a general-purpose computer device or a special-purpose computer device. In specific implementations, the computer device can be a desktop computer, a portable computer, a network server, a handheld digital assistant (PDA), a mobile phone, a tablet computer, a wireless terminal device, a communication device, or an embedded device. This application does not limit the type of computer device.

[0124] In addition, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described time-series-based sensor data compression and transmission method.

[0125] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0126] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A time-series-based sensor data compression and transmission method, characterized in that, Includes the following steps: By collecting current signals from power distribution network lines using sensors, continuous time-series data of transient current waveforms can be obtained. Compressed sensing projection is performed on the time series data within each time window to obtain a detailed data layer that preserves the sparse structure of the signal. Statistical feature extraction is performed on the detailed data layer to obtain the compressed domain features of the current state within each time window. The sensor reads the compressed domain features to determine the conditions that trigger the transient disturbance event. If no transient disturbance event is triggered, only the compressed domain features of the current time window are sent to the receiving end; otherwise, the context transmission mode is triggered, and the detailed data layer and the corresponding compressed domain features for a predetermined time length before and after the event point are sent with high priority. The receiving end monitors the characteristics of the received compressed domain. When it receives a complete data packet triggered by a transient disturbance event, it decompresses and reconstructs the relevant detailed data layer to backtrack and analyze the entire event process. Specifically, performing compressed sensing projection on the time-series data within each time window to obtain a detailed data layer that preserves the sparse structure of the signal includes: A random measurement matrix is ​​generated using a random seed shared with the receiver; a sparse transformation is applied to the time-series data within each time window to obtain multiple sparse vectors; the measurement matrix is ​​used to compress and project all the sparse vectors to obtain low-dimensional measurement value vectors for each time window; all the measurement value vectors are used as detail data layers that preserve the sparse structure of the signal within the corresponding time window and stored in the sliding buffer of the sensor in chronological order. Specifically, the conditions for the sensor to read the compressed domain features to determine the current triggering of a transient disturbance event include: The baseline energy range and baseline sparsity range of the current waveform established by the historical time window are obtained; the sensor reads the compressed domain features and compares the energy features in the compressed domain features of the current window with the baseline energy range; the sparsity estimate in the compressed domain features of the current window is compared with the baseline sparsity range; when the energy features in the compressed domain features of the current window exceed the baseline energy range and the sparsity estimate exceeds the baseline sparsity range, it is determined that a transient disturbance event is triggered in the current time window.

2. The time-series-based sensor data compression and transmission method as described in claim 1, characterized in that, The acquisition of current signals from power distribution network lines using sensors yields continuous time-series data of transient current waveforms, specifically including: The current signal of the power distribution network line is continuously sampled by a sensor and then converted from analog to digital to obtain a current sequence. The current sequence is continuously divided according to a preset time window length to form time-series data of the transient current waveform in continuous time.

3. The time-series-based sensor data compression and transmission method as described in claim 1, characterized in that, Statistical feature extraction is performed on the detailed data layer to obtain the compressed domain features of the current state within each time window, specifically including: For the measurement value vector in the detailed data layer, without reconstructing the original signal, its element-squared sum is directly calculated as the energy feature of the current waveform within the current window; Determine the sparsity estimate of the measurement vector in the detailed data layer; The energy characteristics and the sparsity estimates are used together as the compressed domain characteristics of the current state within each time window.

4. The time-series-based sensor data compression and transmission method as described in claim 1, characterized in that, If no transient disturbance event is triggered, only the compressed domain features of the current time window are sent to the receiving end, specifically including: Extract the compressed domain features of the current time window and the random seed shared with the receiver; The compressed domain features of the current time window are encapsulated into a lightweight state packet, and the random seed is placed in a fixed field in the packet header; Only the lightweight status packet is sent to the receiving end.

5. The time-series-based sensor data compression and transmission method as described in claim 1, characterized in that, Triggering the context transmission mode, the detailed data layer and corresponding compressed domain features for a predetermined time period before and after the event point are sent with high priority, specifically including: Extract all detailed data layers within a predetermined time period prior to the event from the sensor's sliding buffer. Get the detailed data layer of the current time window and concatenate it with all the detailed data layers within the predetermined time length in chronological order; The random seed, the spliced ​​detail data layer, and the corresponding compressed domain features are encapsulated into a high-priority data frame, which is then transmitted to the receiving end via an independent control channel or a priority transmission queue.

6. A time-series-based sensor data compression and transmission system, comprising performing time-series-based sensor data compression and transmission using the method described in any one of claims 1 to 5, characterized in that, include: The acquisition module is used to acquire current signals from power distribution network lines through sensors to obtain continuous time-series data of current transient waveforms. The compression module is used to perform compressed sensing projection on the time series data within each time window to obtain a detailed data layer that preserves the sparse structure of the signal, and to extract statistical features from the detailed data layer to obtain the compressed domain features of the current state within each time window. The judgment module is used by the sensor to read the compressed domain features in order to determine the conditions that trigger the transient disturbance event. The transmission module is used to send only the compressed domain features of the current time window to the receiving end if no transient disturbance event is triggered; otherwise, it triggers the context transmission mode and sends the detailed data layer and corresponding compressed domain features for a predetermined time length before and after the event point with high priority. The decompression module is used by the receiving end to monitor the characteristics of the received compressed domain. When a complete data packet triggered by a transient disturbance event is received, it decompresses and reconstructs the relevant detailed data layer to backtrack and analyze the entire event process.

7. A computer device, characterized in that, The computer device includes a memory and a processor, the memory storing code, and the processor being configured to retrieve the code and execute the time-series-based sensor data compression and transmission method as described in any one of claims 1 to 5.

8. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the time-series-based sensor data compression and transmission method as described in any one of claims 1 to 5.

Citation Information

Patent Citations

  • Photovoltaic system direct-current arc fault feature extraction and recognition method and device

    CN108037422A

  • Electric energy meter data acquisition method and system based on Internet of Things

    CN122027669A