Adaptive optimization configuration method and system for power line carrier communication relay node

Abstract

The application provides a method and system for adaptive optimization configuration of a power line carrier communication relay node. The method reduces data deviation by obtaining multi-dimensional data, extracts features from the multi-dimensional data, obtains channel feature vectors, link quality score data, harmonic interference intensity data, and comprehensive health index data, infers channel capacity prediction data, and constructs a power carrier network graph to dynamically adjust the relay node configuration, thereby improving the stability of transmission performance, integrating the interaction between transmission performance, energy efficiency, and robustness, reducing the risk of decision-making, and performing closed-loop feedback to ensure that the relay node configuration scheme meets the long-term operation requirements, thereby reducing manual intervention and improving the stability and efficiency of communication transmission through the adaptive optimization configuration method of the power line carrier communication relay node.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of electric communication, in particular to a method and system for adaptive optimization configuration of a relay node based on power line carrier communication. BACKGROUND

[0002] With the rapid development of smart grid, smart home and other fields, power line carrier communication has become one of the key technologies for realizing device interconnection due to its advantages such as no need for re-wiring and wide coverage. However, PLC (Power Line Communication) networks often face problems such as strong power grid harmonic interference, serious channel attenuation, and dynamic changes in node state, which result in defects such as unstable transmission performance and low energy efficiency of the traditional fixed relay node configuration method.

[0003] Therefore, there is an urgent need for a method for dynamically optimizing relay node configuration to improve the stability and efficiency of PLC network data transmission. SUMMARY

[0004] In view of the above-mentioned problems, in combination with the first aspect of the present application, the embodiments of the present application provide a method for adaptive optimization configuration of a relay node based on power line carrier communication, which comprises:

[0005] S1: acquiring channel probe data, performance metric data, power grid operation state data and device ontology state data and performing data preprocessing to obtain a power carrier data set;

[0006] S2: performing feature extraction and fusion on the power carrier data set to obtain channel feature vectors, link quality score data, harmonic interference intensity data and comprehensive health index data;

[0007] S3: inferring based on the channel feature vectors, link quality score data and harmonic interference intensity data to obtain channel capacity prediction data, and constructing a power carrier network graph in combination with the comprehensive health index data;

[0008] S4: constructing a revenue function based on the power carrier network graph and the relay node configuration problem, and solving the revenue function to obtain a relay node configuration optimal data set, making a decision on the relay node configuration optimal data set to obtain relay node configuration optimal data;

[0009] S5: issuing the relay node configuration optimal data to the corresponding relay node for feedback verification to obtain stable relay node configuration data.

[0010] As a further scheme of the present application, acquiring channel probe data, performance metric data, power grid operation state data and device ontology state data and performing data preprocessing to obtain a power carrier data set comprises:

[0011] Obtain channel impulse response data and signal-to-noise ratio spectrum data of each link in the PLC network and form channel sounding data;

[0012] Obtain message success rate data and network average delay data of each node in the PLC network and form performance metric data;

[0013] Obtain real-time load current data as grid operating state data;

[0014] Obtain remaining power data, CPU load rate data, and historical average fault interval event data and form device intrinsic state data;

[0015] Perform data preprocessing operations on the channel sounding data, performance metric data, grid operating state data, and device intrinsic state data to obtain a power line carrier dataset.

[0016] As a further scheme of the present application, feature extraction and fusion are performed on the power line carrier dataset to obtain channel feature vectors, link quality score data, harmonic interference intensity data, and comprehensive health index data, including:

[0017] Perform principal component analysis and feature extraction on the channel sounding data to obtain channel feature vectors;

[0018] Perform weighted average calculation on the message success rate data and network average delay data in the performance metric data to obtain link quality score data;

[0019] Perform Fourier transform operations on the real-time load current data in the grid operating state data to obtain harmonic interference data;

[0020] Perform discreteness and importance analysis on the device intrinsic state data based on an entropy weight method to obtain objective weight coefficients;

[0021] Fuse the remaining power data, CPU load rate data, and historical average fault interval event data in the device intrinsic state data based on the objective weight coefficients to obtain comprehensive health index data.

[0022] As a further scheme of the present application, principal component analysis and feature extraction are performed on the channel sounding data to obtain channel feature vectors, including:

[0023] Perform principal component analysis and feature extraction operations on the channel impulse response data and signal-to-noise ratio spectrum data in the channel sounding data to obtain projection coefficients of the channel impulse response data dimension, time distribution characteristics of dominant multipath components, principal component contribution degrees, and distribution variances of the signal-to-noise ratio spectrum data in the principal component space;

[0024] The dominant multipath component amplitude is obtained based on the projection coefficient of the channel impulse response data and its dimension, the time delay spread is obtained based on the time distribution characteristics of the dominant multipath component and the principal component contribution degree, the frequency selective fading factor is obtained based on the distribution variance of the signal-to-noise ratio spectrum data in the principal component space, and the dominant multipath component amplitude, the time delay spread and the frequency selective fading factor are combined to form a channel feature vector.

[0025] The principal component analysis retains the principal components with a cumulative contribution rate greater than 90% to obtain the channel feature vector.

[0026] As a further scheme of the application, based on the channel feature vector, the link quality score data and the harmonic interference intensity data, channel capacity prediction data is obtained by inference, and a power carrier network graph is constructed in combination with the comprehensive health index data, including:

[0027] The channel feature vector, the link quality score data and the harmonic interference intensity data are taken as inputs, and based on the pre-constructed lightweight neural network model, inference is performed to obtain channel capacity prediction data.

[0028] All communication nodes in the PLC network and corresponding comprehensive health index data are obtained, each communication node is identified by ID, and the comprehensive health index data is taken as the point weight of the corresponding communication node, and a communication node set with point weight is obtained.

[0029] The communication links between nodes and corresponding channel capacity prediction data are obtained, each communication link is identified by a directed attribute based on the bidirectional communication relationship, and the channel capacity prediction data is taken as the edge weight of the corresponding communication link, and a communication link set with edge weight is obtained.

[0030] The time granularity and timestamp information are set, the communication node set with point weight and the communication link set with edge weight are periodically updated and synchronously marked with timestamps, and a node link set is obtained.

[0031] The node link set is integrated into a graph structure to obtain a power carrier network graph.

[0032] As a further scheme of the application, based on the power carrier network graph and the relay node configuration problem, a revenue function is constructed, and the revenue function is solved to obtain a relay node configuration optimal data set, and the relay node configuration optimal data set is decided to obtain relay node configuration optimal data, including:

[0033] The network topology and traffic demand data are obtained, the network basic maximum flow without relay node configuration is obtained in combination with the power carrier network graph, the network maximum flow after relay node configuration is obtained in combination with different relay node configuration schemes, the network maximum flow improvement degree is obtained based on the network basic maximum flow and the network maximum flow, and the first revenue function is constructed based on the network maximum flow improvement degree.

[0034] obtaining the relay nodes and corresponding power consumption parameters, obtaining total power consumption of activated relays in the entire network based on the number of activated relay nodes in the power carrier network graph and the power consumption parameters of each relay node, and constructing a second benefit function based on the inverse of the total power consumption of activated relays in the entire network;

[0035] The activated relay nodes represent the relay nodes in the working state.

[0036] obtaining the comprehensive health index data of the activated relay nodes in the power carrier network graph and adding them up to obtain total health data, and constructing a third benefit function based on the total health data;

[0037] constructing a multi-benefit model based on the first, second and third benefit functions, and solving the multi-benefit model to obtain an optimal data set of relay node configuration;

[0038] Based on the fusion regret theory and Choquet integral, the optimal data set of relay node configuration is screened to obtain the optimal data of relay node configuration.

[0039] As a further scheme of the present application, a multi-benefit model is constructed based on the first, second and third benefit functions, and the multi-benefit model is solved to obtain an optimal data set of relay node configuration, comprising:

[0040] obtaining the balance weight parameters of the first, second and third benefit functions, and constructing an Euler-Lagrange equation based on the first, second and third benefit functions, and obtaining an initial multi-benefit model based on the Euler-Lagrange equation;

[0041] obtaining each relay node configuration scheme and relay node data n, mapping each relay node configuration scheme to a high-dimensional state space to obtain a state point set in the n-dimensional state space;

[0042] obtaining the Onsager dissipation function and the small perturbation parameter, and introducing them into the initial multi-benefit model to obtain the multi-benefit model;

[0043] solving the multi-benefit model based on the variable step size Runge-Kutta method to obtain a state point on a stable manifold;

[0044] Based on the topological data analysis, the state points are identified by the Pareto attractor to obtain an optimal data set of relay node configuration.

[0045] As a further scheme of the present application, based on the fusion regret theory and Choquet integral, the optimal data set of relay node configuration is screened to obtain the optimal data of relay node configuration, comprising:

[0046] The perception effect and the regret value of each relay node configuration optimal data in the relay node configuration optimal data set are obtained based on the fusion regret theory;

[0047] The target subsets are obtained based on the first, second and third benefit functions, the interaction degree of the target subsets is quantified based on a lambda-fuzzy measure, the interaction coefficients of the target subsets are obtained, and the interaction coefficients of the target subsets are obtained.

[0048] The utility-regret comprehensive values are obtained by fusing the perception effect and the regret value, the relay node configuration schemes with utility-regret comprehensive values less than 0.95 are removed, the marginal contribution degrees of the target subsets are obtained based on the interaction coefficients of the target subsets, and the global comprehensive evaluation values are obtained based on the perception effect and the marginal contribution degrees and combined with Choquet integral.

[0049] The data corresponding to the highest score is selected from all global comprehensive evaluation values, and the relay node configuration optimal data is obtained.

[0050] As a further scheme of the application, the relay node configuration optimal data is fed back to the corresponding relay node for verification to obtain stable relay node configuration data, including:

[0051] The relay node configuration optimal data is fed as a configuration instruction to the corresponding relay node for work, and a power carrier data set is collected.

[0052] A power carrier threshold is set, if the collected power carrier data set meets the threshold requirement, it is determined that the current configuration system is stable, and stable relay node configuration data is obtained.

[0053] If the collected power carrier data set does not meet the threshold requirement, return to S2 for cyclic optimization until stable relay node configuration data is obtained.

[0054] In another aspect, the embodiment of the application further provides a power line carrier communication relay node adaptive optimization configuration system, including:

[0055] An acquisition module is configured to acquire channel detection data, performance measurement data, power grid operation state data and device ontology state data.

[0056] A prediction module is configured to predict the channel detection data, the performance measurement data, the power grid operation state data and the device ontology state data to obtain a power carrier data set.

[0057] An extraction module is configured to extract and fuse features of the power carrier data set to obtain channel feature vectors, link quality score data, harmonic interference intensity data and comprehensive health index data.

[0058] an inference module that infers based on the channel feature vector, the link quality score data, and the harmonic interference intensity data to obtain channel capacity prediction data,

[0059] a construction module that constructs a power line carrier network graph based on the channel capacity prediction data and the comprehensive health index data;

[0060] a calculation module that constructs a revenue function based on the relay node configuration problem and solves to obtain a relay node configuration optimal data set, makes a decision on the relay node configuration optimal data set, and obtains relay node configuration optimal data;

[0061] a feedback module that feeds down the relay node configuration optimal data to the corresponding relay node for feedback verification to obtain stable relay node configuration data.

[0062] Based on the above aspects, the embodiments of the present application realize multi-dimensional data fusion, reduce configuration bias caused by a single data dimension, construct a multi-element revenue model, predict channel capacity through a lightweight neural network, and solve the multi-element revenue model in combination with the variable step-size Runge-Kutta method to dynamically adjust the relay node configuration, thereby improving the stability of transmission performance. The interactive relationship between transmission performance, energy efficiency, and robustness is fused through Choquet integration to reduce decision risk, and closed-loop feedback is used to ensure stable operation of the configuration scheme and reduce manual intervention. The method effectively solves the problems of unstable transmission performance and low energy efficiency in traditional relay node configuration, provides an adaptive optimization relay node configuration scheme for power line carrier communication, and thereby improves the stability and efficiency of communication transmission. BRIEF DESCRIPTION OF DRAWINGS

[0063] Figure 1 is an execution flow schematic diagram of a relay node adaptive optimization configuration method based on power line carrier communication provided by an embodiment of the present application.

[0064] Figure 2 is a schematic diagram of a relay node adaptive optimization configuration system based on power line carrier communication provided by an embodiment of the present application. DETAILED DESCRIPTION

[0065] The present application will be described in detail below with reference to the accompanying drawings, Figure 1 is an execution flow schematic diagram of a relay node adaptive optimization configuration method based on power line carrier communication provided by an embodiment of the present application. The relay node adaptive optimization configuration method based on power line carrier communication will be described in detail below.

[0066] Specifically, the relay node adaptive optimization configuration method based on power line carrier communication comprises:

[0067] Step S1, obtaining channel sounding data, performance metric data, power grid operating state data and device ontology state data and performing data preprocessing to obtain a power carrier dataset.

[0068] In this embodiment, step S1 includes:

[0069] Step S11, obtaining channel impulse response data and signal-to-noise ratio spectrum data of each link in a PLC (Power Line Communication) network and composing channel sounding data.

[0070] It should be noted that the channel impulse response data reflects the multipath effect of the signal when propagating in the link, and the multipath effect represents the phenomenon that the signal propagates through multiple paths to the receiving end, causing signal distortion or interference. The signal-to-noise ratio spectrum data reflects the intensity relationship between signal and noise at different frequencies.

[0071] For example, taking a certain intelligent building PLC network as an example, assuming that the network includes 10 communication nodes and 15 communication links, and numbering, each link can be detected by an orthogonal frequency division multiplexing signal generator, with a frequency of 10 seconds per time. The channel impulse response data of link 1-2 is represented as a sequence of signal amplitudes at different time delays, such as [0.8, 0.3, 0.1…], and the signal-to-noise ratio spectrum data is represented as the signal-to-noise ratio values in the 50kHz-500kHz frequency band, such as [25dB, 28dB, …, 22dB], and the channel impulse response data and the signal-to-noise ratio spectrum data are composed into channel sounding data.

[0072] Step S12, obtaining message success rate data and network average delay data of each node in the PLC network and composing performance metric data.

[0073] It should be noted that the message success rate data is obtained by the ratio of the number of successfully transmitted messages in a unit of time to the total number of sent messages, and the network average delay data is obtained by the average value of the transmission time of all messages from the sending end to the receiving end. The communication quality of the PLC network is reflected by the message success rate data and the network average delay data.

[0074] For example, continuing the above-mentioned intelligent building PLC network, the message transmission situation of each communication node in a certain time can be counted through the gateway of the PLC network. Assuming that 1 minute is a statistical period, within 1 minute, node sends 1000 messages, successfully receives 980 messages, and the message success rate data is 98%. The transmission time of all messages of node 3 and other nodes is 20ms, 22ms, 18ms…, and the average delay data is calculated to be 21ms. The message success rate data and the network average delay data are composed into performance metric data.

[0075] Step S13, obtaining real-time load current data as power grid operation state data.

[0076] It should be noted that the current load condition of the power grid is reflected through the real-time load current data.

[0077] For example, the real-time load current values of each line, such as [10A, 10.2A, 9.8A…], can be obtained through smart meters, sensors and other devices in the power grid, and the real-time load current data is taken as the power grid operation state data.

[0078] Step S14, obtaining remaining power data, CPU load rate data and historical average fault interval event data and forming device body state data.

[0079] It should be noted that the remaining power of the battery-powered node is reflected through the remaining power data, the node processor busy degree is reflected through the CPU load rate data, and the historical average fault interval event data is obtained through the time interval average value of past node faults.

[0080] For example, the device body state data can be collected through power monitoring chips, CPU load sampling circuits, clock modules and microcontrollers, and the intelligent building network is still continued. Assuming that node 5 is a battery-powered node, the remaining power is 75%, the current CPU load rate is 30%, the past fault interval times are 64 days, 62 days and 60 days, the historical average fault interval event data is 62 days, and the remaining power data, CPU load rate data and historical average fault interval event data are combined into device body state data.

[0081] Step S15, performing data preprocessing operation on channel probe data, performance metric data, power grid operation state data and device body state data to obtain power carrier data set.

[0082] Specifically, the outliers in the above data are removed by 3σ criterion, the missing data are filled by linear interpolation method, and all data are standardized. Finally, the preprocessed channel probe data, performance metric data, power grid operation state data and device body state data are associated according to the time stamp to form the power carrier data set.

[0083] In some possible embodiments, the mean value of the signal-to-noise ratio data is calculated to be 25dB, it is found through monitoring that the signal-to-noise ratio data of link 4-5 has a value of 35dB, which is removed by 3σ criterion, the network average delay of node 7 in the performance metric data is missing one value, which is filled by linear interpolation method, all data are normalized by min-max standardization value 0-1 interval range, and all data are associated by time stamp to form the power carrier data set.

[0084] Step S2, feature extraction and fusion are performed on the power carrier data set to obtain channel feature vectors, link quality score data, harmonic interference intensity data, and comprehensive health index data.

[0085] In this embodiment, step S2 includes:

[0086] Step S21, principal component analysis and feature extraction are performed on the channel sounding data to obtain channel feature vectors.

[0087] In this embodiment, step S21 includes:

[0088] Step S211, principal component analysis and feature extraction are performed on the channel impulse response data and the signal-to-noise ratio spectrum data in the channel sounding data to obtain projection coefficients of the channel impulse response data dimension, time distribution characteristics of the dominant multipath components, principal component contribution degrees, and distribution variances of the signal-to-noise ratio spectrum data in the principal component space.

[0089] The principal component analysis retains principal components with an accumulated contribution rate greater than 90% to obtain channel feature vectors.

[0090] Specifically, the channel impulse response data and the signal-to-noise ratio spectrum data in the channel sounding data are combined into a high-dimensional matrix, and a covariance matrix is calculated to solve for eigenvalues. The eigenvalues are sorted by size to select principal components, and components with an accumulated contribution degree greater than 90% are defined as principal components. Finally, projection coefficients, time distribution characteristics of dominant multipath components, principal component contribution degrees, and distribution variances of signal-to-noise ratio spectrum data in the principal component space are extracted from the principal components.

[0091] It should be noted that the projection coefficients are represented as the projection weights of each original data on the principal components, the time distribution characteristics of the dominant multipath components are represented as the time delay sequences corresponding to the principal components, the principal component contribution degrees are represented as the proportion of the eigenvalues of each principal component, and the distribution variances of the signal-to-noise ratio spectrum data in the principal component space are represented as the degree of dispersion of the signal-to-noise ratio data in the principal component dimension.

[0092] In some possible embodiments, taking the channel sounding data of the above intelligent building network as an example, the channel impulse responses and the signal-to-noise ratio spectrums of 15 links form a 15*30 high-dimensional matrix, wherein each of the channel impulse responses contains 20 time delay points, and each of the signal-to-noise ratio spectrums contains 10 time delay points. The covariance matrix is calculated and solved to obtain eigenvalues λ1=12.5, λ2=8.3, λ3=4.2, λ4=2.1,..., and the total eigenvalue is 28. The cumulative contribution degree of λ1, λ2, and λ3 is 89.3%, which is less than 90%, and therefore λ4 is also a principal component, and the cumulative contribution degree is calculated to be 96.8%. The four components are defined as principal components, and the projection coefficients on the four principal components, the time distribution characteristics of the dominant multipath components, such as the time delay sequence corresponding to the first principal component [0.1 ms, 0.3 ms, 0.5 ms,...], the principal component contribution degrees, such as λ1 accounting for 44.6% and λ2 accounting for 29.6%, and the distribution variance of the signal-to-noise ratio spectrum in the principal component space, such as the variance in the λ1 principal component dimension being 0.08, are extracted.

[0093] In step S212, the dominant multipath component amplitude is obtained based on the channel impulse response data and the projection coefficients of the dimensions thereof, the time delay spread is obtained based on the time distribution characteristics of the dominant multipath components and the principal component contribution degrees, the frequency selective fading factor is obtained based on the distribution variance of the signal-to-noise ratio spectrum in the principal component space, and the dominant multipath component amplitude, the time delay spread, and the frequency selective fading factor are combined to form a channel feature vector.

[0094] Specifically, the dominant multipath component amplitude is obtained by summing the products of each channel impulse response data and the projection coefficients of the channel impulse response data, the time delay spread is obtained by calculating the root mean square time delay based on the time distribution characteristics of the dominant multipath components, the frequency selective fading factor is obtained by multiplying the distribution variance of the signal-to-noise ratio spectrum in the principal component space and the principal component contribution degrees, and the dominant multipath component amplitude, the time delay spread, and the frequency selective fading factor are combined to form a channel feature vector.

[0095] In some possible embodiments, taking link 1-2 as an example, assuming that the obtained projection coefficient is [0.9, 0.1, 0, 0], the original channel impulse response data is [0.8, 0.3, 0.1, 0.05], the product sum is (0.9 x 0.8) + (0.1 x 0.3) + (0 x 0.1) + (0 x 0.05) = 0.75, that is, the dominant multipath component amplitude is 0.75, if the time distribution characteristics of the obtained dominant multipath component is [0.1 ms, 0.3 ms, 0.5 ms], and the root mean square delay is calculated as 0.32 ms through the time distribution characteristics, the time distribution characteristics are defined as the delay spread, the distribution variance is 0.08 and the principal component contribution degree is 44.6%, the frequency selective fading factor is calculated as 0.036 through the product of the two, and the three data are combined into a channel feature vector [0.75, 0.32 ms, 0.036], which is the channel feature vector of link 1-2, and the channel feature vectors of all links are obtained through the method.

[0096] In step S22, the weighted average calculation is performed on the packet success rate data and the network average delay data in the performance metric data to obtain the link quality score data.

[0097] For example, continuing the performance metric data of node 3, the packet success rate is 98%, that is, 0.98, the network average delay is 21 ms, and the maximum delay is 30 ms and the minimum delay is 15 ms in the network, the average delay 21 ms is normalized to be converted into 0.6, assuming that the weight of the packet success rate is 0.6 and the weight of the network average delay is 0.4, the link quality score data is obtained as 0.828 according to the weighted average.

[0098] It can be understood that the link quality is evaluated through the link quality score data.

[0099] In step S23, the Fourier transform operation is performed on the real-time load current data in the power grid operation state data to obtain the harmonic interference data.

[0100] Specifically, the time domain current signal is converted into the frequency domain current signal through the Fourier transform, and the amplitude of each harmonic is obtained, the main harmonics such as 3rd, 5th and 7th are selected, and the root mean square value of the amplitudes of these frequencies is calculated as the harmonic interference intensity.

[0101] It should be noted that because the content of 3rd, 5th and 7th harmonics in the power grid is high, the root mean square value of these data is selected as the harmonic interference intensity, and the larger the value is, the more serious the power grid harmonic interference is, and the greater the influence on the PLC communication is.

[0102] In some possible embodiments, taking the real-time load current data of the A-phase line of the smart building power grid as an example, Fourier transform is performed on the real-time load current data, a frequency domain spectrum is obtained, the 3rd harmonic amplitude 1.2A, the 5th harmonic amplitude 0.8A, and the 7th harmonic amplitude 0.5A are extracted, the root mean square value 0.88A of these orders is calculated, and the value is regarded as the current harmonic interference intensity data.

[0103] In step S24, the entropy weight method is used to analyze the dispersion and importance of the equipment body state data, and an objective weight coefficient is obtained.

[0104] Specifically, the information entropy of the remaining power data, the CPU load rate data, and the historical average fault interval event data in the equipment body state data is obtained, and an objective weight coefficient is calculated according to an objective weight coefficient formula, which is expressed as wherein, is the objective weight coefficient, is the information entropy of each index.

[0105] It should be noted that the entropy weight method obtains the weight through the dispersion degree of data. The greater the dispersion degree, the greater the influence of the index on the result, and the higher the weight.

[0106] In some possible embodiments, it is assumed that the information entropy of the remaining power data in the equipment body state data of the five nodes is , the information entropy of the CPU load rate data is , and the information entropy of the historical average fault interval event data is An objective weight coefficient is obtained through an objective weight coefficient formula , , .

[0107] In step S25, the remaining power data, the CPU load rate data, and the historical average fault interval event data in the equipment body state data are fused based on the objective weight coefficient, and comprehensive health degree index data is obtained.

[0108] Specifically, the remaining power data, the CPU load rate data, and the historical average fault interval event data are standardized, and the standardized data and the objective weight coefficient are weighted and fused to obtain the comprehensive health degree index data.

[0109] In some possible embodiments, the above example of step S24 is continued, and it is assumed that the standardized remaining power data 0.75, the CPU load rate data 0.7, and the historical average fault interval event data 0.65 of node 5 are obtained, and the comprehensive health degree index data of node 5 is obtained by fusing the objective weight coefficient obtained in step S24.

[0110] Step S3, based on the channel feature vector, link quality score data, harmonic interference intensity data, inference is carried out to obtain channel capacity prediction data, and a power carrier network diagram is constructed in combination with the comprehensive health index data.

[0111] In this embodiment, step S3 includes:

[0112] Step S31, taking the channel feature vector, link quality score data, and harmonic interference intensity data as input, inference is carried out based on the pre-constructed lightweight neural network model to obtain channel capacity prediction data.

[0113] Specifically, the historical data is used to train the lightweight neural network model, and the accuracy of the lightweight neural network model is optimized by adjusting parameters such as network layer number and neuron number. After training, the real-time channel feature vector, link quality score data, and harmonic interference intensity data are input into the model, and the channel capacity prediction data is output.

[0114] It should be noted that the historical data includes historical channel feature vectors, historical link quality score data, and historical harmonic interference intensity data in the past 3 months.

[0115] In some possible embodiments, a 3-layer BP neural network can be used as a lightweight model, the input layer has 3 neurons corresponding to the channel feature vector, the link quality score data, and the harmonic interference intensity data, the hidden layer has 10 neurons, and the output layer has 1 neuron corresponding to the channel capacity prediction data. The historical data of the intelligent building network in the past 3 months is used to train the model, and the trained lightweight neural network model is obtained. The channel feature vector [0.75, 0.32 ms, 0.036] of link 1-2, the link quality score data 8.28, and the harmonic interference intensity data 0.88A are input into the lightweight neural network model, and the channel capacity prediction data 2.5 Mbps is output.

[0116] Step S32, all communication nodes in the PLC network and corresponding comprehensive health index data are obtained, each communication node is ID identified, and the comprehensive health index data is given as the point weight of the corresponding communication node, and a communication node set with point weight is obtained.

[0117] Specifically, each node is assigned a unique ID, and the comprehensive health index data of each node obtained in step S25 is given as the point weight of the corresponding communication node, and a communication node set with point weight is obtained.

[0118] It should be noted that the communication node includes a collection node, a control node, a routing node and other nodes with communication function, and the step reflects the health state of the node through the point weight, and the higher the health degree of the node, the higher the adaptability of the relay node.

[0119] In some possible embodiments, assuming that 10 communication nodes of the intelligent building network are assigned IDs Node1 to Node 10 , and the comprehensive health degree index data are 0.62, 0.72, 0.68, …, 0.73, 0.67 respectively, and the data are combined as [(Node1, 0.62), (Node2, 0.72), (Node3, 0.68), …, (Node9, 0.73), (Node 10 , 0.67)], the communication node set with the point weight is obtained.

[0120] In step S33, the communication link between nodes and the corresponding channel capacity prediction data are obtained, the directed attribute of each communication link is identified based on the bidirectional communication relationship, and the channel capacity prediction data is given as the edge weight of the corresponding communication link, and the communication link set with the edge weight is obtained.

[0121] It can be understood that the link channel capacity prediction data obtained in step S31 is given as the edge weight of the corresponding link, and the step reflects the transmission capacity of the link through the edge weight, and the larger the capacity, the better the transmission performance of the link.

[0122] It should be noted that the communication link set includes all communication links between nodes, and each link has bidirectional communication characteristics, and needs to identify the forward link or the reverse link.

[0123] For example, in the intelligent building network, Node1 and Node2 have a characteristic link, the forward link is identified as L 12 , and the reverse link is identified as L 21 , and the channel capacity prediction data of both is 2.5Mbps, the link between Node2 and Node3 is identified as L 23 and L 32 , and the capacity is 2.3Mbps, and the combination is [(L 12 , 2.5Mbps), (L 21 , 2.5Mbps), (L 23 , 2.3Mbps), …], and the communication link set with the edge weight is obtained.

[0124] In step S34, the time granularity and the timestamp information are set, the communication node set with the point weight and the communication link set with the edge weight are periodically updated and synchronously marked with the timestamp, and the node link set is obtained.

[0125] For example, the time granularity is set to 5 minutes, the communication node set with point weight and the communication link set with edge weight are updated at 10:00, and the timestamp is marked as 10:00. The timestamp is updated again at 10:05 and marked as 10:05, and the communication node set and the communication link set corresponding to each timestamp are combined to obtain the node link set.

[0126] In step S35, the node link set is integrated into a graph structure to obtain the power carrier network graph.

[0127] Specifically, the node link set is integrated by a directed graph or an undirected graph, wherein the node corresponds to the communication node with point weight, the edge corresponds to the communication link with edge weight, the attribute of the node is ID and point weight, the attribute of the edge is link identifier and edge weight, and the timestamp information is associated to obtain the power carrier network graph.

[0128] In some possible embodiments, taking the node link set at 10:00 as an example, a directed graph is constructed in this embodiment. The attribute of the node Node1 includes ID Node1 and point weight 0.65. The attribute of the edge L12 includes identifier L12 and edge weight 2.5 Mbps. All nodes and edges are combined according to the actual connection relationship to obtain the power carrier network graph. Through the power carrier network graph, it can be directly and clearly seen that Node1 communicates with Node2 through the L12 link at a capacity of 2.5 Mbps, and the comprehensive health index data of Node1 is 0.65. 12 12

[0129] In step S4, a revenue function is constructed based on the power carrier network graph and the relay node configuration problem, and the revenue function is solved to obtain an optimal data set of relay node configuration. The optimal data set of relay node configuration is decided to obtain the optimal data of relay node configuration.

[0130] In this embodiment, step S4 includes:

[0131] In step S41, the network topology and traffic demand data are obtained, the network basic maximum flow without relay node configuration is obtained in combination with the power carrier network graph, the network maximum flow after relay node configuration is obtained in combination with different relay node configuration schemes, the full-network maximum flow improvement degree is obtained based on the network basic maximum flow and the network maximum flow, and the first revenue function is constructed based on the full-network maximum flow improvement degree.

[0132] ​​Specifically, the network topology represented by node connection relationship and the traffic demand data represented by service traffic demand between nodes are extracted from the power carrier network graph, the network base maximum flow is calculated according to the Ford-Fulkerson algorithm when no relay node is configured, then the link capacity of the network graph is updated according to different relay node configuration schemes, the network maximum flow after the relay node configuration is calculated, finally the network maximum flow improvement degree is obtained through the network maximum flow improvement degree formula, and the first benefit function is constructed by the network maximum flow improvement degree, and the network maximum flow improvement degree formula is represented as network maximum flow improvement degree = (configured network maximum flow-network base maximum flow) / network base maximum flow.

[0133] It should be noted that the Ford-Fulkerson algorithm is a greedy algorithm for solving network maximum flow, which finds an augmented path from the source node to the sink node, and increases the traffic in the network through the augmented path until there is no augmented path in the network, at which time the traffic obtained is the network base maximum flow.

[0134] It can be understood that the embodiment constructs the power carrier graph in the above steps, and the Ford-Fulkerson algorithm can be directly calculated through the link capacity and topological relationship in the power carrier graph to obtain the network base maximum flow and the configured network maximum flow.

[0135] In some possible embodiments, continuing the above example of the intelligent building network, when the intelligent building network has no relay, the network base maximum flow obtained by the Ford-Fulkerson algorithm is 5 Mbps, when Node2 and Node4 are configured as relay nodes, the configured network maximum flow is 8 Mbps after updating the link capacity, and the network maximum flow improvement degree is 0.6 according to the network maximum flow improvement degree formula, then the first benefit function f1 = 0.6.

[0136] In step S42, the relay nodes and corresponding power consumption parameters are obtained, the total power consumption of all activated relays in the network is obtained based on the number of activated relay nodes in the power carrier network graph and the power consumption parameter of each relay node, and the second benefit function is constructed based on the reciprocal of the total power consumption of all activated relays in the network.

[0137] The activated relay node is a relay node in an active state.

[0138] For example, the power consumption data of each relay node is obtained, assuming that Node2 and Node4 are activated relay nodes, the power consumption data of Node2 is 1.2W, the power consumption data of Node4 is 1.0W, and the total power data is 2.2W, the second benefit function is obtained by taking the reciprocal of the total power, and the second benefit function f2 = 0.45.

[0139] Step S43, the comprehensive health index data of the activated relay nodes in the power carrier network graph are obtained and added to obtain total health data, and a third benefit function is constructed based on the total health data.

[0140] For example, the comprehensive health index data of the activated relay point Node2 is 0.72, and the comprehensive health index data of the activated relay point Node4 is 0.80, so the total health data is 1.52, and the third benefit function is obtained through the total health data, that is, the third benefit function f3=1.52.

[0141] Step S44, a multi-benefit model is constructed based on the first, second and third benefit functions, and the multi-benefit model is solved to obtain an optimal relay node configuration data set.

[0142] In this embodiment, step S44 includes:

[0143] Step S441, balance weight parameters of the first, second and third benefit functions are obtained, and an Euler-Lagrange equation is constructed in combination with the first, second and third benefit functions, and an initial multi-benefit model is obtained based on the Euler-Lagrange equation.

[0144] Specifically, the balance weight parameters of the first, second and third benefit functions are obtained according to network performance, energy efficiency and robustness, and a Lagrange equation is constructed, and the Lagrange equation is , wherein respectively represent the balance weight parameters of the first, second and third benefit functions, and then the derivative of the Lagrange equation is obtained to obtain the Euler-Lagrange equation, and the initial multi-benefit model is constructed through the Euler-Lagrange equation.

[0145] Further, when the derivative of the Lagrange equation is taken, the derivative of the Lagrange equation with respect to each variable is obtained by taking the relay point activation state vector as a variable, and the constraint condition that the Lagrange multiplier is less than or equal to 0 is introduced into the equation, and finally the Euler-Lagrange equation is obtained, and the Euler-Lagrange equation is expressed as ; ; wherein, represents the relay point, represents the first relay point in the activated state, represents the first relay point in the unactivated state, represents the Lagrange multiplier, and the initial multi-benefit model is constructed based on the Euler-Lagrange equation.

[0146] Step S442, obtaining each relay node configuration scheme and relay node data n, mapping each relay node configuration scheme to a high-dimensional state space, and obtaining a set of state points in the n-dimensional state space.

[0147] Specifically, the relay node data n is set by the number of communication nodes with relay function in the PLC network, an n-dimensional state space is constructed, each dimension corresponds to the activation state of a candidate relay node, where 0 represents inactivation and 1 represents activation; each relay node configuration scheme corresponds to an n-dimensional binary vector, the vector is mapped to the high-dimensional state space as a state point, and the state points corresponding to all configuration schemes form a set of state points.

[0148] For example, in the intelligent building network, the candidate relay nodes are Node2, Node3, Node4, and Node7, a 4-dimensional state space is constructed, the corresponding vector [1, 0, 1, 0] of the configuration scheme activating Node2 and Node4 is mapped to the 4-dimensional space as the state point (1, 0, 1, 0), the corresponding vector [0, 1, 0, 1] of the configuration scheme activating Node3 and Node7 is mapped to the 4-dimensional space as the state point (0, 1, 0, 1), and all possible configuration scheme vectors form a set of state points.

[0149] Step S443, obtaining the Lyapunov dissipation function and the small perturbation parameter, and introducing them into the initial multi-element yield model to obtain a multi-element yield model.

[0150] It can be understood that the Lyapunov dissipation function is used to simulate the energy dissipation of the system, and the small perturbation parameter is used to simulate the random interference in the actual network. The Lyapunov dissipation function parameter and the small perturbation parameter are introduced into the Euler-Lagrange equation of the initial multi-element yield model.

[0151] For example, the Lyapunov dissipation function is represented as wherein, is the Lyapunov damping coefficient, which is usually set to 0.1, is the change rate of the corresponding state point, and the small perturbation is represented as wherein is the standard normal distribution, and the Lyapunov dissipation function and the small perturbation parameter are introduced into the Euler-Lagrange equation.

[0152] Step S444, solving the multi-element yield model based on the variable step size Runge-Kutta method to obtain the state points on the stable manifold.

[0153] Specifically, the change rate in the Euler-Lagrange equation is adjusted in adaptive step length by the variable step length Runge-Kutta method; an initial step length and a convergence threshold value are obtained, the Euler-Lagrange equation is simulated by numerical integration, and when the change amount of a plurality of state points is less than the convergence threshold value, the state points are output and combined to form state points on a stable manifold.

[0154] It should be noted that the Euler-Lagrange equation introduces the Rayleigh dissipation function and a small perturbation, and the function changes according to the evolution of the state point, and the change rate of different state points is quite different. The variable step length Runge-Kutta method has the ability to adjust the adaptive step length, so the variable step length Runge-Kutta method is used to adjust the adaptive step length of the change rate in the Euler-Lagrange equation.

[0155] In some possible embodiments, it is assumed that the initial step length 1e-3 and the convergence threshold value 1e-6 are obtained, and the evolution process of the state point is simulated by numerical integration of the Euler-Lagrange equation. After a plurality of iterations, the change amount of the initial state point (1, 0, 1, 0) is less than the convergence threshold value 1e-6, indicating that the state point tends to be stable. The remaining state points are also converged by the same method, and all state points on the stable manifold are obtained.

[0156] In step S445, the state points are identified as Pareto attractors based on the topological data analysis, and an optimal data set of relay node configurations is obtained.

[0157] Specifically, the Vietoris-Rips complex is constructed by calculating the distance between the state points, and the topological structure is analyzed to identify the longest-lived homogenous source, i.e., the Pareto attractor, which contains the state points corresponding to the relay node configuration scheme, and the optimal data set of relay node configurations is formed by the relay node configuration scheme.

[0158] It should be noted that the Vietoris-Rips complex is a set composed of points, line segments, triangles and other topological structures, the homogenous source is a connected component or a hole structure with the same topological adjustment in the complex, and the life span is the distance from the first appearance to the disappearance of the topological structure. The longer the life span, the more stable the topological association of the corresponding state point set, i.e., the Pareto attractor.

[0159] In some possible embodiments, the state points on the stable manifold in step S444 are subjected to topological data analysis, the Euclidean distances between the state points are obtained, a Vietoris-Rips complex is constructed and analyzed, and state points (1, 0, 1, 0), (0, 1, 0, 1), and (1, 0, 0, 1) contained in the same source with the longest lifetime are obtained, which correspond to relay node configuration schemes of activating Node2 and Node4, activating Node3 and Node7, and activating Node2 and Node7, respectively. The three schemes are combined to form an optimal relay node configuration data set.

[0160] In step S45, the optimal relay node configuration data set is screened based on the fused regret theory and Choquet integral, and optimal relay node configuration data are obtained.

[0161] In this embodiment, step S45 includes:

[0162] In step S451, the perceived effect and the regret value of each optimal relay node configuration data in the optimal relay node configuration data set are obtained based on the fused regret theory.

[0163] Specifically, the perceived effect of each scheme is obtained by mapping the value of the revenue function of each scheme to 0-1 through a preset perceived utility function, and the regret value of each scheme is calculated based on the perceived utility function and the perceived effect of each scheme.

[0164] In some possible embodiments, it is assumed that the revenue values of the three schemes obtained in step S44 are scheme 1 (f1=0.6, f2=0.45, f3=1.52), scheme 2 (f1=0.55, f2=0.48, f3=1.45), and scheme 3 (f1=0.58, f2=0.42, f3=1.55), respectively. The perceived utility function can be represented as wherein f represents the revenue function in each scheme, fmax represents the maximum value of each revenue function, it is assumed that the maximum value of the revenue function f1 is 0.6, the maximum value of the revenue function f2 is 0.48, and the maximum value of the revenue function f3 is 1.55, the perceived utility of scheme 1 obtained through the perceived utility function is (1, 0.9375, 0.9806), and the regret value is obtained through a regret value function and can be represented as wherein f represents the regret value, the regret value of scheme 1 obtained according to the regret value function is (0, 0.625, 0.0194), and fusion is performed as 0.4x0+0.3x0.0625+0.3x0.0194=0.0246.

[0165] Step S452, obtaining target subsets based on the first, second and third benefit functions, and quantifying the interaction degree of the target subsets based on the λ-fuzzy measure to obtain the interaction coefficients of the target subsets.

[0166] Specifically, the target subsets are obtained by the first, second and third benefit functions, such as {f1}, {f2}, {f3}, {f1, f2}, {f1, f3}, {f2, f3}, {f1, f2, f3}, λ is initialized, and the λ-fuzzy measure is iteratively optimized by, for example, a genetic algorithm, so that the λ-fuzzy measure satisfies the additivity constraint, and the λ-fuzzy measure value of each target subset is obtained. The λ-fuzzy measure value represents the interaction degree of the target subset, and the greater the value, the stronger the interaction of the targets in the subset.

[0167] It should be noted that the λ-fuzzy measure is an uncertainty measure method for quantifying the importance and interaction degree of a set. The interaction relationship between subsets is adjusted by λ to satisfy the additivity constraint. The additivity constraint is represented as μ(A∪B)=μ(A)+μ(B)+λμ(A)μ(B).

[0168] In some possible embodiments, assuming that λ is initialized as 0.5, the interaction coefficients of the subsets are obtained after multiple iterations by the genetic algorithm, such as μ({f1})=0.35, μ({f2})=0.25, μ({f3})=0.30, μ({f1,f2})=0.55, μ({f1,f3})=0.60, μ({f2,f3})=0.50, and μ({f1,f2,f3})=0.85.

[0169] Step S453, fusing the perception effect and the regret value to obtain a utility-regret comprehensive value, eliminating a relay node configuration scheme whose utility-regret comprehensive value is less than 0.95, obtaining the marginal contribution degrees of the target subsets based on the interaction coefficients of the target subsets, and obtaining a global comprehensive evaluation value based on the perception effect and the marginal contribution degrees and in combination with Choquet integral.

[0170] Specifically, the weights of the perception effect and the regret value are set, the utility-regret comprehensive value is obtained in combination with the perception effect and the regret value, if the obtained utility-regret comprehensive value is less than 0.95, it is proved that the regret degree of the relay node configuration scheme corresponding to the data is high, and then the scheme is eliminated. The benefit functions are sorted based on the importance of the benefit functions in the system, the marginal contribution degrees of the target subsets are calculated in combination with the interaction coefficients of the target subsets, and the global comprehensive evaluation value of each relay node configuration scheme is calculated based on Choquet integral and in combination with the perception effect and the marginal contribution degrees.

[0171] It can be understood that the Choquet integral is a fuzzy degree-based nonlinear integral method, which sorts the perception effects of each target according to importance, combines the marginal contribution of the target subset, and obtains the global comprehensive evaluation value of each relay node configuration scheme through the product of the perception utility and the corresponding subset marginal contribution. Compared with the traditional weighted average method, the Choquet integral not only considers the importance of a single target, but also considers the interaction relationship between multiple targets such as synergy and redundancy.

[0172] It should be noted that if the utility-regret comprehensive value obtained is less than 0.95, it means that either the perception utility is low or the regret degree is high, so the scheme with a high decision risk whose utility-regret comprehensive value is less than 0.95 is eliminated.

[0173] In some possible embodiments, the weights of the perception effect and the regret value are set to 0.6 and 0.4 respectively, the perception effect of scheme 1 is (1, 0.9375, 0.9806), the regret value is 0.0246, and the utility-regret comprehensive value is calculated as 0.6×[(1+0.9375+0.9806) / 3]+0.4×(1-0.0246)=0.9738, which is greater than 0.95, so the relay node configuration scheme is retained; f1 corresponds to the network transmission performance, f2 corresponds to the energy efficiency, and f3 corresponds to the network robustness. In the intelligent building PLC network, the important priority of the network transmission performance is the highest, the important priority of the network robustness is the second, and the priority of the energy efficiency is the lowest. Therefore, the comprehensive ranking of the benefit function is f1>f3>f2. The marginal contribution of each target subset is calculated through the interaction coefficient of each target subset, μ({f1})=0.35, μ({f1,f3})-μ({f1})=0.25, μ({f1,f2,f3})-μ({f1,f3})=0.25, and finally the global comprehensive evaluation value of scheme 1 is calculated by Choquet integral as 0.35×1+0.25×0.9806+0.25×0.9375=0.8296. The global comprehensive evaluation values of scheme 2 and scheme 3 are obtained by the same method as in this step, which are 0.7921 and 0.8153 respectively.

[0174] Step S454: selecting the data corresponding to the highest score from all global comprehensive evaluation values to obtain the optimal relay node configuration data.

[0175] Specifically, all global comprehensive evaluation values are sorted in descending order, and the relay node configuration scheme with the highest score is selected as the relay node configuration scheme with the best comprehensive performance. The relay node activation state corresponding to the scheme is the optimal relay node configuration data.

[0176] In some possible embodiments, the global comprehensive evaluation values of the schemes 1, 2 and 3 are 0.8296, 0.7921 and 0.8153 respectively, and the scheme 1 corresponding configuration is selected to activate Node2 and Node4 as the relay node configuration optimal data.

[0177] In step S5, the relay node configuration optimal data is fed to the corresponding relay node for feedback verification, and stable relay node configuration data is obtained.

[0178] In this embodiment, step S5 includes:

[0179] In step S51, the relay node configuration optimal data is fed to the corresponding relay node as a configuration instruction, and power carrier data set is collected.

[0180] For example, the relay node configuration optimal data obtained in step S454 is fed to the corresponding relay node by a communication protocol format such as DL / T645 protocol, and in the following time, power carrier data set is collected in each minute as a collection period, such as message success rate 99%, network average time delay data 18ms, channel capacity prediction data 2.6Mbps and other data.

[0181] In step S52, a power carrier threshold is set, and if the collected power carrier data set meets the threshold requirement, it is determined that the current configuration system is stable, and stable relay node configuration data is obtained.

[0182] For example, the power carrier threshold includes message success rate greater than or equal to 98%, network time delay less than or equal to 20ms, and channel capacity prediction data greater than or equal to 2.4Mbps. It is found by the collected data that the data of t1-t3 continuous three periods meet the threshold, and it is determined that the activated Node2 and Node4 are stable relay node configuration data.

[0183] In step S53, if the collected power carrier data set does not meet the threshold requirement, return to S2 for cyclic optimization until stable relay node configuration data is obtained.

[0184] For example, the message success rate collected at t2 is 97%, which is lower than the set threshold 98%, and the power carrier data set collected at t2 is fed back to step S2 for re-execution of adjustment extraction into a cycle until stable relay node configuration data is obtained.

[0185] Figure 2 A schematic diagram of a power line carrier communication relay node adaptive optimization configuration system is shown, which can realize the idea of the present application.

[0186] Specifically, the power line carrier communication relay node adaptive optimization configuration system includes:

[0187] an acquisition module configured to acquire channel sounding data, performance metric data, power grid operating state data, and device ontology state data;

[0188] a prediction module configured to perform data prediction on the channel sounding data, the performance metric data, the power grid operating state data, and the device ontology state data, and acquire a power carrier dataset;

[0189] an extraction module configured to perform feature extraction and fusion on the power carrier dataset, and acquire channel feature vectors, link quality score data, harmonic interference intensity data, and comprehensive health index data;

[0190] an inference module configured to perform inference based on the channel feature vectors, the link quality score data, and the harmonic interference intensity data, and acquire channel capacity prediction data,

[0191] a construction module configured to construct a power carrier network graph based on the channel capacity prediction data and the comprehensive health index data;

[0192] a calculation module configured to construct a revenue function based on a relay node configuration problem, and solve and acquire optimal relay node configuration data, make a decision on the optimal relay node configuration data, and acquire optimal relay node configuration data;

[0193] a feedback module configured to feed back the optimal relay node configuration data to corresponding relay nodes for feedback verification, and acquire stable relay node configuration data.

[0194] The specific use and role of the embodiment are described as follows:

[0195] First, multi-dimensional data is acquired in step S1, including channel sounding data, performance metric data, power grid operating state data, and device ontology state data, which reduces data bias. Then, in step S2, feature extraction is performed on the multi-dimensional data to acquire channel feature vectors, link quality score data, harmonic interference intensity data, and comprehensive health index data. In step S3, channel capacity prediction data is inferred and a power carrier network graph is constructed, and the relay node configuration is dynamically adjusted to improve the stability of transmission performance. In step S4, the interaction relationship between transmission performance, energy efficiency, and robustness is fused to reduce the risk of decision-making. Finally, in step S5, a closed-loop feedback is performed to ensure that the relay node configuration scheme meets the operating requirements, thereby reducing manual intervention. The power line carrier communication relay node adaptive optimization configuration method improves the stability and efficiency of communication transmission.

[0196] In addition, the embodiment of the present application further provides an electronic device, comprising:

[0197] at least one processor; and a memory connected with the at least one processor in communication; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method provided in the embodiment of the present application.

[0198] The various constituent components of the electronic device are specifically introduced as follows:

[0199] The processor is the control center of the electronic device, and can be one processor or a collective term of multiple processing elements. The processor can perform various functions of the electronic device by running or executing software programs stored in the memory and calling data stored in the memory.

[0200] The memory is used to store software programs for executing the scheme of the present application, and is controlled by the processor to perform the execution. The specific implementation manner can refer to the method embodiments described above, and will not be described here again.

[0201] The above embodiments can be realized wholly or partially by software, hardware (such as a circuit), firmware or any other combination. When realized by software, the above embodiments can be realized wholly or partially in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the flow or function described in the embodiments of the present application is wholly or partially generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions can be transferred from one website, computer, server or data center to another website, computer, server or data center by limited (for example, infrared, wireless, microwave, etc.) mode. The computer-readable storage medium can be any available medium accessible by a computer or a data storage device such as a server, data center and the like containing one or more available medium sets. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a magnetic tape), an optical medium (for example, a DVD) or a semiconductor medium. The semiconductor medium can be a solid-state disk.

[0202] It should be understood that the term "and / or" in this text only describes the association relationship of the associated objects, which means that there can be three relationships, for example, and / or I frame / P frame interval offset rate, which means that I frame / P frame interval offset rate alone, I frame / P frame interval offset rate and I frame / P frame interval offset rate exist at the same time, and I frame / P frame interval offset rate alone, where I frame / P frame interval offset rate can be singular or plural. In addition, the character " / " in this text generally indicates that the associated objects before and after are an "or" relationship, but it can also mean an "and / or" relationship, which can be understood in the context.

[0203] It should be understood that in the embodiments of the present application, the size of the sequence number of the above-mentioned processes does not mean the order of execution, and the execution order of the processes should be determined according to its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0204] The above-described embodiments are only used to illustrate the technical solutions of the present application, but not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that the technical solutions recorded in the foregoing embodiments can still be modified, or some technical features can be replaced by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A method for adaptive optimization configuration of a power line carrier communication relay node, characterized in that, The method comprises the following steps: S1: acquiring channel sounding data, performance metric data, power grid operating state data, and device body state data, and performing data preprocessing to obtain a power carrier dataset; S2: performing feature extraction and fusion on the power carrier dataset to obtain channel feature vectors, link quality score data, harmonic interference intensity data, and comprehensive health index data; S3: performing inference based on the channel feature vectors, link quality score data, and harmonic interference intensity data to obtain channel capacity prediction data, and combining the comprehensive health index data to construct a power carrier network graph; S4: constructing a revenue function based on the power carrier network graph and the relay node configuration problem, and solving the revenue function to obtain an optimal relay node configuration dataset, and making a decision on the optimal relay node configuration dataset to obtain optimal relay node configuration data; S5: issuing the optimal relay node configuration data to the corresponding relay nodes for feedback verification to obtain stable relay node configuration data; The step S4 comprises: acquiring network topology and traffic demand data, combining the power carrier network graph to obtain a network base maximum flow without relay node configuration, combining different relay node configuration schemes to obtain a network maximum flow after relay node configuration, obtaining a network maximum flow improvement degree based on the network base maximum flow and the network maximum flow, and constructing a first revenue function based on the network maximum flow improvement degree; acquiring relay nodes and corresponding power consumption parameters, obtaining a total active relay power consumption of the entire network based on the number of active relay nodes in the power carrier network graph and the power consumption parameters of each relay node, and constructing a second revenue function based on the reciprocal of the total active relay power consumption of the entire network; acquiring comprehensive health index data of the active relay nodes in the power carrier network graph and adding them together to obtain total health data, and constructing a third revenue function based on the total health data; constructing a multi-element revenue model based on the first, second, and third revenue functions, and solving the multi-element revenue model to obtain an optimal relay node configuration dataset; screening the optimal relay node configuration dataset based on the integrated regret theory and Choquet integral to obtain optimal relay node configuration data; The active relay node represents a relay node in an operating state.

2. The method of claim 1, wherein, The step S1 comprises: acquiring channel impulse response data and signal-to-noise ratio spectrum data of each link in the PLC network and combining them to form channel sounding data; acquiring message success rate data and network average delay data of each node in the PLC network and combining them to form performance metric data; acquiring real-time load current data as power grid operating state data; acquiring residual power data, CPU load rate data, and historical average fault interval event data and combining them to form device body state data; performing data preprocessing operations on the channel sounding data, performance metric data, power grid operating state data, and device body state data to obtain a power carrier dataset.

3. The method of claim 1, wherein, The step S2 comprises: performing principal component analysis and feature extraction on the channel sounding data to obtain channel feature vectors; performing weighted average calculation on the message success rate data and network average delay data in the performance metric data to obtain link quality score data; Performing Fourier transform operation on real-time load current data in power grid operation state data to obtain harmonic interference data; Performing discreteness and importance analysis on equipment ontology state data based on entropy weight method to obtain objective weight coefficient; Fusing residual power data, CPU load rate data and historical average fault interval event data in equipment ontology state data based on objective weight coefficient to obtain comprehensive health degree index data.

4. The method of claim 3, wherein, Performing principal component analysis and feature extraction on channel sounding data to obtain channel feature vector, including: Performing principal component analysis and feature extraction operation on channel impulse response data and signal-to-noise ratio spectrum data in channel sounding data to obtain projection coefficient of channel impulse response data dimension, time distribution characteristics of dominant multipath component, principal component contribution degree and distribution variance of signal-to-noise ratio spectrum data in principal component space; Based on channel impulse response data and its dimension projection coefficient, the dominant multipath component amplitude is obtained, based on the time distribution characteristics of the dominant multipath component and the principal component contribution degree, the delay spread is obtained, based on the distribution variance of the signal-to-noise ratio spectrum data in the principal component space, the frequency selective fading factor is obtained, and the dominant multipath component amplitude, the delay spread and the frequency selective fading factor are combined to form the channel feature vector. The principal component analysis retains the principal components with cumulative contribution rate greater than 90% to obtain the channel feature vector.

5. The method of claim 1, wherein, The step S3 includes: Taking the channel feature vector, the link quality score data and the harmonic interference intensity data as input, performing inference based on the pre-constructed lightweight neural network model to obtain channel capacity prediction data; Obtain all communication nodes in the PLC network and corresponding comprehensive health degree index data, identify each communication node by ID and assign the comprehensive health degree index data as point weight to the corresponding communication node, and obtain the communication node set with point weight; Obtain the communication link between nodes and the corresponding channel capacity prediction data, identify each communication link by directional attribute based on the bidirectional communication relationship and assign the channel capacity prediction data as edge weight to the corresponding communication link, and obtain the communication link set with edge weight; Set the time granularity and timestamp information, periodically update the communication node set with point weight and the communication link set with edge weight and synchronize the timestamp, and obtain the node link set; Integrate the node link set into a graph structure to obtain a power carrier network graph.

6. The method of claim 1, wherein, Based on the first, second and third revenue functions, a multi-revenue model is constructed, and the multi-revenue model is solved to obtain an optimal relay node configuration data set, including: Obtain the balance weight parameters of the first, second and third revenue functions, and construct the Euler-Lagrange equation combined with the first, second and third revenue functions, and obtain the initial multi-revenue model based on the Euler-Lagrange equation; Obtain each relay node configuration scheme and relay node data n, map each relay node configuration scheme to a high-dimensional state space, and obtain a state point set in the n-dimensional state space; Obtain the Lyapunov dissipation function and the small perturbation parameter, and introduce them into the initial multi-revenue model to obtain the multi-revenue model; Solving the parameter-dependent multi-utility model based on the variable step size Runge-Kutta method to obtain state points on the stable manifold; Performing Pareto attractor identification on the state points based on topological data analysis to obtain a relay node configuration optimal data set.

7. The method of claim 1, wherein, Screening the relay node configuration optimal data set based on the fused regret theory and Choquet integral to obtain relay node configuration optimal data, including: Obtaining the perception effect and regret value of each relay node configuration optimal data in the relay node configuration optimal data set based on the fused regret theory; Obtaining a target subset based on the first, second and third utility functions, and quantifying the interaction degree of the target subset based on a λ-fuzzy measure to obtain an interaction coefficient of each target subset; Fusing the perception effect and regret value to obtain a utility-regret comprehensive value, eliminating a relay node configuration scheme whose utility-regret comprehensive value is less than 0.95, obtaining a marginal contribution degree of each target subset based on the interaction coefficient of each target subset, and obtaining a global comprehensive evaluation value based on the perception effect and the marginal contribution degree in combination with the Choquet integral; Selecting data corresponding to the highest score from all global comprehensive evaluation values to obtain relay node configuration optimal data.

8. The method of claim 1, wherein, The step S5 includes: issuing the relay node configuration optimal data as a configuration instruction to a corresponding relay node for work, and collecting a power carrier data set; Setting a power carrier threshold value, and if the collected power carrier data set meets the threshold value requirement, determining that the current configuration system is stable, and obtaining stable relay node configuration data; If the collected power carrier data set does not meet the threshold value requirement, returning to S2 for cyclic optimization until stable relay node configuration data is obtained.

9. A system for adaptive optimization configuration of a power line carrier communication relay node for implementing the method of any of claims 1 to 8, characterized in that It includes: An acquisition module for acquiring channel probe data, performance metric data, power grid operating state data and device ontology state data; A prediction module for predicting channel probe data, performance metric data, power grid operating state data and device ontology state data to obtain a power carrier data set; An extraction module for feature extraction and fusion of the power carrier data set to obtain channel feature vectors, link quality score data, harmonic interference intensity data and comprehensive health index data; An inference module for inferring based on channel feature vectors, link quality score data and harmonic interference intensity data to obtain channel capacity prediction data, A construction module for constructing a power carrier network graph based on channel capacity prediction data and comprehensive health index data; A calculation module for constructing an utility function based on a relay node configuration problem and solving to obtain a relay node configuration optimal data set, making a decision on the relay node configuration optimal data set to obtain relay node configuration optimal data; A feedback module for issuing the relay node configuration optimal data to a corresponding relay node for feedback verification to obtain stable relay node configuration data.

Images