Multi-source data fusion lightweight design method for battery power supply

Abstract

The invention provides a battery power supply-oriented multi-source data fusion lightweight design method, which comprises a cloud side end collaborative compression network architecture, the cloud side end collaborative compression network architecture is composed of multiple types of sensor nodes and edge agents, the sensor nodes are deployed at each key position of a power facility, and the edge agents are deployed at each key position of the power facility. The acquisition module is responsible for acquiring environment and equipment operation data; the sensor nodes transmit sensed data to adjacent edge agent nodes in a single-hop or multi-hop mode; single-hop communication is generally suitable for a scene in which the distance between a sensor node and an edge agent is relatively close, and under the condition of relatively wide distribution of the sensor nodes or complex environment, a multi-hop communication mode is adopted to gradually transmit data to the edge agent through a plurality of relay nodes; according to the technical scheme, the overall perception capability and the data processing capability of the power transmission Internet of Things system can be remarkably enhanced under the condition that the hardware burden is not increased.

Description

Technical Field

[0001] The invention relates to the fields of artificial intelligence and computer technology, and in particular to a multi-source data fusion lightweight design method for battery power supply. Background Art

[0002] With the in-depth construction of the power transmission Internet of Things, large-scale and various types of transmission line sensing terminals are widely deployed. The sensing terminals on the transmission lines have the characteristics of a large number of applications, complex field environments, and the coexistence of wideband and narrowband data. Facing the demand for massive data collection, higher requirements are placed on the reliability of the sensing data return. In order to reduce the pressure on the master station and communication, it is urgent to build a unified access method for the sensing layer of the power transmission Internet of Things, reduce repeated investment, and achieve comprehensive perception of the transmission line.

[0003] Large-scale and various types of transmission lines are widely deployed with sensing terminals, including micro-meteorological devices, breeze vibration, conductor temperature, etc. The comprehensive perception of transmission lines involves the perception of various types of data such as electrical measurement, mechanical measurement, temperature measurement, etc., and supports various business applications such as dynamic capacity expansion, icing, dancing, video front-end recognition, etc. of conductors. The frequency of collection is getting higher and higher, and the time interval of collection is even at the millisecond level, and the amount of data generated is getting larger and larger. Due to the wide distribution range of sensing terminals, complex on-site environments, and diverse network communication conditions, how to reduce the pressure on the main station and communication network while ensuring the reliability of data transmission has become an urgent problem to be solved. In addition, sensing terminals are mostly battery-powered and have limited energy. Therefore, how to improve data processing efficiency, reduce energy consumption, and extend the operating time of equipment under limited energy conditions is also a technical challenge currently faced. Summary of the invention

[0004] In view of this, the purpose of the present invention is to provide a lightweight design method for multi-source data fusion for battery power supply, which significantly enhances the overall perception ability and data processing ability of the power transmission Internet of Things system without increasing the hardware burden.

[0005] To achieve the above-mentioned purpose, the present invention adopts the following technical scheme: a lightweight design method for multi-source data fusion for battery power supply, including a cloud-edge-end collaborative compression network architecture, wherein the cloud-edge-end collaborative compression network architecture is composed of various types of sensor nodes and edge agents, and the sensor nodes are deployed at various key locations of power facilities, responsible for collecting operating data of the environment and equipment; the sensor nodes transmit the sensed data to the adjacent edge agent nodes through a single hop or multi-hop method; single-hop communication is generally suitable for scenarios where the distance between the sensor node and the edge agent is relatively close, while in the case where the sensor nodes are widely distributed or the environment is complex, a multi-hop communication mode is adopted to gradually transmit the data to the edge agent through multiple relay nodes;

[0006] After the data sent by the sensor node is transmitted to the edge agent, the edge agent first pre-processes the data, including data format conversion, preliminary filtering and compression operations;

[0007] The edge agent will learn and extract features from the received multi-source data, identify the valuable information in the data, and fuse the data from different sources; after completing the data fusion, the edge agent will transmit the processed data to the cloud data center for subsequent analysis and decision-making; the data center will restore the received data according to the needs of different application scenarios, thereby providing accurate real-time data support for the monitoring and management of the power system.

[0008] In a preferred embodiment, the raw data collected by the sensor nodes is sparsely represented as follows:

[0009] X j =Z C +Z j ,j∈{1,2,...,J} (1)

[0010] Among them, J is the size of the data set, that is, the number of sensor nodes in the area; Z C =ΨΘ C ,||Θ C ||0=K C , Z j =ΨΘ j ,||Θ j ||0=K j , signal Z C All X j The common components of C Sparse; signal Z j Yes X j The independent components of j sparse;

[0011] In the initial state, the sensor nodes perform distributed compressed sensing on the collected data, and the obtained data set is expressed by the following formula:

[0012]

[0013] Among them, y j is the M×1-dimensional randomly compressed observation projection vector of perceptual data; j is the perception data of sensor node j, with a size of N×1; Φ j is the random Gaussian observation matrix of node j, with a size of M×N; Ψ is the public sparse basis of N×N dimensions; Θ C and θ jThey are the common part and independent part of the sparse support set of node j data, which are also the feature set of data. Then, each node transmits its own data to the edge agent. Since the node random observation matrix Φ j It is generated by the random seed shared by each node and the edge agent. The random compression observation process of the signal is also equivalent to the encryption process. For other nodes that do not know the random seed, the compressed observation signal is a meaningless noise signal.

[0014] After receiving the observation data from all sensor nodes in the area, the edge agent performs a joint sparse representation on the data set and obtains the sparse components of the data set by solving the optimization problem and linear equations:

[0015] [Z C ; Z1; Z2; LZ J ](3).

[0016] In a preferred embodiment, before data fusion, the edge agent performs two-dimensional sparse support set feature M={Θ C ,Θ P} to extract and obtain the spectrum distribution characteristics of the common part of the data set and the independent part of each data in the sparse transform domain;

[0017] First, the collected and compressed data needs to be reconstructed in phase space; assuming that the attribute sets of compressed data samples of any two sets of sensors are i, j∈S, the equivalent semantic mapping relationship satisfies:

[0018]

[0019] Data sample y to be mined i The associated feature distribution is:

[0020]

[0021] a mn Yes i The envelope amplitude value of the associated information, g mn (t) is the statistical average of the data, and n(t) is the interference noise. The phase space of the sensor data is reconstructed, and the phase trajectory of the data attribute set distribution in the reconstructed phase space is obtained as follows:

[0022]

[0023] The objective function of big data semantic correlation feature fusion is constructed by using the correlation dimension feature extraction method, which is expressed as:

[0024]

[0025] Where p is the Euclidean distance between the reconstructed phase trajectories; for the semantic relevance feature fusion objective function, all sensor data are averaged to obtain the common part of the data features:

[0026]

[0027] For each sensor data, the semantic correlation feature fusion objective function is combined to perform iterative optimization processing to obtain the iterative formula of the independent part of the feature:

[0028]

[0029] In the data fusion stage, the edge agent performs additive fusion on different sensor data in sequence according to the number of steps, and records and stores the data features of each fusion step:

[0030]

[0031] Finally, the obtained fused data is expressed as:

[0032]

[0033] In a preferred embodiment, the fused data reconstruction is specifically as follows: the edge agent converts the fused data Upload to the data center, the data center will fusion data according to the linear properties of distributed compressed sensing and data characteristics as needed Split and mine the required data; the linear property of distributed compressed sensing is expressed as:

[0034]

[0035] Among them, w j is the weight, v is the data additive fusion value;

[0036] The structural expression of fusion data splitting is:

[0037]

[0038] Among them A CS is a compressed sensing information operator; in the process of data splitting, it is first necessary to determine the number and location of the data source sensor based on the source of the required information; then, based on the above results, retrieve the feature vector data set M to obtain the feature set {Θ C ,Θ i}; Finally, based on the feature set of the data, a subtractive operation is performed on the fused data to obtain the required data, which is expressed as:

[0039]

[0040] in and It is the starting and ending position of the data extracted based on the support set features;

[0041] After the split is completed, the edge agent performs inverse sparse transformation to decompress the data and obtain the original data fusion value. For further use:

[0042]

[0043] where F -1 Represents the inverse sparse transform.

[0044] Compared with the prior art, the present invention has the following beneficial effects: The present invention proposes a lightweight data fusion design method, which aims to perform effective data compression and fusion processing in the primary stage of data collection through the edge intelligent gateway, reduce redundant data, and optimize the data transmission path, thereby reducing the burden on the network and the main station. At the same time, low-power design technology is adopted to ensure that the perception terminal can operate efficiently and stably under limited energy conditions. This method not only improves the efficiency and reliability of data transmission, but also significantly reduces the power consumption of the equipment, providing a guarantee for the sustainable operation of the Internet of Things perception system of the transmission line. Ultimately, the implementation of the present invention will realize the comprehensive perception of the transmission line, improve the real-time and accuracy of the perception data, and promote the development of smart grids. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 A schematic diagram of a network architecture of a preferred embodiment of the present invention;

[0046] Figure 2 A schematic diagram of a feature mining model according to a preferred embodiment of the present invention; DETAILED DESCRIPTION

[0047] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0048] It should be noted that the following detailed descriptions are illustrative and are intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present application belongs.

[0049] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application; as used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or their combinations.

[0050] A lightweight design method for multi-source data fusion for battery power supply, reference Figure 1-2 , including a cloud-edge-end collaborative compression network architecture, which is composed of various types of sensor nodes and edge agents. The sensor nodes are deployed at various key locations of power facilities and are responsible for collecting environmental and equipment operation data; the sensor nodes transmit the sensed data to the adjacent edge agent nodes through a single hop or multi-hop method; single-hop communication is usually suitable for scenarios where the distance between sensor nodes and edge agents is relatively close, while in the case where sensor nodes are widely distributed or the environment is complex, a multi-hop communication mode is adopted to gradually transmit the data to the edge agent through multiple relay nodes;

[0051] After the data sent by the sensor node is transmitted to the edge agent, the edge agent first pre-processes the data, including data format conversion, preliminary filtering and compression operations;

[0052] The edge agent will learn and extract features from the received multi-source data, identify the valuable information in the data, and fuse the data from different sources; after completing the data fusion, the edge agent will transmit the processed data to the cloud data center for subsequent analysis and decision-making; the data center will restore the received data according to the needs of different application scenarios, thereby providing accurate real-time data support for the monitoring and management of the power system.

[0053] The sparse representation of the original data collected by the sensor node is:

[0054] X j =Z C +Z j , j∈{1,2,...,J} (1)

[0055] Among them, J is the size of the data set, that is, the number of sensor nodes in the area; Z C =ΨΘ C ,||Θ C ||0=K C , Z j =ΨΘ j ,||Θ j ||0=K j , signal Z C All X j The common components of C Sparse; signal Z j Yes X j The independent components of j sparse;

[0056] In the initial state, the sensor nodes perform distributed compressed sensing on the collected data, and the obtained data set is expressed by the following formula:

[0057]

[0058] Among them, y j is the M×1-dimensional randomly compressed observation projection vector of perceptual data; j is the perception data of sensor node j, with a size of N×1; Φ j is the random Gaussian observation matrix of node j, with a size of M×N; Ψ is the public sparse basis of N×N dimensions; Θ C and θ j They are the common part and independent part of the sparse support set of node j data, which are also the feature set of data. Then, each node transmits its own data to the edge agent. Since the node random observation matrix Φ j It is generated by the random seed shared by each node and the edge agent. The random compression observation process of the signal is also equivalent to the encryption process. For other nodes that do not know the random seed, the compressed observation signal is a meaningless noise signal.

[0059] After receiving the observation data from all sensor nodes in the area, the edge agent performs a joint sparse representation on the data set and obtains the sparse components of the data set by solving the optimization problem and linear equations:

[0060] [Z C ; Z1; Z2; LZ J ](3).

[0061] Before data fusion, the edge agent performs two-dimensional sparse support set feature M = {Θ C ,Θ P} to extract and obtain the spectrum distribution characteristics of the common part of the data set and the independent part of each data in the sparse transform domain;

[0062] First, the collected and compressed data needs to be reconstructed in phase space; assuming that the attribute sets of compressed data samples of any two sets of sensors are i, j∈S, the equivalent semantic mapping relationship satisfies:

[0063]

[0064] Data sample y to be mined i The associated feature distribution is:

[0065]

[0066] a mn Yes i The envelope amplitude value of the associated information, gmn (t) is the statistical average of the data, and n(t) is the interference noise. The phase space of the sensor data is reconstructed, and the phase trajectory of the data attribute set distribution in the reconstructed phase space is obtained as follows:

[0067]

[0068] The objective function of big data semantic correlation feature fusion is constructed by using the correlation dimension feature extraction method, which is expressed as:

[0069]

[0070] Where p is the Euclidean distance between the reconstructed phase trajectories; for the semantic relevance feature fusion objective function, all sensor data are averaged to obtain the common part of the data features:

[0071]

[0072] For each sensor data, the semantic correlation feature fusion objective function is combined to perform iterative optimization processing to obtain the iterative formula of the independent part of the feature:

[0073]

[0074] In the data fusion stage, the edge agent performs additive fusion on different sensor data in sequence according to the number of steps, and records and stores the data features of each fusion step:

[0075]

[0076] Finally, the obtained fused data is expressed as:

[0077]

[0078] The fusion data reconstruction is as follows: the edge agent converts the fused data Upload to the data center, the data center will fusion data according to the linear properties of distributed compressed sensing and data characteristics as needed Split and mine the required data; the linear property of distributed compressed sensing is expressed as:

[0079]

[0080] Among them, w j is the weight, v is the data additive fusion value;

[0081] The structural expression of fusion data splitting is:

[0082]

[0083] Among them A CSis a compressed sensing information operator; in the process of data splitting, it is first necessary to determine the number and location of the data source sensor based on the source of the required information; then, based on the above results, retrieve the feature vector data set M to obtain the feature set {Θ C ,Θ i}; Finally, based on the feature set of the data, a subtractive operation is performed on the fused data to obtain the required data, which is expressed as:

[0084]

[0085] in and It is the starting and ending position of the data extracted based on the support set features;

[0086] After the split is completed, the edge agent performs inverse sparse transformation to decompress the data and obtain the original data fusion value. For further use:

[0087]

[0088] where F -1 Represents the inverse sparse transform.

[0089] The method proposed in the present invention is mainly used to achieve efficient fusion and processing of transmission line perception data, and support rapid response and real-time decision-making for critical states. The method is based on the existing data processing architecture and sensor network, and only requires limited algorithm optimization and software adjustment to significantly improve the performance of the system; in addition, the method has the characteristics of low power consumption and high efficiency, and will not have a significant impact on the battery life of the perception terminal, while maintaining the data processing speed of the edge device and the stability of the system. Through this lightweight design, the present invention can significantly enhance the overall perception and data processing capabilities of the power transmission Internet of Things system without increasing the hardware burden.

Claims

1. A lightweight design method for multi-source data fusion for battery power supply, characterized in that The invention comprises a cloud-edge collaborative compression network architecture, which is composed of various types of sensor nodes and edge agents. The sensor nodes are deployed at various key locations of power facilities and are responsible for collecting operating data of the environment and equipment. The sensor nodes transmit the sensed data to the adjacent edge agent nodes through a single hop or multi-hop method. Single-hop communication is usually suitable for scenarios where the distance between the sensor node and the edge agent is relatively close. In the case where the sensor nodes are widely distributed or the environment is complex, a multi-hop communication mode is adopted to gradually transmit the data to the edge agent through multiple relay nodes. After the data sent by the sensor node is transmitted to the edge agent, the edge agent first pre-processes the data, including data format conversion, preliminary filtering and compression operations; The edge agent will learn and extract features from the received multi-source data, identify the valuable information in the data, and fuse the data from different sources; after completing the data fusion, the edge agent will transmit the processed data to the cloud data center for subsequent analysis and decision-making; the data center will restore the received data according to the needs of different application scenarios, thereby providing accurate real-time data support for the monitoring and management of the power system.

2. According to claim 1, a multi-source data fusion lightweight design method for battery power supply is characterized in that: The sparse representation of the original data collected by the sensor node is: X j =Z C +Z j ,j∈{1,2,...,J} (1) Among them, J is the size of the data set, that is, the number of sensor nodes in the area; Z C =ΨΘ C ,||Θ C ||0=K C , Z j =ΨΘ j ,||Θ j ||0=K j , signal Z C All X j The common components of C Sparse; signal Z j Yes X j The independent components of j sparse; In the initial state, the sensor nodes perform distributed compressed sensing on the collected data, and the obtained data set is expressed by the following formula: Among them, y j is the M×1-dimensional randomly compressed observation projection vector of perceptual data; j is the perception data of sensor node j, with a size of N×1; Φ j is the random Gaussian observation matrix of node j, with a size of M×N; Ψ is the public sparse basis of N×N dimensions; Θ C and θ j They are the common part and independent part of the sparse support set of node j data, which are also the feature set of data. Then, each node transmits its own data to the edge agent. Since the node random observation matrix Φ j It is generated by the random seed shared by each node and the edge agent. The random compression observation process of the signal is also equivalent to the encryption process. For other nodes that do not know the random seed, the compressed observation signal is a meaningless noise signal. After receiving the observation data from all sensor nodes in the area, the edge agent performs a joint sparse representation on the data set and obtains the sparse components of the data set by solving the optimization problem and linear equations: [Z C ;Z1;Z2;LZ J ](3)。 3. The multi-source data fusion lightweight design method for battery power supply according to claim 2 is characterized in that: Before data fusion, the edge agent performs two-dimensional sparse support set feature M = {Θ C ,Θ P },Θ P Extract the independent part set in the sparse support set of the data, and obtain the spectrum distribution characteristics of the common part of the data set and the independent part of each data in the sparse transform domain; First, the collected and compressed data needs to be reconstructed in phase space; assuming that the attribute set of compressed data samples of any two sets of sensors is i, j∈S, Y is the data sample set, Y T is the transpose of the data sample set, is a random Gaussian measurement matrix, and the equivalent semantic mapping relationship satisfies: Data sample y to be mined i The associated feature distribution is: a mn Yes i The envelope amplitude value of the associated information, g mn (t) is the statistical average of the data, and n(t) is the interference noise. The phase space of the sensor data is reconstructed, and the phase trajectory of the data attribute set distribution in the reconstructed phase space is obtained as follows: The objective function of big data semantic correlation feature fusion is constructed by using the correlation dimension feature extraction method, which is expressed as: Where p is the Euclidean distance between the reconstructed phase trajectories; for the semantic relevance feature fusion objective function, all sensor data are averaged to obtain the common part of the data features: For each sensor data, the semantic correlation feature fusion objective function is combined to perform iterative optimization processing to obtain the iterative formula of the independent part of the feature, Θ i is an element of the feature independence set: In the data fusion stage, the edge agent performs additive fusion on different sensor data in sequence according to the number of steps, and records and stores the data features of each fusion step: Finally, the obtained fused data is expressed as:

4. The multi-source data fusion lightweight design method for battery power supply according to claim 3 is characterized in that: The fusion data reconstruction is as follows: the edge agent converts the fused data Upload to the data center, the data center will fusion data according to the linear properties of distributed compressed sensing and data characteristics as needed Split and mine the required data; the linear property of distributed compressed sensing is expressed as: Among them, w j is the weight, v is the data additive fusion value; The structural expression of fusion data splitting is: Among them A CS is a compressed sensing information operator; in the process of data splitting, it is first necessary to determine the number and location of the data source sensor based on the source of the required information; then, based on the above results, retrieve the feature vector data set M to obtain the feature set {Θ C ,Θ i }; Finally, based on the feature set of the data, a subtractive operation is performed on the fused data to obtain the required data, which is expressed as: in and It is the starting and ending position of the data extracted based on the support set features; After the split is completed, the edge agent performs inverse sparse transformation to decompress the data and obtain the original data fusion value. For further use, l is the total number of sensors, GSS refers to the total capacity value of sensors connected to the gateway, and refers to the sum of the public part and the private part after sparsification: Among them, F -1 Represents the inverse sparse transform.

Images