Channel feature extraction method of MIMO-PLC (Multiple Input Multiple Output-Programmable Logic Controller)
By using the channel feature extraction method in the MIMO-PLC system, noise is identified and eliminated, the accuracy of channel estimation is improved, the problem of noise interference and line mutually coupled noise cancellation is solved, and the high reliability of data communication is achieved.
Patent Information
- Application Number
- CN202410930476.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-11
- Publication Date
- 2024-11-19
AI Technical Summary
In MIMO-PLC systems, noise interference, line mutually coupled noise cancellation and channel estimation accuracy are difficult to achieve, which affects the high reliability of data communication.
A channel feature extraction method for MIMO-PLC is proposed. By synchronously collecting noise time series in a three-phase power line noise environment, the noise type is identified by power spectral density estimation analysis, short-time Fourier transform and time-frequency analysis, and the noise characteristics are extracted. parameters and channel estimation is performed after major noise is eliminated.
Improve the accuracy and reliability of channel estimation, ensuring high reliability of data communication.
Smart Images

Figure CN118984173A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of MIMO-PLC communication, and in particular to a channel feature extraction method of MIMO-PLC. Background Art
[0002] Power Line Communication (PLC) is the use of existing power line transmission lines as a transmission medium for data communication. Affected by historical, regional and engineering reasons, power line circuits have complex line structures, impedance characteristics, attenuation characteristics, noise interference, multipath fading and other multiplicative and additive channel transmission characteristics, which greatly affect the quality and capacity of power line communication.
[0003] Traditional PLC is a single-input single-out system that uses differential signals of two wires for data communication. In fact, most household power lines in my country use three-wire power lines including phase line, neutral line, and ground line as the standard. As multi-wire power cables gradually become popular in low-voltage and medium-voltage distribution networks, conditions are provided for multi-input multi-output (MIMO) technology. Compared with SISO, MIMO technology uses multiple transmitting and receiving ports to achieve spatial diversity and multiplexing, and feeds multiple signals at the same time, providing higher transmission rates and system capacity without increasing additional bandwidth and transmission power.
[0004] The main idea of MIMO-PLC technology is to regard the three-phase power lines as multiple transmission channels of the MIMO system and transmit information simultaneously at each channel port. Although my country has not yet formulated relevant technical standards for MIMO-PLC, ITU-T.9963 and HomePlug AV2 standards have incorporated MIMO technology into the technical standards as a key technology for the next generation of PLC. MIMO-PLC also provides an effective solution to alleviate the adverse effects of multipath effects, frequency selective fading, low throughput and other problems in power line communication systems.
[0005] Among them, noise interference elimination, line mutual coupling noise elimination and channel estimation are crucial parts of the MIMO-PLC system. How to effectively estimate the accurate channel frequency response of the common and mutual channels in MIMO-PLC plays a vital role in ensuring the high reliability of data communication and is a topic worthy of study in the MIMO-PLC system. Summary of the invention
[0006] Based on the MIMO-PLC system, noise interference elimination, line mutual coupling noise elimination and channel estimation are crucial parts of the MIMO-PLC system. The present invention proposes a channel feature extraction method for MIMO-PLC, which can effectively identify the inherent noise type, extract noise feature parameters, and perform channel estimation after eliminating the main noise, thereby improving the accuracy and reliability of the estimation.
[0007] To achieve the above object, the technical solution of the present invention is as follows:
[0008] A method for extracting channel characteristics of MIMO-PLC comprises the following steps:
[0009] S1. Synchronously collect noise time series of three channels corresponding to the three-phase power line in a three-phase power line noise environment, and perform data preprocessing;
[0010] S2, using power spectrum density estimation analysis, short-time Fourier transform and time-frequency analysis methods to calculate and process the noise time series data collected in step S1 to obtain time-frequency characteristics; based on the characteristics, the noise category is identified to obtain the noise type of the identification judgment, and the noise characteristic parameters are estimated for different categories to obtain the environmental inherent noise model;
[0011] S3. In a three-phase power line environment, a known signal is sent according to a predetermined time-frequency rule to obtain a signal time acquisition sequence of three channels, and perform data preprocessing;
[0012] S4, using power spectrum density estimation analysis, short-time Fourier transform and time-frequency analysis methods to calculate and process the signal time acquisition sequence data obtained in step S3 to obtain time-frequency characteristics; based on the time-frequency characteristics, the signal acquisition sequence characteristics are matched with the noise type identified in step S2 to eliminate inherent noise;
[0013] S5. Perform channel estimation on the signal time acquisition sequence after noise elimination and the local known signal to obtain a channel transmission estimation model.
[0014] Step S1 specifically includes:
[0015] S1.1 The acquisition between the three power line channels is independent of each other, corresponding to the A\B\C three-phase live wires in the three-phase power line. The acquisition between the three power line channels must be synchronized in time, and the time length of each acquisition data sequence must be at least greater than one power frequency working cycle to obtain complete cycle characteristics.
[0016] S1.2 Data preprocessing includes aligning the energy gains on the three acquisition links and eliminating discontinuous, over-saturated or distorted data samples introduced by the acquisition links.
[0017] Step S2 specifically includes:
[0018] S2. Use power spectral density estimation analysis, short-time Fourier transform, and time-frequency analysis methods to calculate and process the noise time acquisition sequence data to obtain time-frequency features; identify the noise category based on the features to obtain the identified noise type of the decision, and estimate the noise characteristic parameters for different categories to obtain the environmental inherent noise model;
[0019] S2.1 Segment and extract data samples after energy alignment and send them into different calculation branches for calculation in sequence. Define the extraction length as Nk, use Pk as the extraction start position, use Pgap as the extraction position jump step, and represent the number of data segments calculated each time as Nm. In different calculation branches, the parameter values may be the same or different. In different parameter cases, the resolution of frequency-domain processing needs to be normalized.
[0020] S2.2 Power spectral density estimation analysis is to take Nm segments of data with a length of Nk, and the extraction start positions are Pk = k * Pgap, k = 0, 1, 2,..., m; Pgap < Nk. Generally, Pgap takes 1 / 8, 1 / 4, 1 / 2 of Nk. Different Pgap values will result in different power spectrum smoothing effects in time observation. Align and average the Nm segments of data and perform Fourier transform to obtain the power spectral density distribution feature map of the complete data sample.
[0021] S2.3 Short-time Fourier transform is to take Nm segments of data with a length of Nk, and the extraction start positions are Pk = k * Pgap, k = 0, 1, 2,..., m; Pgap = Nk. Perform short-time Fourier transform on each segment of the extracted data, where the length of the Fourier transform is greater than the extraction length Nk to improve the resolution, so as to obtain Nm short-time Fourier transform outputs, corresponding to forming a two-dimensional time-frequency distribution feature map with a time resolution of Nk * sampling interval and a frequency-domain resolution of sampling interval / Fourier transform length.
[0022] S2.4 Time-frequency analysis is mainly based on correlation calculation, Fourier transform, wavelet transform and filter to further extract the main distribution characteristics. Mark the frequency interval where high energy is located in the power spectrum density distribution characteristics of the complete data sample; extract the mutation amount and high energy distribution position under the time-frequency distribution coordinates from the two-dimensional time-frequency distribution characteristics and mark them; select the scale parameter range of the wavelet transform and the wavelet basis design parameter value range in the time-frequency analysis through the marked time-frequency position and resolution; at the same time, design the filter based on the marked frequency domain characteristics to filter out the part outside the marked interval of the frequency domain distribution; for the filtered data, re-take the data segment in the feature mark interval for wavelet transform, take the median as the initial value within the parameter value range, and the left and right dichotomy as the second and third parameter values, and obtain the wavelet transform matching gradient relationship, and take the values in the parameter range in the minimum gradient direction to perform wavelet transform until the best matching parameters are obtained.
[0023] S2.5 identifies noise categories based on features and classifies noise in time-frequency analysis. The main categories include but are not limited to: periodic noise of 50 / 60Hz power frequency, periodic pulse noise, non-periodic pulse noise, and non-periodic pulse extended short-term noise.
[0024] S2.6 Based on the empirical expression of the noise category, the noise type and parameters identified by the time-frequency analysis are normalized to give a set of normalized expressions of the inherent noise of the channel. The normalized expression and the noise type sampling segment are used to perform correlation calculations and Fourier transforms to obtain the power spectrum and energy spectrum of the identified noise type, and extract the eigenvalue set of the corresponding noise type. The eigenvalue set includes but is not limited to the noise peak energy value, time-frequency position, time-frequency period, and duration.
[0025] S2.7 The calculation and processing methods of the noise time acquisition sequences of the three power line channels are consistent, and the calculation and processing are independent of each other. Compare the noise extraction types and characteristic value sets of the three power line channels, and mark the coupling characteristics between the lines. The coupling feature marking includes but is not limited to the correlation coefficient, energy ratio, peak time-frequency position deviation, and peak time-frequency period deviation of the noise frequency domain data.
[0026] Step S3 specifically includes:
[0027] The known signal sent by S3.1 is an OFDM signal, which is filled with a known data sequence on the carrier frequency of the set bandwidth, and the known data sequence in the time domain is obtained through inverse Fourier transform. This data sequence presents a continuous and complete periodic characteristic in the time domain, and the copy and splicing of the data sequence also has a continuous and complete periodic characteristic.
[0028] S3.2 Fixed-time and fixed-frequency rule transmission means that both the frequency-domain and time-domain data sequences of the signal are known, and it exhibits periodic characteristics in the time domain; it is sent in integer multiples of the cycle length on the three-phase circuit, rotated and sent on the three-phase lines in turn, and then a non-transmission cycle is inserted; looking from the time perspective, at the same transmission moment, there are only two situations: no signal is transmitted on all lines and there is a signal on only one of the three-phase lines.
[0029] S3.3 The acquisitions between the three power line channels are independent of each other, corresponding to the A, B, and C phase live wires in the three-phase power line respectively. The acquisitions between the three power line channels must be synchronized in time, and the time length of each acquired data sequence needs to be at least greater than one power frequency working cycle, and it needs to have a certain multiple of the known signal cycle to obtain a complete cycle characteristic.
[0030] S3.4 Data preprocessing includes first making a signal synchronization mark for the signal acquisition sequence. When the data enters each analysis and calculation branch, the data segmentation should start from the synchronization mark, and the segmentation length should be exactly an integer multiple of the known periodic signal cycle.
[0031] S3.5 Data preprocessing also includes performing segmented alignment processing on the energy gains of the known signal received data segments and noise data segments on the three acquisition links, and剔除 non - continuous, over - saturated or distorted data samples introduced by the acquisition links.
[0032] Step S4 specifically includes:
[0033] S4.1 Take data samples in segments from the energy - aligned data, and send them to different calculation branches for calculation in turn. Define the fetching length as Nk, the fetching start position as Pk, the fetching position jump step as Pgap, and the number of data segments calculated each time as Nm. In different calculation branches, the parameter values may be the same or different. In different parameter cases, the resolution of the frequency - domain processing needs to be normalized. The known signal received data segments and noise data segments need to be grouped for processing.
[0034] S4.2 Power spectral density estimation analysis is to take Nm segments of data with a length of Nk, and the fetching start position is Pk = k * Pgap, k = 0, 1, 2,..., m; Pgap < Nk. Generally, Pgap takes 1 / 8, 1 / 4, or 1 / 2 of Nk. Different values of Pgap will result in different power spectrum smoothing effects in time observation. Align and average the Nm segments of data, and perform a Fourier transform to obtain the power spectral density distribution characteristic diagram of the complete data sample.
[0035] S4.3 Short-time Fourier transform is to take data of Nm segments with a length of Nk, and the starting position of the data is Pk=k*Pgap, k=0,1,2,...,m; Pgap=Nk. Perform short-time Fourier transform on each segment, where the length of Fourier transform is greater than the length of data acquisition Nk to improve the resolution, so that Nm short-time Fourier transform outputs can be obtained, corresponding to the formation of a two-dimensional time-frequency distribution feature map with Nk*sampling interval as the time resolution and sampling interval / Fourier transform length as the frequency domain resolution.
[0036] S4.4 Time-frequency analysis is mainly based on correlation calculation, Fourier transform, wavelet transform and filter to further extract the main distribution characteristics. Mark the frequency interval where high energy is located in the power spectrum density distribution characteristics of the complete data sample; extract the mutation amount and high energy distribution position under the time-frequency distribution coordinates from the two-dimensional time-frequency distribution characteristics and mark them; design a filter based on the marked frequency domain characteristics to filter out the part outside the marked interval of the frequency domain distribution; for the filtered data sequence, use the normalized expression of the inherent noise and the noise type given in S2 to take the noise segment for correlation calculation and Fourier transform to determine the noise type and energy in the data sequence.
[0037] S4.5 uses the identified noise type to match the signal acquisition sequence characteristics, mainly including sorting the noise types based on the energy and correlation coefficient of the noise type; copying the noise estimation sequence in the data sequence in sequence according to the sorting, adding a periodic configuration of periodic update during the estimation to reduce the estimation error introduced by the excessive energy of the signal data segment, and obtaining all the noise estimation sequences in the acquisition sequence; subtracting all the noise estimation sequences from the acquisition sequence to obtain the data sequence after eliminating the inherent noise.
[0038] S4.6 Sorting noise types based on their energy and correlation coefficient means sorting the energy and correlation coefficient of the noise types in descending order; the noise type with the largest energy and correlation coefficient is sorted to the first place and determined as the strongest noise; the noise type with the smallest energy and correlation coefficient is sorted to the last place and determined as the weakest noise; if the order of noise types is inconsistent in the energy sorting and correlation coefficient sorting, the weighted coefficients (a1, a2) are multiplied and added to (normalized energy ratio, correlation coefficient) for sorting, where the normalized energy ratio refers to the ratio of all sorted energies divided by the maximum energy.
[0039] S4.7 is based on the coupling characteristics between three-phase power lines. It can directly use the normalized expression of inherent noise and the noise type to take the noise segment for correlation calculation and Fourier transform to determine the noise type and energy in the data sequence. It can omit the power spectrum density estimation analysis, short-time Fourier transform calculation processing, and simplify the time-frequency analysis calculation processing.
[0040] Step S5 specifically includes:
[0041] Channel estimation of the signal time acquisition sequence after noise elimination and the locally known signal includes performing time sliding correlation on the signal time acquisition sequence after noise elimination with the locally known time domain sequence in the time domain to obtain the energy distribution after correlation; at the same time, the signal time acquisition sequence after noise elimination is further denoised by averaging the periodic segments, and then Fourier transform is performed to perform minimum mean square error channel estimation with the locally known frequency domain sequence, and the effective path and maximum multipath delay are determined in combination with the time domain correlation path energy distribution for denoising to obtain a channel transmission estimation model.
[0042] Beneficial effects of the present invention:
[0043] The present invention proposes a MIMO-PLC channel feature extraction method, which can effectively identify the inherent noise type, extract noise feature parameters, and perform channel estimation after eliminating the main noise, thereby improving the accuracy and reliability of the estimation. BRIEF DESCRIPTION OF THE DRAWINGS
[0044] Figure 1 Flow chart of a method for extracting channel characteristics of MIMO-PLC according to an embodiment of the present invention;
[0045] Figure 2 A schematic diagram of sending a certain rule according to an embodiment of the present invention. DETAILED DESCRIPTION
[0046] In order to make the technical problems, technical solutions and advantages of the embodiments of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings. It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.
[0047] The present invention provides a MIMO-PLC channel feature extraction method, the flow chart is as follows Figure 1 , including the following steps:
[0048] S1. Obtain the noise time acquisition sequence of three channels in a three-phase power line noise environment and perform data preprocessing;
[0049] The step S1 also includes:
[0050] S1.1 The acquisition between the three power line channels is independent of each other, corresponding to the A\B\C three-phase live wires in the three-phase power line. The acquisition between the three power line channels must be synchronized in time, and the time length of each acquisition data sequence must be at least greater than one power frequency working cycle to obtain complete cycle characteristics.
[0051] S1.2 Data preprocessing includes aligning the energy gains on the three acquisition links and removing discontinuous, oversaturated, or distorted data samples introduced by the acquisition links.
[0052] S2. Use power spectral density estimation analysis, short-time Fourier transform, and time-frequency analysis methods to calculate the noise time acquisition sequence data to obtain time-frequency features; identify the noise category based on the features to obtain the identified noise type of the decision, and estimate the noise characteristic parameters for different categories to obtain the environmental inherent noise model;
[0053] In the step S2, it also includes
[0054] S2.1 Segment and extract data samples from the energy-aligned data, and send them to different calculation branches for calculation in sequence. Define the extraction length as Nk, the extraction start position as Pk, and the extraction position jump step as Pgap. Each time the number of data segments for calculation is represented by Nm. In different calculation branches, the parameter values may be the same or different. In different parameter cases, the resolution of the frequency-domain processing needs to be normalized.
[0055] S2.2 Power spectral density estimation analysis is to extract data with a length of Nk for Nm segments, with the extraction start position Pk = k * Pgap, k = 0, 1, 2,..., m; Pgap < Nk. Generally, Pgap takes 1 / 8, 1 / 4, or 1 / 2 of Nk. Different Pgap values will result in different power spectral smoothing effects in time observation. Align and average the Nm segments of data, and perform Fourier transform to obtain the power spectral density distribution feature map of the complete data sample.
[0056] S2.3 Short-time Fourier transform is to extract data with a length of Nk for Nm segments, with the extraction start position Pk = k * Pgap, k = 0, 1, 2,..., m; Pgap = Nk. Perform short-time Fourier transform on each extracted segment, where the length of the Fourier transform is greater than the extraction length Nk to improve the resolution, so as to obtain Nm short-time Fourier transform outputs, corresponding to forming a two-dimensional time-frequency distribution feature map with a time resolution of Nk * sampling interval and a frequency-domain resolution of sampling interval / Fourier transform length.
[0057] S2.4 Time-frequency analysis is mainly based on correlation calculation, Fourier transform, wavelet transform and filter to further extract the main distribution characteristics. Mark the frequency interval where high energy is located in the power spectrum density distribution characteristics of the complete data sample; extract the mutation amount and high energy distribution position under the time-frequency distribution coordinates from the two-dimensional time-frequency distribution characteristics and mark them; select the scale parameter range of the wavelet transform and the wavelet basis design parameter value range in the time-frequency analysis through the marked time-frequency position and resolution; at the same time, design the filter based on the marked frequency domain characteristics to filter out the part outside the marked interval of the frequency domain distribution; for the filtered data, re-take the data segment in the feature mark interval for wavelet transform, take the median as the initial value within the parameter value range, and the left and right dichotomy as the second and third parameter values, and obtain the wavelet transform matching gradient relationship, and take the values in the parameter range in the minimum gradient direction to perform wavelet transform until the best matching parameters are obtained.
[0058] The energy spectrum characteristics are expressed in the following calculation formula in this specific implementation:
[0059]
[0060] Among them, S E (f,I i ) is the first i The energy distribution of a data sample in the frequency domain coordinates; N corresponds to the number of sample points in the symbol period; x(k,I i ) is the first i The kth sample value of a data sample sequence; N cap is the number of data samples.
[0061] The power spectrum density feature is expressed by the following calculation formula in this specific implementation:
[0062]
[0063] Among them, S E (f,I i ) is the first i The energy distribution of a data sample in the frequency domain coordinates; N corresponds to the number of sample points in the symbol period; x(k,I i ) is the first i The kth sample value of a data sample sequence.
[0064] The two-dimensional time-frequency distribution feature is expressed by the following calculation formula in this specific implementation:
[0065]
[0066] Among them, S x (t,f,I i ) is the first iThe energy distribution of a data sample in the two-dimensional time-frequency coordinate (t, f); N corresponds to the number of sample points in the symbol period; x(t, I i ) is the first i data sample sequence; ω is the resolution in the frequency domain.
[0067] In the present specific implementation, the frequency interval where high energy is located in the power spectrum density distribution characteristics of the complete data sample is marked, and a value higher than 1.2 times of the peak energy of the noise is set for marking.
[0068] S2.5 identifies noise categories based on features and classifies noise in time-frequency analysis. The main categories include but are not limited to: periodic noise of 50 / 60Hz power frequency, periodic pulse noise, non-periodic pulse noise, and non-periodic pulse extended short-term noise.
[0069] S2.6 Based on the empirical expression of the noise category, the noise type and parameters identified by the time-frequency analysis are normalized to give a set of normalized expressions of the inherent noise of the channel. The normalized expression and the noise type sampling segment are used to perform correlation calculations and Fourier transforms to obtain the power spectrum and energy spectrum of the identified noise type, and extract the eigenvalue set of the corresponding noise type. The eigenvalue set includes but is not limited to the noise peak energy value, time-frequency position, time-frequency period, and duration.
[0070] S2.7 The calculation and processing methods of the noise time acquisition sequences of the three power line channels are consistent, and the calculation and processing are independent of each other. Compare the noise extraction types and characteristic value sets of the three power line channels, and mark the coupling characteristics between the lines. The coupling feature marking includes but is not limited to the correlation coefficient, energy ratio, peak time-frequency position deviation, and peak time-frequency period deviation of the noise frequency domain data.
[0071] S3. In a three-phase power line environment, a known signal is sent according to a certain time-frequency rule to obtain a signal time acquisition sequence of three channels and perform data preprocessing;
[0072] The step S3 also includes:
[0073] The known signal sent by S3.1 is an OFDM signal, which is filled with a known data sequence on the carrier frequency of the set bandwidth, and the known data sequence in the time domain is obtained through inverse Fourier transform. This data sequence presents a continuous and complete periodic characteristic in the time domain, and the copy and splicing of the data sequence also has a continuous and complete periodic characteristic.
[0074] S3.2 Fixed-time and frequency rule transmission means that both the frequency-domain and time-domain data sequences of the signal are known, and it exhibits periodic characteristics in the time domain; at integer multiples of the cycle length in a three-phase circuit, it is sequentially transmitted on the three-phase lines and then a non-transmission cycle is inserted; looking at the time, at the same transmission moment, there are only two situations: no signal is transmitted on all lines and there is a signal on only one of the three-phase lines.
[0075] The acquisitions among the three power line channels are independent of each other, corresponding to the three live wires A, B, and C in the three-phase power lines respectively. The acquisitions among the three power line channels must be synchronized in time, and the time length of each acquisition data sequence should be at least greater than one power frequency working cycle, and it is required to have a certain multiple of the known signal cycle to obtain complete cycle characteristics.
[0076] S3.4 Data preprocessing includes first making signal synchronization marks on the signal acquisition sequence. When the data enters each analysis and calculation branch, the data segmentation should start from the synchronization mark, and the segmentation length should be exactly an integer multiple of the known periodic signal cycle.
[0077] S3.5 Data preprocessing also includes performing segmented alignment processing on the energy gains of the known signal received data segments and noise data segments on the three acquisition links, and eliminating discontinuous, over-saturated or distorted data samples introduced by the acquisition links.
[0078] S4. Use power spectral density estimation analysis, short-time Fourier transform, and time-frequency analysis methods to perform calculation processing on the signal time acquisition sequence data to obtain time-frequency characteristics; based on the characteristics, use the identified noise types to match the signal acquisition sequence characteristics to eliminate inherent noise;
[0079] In the step S4, it also includes
[0080] S4.1 Take data in segments from the energy-aligned data samples and send them into different calculation branches for calculation in turn. Define the data-taking length as Nk, with Pk as the data-taking starting position, Pgap as the data-taking position jump step, and the number of data segments for each calculation is represented by Nm. In different calculation branches, the parameter values may be the same or different. In different parameter cases, the resolution of the frequency-domain processing needs to be normalized. The known signal received data segments and noise data segments need to be grouped for processing.
[0081] S4.2 Power spectral density estimation analysis is to take Nm segments of data with a length of Nk, and the data-taking starting positions are Pk = k * Pgap, k = 0, 1, 2,..., m; Pgap < Nk. Generally, Pgap takes 1 / 8, 1 / 4, or 1 / 2 of Nk. Different Pgap values will result in different power spectral smoothing effects in time observation. Align and average the Nm segments of data and perform Fourier transform to obtain the power spectral density distribution characteristic diagram of the complete data sample.
[0082] S4.3 Short-time Fourier transform is to take data of Nm segments with a length of Nk, and the starting position of the data is Pk=k*Pgap, k=0,1,2,...,m; Pgap=Nk. Perform short-time Fourier transform on each segment, where the length of Fourier transform is greater than the length of data acquisition Nk to improve the resolution, so that Nm short-time Fourier transform outputs can be obtained, corresponding to the formation of a two-dimensional time-frequency distribution feature map with Nk*sampling interval as the time resolution and sampling interval / Fourier transform length as the frequency domain resolution.
[0083] S4.4 Time-frequency analysis is mainly based on correlation calculation, Fourier transform, wavelet transform and filter to further extract the main distribution characteristics. Mark the frequency interval where high energy is located in the power spectrum density distribution characteristics of the complete data sample; extract the mutation amount and high energy distribution position under the time-frequency distribution coordinates from the two-dimensional time-frequency distribution characteristics and mark them; design a filter based on the marked frequency domain characteristics to filter out the part outside the marked interval of the frequency domain distribution; for the filtered data sequence, use the normalized expression of the inherent noise and the noise type given in S2 to take the noise segment for correlation calculation and Fourier transform to determine the noise type and energy in the data sequence.
[0084] S4.5 uses the identified noise type to match the signal acquisition sequence characteristics, mainly including sorting the noise types based on the energy and correlation coefficient of the noise type; copying the noise estimation sequence in the data sequence in sequence according to the sorting, adding a periodic configuration of periodic update during the estimation to reduce the estimation error introduced by the excessive energy of the signal data segment, and obtaining all the noise estimation sequences in the acquisition sequence; subtracting all the noise estimation sequences from the acquisition sequence to obtain the data sequence after eliminating the inherent noise.
[0085] The two-dimensional time-frequency distribution characteristics based on the inherent noise are expressed by the following calculation formula in this specific implementation:
[0086]
[0087] Among them, S x (t,f,I j ,s j ) is the first j The data samples are compared with the noise reference sequence s in the two-dimensional time-frequency coordinate (t,f). j (nT,I0) is used to match the possible energy distribution; N corresponds to the number of sample points in the symbol period; ω is the resolution in the frequency domain.
[0088] S4.6 Sorting noise types based on their energy and correlation coefficient means sorting the energy and correlation coefficient of the noise types in descending order; the noise type with the largest energy and correlation coefficient is sorted to the first place and determined as the strongest noise; the noise type with the smallest energy and correlation coefficient is sorted to the last place and determined as the weakest noise; if the order of noise types is inconsistent in the energy sorting and correlation coefficient sorting, the weighted coefficients (a1, a2) are multiplied and added to (normalized energy ratio, correlation coefficient) for sorting, where the normalized energy ratio refers to the ratio of all sorted energies divided by the maximum energy.
[0089] S4.7 is based on the coupling characteristics between three-phase power lines. It can directly use the normalized expression of inherent noise and the noise type to take the noise segment for correlation calculation and Fourier transform to determine the noise type and energy in the data sequence. It can omit the power spectrum density estimation analysis, short-time Fourier transform calculation processing, and simplify the time-frequency analysis calculation processing.
[0090] S5. Perform channel estimation on the signal time acquisition sequence after noise elimination and the local known signal to obtain a channel transmission estimation model.
[0091] The step S5 also includes:
[0092] Channel estimation of the signal time acquisition sequence after noise elimination and the locally known signal includes performing time sliding correlation on the signal time acquisition sequence after noise elimination and the locally known time domain sequence in the time domain to obtain the correlated path energy distribution after correlation; at the same time, the signal time acquisition sequence after noise elimination is averaged by periodic segments for further noise reduction, and then Fourier transform is performed to perform minimum mean square error channel estimation with the locally known frequency domain sequence, and combined with the time domain correlated path energy distribution, the effective path and the maximum multipath delay are determined for denoising to obtain a channel transmission estimation model.
[0093] The background section of the present invention may contain background information about the problem or environment of the present invention, but does not necessarily describe the prior art. Therefore, the content included in the background section is not an admission by the applicant that the prior art is available.
[0094] The above content is a further detailed description of the present invention in combination with specific / preferred embodiments, and it cannot be determined that the specific implementation of the present invention is limited to these descriptions. For ordinary technicians in the technical field to which the present invention belongs, without departing from the concept of the present invention, it can also make several substitutions or modifications to these described embodiments, and these substitutions or modifications should be regarded as belonging to the protection scope of the present invention. In the description of this specification, the description of the reference terms "an embodiment", "some embodiments", "preferred embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in combination with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily target the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In the absence of mutual contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples. Although the embodiments of the present invention and their advantages have been described in detail, it should be understood that various changes, substitutions and modifications can be made herein without departing from the scope of protection of the patent application.
Claims
1. A method for extracting channel characteristics of MIMO-PLC, comprising the following steps: S1. Synchronously collect the noise time series of three channels corresponding to the three-phase power line in the three-phase power line noise environment, and perform data preprocessing; S2, using power spectrum density estimation analysis, short-time Fourier transform and time-frequency analysis methods to calculate and process the noise time series data collected in step S1 to obtain time-frequency characteristics; Based on the characteristics, identify the noise category, obtain the identified noise type, and estimate the noise characteristic parameters for different categories to obtain the environmental inherent noise model; S3. In the three-phase power line environment, send a known signal according to a predetermined time-frequency rule, obtain the signal time acquisition sequence of the three channels, and perform data preprocessing; S4. Use power spectral density estimation analysis, short-time Fourier transform, and time-frequency analysis methods to calculate and process the signal time acquisition sequence data obtained in step S3 to obtain time-frequency characteristics; based on the time-frequency characteristics, match the signal acquisition sequence characteristics with the noise type identified in step S2 to eliminate the inherent noise; S5. Perform channel estimation on the signal time acquisition sequence after noise elimination and the local known signal to obtain a channel transmission estimation model.
2. The channel feature extraction method according to claim 1, wherein: In the said step S1: S1.1 The acquisitions between the three power line channels are independent of each other, corresponding to the A, B, and C phase live wires in the three-phase power line respectively. The time length of each acquisition data sequence of the three power line channels is at least greater than one power frequency working cycle to obtain complete cycle characteristics; S1.2 The data preprocessing includes aligning the energy gains on the three acquisition links and剔除 the discontinuous, oversaturated, or distorted data samples introduced by the acquisition links.
3. The channel feature extraction method according to claim 1 or 2, characterized in that: In the said step S2: S2.1 Segment and extract data samples from the energy-aligned data samples, and send them to different calculation branches for calculation in turn; define the extraction length as Nk, use Pk as the extraction start position, use Pgap as the extraction position jump step, and use Nm to represent the number of data segments calculated each time; under different parameter conditions, normalize the resolution of the frequency domain processing; S2.2 Power spectral density estimation analysis is to extract data from Nm segments of data with a length of Nk, with the extraction start position Pk = k*Pgap, k = 0, 1, 2,..., m; Pgap < Nk; align and average the Nm segments of data, and perform Fourier transform to obtain the power spectral density distribution characteristic map of the complete data sample; S2.3 Short-time Fourier transform is to extract data from Nm segments of data with a length of Nk, with the extraction start position Pk = k*Pgap, k = 0, 1, 2,..., m; Pgap = Nk; perform short-time Fourier transform on each extracted segment, where the length of the Fourier transform is greater than the extraction length Nk to improve the resolution, so as to obtain Nm short-time Fourier transform outputs, corresponding to forming a two-dimensional time-frequency distribution characteristic map with a time resolution of Nk*sampling interval and a frequency domain resolution of sampling interval / Fourier transform length; S2.4 Time-frequency analysis is based on correlation calculation, Fourier transform, wavelet transform and filter to further extract the main distribution characteristics; mark the frequency interval where high energy is located in the power spectrum density distribution characteristics of the complete data sample; extract the mutation amount and high energy distribution position under the time-frequency distribution coordinates from the two-dimensional time-frequency distribution characteristics and mark them; select the scale parameter range of the wavelet transform and the wavelet basis design parameter value range in the time-frequency analysis through the marked time-frequency position and resolution; design a filter based on the marked frequency domain characteristics to filter out the part outside the marked interval of the frequency domain distribution; For the filtered data, re-take the data segment within the feature mark interval for wavelet transform, take the median value as the initial value within the parameter value range, and use the left and right dichotomy as the second and third parameter values to obtain the wavelet transform matching gradient relationship, and take the minimum gradient direction in the parameter range to perform wavelet transform until the best matching parameters are obtained; S2.5 Identify noise categories based on features and classify noise in time-frequency analysis; S2.6 Based on the empirical expression of noise categories, the noise types and parameters identified by time-frequency analysis are normalized to give a set of normalized expressions of channel inherent noise; Using the normalized expression and the noise type sampling segment to perform correlation calculation and Fourier transform, the power spectrum and energy spectrum of the identified noise type are obtained, and the eigenvalue set corresponding to the noise type is extracted; S2.7 The calculation and processing methods of the noise time acquisition sequences of the three power line channels are consistent, and the calculation and processing are independent of each other. The power line channel noise extraction types and characteristic value sets between the three channels are compared, and the coupling characteristics between the lines are marked.
4. The channel feature extraction method according to claim 3, wherein: In step S2.2, Pgap is 1 / 8, 1 / 4 or 1 / 2 of Nk.
5. The channel feature extraction method according to claim 3, wherein: In step S2.5, the categories include: periodic noise of 50 / 60 Hz power frequency, periodic pulse noise, non-periodic pulse noise, and non-periodic pulse extended short-term noise; In step S2.6, the characteristic value set includes noise peak energy value, time-frequency position, time-frequency period, and duration; In step S2.7, the coupling characteristic marks include the correlation coefficient, energy ratio, peak time-frequency position deviation, and peak time-frequency period deviation of the noise frequency domain data.
6. The channel feature extraction method according to any one of claims 1 to 5, characterized in that: In step S3: The known signal sent by S3.1 is an OFDM signal, which is filled with a known data sequence on the carrier frequency of the set bandwidth, and the time domain known data sequence is obtained by inverse Fourier transform; S3.2 Certain time-frequency rule transmission means that the signal frequency domain and time domain data sequences are known and present periodic characteristics in the time domain; after the three-phase circuit is transmitted in turn on the three-phase line with an integer multiple of the period length, a non-transmission period is inserted; from the perspective of time, at the same transmission time, there are only two situations: no transmission signal on all lines and signal on only one phase line; S3.3 The acquisition between the three power line channels is independent of each other, corresponding to the A\B\C three-phase live wires in the three-phase power line respectively. The acquisition between the three power line channels is synchronized in time, and the time length of each acquisition data sequence is at least greater than one power frequency working cycle, and has a certain multiple of the known signal cycle to obtain complete periodic characteristics; S3.4 Data preprocessing includes first making signal synchronization marks for the signal acquisition sequence. When data enters each analysis and calculation branch, data segmentation should start from the synchronization marks, and the segmentation length should be exactly an integer multiple of the known periodic signal period. S3.5 Data preprocessing also includes performing segmented alignment processing on the energy gains of the known signal received data segments and noise data segments on the three acquisition links, and eliminating discontinuous, oversaturated, or distorted data samples introduced by the acquisition links.
7. The channel feature extraction method according to any one of claims 1 to 6, characterized in that: In the said step S4: S4.1 Take data samples in segments from the energy-aligned data, and send them to different calculation branches for calculation in sequence; define the fetch length as Nk, use Pk as the fetch start position, use Pgap as the fetch position jump step, and use Nm to represent the number of data segments calculated each time; under different parameter conditions, perform normalization processing on the resolution of frequency domain processing; group and process the known signal received data segments and noise data segments. S4.2 Power spectral density estimation analysis is to take data of length Nk in Nm segments, with the fetch start position Pk = k * Pgap, k = 0, 1, 2,..., m; Pgap < Nk; align and average the Nm segments of data, perform Fourier transform, and obtain the power spectral density distribution characteristic map of the complete data sample. S4.3 Short-time Fourier transform is to take data of length Nk in Nm segments, with the fetch start position Pk = k * Pgap, k = 0, 1, 2,..., m; Pgap = Nk; perform short-time Fourier transform on each fetched segment, where the length of the Fourier transform is greater than the fetch length Nk to improve the resolution, so as to obtain Nm short-time Fourier transform outputs, corresponding to forming a two-dimensional time-frequency distribution characteristic map with Nk * sampling interval as the time resolution and sampling interval / Fourier transform length as the frequency domain resolution. S4.4 Time-frequency analysis is mainly based on correlation calculation, Fourier transform, wavelet transform, and filters to further extract the main distribution characteristics; mark the frequency interval where the high energy is located in the power spectral density distribution characteristic of the complete data sample; extract the mutation amount and high energy distribution position under the time-frequency distribution coordinates from the two-dimensional time-frequency distribution characteristics for marking; design a filter based on the marked frequency domain characteristics to filter out the part outside the marked interval of the frequency domain distribution. For the filtered data sequence, use the inherent noise normalization expression and noise type given in S2 to take the noise number segments for correlation calculation and Fourier transform to determine the noise type and energy in the data sequence. S4.5 Matching the signal acquisition sequence characteristics with the identified noise type mainly includes sorting the noise types based on the energy and correlation coefficient magnitude of the noise type; sequentially copy the noise estimation sequences in the data sequence according to the sorting, and add a period configuration with periodic update during the estimation to reduce the estimation error introduced by the excessive energy of the signal data segment, so as to obtain all the noise estimation sequences in the acquisition sequence. Subtract all the noise estimation sequences from the acquisition sequence to obtain the data sequence after eliminating the inherent noise. S4.6 Sorting noise types based on the energy and correlation coefficient of the noise types refers to sorting the energy and correlation coefficient of the noise types in descending order; the noise type with the largest energy and correlation coefficient is sorted to the first rank and determined as the strongest noise; the noise type with the smallest energy and correlation coefficient is sorted to the last rank and determined as the weakest noise; if the energy sorting and correlation coefficient sorting show inconsistent order of noise types, the weighted coefficients a1 and a2 are multiplied and added to the normalized energy ratio and the correlation coefficient respectively for sorting, where the normalized energy ratio refers to the ratio obtained by dividing all sorted energies by the maximum energy; S4.7 is based on the coupling characteristics between the three-phase power lines, uses the normalized expression of the inherent noise and the noise type to take the noise segment for correlation calculation and Fourier transform to determine the noise type and energy in the data sequence.
8. The channel feature extraction method according to claim 7, wherein: Pgap is 1 / 8, 1 / 4 or 1 / 2 of Nk.
9. The channel feature extraction method according to any one of claims 1 to 8, characterized in that: In step S5: Channel estimation of the signal time acquisition sequence after noise elimination and the locally known signal includes performing time sliding correlation on the signal time acquisition sequence after noise elimination and the locally known time domain sequence in the time domain to obtain the correlated path energy distribution after correlation; at the same time, the signal time acquisition sequence after noise elimination is averaged by periodic segments for further noise reduction, and then Fourier transform is performed to perform minimum mean square error channel estimation with the locally known frequency domain sequence, and combined with the time domain correlated path energy distribution, the effective path and the maximum multipath delay are determined for denoising to obtain a channel transmission estimation model.
10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the channel feature extraction method according to any one of claims 1 to 9 is implemented.
Citation Information
Cited By
Smart grid communication noise analysis and ground noise visualization method
CN121842019A
