A time series data compression storage method and system for industrial internet of things

Abstract

This invention discloses a method and system for time-series data compression and storage for the Industrial Internet of Things (IIoT), relating to the field of time-series data processing technology. The method includes extracting and retaining operator matrices and perturbation summaries based on stability operator matrices and perturbation strength, forming compressed records with system state vectors, constructing an index table and managing hot and cold storage, deserializing and reconstructing sparse matrices on the compressed records to obtain decompressed compressed records. Through rasterization and neighborhood similarity measurement, efficient data compression and storage are achieved, reducing redundant information. Euclidean distance and displacement matrices are used to accurately calculate the similarity of time-series data, optimizing the data compression process. Edge computing and real-time compression strategies are introduced to ensure low-latency access and processing of time-series data in the IIoT environment.

Description

Technical Field

[0001] This invention relates to the field of time-series data processing technology, and in particular to a method and system for time-series data compression and storage for the Industrial Internet of Things. Background Technology

[0002] Time-series data from the Industrial Internet of Things (IIoT) is widely used in monitoring systems, sensors, controllers, and other devices, generating massive amounts of data. Currently, IIoT systems generally adopt network interfaces or edge acquisition interfaces to uniformly access and store sampled data from different devices and channels, and rely on cloud or edge data analysis platforms to complete subsequent processing. In this process, the acquisition frequency of time-series data is constantly increasing, the data dimensions are gradually expanding, and the data scale is growing exponentially, which puts higher demands on the capacity, throughput, and long-term stability of storage systems. To reduce storage pressure and improve system scalability, related technical fields have gradually introduced technical solutions such as time-series data compression, hierarchical storage, and cold and hot data migration, which are widely used in industrial databases, time-series databases, and edge computing frameworks.

[0003] However, existing technologies still have shortcomings. Current time-series data compression technologies use timestamps or sequential indexes as the only organization method, lacking unified modeling of sampling intervals, window scales, and multi-channel synchronization relationships. This results in cross-channel data not being able to be correlated and compressed under a unified structure, leading to high redundancy and difficulty in supporting fine-grained reconstruction. Existing industrial IoT time-series data storage solutions rely on offline batch processing or full model recalculation, lacking lightweight real-time compression and cold / hot tiered storage mechanisms for the edge side. It is difficult to balance real-time response and long-term storage efficiency under conditions of high sampling frequency and large-scale device access. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a time-series data compression and storage method and system for the Industrial Internet of Things (IIoT). It solves the problems of existing time-series data compression technologies that use timestamps or sequential indexes as the only organization method, lack unified modeling of sampling intervals, window scales, and multi-channel synchronization relationships, resulting in cross-channel data being unable to be correlated and compressed under a unified structure, high redundancy, and difficulty in supporting fine-grained reconstruction. Existing IIoT time-series data storage solutions rely on offline batch processing or full model recalculation, lacking lightweight real-time compression and cold / hot tiered storage mechanisms for the edge side, making it difficult to balance real-time response and long-term storage efficiency under high sampling frequency and large-scale device access conditions.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a method for time-series data compression and storage for the Industrial Internet of Things, comprising the following steps:

[0008] The system obtains the standard sampling frame, sampling interval, and window length through the API interface, generates a raster index, extracts the sampling value and channel index from the standard sampling frame, and constructs a system state vector by combining the raster index.

[0009] A candidate neighborhood index set is constructed with the grid index as the center. The Euclidean distance of the system state vector is calculated to filter the neighborhood set. The displacement vector and observation velocity are calculated based on the neighborhood set. The regularization intensity is determined based on the minimum channel resolution obtained from the API interface. The displacement matrix is ​​constructed using the displacement vector, observation velocity, and regularization intensity, and the coefficient column vector is solved to calculate the consistency velocity.

[0010] The disturbance intensity is calculated based on the observation velocity and the consistency velocity, and a local state transition operator matrix is ​​constructed based on the projection coefficient scalar in the coefficient column vector. The local state transition operator matrix is ​​interpolated and predicted. The responsibility degree is calculated in combination with the disturbance intensity to obtain the stability operator matrix and generate the working condition label sequence and working condition segment.

[0011] Based on the stability operator matrix and the perturbation intensity, the retained operator matrix and perturbation summary are extracted respectively, and a compressed record is formed with the system state vector. An index table is constructed and hot storage and cold storage management are performed. The compressed record is deserialized and sparse matrix reconstructed to obtain the decompressed compressed record.

[0012] As a preferred embodiment of the time-series data compression and storage method for the Industrial Internet of Things described in this invention, the step of interpolating and predicting the local state transition operator matrix, calculating the responsibility degree in conjunction with the disturbance intensity, obtaining the stable operator matrix, and generating the operating condition label sequence and operating condition segment includes:

[0013] The number of working condition models is obtained through the API interface, the local state transition operator matrix is ​​inverted, and the interpolation prediction state of the model in the grid index is calculated.

[0014] Evidence is constructed based on interpolated predicted states, system state vectors, and disturbance strengths, and the evidence is normalized to obtain the degree of responsibility.

[0015] The stability operator matrix is ​​calculated based on the degree of responsibility and the local state transition operator matrix.

[0016] Sort the responsibility levels in descending order, filter the model numbers corresponding to the maximum values ​​to obtain the working condition labels, generate the working condition label sequence, and set the working condition segments.

[0017] As a preferred embodiment of the time-series data compression and storage method for the Industrial Internet of Things described in this invention, the step of extracting and retaining the operator matrix and the perturbation summary based on the stability operator matrix and the perturbation strength, respectively, and forming a compressed record with the system state vector, includes:

[0018] The absolute values ​​of matrix elements are extracted from the stable operator matrix. Elements whose absolute values ​​are greater than or equal to the unified upper bound of the system and their corresponding row and column indices are retained to obtain the retained operator matrix. A perturbation summary is constructed based on the perturbation intensity.

[0019] Compressed records are obtained by horizontally arranging the operating condition segments, system state vectors, preserved operator matrices, and disturbance summaries.

[0020] As a preferred embodiment of the time-series data compression and storage method for the Industrial Internet of Things described in this invention, the steps of constructing an index table and managing hot and cold storage, and deserializing and reconstructing the compressed records to obtain the decompressed compressed records include:

[0021] Extract equipment identifiers, operating condition sections, operating condition labels, and intensity labels, arrange them horizontally, and set them as index items for compressed records. Construct an index table and write the index items into storage blocks.

[0022] The system obtains query requests through the API interface, filters records in the index table that have an intersection between the raster index and the query request's start and end raster, extracts the storage block corresponding to the index item in the record, restores the compressed record using deserialization, and reconstructs the preserved operator matrix using the sparse matrix reconstruction method to obtain the decompressed compressed record.

[0023] As a preferred embodiment of the time-series data compression and storage method for industrial IoT described in this invention, the step of extracting sampled values ​​and channel indices from standard sampled frames and constructing a system state vector by combining them with a raster index includes:

[0024] Obtain the standard sampling frame, sampling interval, and window length through the API interface. Divide the window length by the sampling interval to obtain the end index of the raster and generate the raster index.

[0025] Extract sampled values ​​and channel indices from standard sampled frames, and arrange them vertically in combination with grid indices to obtain the system state vector.

[0026] As a preferred embodiment of the time-series data compression and storage method for the Industrial Internet of Things described in this invention, the step of constructing a displacement matrix and solving for the coefficient column vector using displacement vector, observation velocity, and regularization intensity to calculate the consistency velocity includes:

[0027] A candidate neighborhood index set is constructed for the raster index. The Euclidean distance between the system state vectors of the raster index and the candidate raster indexes is calculated and filtered to obtain the neighborhood set. The displacement vector and observed velocity are calculated within the neighborhood set.

[0028] The minimum resolution of each channel is obtained through the API interface, and a unified upper bound is set for the system to calculate the regularization intensity.

[0029] Each displacement vector is set as a column and arranged vertically to obtain the displacement matrix;

[0030] The coefficient column vector is solved based on the regularity intensity, displacement matrix, identity matrix, and observed velocity, and the consistency velocity is calculated by combining the displacement vector.

[0031] As a preferred embodiment of the time-series data compression and storage method for the Industrial Internet of Things described in this invention, wherein: the construction of the local state transition operator matrix based on the projection coefficient scalar in the coefficient column vector includes:

[0032] The residual velocity is obtained by subtracting the consistency velocity from the observation velocity. The disturbance intensity is obtained by multiplying the transpose of the residual velocity, the residual velocity, and the reciprocal of the number of channels.

[0033] The directed transition probabilities are extracted from the projection coefficient scalars and normalized to obtain the directed weights. The local state transition operator matrix is ​​then calculated.

[0034] Secondly, the present invention provides a time-series data compression and storage system for the Industrial Internet of Things, comprising:

[0035] The standard sampling frame acquisition module is used to acquire sampling frames through the API interface and perform verification processing.

[0036] The grid index generation module is used to generate grid indexes based on sampling intervals and window lengths, calculate grid time, and vertically arrange sampled values ​​and channel indices to obtain the system state vector.

[0037] The neighborhood calculation module is used to extract the neighborhood range centered on the grid index, construct the neighborhood set, and perform vector operations.

[0038] The disturbance intensity calculation module is used to calculate the disturbance intensity using displacement vector, observation velocity, and quantization error, and to solve for the coefficient column vector based on the canonical intensity and displacement matrix.

[0039] The compressed record storage module is used to extract and arrange equipment identifiers, operating conditions, and tags, build an index table, and perform data storage and retrieval between hot storage and cold storage.

[0040] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, it implements any step of the time-series data compression and storage method for the Industrial Internet of Things as described in the first aspect of the present invention.

[0041] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the time-series data compression and storage method for the Industrial Internet of Things as described in the first aspect of the present invention.

[0042] The beneficial effects of this invention are as follows: This invention extracts and retains the operator matrix and perturbation summary based on the stability operator matrix and perturbation intensity, respectively, forms a compressed record with the system state vector, constructs an index table and manages hot and cold storage, and performs deserialization and sparse matrix reconstruction on the compressed record to obtain the decompressed compressed record; it achieves efficient data compression and storage, reduces redundant information, optimizes the data compression process, and ensures low-latency access and processing of time-series data in the industrial Internet of Things environment. Attached Figure Description

[0043] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0044] Figure 1 This is a flowchart illustrating the operation of the time-series data compression and storage method for the Industrial Internet of Things in Example 1.

[0045] Figure 2 This is a schematic diagram of the time-series data compression and storage system for industrial IoT in Example 1. Detailed Implementation

[0046] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0047] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0048] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0049] Example 1, referring to Figure 1 This is the first embodiment of the present invention, which provides a method for time-series data compression and storage for industrial Internet of Things, including the following steps:

[0050] S1. Obtain the standard sampling frame, sampling interval, and window length through the API interface, generate a raster index, extract the sampling value and channel index from the standard sampling frame, and construct the system state vector by combining the raster index;

[0051] Specifically, sampled values ​​and channel indices are extracted from standard sampled frames, and combined with raster indexes to construct a system state vector, including:

[0052] Obtain the standard sampling frame through the API interface, including device ID, channel index, sample value, device-side timestamp, edge access timestamp, and frame verification flag (if it passes, mark it as 1 for subsequent operations; otherwise, discard the standard sampling frame).

[0053] The sampling interval and window length are obtained through the API interface. The window length is divided by the sampling interval to obtain the end index of the raster and a raster index is generated. This indicates that the raster index t increases from 0 to the end index of the raster. The raster time is set on the raster index (obtained by multiplying the raster index by the sampling interval).

[0054] Extract sampled values ​​and channel indices from the standard sampled frame, and arrange them vertically in combination with the grid index to obtain the system state vector, such as the sampled value of channel 1 at grid index t and the sampled value of channel 2 at grid index t.

[0055] By organizing and aligning standard sampling frames from different devices and channels under a unified grid indexing system, this invention achieves a structured representation of multi-source time-series data, enabling discrete sampling data to have a consistent reference benchmark on the time scale. This fundamentally avoids the alignment drift problem introduced by traditional direct splicing based on original timestamps. This approach naturally gives the system state vector a consistent expression capability across channels and devices, providing a stable input foundation for subsequent modeling and compression. By introducing a frame verification and filtering mechanism at the sampling stage, abnormal or incomplete data can be effectively blocked from entering the compression process, reducing the interference of invalid information on subsequent model solving. Compared with the existing technology that separates compression and cleaning, this invention completes quality constraints at the state construction stage, making the overall compression process more compact and robust, and is particularly suitable for high-frequency continuous acquisition scenarios in the Industrial Internet of Things.

[0056] S2. Construct a candidate neighborhood index set centered on the grid index, calculate the Euclidean distance of the system state vector to filter the neighborhood set, calculate the displacement vector and observation velocity based on the neighborhood set, determine the regularization intensity based on the minimum channel resolution obtained from the API interface, construct the displacement matrix using the displacement vector, observation velocity and regularization intensity and solve the coefficient column vector, and calculate the consistency velocity.

[0057] Specifically, the displacement matrix is ​​constructed using the displacement vector, observed velocity, and canonical intensity, and the coefficient column vector is solved to calculate the uniform velocity, including:

[0058] Centered on raster index t, extract W (half-width of the neighborhood window, set based on the system time constant) raster indices forward and backward respectively, and set the extracted range as the candidate range;

[0059] Construct a candidate neighborhood index set for raster index t, representing all raster indices from tW to t+W;

[0060] Based on the grid index t, candidate grid indices s are selected sequentially from the candidate neighborhood index set. The Euclidean distance between the system state vectors of grid index t and candidate grid index s is calculated. The candidate grid indices s are sorted in ascending order according to the magnitude of the Euclidean distance. The top k candidate grid indices are selected and arranged horizontally to obtain the neighborhood set.

[0061] Based on the grid index t, traverse each neighbor grid index j in the neighborhood set, and subtract the system state vector of grid index t from the system state vector of neighbor grid index j to obtain the displacement vector;

[0062] The system state vector at grid index t is subtracted from the system state vector at grid index t+1 to obtain the vector difference. The observation velocity is obtained by dividing the vector difference by the sampling interval.

[0063] The minimum resolution of each channel is obtained through the API interface. The minimum resolution is divided by 2 to obtain the upper bound of the quantization error. The upper bounds of the quantization error of all channels are sorted in descending order, and the maximum value is set as the unified upper bound of the system.

[0064] The regularization intensity is obtained by dividing the system's unified upper bound by the sampling interval and taking the square of the division result.

[0065] Each displacement vector is set as a column and arranged vertically to obtain the displacement matrix;

[0066] Construct an identity matrix 1 based on the first k candidate raster indices, where all elements except the diagonal are 0, the diagonal elements are 1, and the number of diagonal elements is the same as k.

[0067] The coefficient column vector is calculated based on the canonical intensity, displacement matrix, identity matrix, and observed velocity, using the following formula:

[0068] ,

[0069] in, For raster index The coefficient column vector, For raster index The regularity strength, It is the identity matrix. For raster index The displacement matrix, For transpose, For raster index The observation speed;

[0070] The r-th element (where r is the row index, representing the element in the r-th row of the extracted coefficient column vector) is set as the projection coefficient scalar. Combined with the displacement vector, the uniform velocity is calculated using the following formula:

[0071] ,

[0072] in, For raster index Consistency speed, For projection coefficients scalars, Indexing for neighbors, For raster index The neighborhood set, This is the displacement vector.

[0073] By introducing a neighborhood selection mechanism based on state similarity, the system focuses only on historical samples that are most representative of the current state within a local time range. This avoids the redundant computation and noise amplification problems caused by traditional fixed window or full participation in modeling. By constructing regularization strength through unified quantization error constraints, the coefficient solution process is kept consistent with the actual sensor resolution, effectively suppressing unstable solutions caused by differences in measurement accuracy. This design enables the consistency velocity to truly reflect the dominant trend of the system in short-term evolution, rather than being driven by occasional disturbances. Compared with existing compression methods that only rely on difference or prediction residuals, this invention introduces velocity consistency modeling under structural constraints, enabling the compression results to have both dynamic interpretability and engineering stability.

[0074] S3. Calculate the disturbance intensity based on the observation velocity and the consistency velocity, and construct the local state transition operator matrix based on the projection coefficient scalar in the coefficient column vector. Perform interpolation prediction on the local state transition operator matrix, calculate the degree of responsibility in combination with the disturbance intensity, obtain the stability operator matrix, and generate the working condition label sequence and working condition segment.

[0075] Specifically, the local state transition operator matrix is ​​constructed based on the projection coefficient scalars in the coefficient column vector, including:

[0076] The residual velocity is obtained by subtracting the consistency velocity from the observation velocity.

[0077] The disturbance intensity is obtained by multiplying the transpose of the residual velocity, the residual velocity, and the reciprocal of the number of channels (obtained by counting channel indices).

[0078] The directed transition probabilities are extracted from the projection coefficient scalars and normalized to obtain the directed weights. The local state transition operator matrix is ​​then calculated using the following formula:

[0079] ,

[0080] ,

[0081] ,

[0082] in, For raster index Pointing to neighbor grid index The directed transition probability, For raster index Pointing to neighbor grid index The directed weights, For raster index Pointing to neighbor grid index The directed transition probability, For raster index The local state transition operator matrix, indexing neighboring grids The system state vector, For raster index The system state vector, This sets a unified upper bound for the quantization error of the system.

[0083] By explicitly characterizing the difference between observed behavior and consistent behavior, this invention separates system disturbances from state changes, enabling the compression process to not only retain the dominant chemical structure but also quantify the intensity of uncertainty. By constructing directional transition relationships using projection coefficients, the local state transition operator can reflect the biased migration characteristics of the system in the state space, thereby avoiding the neglect of evolution direction information by traditional symmetric modeling methods. This operator provides a compact and information-dense representation for subsequent interpolation prediction and operating condition analysis. Compared with existing compression methods that rely on statistics or low-rank decomposition, this invention embeds dynamic structural information during the compression stage, giving the stored data higher reconstruction value and analytical depth.

[0084] S4. Based on the stable operator matrix and the perturbation intensity, extract the retained operator matrix and perturbation summary respectively, form a compressed record with the system state vector, construct an index table and perform hot storage and cold storage management, deserialize the compressed record and reconstruct the sparse matrix to obtain the decompressed compressed record;

[0085] Specifically, interpolation prediction is performed on the local state transition operator matrix, and the responsibility degree is calculated in conjunction with the disturbance intensity to obtain the stability operator matrix and generate the working condition label sequence and working condition segment, including:

[0086] The number M of the operating condition (in an industrial IoT system, the operating state mode in which the system state vector evolves according to the same type of state transition operator in a continuous grid time) model is obtained through the API interface, and the model number is set from m=1 to m=M.

[0087] The local state transition operator matrix is ​​inverted, and the interpolated predicted state of model m at raster index t is calculated using the following formula:

[0088] ,

[0089] in, For interpolation prediction of state, For model indexing, For raster index The local state transition operator matrix, For raster index The system state vector, For raster index The system state vector;

[0090] Evidence is constructed based on interpolated predicted states, system state vectors, and disturbance strengths. This evidence is then normalized to obtain the degree of responsibility, as shown in the formula:

[0091] ,

[0092] in, As evidence, The disturbance intensity;

[0093] Based on the degree of responsibility and the local state transition operator matrix, the stability operator matrix is ​​calculated using the following formula:

[0094] ,

[0095] in, For a stable operator matrix, For the end index of the raster, For the degree of responsibility;

[0096] Sort the responsibility levels in descending order, filter the model numbers corresponding to the maximum values ​​to obtain the working condition labels (e.g., m=1 represents steady state, climbing, and fluctuation), and generate a working condition label sequence.

[0097] Based on the working condition label sequence, starting from raster index t=1, if the working condition labels at raster index t and raster index t-1 are inconsistent, then a working condition segment is defined, using the following formula:

[0098] ,

[0099] in, For the first Each working condition section For raster indexes where the labels start to remain unchanged, The raster index where the label last remained unchanged.

[0100] By employing a multi-model interpolation prediction and responsibility assessment mechanism, the system achieves automatic identification and segmentation of operating conditions, enabling compressed data to form a one-to-one correspondence with actual operating stages. This approach avoids the lack of adaptability inherent in traditional methods that rely on manually set thresholds or single indicators for operating condition identification. It allows the system to maintain stable discrimination capabilities during complex operating condition transitions. By dividing continuous grids into operating condition segments, data compression is no longer based on single points but on operating stages as the basic organizational structure, thereby significantly improving the efficiency of subsequent queries, backtracking, and analysis. Compared to existing technical solutions that only compress the original time series, this invention simultaneously completes semantic-level structural annotation during the compression process, enhancing the engineering usability of the data.

[0101] Furthermore, based on the stability operator matrix and the perturbation intensity, the preserved operator matrix and perturbation summary are extracted respectively, and combined with the system state vector to form a compressed record, including:

[0102] The absolute values ​​of the matrix elements are extracted from the stable operator matrix. Elements whose absolute values ​​are greater than or equal to the system's unified upper bound and their corresponding row and column indices are retained to obtain the retained operator matrix.

[0103] The perturbation summary is constructed based on the perturbation intensity, using the following formula:

[0104] ,

[0105] in, For the first Disturbance summary for each operating condition segment;

[0106] Compressed records are obtained by horizontally arranging the operating condition segments, system state vectors, preserved operator matrices, and disturbance summaries.

[0107] By extracting key structural elements from the stable operator matrix and summarizing the system's unstable behavior using perturbation summaries, this invention simultaneously preserves system state, evolutionary structure, and uncertainty intensity information in a single compressed record. This results in a high information density and interpretability of the compressed result. This approach avoids the problem of traditional compression methods that only retain numerical approximations while losing structural features. The compressed record can be used for storage and can also directly support subsequent analysis and reconstruction. By generating records based on operating condition segments, redundant storage is further reduced, and the compression ratio is improved. Compared with existing sampling-point-level compression schemes, this invention achieves structured, high-value information retention without significantly increasing computational complexity.

[0108] Furthermore, an index table is constructed and hot / cold storage management is implemented. The compressed records are deserialized and reconstructed using a sparse matrix to obtain the decompressed compressed records, including:

[0109] Extract the equipment identifier, operating condition section, operating condition label, and intensity label, arrange them horizontally, and set them as the index items for compressed records. Construct an index table, where each index item corresponds to one compressed record.

[0110] Write the index item into the hot storage block, extract the end-of-block raster index of the hot storage block, multiply it by the sampling interval to get the end time of the block, obtain the current system time through the API interface, subtract the end time of the block from the current system time to get the difference, if it is greater than or equal to the hot storage retention time threshold, then copy the hot storage block from the hot storage location to the cold storage location to obtain the cold storage block.

[0111] Obtain the query request through the API interface, including the device identifier, query start and end grid indexes, and output type identifier;

[0112] In the index table, filter records where the raster index intersects with the query start and end raster. Extract the storage blocks (hot and cold blocks) corresponding to the index items in the record. Use deserialization to restore the compressed records in the record. Use the sparse matrix reconstruction method to reconstruct the preserved operator matrix to obtain the decompressed compressed records.

[0113] By constructing an index system centered on device identifiers and operating conditions, and introducing a hierarchical management strategy of hot and cold storage, this invention achieves efficient scheduling of compressed data throughout its entire lifecycle. This allows high-frequency access data and historically accumulated data to obtain optimal storage and response strategies, respectively. During the decompression stage, by combining deserialization and sparse matrix reconstruction methods, only the compressed records hit by the index are partially restored, significantly reducing decompression computing power consumption and access latency. Compared with existing storage systems that require full scanning or batch decompression, this invention has significant advantages in query efficiency, system scalability, and edge deployment adaptability, and can effectively support the needs of large-scale device access and long-term operation in the industrial IoT environment.

[0114] Example 2, refer to Figure 2 As a second embodiment of the present invention, a time-series data compression and storage system for industrial Internet of Things includes:

[0115] The standard sampling frame acquisition module is used to acquire sampling frames through the API interface and perform verification processing.

[0116] The grid index generation module is used to generate grid indexes based on sampling intervals and window lengths, calculate grid time, and vertically arrange sampled values ​​and channel indices to obtain the system state vector.

[0117] The neighborhood calculation module is used to extract the neighborhood range centered on the grid index, construct the neighborhood set, and perform vector operations.

[0118] The disturbance intensity calculation module is used to calculate the disturbance intensity using displacement vector, observation velocity, and quantization error, and to solve for the coefficient column vector based on the canonical intensity and displacement matrix.

[0119] The compressed record storage module is used to extract and arrange equipment identifiers, operating conditions, and tags, build an index table, and perform data storage and retrieval between hot storage and cold storage.

[0120] This embodiment also provides a computer device applicable to a time-series data compression and storage method for industrial Internet of Things (IoT), comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the time-series data compression and storage method for industrial IoT as proposed in the above embodiment.

[0121] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0122] This embodiment also provides a storage medium on which a computer program is stored. When executed by a processor, the program implements the time-series data compression storage method for industrial Internet of Things as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0123] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for time-series data compression and storage for the Industrial Internet of Things, characterized in that: Includes the following steps: The system obtains the standard sampling frame, sampling interval, and window length through the API interface, generates a raster index, extracts the sampling value and channel index from the standard sampling frame, and constructs a system state vector by combining the raster index. A candidate neighborhood index set is constructed with the grid index as the center. The Euclidean distance of the system state vector is calculated to filter the neighborhood set. The displacement vector and observation velocity are calculated based on the neighborhood set. The regularization intensity is determined based on the minimum channel resolution obtained from the API interface. The displacement matrix is ​​constructed using the displacement vector, observation velocity, and regularization intensity, and the coefficient column vector is solved to calculate the consistency velocity. The disturbance intensity is calculated based on the observation velocity and the consistency velocity, and a local state transition operator matrix is ​​constructed based on the projection coefficient scalar in the coefficient column vector. The local state transition operator matrix is ​​interpolated and predicted. The responsibility degree is calculated in combination with the disturbance intensity to obtain the stability operator matrix and generate the working condition label sequence and working condition segment. Based on the stability operator matrix and the perturbation intensity, the retained operator matrix and perturbation summary are extracted respectively, and a compressed record is formed with the system state vector. An index table is constructed and hot storage and cold storage management are performed. The compressed record is deserialized and sparse matrix reconstructed to obtain the decompressed compressed record.

2. The time-series data compression and storage method for the Industrial Internet of Things as described in claim 1, characterized in that: The process of interpolating and predicting the local state transition operator matrix, calculating the degree of responsibility based on the disturbance intensity, obtaining the stable operator matrix, and generating the operating condition label sequence and operating condition segment includes: The number of working condition models is obtained through the API interface, the local state transition operator matrix is ​​inverted, and the interpolation prediction state of the model in the grid index is calculated. Evidence is constructed based on interpolated predicted states, system state vectors, and disturbance strengths, and the evidence is normalized to obtain the degree of responsibility. The stability operator matrix is ​​calculated based on the degree of responsibility and the local state transition operator matrix. Sort the responsibility levels in descending order, filter the model numbers corresponding to the maximum values ​​to obtain the working condition labels, generate the working condition label sequence, and set the working condition segments.

3. The time-series data compression and storage method for the Industrial Internet of Things as described in claim 2, characterized in that: The process of extracting and preserving the operator matrix and perturbation summary based on the stability operator matrix and perturbation intensity, respectively, and forming a compressed record with the system state vector, includes: The absolute values ​​of matrix elements are extracted from the stable operator matrix. Elements whose absolute values ​​are greater than or equal to the unified upper bound of the system and their corresponding row and column indices are retained to obtain the retained operator matrix. A perturbation summary is constructed based on the perturbation intensity. Compressed records are obtained by horizontally arranging the operating condition segments, system state vectors, preserved operator matrices, and disturbance summaries.

4. The time-series data compression and storage method for the Industrial Internet of Things as described in claim 3, characterized in that: The process involves constructing an index table and managing hot and cold storage, deserializing compressed records, and reconstructing sparse matrices to obtain decompressed compressed records, including: Extract equipment identifiers, operating condition sections, operating condition labels, and intensity labels, arrange them horizontally, and set them as index items for compressed records. Construct an index table and write the index items into storage blocks. The system obtains query requests through the API interface, filters records in the index table that have an intersection between the raster index and the query request's start and end raster, extracts the storage block corresponding to the index item in the record, restores the compressed record using deserialization, and reconstructs the preserved operator matrix using the sparse matrix reconstruction method to obtain the decompressed compressed record.

5. The time-series data compression and storage method for the Industrial Internet of Things as described in claim 1, characterized in that: The step of extracting sampled values ​​and channel indices from standard sampled frames and constructing a system state vector by combining them with the raster index includes: Obtain the standard sampling frame, sampling interval, and window length through the API interface. Divide the window length by the sampling interval to obtain the end index of the raster and generate the raster index. Extract sampled values ​​and channel indices from standard sampled frames, and arrange them vertically in combination with grid indices to obtain the system state vector.

6. The time-series data compression and storage method for the Industrial Internet of Things as described in claim 1, characterized in that: The process of constructing a displacement matrix using displacement vectors, observed velocities, and canonical strength, and solving for the coefficient column vectors to calculate the consistency velocity includes: A candidate neighborhood index set is constructed for the raster index. The Euclidean distance between the system state vectors of the raster index and the candidate raster indexes is calculated and filtered to obtain the neighborhood set. The displacement vector and observed velocity are calculated within the neighborhood set. The minimum resolution of each channel is obtained through the API interface, and a unified upper bound is set for the system to calculate the regularization intensity. Each displacement vector is set as a column and arranged vertically to obtain the displacement matrix; The coefficient column vector is solved based on the regularity intensity, displacement matrix, identity matrix, and observed velocity, and the consistency velocity is calculated by combining the displacement vector.

7. The time-series data compression and storage method for the Industrial Internet of Things as described in claim 1, characterized in that: The construction of the local state transition operator matrix based on the projection coefficient scalars in the coefficient column vector includes: The residual velocity is obtained by subtracting the consistency velocity from the observation velocity. The disturbance intensity is obtained by multiplying the transpose of the residual velocity, the residual velocity, and the reciprocal of the number of channels. The directed transition probabilities are extracted from the projection coefficient scalars and normalized to obtain the directed weights. The local state transition operator matrix is ​​then calculated.

8. A time-series data compression and storage system for the Industrial Internet of Things (IIoT), used to implement the time-series data compression and storage method for the Industrial Internet of Things as described in any one of claims 1 to 7, characterized in that: include: The standard sampling frame acquisition module is used to acquire sampling frames through the API interface and perform verification processing. The grid index generation module is used to generate grid indexes based on sampling intervals and window lengths, calculate grid time, and vertically arrange sampled values ​​and channel indices to obtain the system state vector. The neighborhood calculation module is used to extract the neighborhood range centered on the grid index, construct the neighborhood set, and perform vector operations. The disturbance intensity calculation module is used to calculate the disturbance intensity using displacement vector, observation velocity, and quantization error, and to solve for the coefficient column vector based on the canonical intensity and displacement matrix. The compressed record storage module is used to extract and arrange equipment identifiers, operating conditions, and tags, build an index table, and perform data storage and retrieval between hot storage and cold storage.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the time-series data compression and storage method for industrial Internet of Things as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the time-series data compression and storage method for industrial Internet of Things as described in any one of claims 1 to 7.

Images