CN-122026957-A - Power line carrier communication channel optimization method based on machine learning

CN122026957ACN 122026957 ACN122026957 ACN 122026957ACN-122026957-A

Abstract

The invention discloses a power line carrier communication channel optimization method based on machine learning, which comprises the following steps of collecting structural response signals of a power line carrier communication channel to generate a channel characteristic sequence, constructing an improved HNN, adding a random disturbance mechanism into the improved HNN to generate the characteristic sequence, executing channel coupling processing based on the characteristic sequence to form a cross-channel coupling characteristic sequence, inputting the multi-time scale characteristic sequence into the improved HNN to perform characteristic updating, applying cross-stage information consistency constraint to form a channel state sequence, constructing an information entropy field, and outputting the optimized power line carrier communication channel state as a channel optimization result. The invention constructs improved HNN and merges a multi-channel, multi-scale and entropy field mechanism to realize accurate analysis and stable optimization of channels, and has the advantages of strong noise immunity, good convergence and high reliability.

Inventors

LIU LIN
ZHOU PENG
CHEN HUOBAO
XIE SHUIYUAN
XIE JINGUI
Wen Gongquan
HUANG XIONGJIE
YANG JIA
CHEN AIYI
WEI JIANBO

Assignees

吉安英佳电子科技有限公司

Dates

Publication Date: 20260512
Application Date: 20260305

Claims (8)

1. The power line carrier communication channel optimization method based on machine learning is characterized by comprising the following steps: Collecting structural response signals of a plurality of sensor channels of a power line carrier communication channel, and preprocessing to generate a channel characteristic sequence; Constructing an improved HNN; inputting the channel characteristic sequence into an improved HNN, and adding a random disturbance mechanism into the improved HNN to generate a characteristic sequence containing random disturbance characteristics; Performing channel coupling processing on the characteristics of the plurality of sensor channels based on the characteristic sequences to form cross-channel coupling characteristic sequences; Performing multi-time scale processing on the cross-channel coupling feature sequence to form a multi-time scale feature sequence, and inputting the multi-time scale feature sequence into an improved HNN for feature updating; applying cross-stage information consistency constraint to the feature updating results of different time stages to form a channel state sequence; and constructing an information entropy field based on the channel state sequence, and outputting the optimized power line carrier communication channel state as a channel optimization result according to the convergence condition of the information entropy field.
2. The machine learning based power line carrier communication channel optimization method of claim 1, wherein the preprocessing specifically comprises denoising, data cleaning, normalization, time alignment, feature filtering and feature enhancement.
3. The power line carrier communication channel optimization method based on machine learning according to claim 1, wherein the constructing an improved HNN specifically includes: Establishing an input feature space based on a channel feature sequence, wherein the input feature space comprises an amplitude change feature, a noise disturbance feature, a time sequence feature and a cross-channel initial association feature; an energy structure framework of the improved HNN is established based on the input feature space, and the energy structure framework comprises a basic energy item, a dissipation energy item and an entropy related energy item which are used as energy components of the improved HNN; taking an energy structure frame as input, establishing an improved HNN internal state representation structure, wherein the improved HNN internal state representation structure comprises a state component set, and receiving the characteristics of an input characteristic space through a fixed mapping relation; Taking the internal state representation structure as a parameter setting basis, constructing an internal parameter set which comprises a parameter structure corresponding to a basic energy item, a dissipation energy item and an entropy related energy item; taking the internal parameter set as input to establish an internal feature processing structure comprising an initial change feature structure, a cross-channel correlation feature structure and a multi-time-scale hierarchical feature structure; taking the input feature space, the energy structure frame, the internal state representation structure and the internal feature processing structure as complete initialization contents, and initializing the improved HNN, wherein the initialization comprises internal structure node setting, parameter setting and mapping relation setting; And taking the initialized improved HNN as a structural basis to complete the construction of the improved HNN.
4. The power line carrier communication channel optimizing method based on machine learning according to claim 1, wherein the generating of the feature sequence specifically comprises: receiving a channel characteristic sequence, inputting an internal state representation structure of the improved HNN, and forming an internal state sequence corresponding to the input characteristic in the internal state representation structure; Forming a disturbance intensity set based on noise information contained in the channel feature sequence; Establishing a disturbance amplitude set based on the disturbance intensity set, and establishing a corresponding disturbance amplitude component for each internal state component, wherein the disturbance amplitude set is kept aligned with the internal state sequence; Establishing a random disturbance source set by taking a disturbance amplitude set as input, wherein the random disturbance source set comprises random disturbance quantity corresponding to each moment, the random disturbance quantity comprises a random variation component and a pulse triggering component, and the random disturbance source set and the disturbance amplitude set keep element-by-element correspondence; And carrying out structural superposition on the random disturbance source set and the channel characteristic sequence, and forming the characteristic sequence containing the random disturbance characteristic based on the superposition structure.
5. The power line carrier communication channel optimization method based on machine learning according to claim 1, wherein the generating of the cross-channel coupling feature sequence specifically comprises: Dividing the characteristic sequence according to a plurality of sensor channels on the basis of the characteristic sequence containing random disturbance characteristics to form a channel characteristic set formed by characteristic components of each channel; Performing inter-channel alignment operation on each channel feature component of the channel feature set to generate a channel alignment structure; Establishing a channel association component set according to the channel alignment structure, wherein the channel association component set is formed by the corresponding relation between each channel and other channels; Forming a coupling weight set according to the channel association component set, wherein the coupling weight set consists of corresponding weight components among different channels and is arranged according to the channels; the coupling weight set corresponds to the channel characteristic set to generate a channel coupling characteristic set, and the channel coupling characteristic set consists of coupling characteristic components arranged according to channels; The channel coupling feature set and the channel feature set are structurally combined to construct a cross-channel combined feature set, and the cross-channel combined feature set is formed by stacking combined feature components in the channel direction; and carrying out dimension unification processing on the basis of the cross-channel combined feature set, and generating a cross-channel coupling feature sequence by arranging the dimensions in time sequence.
6. The power line carrier communication channel optimizing method based on machine learning according to claim 1, wherein the feature updating specifically comprises: based on the cross-channel coupling feature sequence, performing scale division to generate a plurality of time slice index sets corresponding to time scales, wherein each time slice index set consists of a plurality of time window structures; Applying the time segment index set to the time sequence feature set, performing segment division on the time sequence feature set, and generating scale feature segment sets, wherein each scale feature segment set consists of features of continuous time segments; Taking the scale feature fragment set as input, and executing fragment aggregation on the features in the fragments to generate a scale aggregation feature set; aligning the scale aggregation feature sets according to the corresponding relation of the time segments to generate a multi-time scale combination feature set; performing feature dimension reorganization based on the multi-time scale combined feature set to generate a multi-time scale feature sequence; And inputting the multi-time scale feature sequence into an improved HNN, and carrying out feature updating to generate a feature updating sequence.
7. The power line carrier communication channel optimization method based on machine learning according to claim 1, wherein the generating of the channel state sequence specifically comprises: Based on the feature updating sequence, carrying out phase division on the sequence according to the time sequence, and generating a phase set consisting of a plurality of adjacent time phases, wherein each phase set corresponds to a fixed time range; Based on the stage set, carrying out stage merging processing on the feature updating results in each stage, arranging according to the stage sequence, and maintaining the structural relationship consistent with time division to form a stage representation vector set; taking the phase representation vector set as input, performing difference extraction on the phase representation vectors between adjacent phases, and generating a cross-phase difference set; Generating a cross-phase constraint set based on the cross-phase difference set, wherein the cross-phase constraint set consists of constraint results of different components of each phase, and an arrangement mode consistent with the cross-phase difference set is maintained; Structural binding is carried out on the cross-stage constraint set and the feature updating sequence to form a constraint labeling feature sequence; performing cross-stage unification processing on the constraint labeling feature sequence, and performing stage boundary unification trimming on feature updating results at each stage boundary to generate a cross-stage unification feature sequence; A sequence of channel states is formed on the basis of the cross-phase unification characteristic sequence.
8. The power line carrier communication channel optimization method based on machine learning according to claim 1, wherein the generating of the channel optimization result specifically comprises: Carrying out time sequence division on the channel state sequence to generate a stage sequence set, and collecting state records corresponding to each stage into a stage data set; On the basis of the stage data set, the data range of each stage is delimited to generate a state interval set; the state interval set is acted on the stage data set, the corresponding processing of the interval is executed on the data of each stage, and a stage interval mapping set is generated; performing phase statistics processing based on the phase interval mapping set to generate a phase statistics set; the method comprises the steps of integrating a cross-phase structure of a phase statistics set to generate an entropy field structure index, wherein the entropy field structure index is composed of a combined structure of a time phase index and an interval index; integrating statistical components of the stage statistics set on the basis of the entropy field structure index to generate an information entropy field structure, wherein the information entropy field structure consists of statistical structures corresponding to each stage and each interval; And extracting continuous fragments of the information entropy field structure along the time sequence, generating a fragment set, and outputting a channel state sequence corresponding to the continuous fragments in the fragment set as a channel optimization result.

Description

Power line carrier communication channel optimization method based on machine learning Technical Field The invention relates to the field of power line carrier communication channel optimization, in particular to a power line carrier communication channel optimization method based on machine learning. Background In order to cope with the complex factors, the prior art generally adopts single-channel signal analysis, fixed-scale feature processing or a prediction method based on a traditional neural network, but the methods generally depend on a single feature source or a linearization processing framework, are difficult to adapt to channel scenes with multiple channels, strong noise and large time span variation, and in addition, the traditional neural network based on an energy function is mostly based on a fixed energy item structure and lacks dynamic response capability to the multiple source fluctuation factors. Under the multi-sensor collaborative acquisition environment, obvious cross-channel correlation, time scale difference and staged change characteristics exist among channel characteristics, the existing method generally cannot uniformly model the structural characteristics, so that the accuracy and reliability of channel state judgment are insufficient, and the existing technology also lacks a mechanism for comprehensively utilizing cross-channel information, multi-time scale structures and cross-stage consistency characteristics, so that the finally output channel state has the problems of unstable convergence, incomplete information structure, sensitivity to burst interference and the like, and the requirement on channel optimization under the complex power line carrier communication environment is difficult to meet. Disclosure of Invention The invention aims to provide a power line carrier communication channel optimization method based on machine learning, which constructs an improved HNN and merges a multi-channel, multi-scale and entropy field mechanism to realize accurate analysis and stable optimization of a channel and has the advantages of strong noise immunity, good convergence and high reliability. According to the embodiment of the invention, the power line carrier communication channel optimization method based on machine learning comprises the following steps: Collecting structural response signals of a plurality of sensor channels of a power line carrier communication channel, and preprocessing to generate a channel characteristic sequence; Constructing an improved HNN; inputting the channel characteristic sequence into an improved HNN, and adding a random disturbance mechanism into the improved HNN to generate a characteristic sequence containing random disturbance characteristics; Performing channel coupling processing on the characteristics of the plurality of sensor channels based on the characteristic sequences to form cross-channel coupling characteristic sequences; Performing multi-time scale processing on the cross-channel coupling feature sequence to form a multi-time scale feature sequence, and inputting the multi-time scale feature sequence into an improved HNN for feature updating; applying cross-stage information consistency constraint to the feature updating results of different time stages to form a channel state sequence; and constructing an information entropy field based on the channel state sequence, and outputting the optimized power line carrier communication channel state as a channel optimization result according to the convergence condition of the information entropy field. Optionally, the preprocessing specifically includes denoising, data cleaning, normalization, time alignment, feature filtering and feature enhancement. Optionally, the construction of the improved HNN specifically includes: Establishing an input feature space based on a channel feature sequence, wherein the input feature space comprises an amplitude change feature, a noise disturbance feature, a time sequence feature and a cross-channel initial association feature; an energy structure framework of the improved HNN is established based on the input feature space, and the energy structure framework comprises a basic energy item, a dissipation energy item and an entropy related energy item which are used as energy components of the improved HNN; taking an energy structure frame as input, establishing an improved HNN internal state representation structure, wherein the improved HNN internal state representation structure comprises a state component set, and receiving the characteristics of an input characteristic space through a fixed mapping relation; Taking the internal state representation structure as a parameter setting basis, constructing an internal parameter set which comprises a parameter structure corresponding to a basic energy item, a dissipation energy item and an entropy related energy item; taking the internal parameter set as input to establish an internal feature processing