A data compression method for wireless sensing networks in power distribution networks

Abstract

A data compression method for wireless sensing networks in power distribution networks, belonging to the field of online monitoring and data interaction technology for power equipment status, includes data DCT sparse transform, simplified measurement matrix construction, sensing end data compression, and base station data reconstruction. The method utilizes the classic LEACH algorithm to cluster sensing nodes. For the clustered wireless sensing network, sensing nodes within each cluster transmit monitoring data x over a period of time. i The data is transmitted to the cluster head, which stores a simplified sparse measurement matrix Ф generated based on low-density parity check and compresses the monitoring data sent by each node to obtain compressed data y. After the compressed data is transmitted to the base station, it is reconstructed based on the classic OMP algorithm. This invention is based on compressed sensing theory and combines the characteristics of limited energy and weak computing power of sensing nodes in wireless sensor networks to reduce the amount of data transmitted and alleviate the network burden.

Description

Technical Field

[0001] The present invention belongs to the technical field of on-line monitoring and data interaction of power equipment, and particularly relates to a data compression method for a wireless sensing network for a distribution network. Background Art

[0002] The "last mile" connection in the power system relies on wireless communication technology and sensor network technology. With the deep integration of 5G technology and various applications, the development of the Internet of Things has entered a new stage, realizing a high-coverage, low-latency distribution network multi-parameter data sensing network, and relying on the base station for edge computing data analysis to achieve further equipment insulation state diagnosis and fault location.

[0003] However, with the further deepening of informatization, multi-type data volumes of monitoring devices are transmitted into the substation control terminal in parallel, thus facing a prominent challenge, namely the problem of limited resources of sensing nodes. Generally, sensing nodes are powered by batteries and are small in size, and the computing power, storage capacity, communication bandwidth and power energy of the nodes are very limited. In the sensing network, energy consumption mainly focuses on the transmission process between the sensing nodes and the base station. If each node directly transmits the original data to the base station, it will not only cause waste of bandwidth resources but also lead to network communication delays, affecting the realization of the functions of the entire system.

[0004] In order to increase the transmission efficiency and survival time of sensing nodes as much as possible and extend the life cycle of the wireless sensing network, it is necessary to compress the underlying raw data collected in the sensing network and then decompress it on the base station side for further utilization. Summary of the Invention

[0005] The technical problem to be solved by the present invention is: to provide a data compression method for a wireless sensing network for a distribution network, which is based on the theory of compressive sensing and combines the characteristics of limited energy and weak computing ability of sensing nodes in the wireless sensor network to reduce the data transmission volume and relieve the network burden.

[0006] A data compression method for a wireless sensing network for a distribution network includes the following steps, and the following steps are carried out sequentially. <O000046>

[0007] Step 1. Sparse transformation of base station data

[0008] Use the classical LEACH algorithm to cluster the sensing nodes, and n sensing nodes in each cluster transmit the data x

[0008] ,

[0007] , ,

[0006] , i , ,

[0005] , , , , ,

[0010] , ,

[0009] (i < n) collected within a period of time to the cluster head;

[0009] Step 2. Construction of an extremely simple measurement matrix

[0010] The base station inputs a measurement matrix, transforms it into a sparse measurement matrix, and performs verification to reduce the correlation between the matrix rows and columns; the generated sparse measurement matrix is ​​then sent to the cluster head for storage via the base station.

[0011] Step 3: Data compression at the sensing end

[0012] The cluster head generates a sparse measurement matrix according to the low-density parity check rule to compress the data, resulting in the compressed data y=ФX;

[0013] Step 4: Base Station Data Reconstruction

[0014] In the base station decompressing compressed data, the base station knows the measurement matrix Ф, the DCT (Discrete Cosine) sparse basis, and the compressed data y sent from the cluster head. It uses the classical OMP algorithm to decompress and reconstruct the compressed data.

[0015] The specific operation method of step one is as follows: the data of a certain sensing node over a period of time is represented as x. i =[x i1 ,x i2 ,…x im ] T To facilitate matrix measurement, the data of n nodes within a cluster can be represented as a single matrix. 3D column vector, X=[x i1 ,x i2 ,…x 1m， x 21, x 22 …,x nm ] T .

[0016] The method for generating the sparse measurement matrix in step two is as follows:

[0017] Step 1: The base station inputs a measurement matrix with M rows, N columns, and a number of 1s in each column, satisfying the following conditions: ; Calculate and confirm the number b of 1s in each row, satisfying ;

[0018] Step 2: Randomly generate a sparse matrix Ф in each column containing a 1s and the rest being 0s. Starting from the first row, determine the number r of 1s in that row. If the number of 1s in that row is r > b, move (rb) 1s from that row to other rows where r < b. Then continue the above operation in the next row until all rows have been traversed.

[0019] Step 3: To reduce the correlation between rows and columns, the generated matrix is ​​validated. If the number of rows in which any two columns of the matrix contain the element 1 is no more than 1, the matrix is ​​retained. If the condition is not met, step 2 is repeated.

[0020] Step 4: Send the generated sparse measurement matrix from the base station to the cluster head for storage.

[0021] Through the above design scheme, the present invention can bring the following beneficial effects: A data compression method for a wireless sensing network for a distribution network, by clustering in combination with the LEACH routing algorithm and collecting data compression at the cluster head, can reduce the redundancy of a large amount of data transmission; transfer the process of sparse transformation from the energy-constrained and weakly computationally capable sensing nodes to the base station, reducing the burden on the nodes; aiming at the problems of high computational complexity and large storage occupied by the random measurement matrix in compressive sensing, a minimalist measurement matrix that is easier to compress and store is designed, reducing the amount of data transmission, improving the compression ratio, and at the same time saving energy consumption and extending the network life cycle. Brief Description of the Drawings

[0022] The following further describes the present invention in conjunction with the drawings and specific embodiments:

[0023] Figure 1 It is a block diagram of a wireless sensing network system for a data compression method for a wireless sensing network for a distribution network.

[0024] Figure 2 It is a block diagram of a data compression model for a data compression method for a wireless sensing network for a distribution network.

[0025] Figure 3 It is a flow chart of constructing a minimalist sparse measurement matrix for a data compression method for a wireless sensing network for a distribution network. Specific Embodiments

[0026] A data compression method for a wireless sensing network for a distribution network, as Figure 1-3 shown, includes the following steps, and the following steps are carried out sequentially,

[0027] Step 1: Sparse transformation of base station data

[0028] Use the classic LEACH algorithm to cluster the sensing nodes, and n sensing nodes in each cluster transmit the data x i (i < n) collected within a period of time to the cluster head;

[0029] Step 2: Construction of a minimalist measurement matrix

[0030] The base station inputs the measurement matrix, converts it into a sparse measurement matrix, and performs verification to reduce the correlation between the rows and columns of the matrix; send the generated sparse measurement matrix from the base station to the cluster head for storage;

[0031] Step 3: Data compression at the sensing end

[0032] The cluster head generates a sparse measurement matrix according to the low-density parity-check rule for data compression, and obtains the compressed data y = ФX;

[0033] Step 4: Base Station Data Reconstruction

[0034] In the base station decompressing compressed data, the base station knows the measurement matrix Ф, the DCT (Discrete Cosine) sparse basis, and the compressed data y sent from the cluster head. It uses the classical OMP algorithm to decompress and reconstruct the compressed data.

[0035] The specific operation method of step one is as follows: the data of a certain sensing node over a period of time is represented as x. i =[x i1 ,x i2 ,…x im ] T To facilitate matrix measurement, the data of n nodes within a cluster can be represented as a single matrix. 3D column vector, X=[x i1 ,x i2 ,…x 1m， x 21, x 22 …,x nm ] T .

[0036] The method for generating the sparse measurement matrix in step two is as follows:

[0037] Step 1: The base station inputs a measurement matrix with M rows, N columns, and a number of 1s in each column, satisfying the following conditions: ; Calculate and confirm the number b of 1s in each row, satisfying ;

[0038] Step 2: Randomly generate a sparse matrix Ф in each column containing a 1s and the rest being 0s. Starting from the first row, determine the number r of 1s in that row. If the number of 1s in that row is r > b, move (rb) 1s from that row to other rows where r < b. Then continue the above operation in the next row until all rows have been traversed.

[0039] Step 3: To reduce the correlation between rows and columns, the generated matrix is ​​validated. If the number of rows in which any two columns of the matrix contain the element 1 is no more than 1, the matrix is ​​retained. If the condition is not met, step 2 is repeated.

[0040] Step 4: The generated sparse measurement matrix is ​​sent from the base station to the cluster head for storage.

[0041] Specifically, the sensing nodes within a cluster send data for a certain period of time to the cluster head, and the data to be compressed in a certain cluster head is [x 1 ,x 2 …x 6 ] TBased on the data size, a 4*6 dimensional LDPC coefficient measurement matrix Ф is generated, where each column contains 2 1s, satisfying √M≤a≤M / 2, and the number of rows where any two columns contain 1s is no greater than 1. The compression process is as follows:

[0042]

[0043] The compressed data y here is [y1, y2, y3, y4]. T Then the cluster head sends the compressed data y to the base station.

[0044] Given the observation matrix Ф, the DCT sparse basis applicable to the data, and the data y sent from the cluster head, the base station can use these data to reconstruct the compressed data using the classic OMP algorithm.

Claims

1. A data compression method for power distribution network wireless sensor network, characterized in that: Comprise the following steps, and the following steps are sequentially carried out, Step one, base station data sparse transformation The classic LEACH algorithm is used to cluster the perception nodes, and the n perception nodes in each cluster transmit the data x collected in a period of time to the cluster head i i < n Step two, simple measurement matrix construction The base station input measurement matrix, which is converted into a sparse measurement matrix, and is checked to reduce the correlation between the matrix rows and columns; the generated sparse measurement matrix is sent by the base station to the cluster head for storage; The step two sparse measurement matrix generation method is, Step 1, the base station inputs the number of rows M and the number of columns N of the measurement matrix, and the number a of 1s contained in each column, satisfying ; calculates the number b of 1s contained in each row, satisfying ; Step 2, randomly generate a sparse matrix Ф containing a 1 in each column and the rest of the elements being 0, start from the first row to confirm the number of elements 1 in the row r, if the number of elements r in the row is greater than b, move the (r-b) elements 1 of the row to other rows with r < b, then continue to execute the above operation in the next row until all rows are traversed; Step 3, to reduce the correlation between the rows and columns, the generated matrix is checked, if the number of rows containing elements 1 in any two columns of the matrix is not greater than 1, the matrix is retained, if not, repeat step 2; Step 4, the generated sparse measurement matrix is sent by the base station to the cluster head for storage; Step three, data compression at the sensing end The cluster head generates a sparse measurement matrix according to the low-density parity-check rule to compress data, and obtains compressed data y=ФX; Step four, base station data reconstruction The base station decompresses the compressed data, the base station knows the measurement matrix Ф, the DCT discrete cosine sparse basis and the compressed data y sent from the cluster head, and uses the classic OMP algorithm to realize the decompression and reconstruction of the compressed data.

2. The data compression method for power distribution network wireless sensor network according to claim 1, characterized in that: The specific operation method of step one is that the data of a certain perception node in a period of time is represented as x i =[x i1 ,x i2 ,…x im ] T , in order to facilitate matrix measurement, the data of n nodes in the cluster is represented as a dimension column vector X=[x i1 ,x i2 ,…x 1m， x 21, x 22 …,x nm ] T .

Images