Power line communication optimization collection method fusing edge computing

Abstract

The present application relates to the technical field of power line communication, and discloses a power line communication optimization collection method combined with edge computing. The present application deploys a collection device in a distribution box, establishes number mapping with each downstream power branch and collects voltage; a voltage data set is constructed using historical sampling, the mean value, signal power and noise power are calculated to obtain the signal-to-noise ratio, and the sampling frequency and sampling period are selected accordingly, and a sampling window and a sampling sequence are generated at the decision moment. The pre-filtering energy threshold is set based on the deviation of the sampling point relative to the decision voltage, the noise is removed to retain the voltage sequence, the signal vector with a fixed length is constructed, and the linear transformation is generated under the orthogonal normalization basis to generate the compression coding coefficient. The bandwidth priority is constructed by combining the compression data amount of each branch, the signal-to-noise ratio and the uplink bandwidth, the cache scheduling is carried out, the upload branch is selected and the structured upload record is generated, so as to reduce redundancy, suppress noise and improve the bandwidth utilization.

Description

Technical Field

[0001] This invention relates to the field of power line communication technology, specifically to an optimized acquisition method for power line communication that integrates edge computing. Background Technology

[0002] Power line communication (PLC) is a commonly used data transmission method in power distribution network environments. It typically relies on concentrators or gateways installed in distribution boxes to uniformly upload monitoring signals such as voltage and current from multiple branches to a host computer or master station system via power lines.

[0003] In existing technologies, PLC acquisition terminals mostly adopt a fixed sampling frequency and a unified sampling strategy. That is, under a pre-set constant sampling period, the signals of each branch are sampled at equal intervals, and then simply packaged and uploaded as a whole. This type of solution is relatively easy to implement, but it has obvious limitations in complex noise environments and scenarios with multiple branches accessing in parallel. On the one hand, the load status and electromagnetic interference of each branch in the distribution network change significantly over time, and the channel signal-to-noise ratio (SNR) of different branches varies significantly. Existing acquisition devices usually do not estimate the instantaneous SNR of each branch in real time at the edge side, and also lack a mechanism to use the SNR results for fine control of the sampling frequency. This results in insufficient sampling of high SNR branches and excessive sampling of low SNR branches. The sampling resources cannot be reasonably allocated according to the channel quality, which increases the amount of redundant data and is not conducive to the high-precision capture of important information. On the other hand, existing PLC data acquisition mostly adopts the mode of "full acquisition first, then unified upload". The edge side generally only performs amplitude limiting, simple averaging or event alarm judgment on the acquired raw waveform, lacking a fine-grained pre-filtering mechanism based on energy changes at the sampling end. A large number of samples that fluctuate slightly around the mean and are near the noise level will be transmitted through the power line along with samples that truly reflect load changes, harmonic disturbances, or fault characteristics. This results in long link occupancy times and low bandwidth utilization, and is more likely to cause data congestion or packet loss in cases of severe noise or limited bandwidth.

[0004] Therefore, this paper aims to propose an optimized power line communication acquisition method that integrates edge computing. On the one hand, it utilizes channel state estimation based on historical data to achieve adaptive sampling, thereby increasing the sampling density when the channel quality is good and reducing the sampling frequency when the channel quality is poor. On the other hand, it constructs a bandwidth priority index on the edge device side to prioritize and schedule uploading tasks for multiple branches, ensuring real-time reporting of data from key branches. Finally, it compresses the signal based on energy distribution through orthogonal transformation, effectively reducing the amount of redundant data. Summary of the Invention

[0005] This invention provides an optimized acquisition method for power line communication that integrates edge computing, thereby helping to solve the problems mentioned in the background art.

[0006] The application provides the following technical scheme: a power line communication optimization collection method based on edge fusion calculation, comprising:

[0007] An edge collection device is deployed, a numbering relationship is established with downstream power branches, a sampling initial time and a fixed sampling period are set, and initial voltage sampling of each downstream power branch is performed;

[0008] A historical voltage sampling data set is constructed based on historical sampling data of each downstream power branch, voltage mean value, instantaneous signal power estimation and noise power estimation are calculated, and an instantaneous signal-to-noise ratio is obtained;

[0009] A sampling frequency level set and a signal-to-noise ratio threshold set are constructed, a sampling frequency level is selected according to an interval in which the instantaneous signal-to-noise ratio is located, and a matched sampling period is set;

[0010] A current sampling time window is set at a statistical and sampling decision time, a segmented and uniform sampling time point sequence is generated in the current sampling time window according to the sampling period, and a corresponding voltage sampling sequence is obtained;

[0011] In the current sampling time window, a local deviation is calculated according to a voltage at each sampling time point and a voltage at the statistical and sampling decision time of the current sampling time window, a pre-filtering energy threshold is constructed, and pre-filtering is performed, so as to form a reserved sampling time point set and a reserved voltage sequence;

[0012] For the reserved voltage sequence of each downstream power branch in the current sampling time window, a fixed-length signal vector is constructed, an orthogonal normalization basis is selected, linear orthogonal transformation is performed, and a compressed coding coefficient vector is generated;

[0013] A bandwidth priority index is constructed according to a compressed data amount, an instantaneous signal-to-noise ratio and an uplink communication bandwidth of each downstream power branch, a total order relationship determined by the bandwidth priority index and the compressed data amount is established in the same edge collection device, buffer scheduling and bandwidth scheduling are performed, and an actual uploading branch set in a current uploading scheduling period is selected;

[0014] For the actual uploading branch set, a structured uploading data record containing an edge collection device identifier, a downstream power branch identifier, a downstream power branch number, a current sampling time window start time and a current sampling time window end time, a reserved sampling time point number in the current sampling time window, and a compressed coding coefficient vector obtained by orthogonal transformation is generated and uploaded.

[0015] Optionally, the edge collection device is deployed, a numbering relationship is established with downstream power branches, a sampling initial time and a fixed sampling period are set, and initial voltage sampling of each downstream power branch is performed, and specifically comprises:

[0016] installing an edge collection device in each distribution box, assigning a unique number to each edge collection device, forming an edge collection device index set, and recording the total number of edge collection devices;

[0017] accessing a plurality of downstream power branches under each edge collection device, generating a unique branch identification for each downstream power branch according to a combination rule of the edge collection device number and the corresponding downstream power branch sequence number of the edge collection device, and recording the number of downstream power branches of each edge collection device;

[0018] installing a voltage signal collection module on each downstream power branch, setting the system startup time as the initial sampling time, and setting a fixed sampling period during the system startup phase as a uniform sampling step on a continuous time axis;

[0019] setting an initial sampling frequency during the system startup phase, the initial sampling frequency and the sampling period being inversely proportional to each other;

[0020] After the modules are enabled, voltage sampling of all downstream power branches is completed at the first sampling time point, the original voltage value at the system startup time is obtained, and the initial voltage sampling record is stored in the edge collection device.

[0021] Optionally, the historical voltage sampling data set is constructed based on the historical sampling data of each downstream power branch, the voltage mean value, the instantaneous signal power estimate and the noise power estimate are calculated, and the instantaneous signal-to-noise ratio is obtained, which specifically includes:

[0022] At the statistical and sampling decision time, the number of effective historical samples participating in the statistics is obtained according to the number of decisions completed since the system startup and the maximum historical sample number target value; when the actual sample number is less than the maximum historical sample number target value, the actual sample number is used; when the actual sample number is not less than the maximum historical sample number target value, the maximum historical sample number target value is used;

[0023] From the statistical and sampling decision time, index each time forward in the order of decisions to construct a set of historical decision times, and arrange them in chronological order;

[0024] Read the voltage sampling values corresponding to each time from the set of historical decision times, and obtain the voltage mean value corresponding to the statistical and sampling decision time by performing an arithmetic average operation on all voltage sampling values;

[0025] Based on the voltage sampling values in the set of historical decision times, perform an arithmetic average operation on the square of each voltage sampling value to obtain the instantaneous signal power estimate value at the statistical and sampling decision time;

[0026] Based on the voltage mean value, the difference between each historical voltage sampling value and the voltage mean value is squared and arithmetically averaged to obtain a noise power estimation value at the sampling decision time;

[0027] When the noise power estimation value is greater than zero, the ratio of the instantaneous signal power estimation value to the noise power estimation value is taken as the instantaneous signal-to-noise ratio, and the ratio is truncated according to a preset maximum effective upper limit of the instantaneous signal-to-noise ratio; when the ratio exceeds the maximum effective upper limit of the instantaneous signal-to-noise ratio, the instantaneous signal-to-noise ratio is limited to the maximum effective upper limit of the instantaneous signal-to-noise ratio;

[0028] When the noise power estimation value is equal to zero and the instantaneous signal power estimation value is greater than zero, the instantaneous signal-to-noise ratio is set to the maximum effective upper limit of the instantaneous signal-to-noise ratio; when the noise power estimation value and the instantaneous signal power estimation value are both equal to zero, the instantaneous signal-to-noise ratio is set to zero.

[0029] Optionally, the sampling frequency level set and the signal-to-noise ratio threshold set are constructed, and the sampling frequency level is selected according to the interval in which the instantaneous signal-to-noise ratio is located, and a matching sampling period is set, specifically including:

[0030] A preset sampling frequency level set is arranged, each sampling frequency level is arranged in ascending order of value to form a sampling frequency level index;

[0031] According to the maximum effective upper limit of the instantaneous signal-to-noise ratio, the signal-to-noise ratio range from zero to the maximum effective upper limit of the instantaneous signal-to-noise ratio is divided into a plurality of continuous subintervals according to equal intervals, a signal-to-noise ratio threshold value is set for each subinterval, and a one-to-one correspondence is established between the signal-to-noise ratio threshold value and the sampling frequency level, so that the subinterval with a lower signal-to-noise ratio corresponds to a lower sampling frequency level, and the subinterval with a higher signal-to-noise ratio corresponds to a higher sampling frequency level;

[0032] At each statistical and sampling decision time, the current instantaneous signal-to-noise ratio is compared with the signal-to-noise ratio threshold value one by one; when the instantaneous signal-to-noise ratio falls into any subinterval, the sampling frequency level corresponding to the falling subinterval is selected as the sampling frequency of the current downstream power branch; when the instantaneous signal-to-noise ratio reaches or exceeds the highest threshold value, the highest sampling frequency is selected from the sampling frequency level set;

[0033] According to the selected sampling frequency level, a sampling period is set, and the sampling period and the sampling frequency are inversely proportional to each other, and voltage sampling is performed according to the selected sampling period in the subsequent sampling time window.

[0034] Optionally, the sampling time window is set at the statistical and sampling decision time, a segmented and uniform sampling time point sequence is generated in the sampling time window according to the sampling period, and a corresponding voltage sampling sequence is obtained, specifically including:

[0035] A sampling time window length is set for each downstream power branch, and a time interval from the statistical and sampling decision time to the statistical and sampling decision time plus the sampling time window length is set as the current sampling time window at the statistical and sampling decision time;

[0036] In the current sampling time window, a sampling time point sequence corresponding to the downstream power branch is generated in a recursive manner according to the sampling period obtained at the statistical and sampling decision time, wherein the first sampling time point is consistent with the statistical and sampling decision time, and each subsequent sampling time point is increased by one sampling period based on the previous sampling time point;

[0037] The sampling time window length is added to the statistical and sampling decision time to obtain the next statistical and sampling decision time, and sampling and decision in the next sampling time window are started at the next statistical and sampling decision time;

[0038] In the current sampling time window, all sampling time points no later than the end time of the sampling time window are filtered to form a sampling time point set and arranged in chronological order, and voltage sampling values at each sampling time point are read to form a voltage sampling sequence corresponding to the sampling time point set.

[0039] Optionally, in the current sampling time window, a local deviation is calculated according to the deviation of the voltage at each sampling time point from the voltage at the statistical and sampling decision time of the current sampling time window, a pre-filtering energy threshold is constructed and pre-filtering is performed to form a retained sampling time point set and a retained voltage sequence, specifically including:

[0040] In each sampling time window, for each sampling time point in the sampling time point set, the difference between the voltage at each sampling time point and the voltage at the statistical and sampling decision time of the current sampling time window is calculated, and the absolute value of the difference is obtained as the local deviation;

[0041] A pre-filtering energy threshold is constructed based on the noise power estimate value of the current sampling time window, and the noise power estimate value is used as a parameter of the pre-filtering energy threshold;

[0042] For each sampling time point in the sampling time point set, the local deviation is compared with the pre-filtering energy threshold; when the local deviation is greater than or equal to the pre-filtering energy threshold, the corresponding sampling time point is retained, and when the local deviation is less than the pre-filtering energy threshold, the corresponding sampling time point is discarded;

[0043] The sampling time points retained under the pre-filtering rule are combined to form a retained sampling time point set, and the voltage sampling values corresponding to the retained sampling time points are combined to form a retained voltage sequence.

[0044] Optionally, the reserved voltage sequence of each downstream power branch in the current sampling time window is constructed into a fixed-length signal vector, an orthogonal normalized basis is selected, a linear orthogonal transformation is performed, and a compressed encoding coefficient vector is generated, specifically including:

[0045] The length of the encoding vector and the number of encoding coefficients are set, and the number of encoding coefficients is not greater than the length of the encoding vector;

[0046] A basis vector set with a length equal to the length of the encoding vector is constructed, in which each basis vector is only one in one component position and zero in the remaining component positions, and the inner product of each basis vector is zero and the inner product of each basis vector is one, forming an orthogonal normalized basis;

[0047] The number of elements of the reserved sampling time point set in the current sampling time window is counted as the number of reserved sampling time points, and the reserved voltage sequence is rearranged in chronological order to obtain a time-ordered reserved voltage sequence;

[0048] According to the relationship between the number of reserved sampling time points and the length of the encoding vector, a signal vector participating in transformation is constructed: when the number of reserved sampling time points is greater than or equal to the length of the encoding vector, the first several reserved voltage values in time order are sequentially filled into each component of the signal vector; when the number of reserved sampling time points is greater than zero and less than the length of the encoding vector, all reserved voltage values are sequentially filled into each component in the front part of the signal vector and the remaining components are filled with zero; when the number of reserved sampling time points is equal to zero, all components of the signal vector are filled with zero;

[0049] For each basis vector in the orthogonal normalized basis, a linear transformation operation is performed, the components of the signal vector are multiplied by the corresponding components of the current basis vector and summed to obtain the corresponding encoding coefficient, and all encoding coefficients are combined into a compressed encoding coefficient vector as the compressed encoding coefficient vector of the corresponding downstream power branch in the current sampling time window.

[0050] Optionally, the bandwidth priority index is constructed according to the compressed data amount, the instantaneous signal-to-noise ratio and the uplink communication bandwidth of each downstream power branch, the total order relationship determined by the bandwidth priority index and the compressed data amount is established in the same edge collection device, the cache scheduling and bandwidth scheduling are performed, and the actual upload branch set in the current upload scheduling period is selected, specifically including:

[0051] At the beginning of each upload scheduling period, the available uplink communication bandwidth of each edge collection device in the current upload scheduling period is obtained, the number of bits occupied by a single encoding coefficient after quantization of the compressed encoding coefficient is obtained, and the length of the current upload scheduling period is set;

[0052] For each downstream power branch, the compression data amount of each downstream power branch in the current upload scheduling period is calculated according to the product of the number of encoded coefficients that each downstream power branch needs to upload in the current upload scheduling period and the number of bits occupied by a single encoded coefficient;

[0053] At the time of statistics and sampling decision, a bandwidth priority index is constructed according to the compression data amount and the instantaneous signal-to-noise ratio of each downstream power branch, so that the downstream power branch with a larger compression data amount and a lower instantaneous signal-to-noise ratio obtains a larger bandwidth priority index, and the downstream power branch with a smaller compression data amount and a higher instantaneous signal-to-noise ratio obtains a smaller bandwidth priority index, and a total order relationship is established in the same edge collection device, which is determined by the bandwidth priority index and the compression data amount.

[0054] For all downstream power branches of the same edge collection device, the total order relationship determined by the bandwidth priority index and the compression data amount is used to arrange the downstream power branches in order from the downstream power branch with a smaller bandwidth priority index; in the case of the same bandwidth priority index, the downstream power branch with a smaller compression data amount is arranged in front, and an ordered downstream power branch sequence is obtained.

[0055] For each edge collection device, the corresponding compression data amount is added in order from the head of the ordered downstream power branch sequence; when the addition result does not exceed the product of the available uplink bandwidth and the time length of the current upload scheduling period, the addition continues; when the next downstream power branch is added and the product of the available uplink bandwidth and the time length of the current upload scheduling period is exceeded, the addition stops, and the number of downstream power branches participating in the addition is recorded as the number of downstream power branches that can be arranged for upload in the current upload scheduling period; when the compression data amount of any single downstream power branch is greater than the product of the available uplink bandwidth and the time length of the current upload scheduling period, the number of downstream power branches that can be arranged for upload is set to zero.

[0056] When the number of downstream power branches that can be arranged for upload is greater than zero, the first several downstream power branches in the ordered downstream power branch sequence that are consistent with the number of downstream power branches that can be arranged for upload are grouped into an actual upload branch set of the current upload scheduling period; when the number of downstream power branches that can be arranged for upload is equal to zero, no data upload is performed in the current upload scheduling period.

[0057] Optionally, for the actual upload branch set, a structured upload data record containing the edge collection device identifier, the downstream power branch identifier, the number of downstream power branches, the start time and the end time of the current sampling time window, the number of reserved sampling time points in the current sampling time window, and the compression encoded coefficient vector obtained by orthogonal transformation is generated and uploaded, specifically including:

[0058] For each edge collection device, traverse the downstream power branches in the actual upload branch set of the current upload scheduling period, and construct a structured upload data record for each downstream power branch, the structured upload data record at least includes the following fields: edge collection device identifier, downstream power branch identifier, downstream power branch quantity, this sampling time window start time, this sampling time window end time, the number of retained sampling time points in this sampling time window, compressed encoding coefficient vector obtained by orthogonal transformation;

[0059] The structured upload data corresponding to all actual upload branches of the same edge collection device in the current upload scheduling period is sent to the data center on the uplink.

[0060] The present application has the following advantages:

[0061] 1. The traditional centralized collection is changed to introduce edge nodes at the distribution box level, and each collection device and its associated multiple downstream branches are accurately corresponded through numbering system, forming a hierarchical management architecture. The present scheme independently numbers and stores the initial voltage value of a single branch, so that subsequent links can be accurately indexed and traced back. The system scalability is improved: new branches or nodes only need to be numbered at the local edge node to quickly access; fault tolerance is enhanced: a single node failure will not affect the entire system, providing a basis for fault location and recovery; the configuration process is simplified: through automatic numbering and the establishment of device index set, the cost of manual configuration and management is reduced.

[0062] 2. The historical sampling data is introduced into the online channel state evaluation, the historical voltage sampling set is constructed, and the optimal sample number target is combined to realize the dynamic estimation of signal mean and noise power, so that the instantaneous signal-to-noise ratio is obtained. The present scheme sets an optimal historical sample threshold, taking into account the calculation complexity and estimation accuracy, effectively alleviating the influence of signal mutation or extreme noise on the estimation result. In addition, the two extreme cases of zero noise power and zero signal power are included in special processing, so that the signal-to-noise ratio can still be reasonably truncated or zeroed in special scenarios, avoiding calculation errors and abnormal data reporting caused by non-zero or numerical divergence.

[0063] 3. The instantaneous signal-to-noise ratio is hierarchically mapped to multiple preset sampling frequency levels, and the sampling period is dynamically adjusted. The present scheme equally divides the signal-to-noise ratio range into several subintervals, and finely matches different quality intervals, so as to realize smoother and more fine-grained sampling frequency switching. The sampling points can be significantly increased when the channel quality is good, thereby improving the system's ability to capture rapidly changing voltage waveforms; the sampling frequency is reduced in time when the channel quality is poor, reducing redundant data generation and invalid transmission; without manual threshold tuning, the entire frequency and threshold set only depends on the upper bound of the system's maximum signal-to-noise ratio, making deployment and maintenance more convenient.

[0064] 4、 Adopted the segmented recursive generation method based on the decision moment and the sampling window length, realized the uniform multi-moment sampling of each branch in the same window. The scheme dynamically determines the starting time and window length at the statistical decision moment, and generates the time sequence point by point in a recursive manner, so that the sampling period can be modified flexibly. It ensures that each sampling point can be accurately positioned at uniform intervals after the sampling frequency changes; it is convenient to dock with pre-filtering and compression algorithm, and the time point sequence can be directly used for data comparison; it improves the adaptability to jitter or packet loss scene, and can smoothly roll to the next window execution when part of the time point sampling fails.

[0065] 5、 Take the deviation of the statistical voltage and the voltage of each sampling time point as the local deviation, and compare it with the energy threshold constructed based on noise estimation, to form a refined sampling point pre-selection strategy. The scheme dynamically assigns the noise power estimation value to the energy threshold, realizes adaptive filtering. The dynamic threshold can adjust in time according to the actual noise level, and more accurately distinguish between effective signals and noise; the node set is mapped to the corresponding voltage data, reducing the redundancy of subsequent compression transmission; it supports automatic shielding of abnormal points in the scene of severe voltage fluctuation, and improves the sampling quality.

[0066] 6、 Map the reserved voltage sequence to a fixed-length signal vector, and perform linear transformation based on the predefined orthogonal normalized basis to generate a compression encoding coefficient vector. The scheme adopts a simple orthogonal basis structure, which has lower calculation amount and is easier to implement in hardware, while ensuring that the components of the vector are orthogonal and independent, and the energy can be completely reconstructed. When the vector length is greater than the length of the reserved sequence, it is automatically filled with zeros, without additional interpolation or redundant data; only a limited number of components in the vector need to be multiplied and accumulated, which is more suitable for edge computing scenarios with limited resources; the coefficient vector reconstructed based on the orthogonal characteristic has both compression efficiency and the ability to accurately recover key signal information during decompression.

[0067] 7、 The compression data amount of each branch and the instantaneous signal-to-noise ratio are jointly included in the bandwidth priority index, and the priority of multiple branches in the current transmission period is sorted by constructing a total order relationship. The scheme considers the time efficiency requirements of branches with large data amount but general signal quality, and branches with small data amount but high signal-to-noise ratio, through double index weighing, so as to improve the overall transmission efficiency. Dynamic scheduling can ensure that high-priority branches are uploaded first when bandwidth is tight; when bandwidth is abundant, more branches can be uploaded concurrently, improving resource utilization; the total order relationship is strictly defined and repeatable, avoiding priority conflicts. BRIEF DESCRIPTION OF DRAWINGS

[0068] Figure 1 The flowchart of the present application. DETAILED DESCRIPTION

[0069] Clearly, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0070] Embodiments, with reference to Figure 1 A power line communication optimization collection method based on edge fusion computing, comprising:

[0071] Deploy an edge collection device, establish a numbering relationship with downstream power branches, set a sampling initial time and a fixed sampling period, and perform initial voltage sampling of each downstream power branch;

[0072] Based on the historical sampling data of each downstream power branch, a historical voltage sampling data set is constructed, the mean voltage, instantaneous signal power estimation and noise power estimation are calculated, and the instantaneous signal-to-noise ratio is obtained;

[0073] A set of sampling frequency levels and a set of signal-to-noise ratio thresholds are constructed, a sampling frequency level is selected according to the interval of the instantaneous signal-to-noise ratio, and a matching sampling period is set;

[0074] Set a sampling time window at the statistical and sampling decision time, generate a segmented and uniform sampling time point sequence according to the sampling period within the sampling time window, and obtain the corresponding voltage sampling sequence;

[0075] Within the sampling time window, calculate the local deviation according to the deviation of the voltage at each sampling time point from the voltage at the statistical and sampling decision time of the sampling time window, construct a pre-filter energy threshold and perform pre-filtering, and form a reserved sampling time point set and a reserved voltage sequence;

[0076] For each downstream power branch, construct a fixed-length signal vector, select an orthogonal and normalized basis, perform linear orthogonal transformation, and generate a compressed and encoded coefficient vector;

[0077] According to the compressed data amount, the instantaneous signal-to-noise ratio and the uplink communication bandwidth of each downstream power branch, a bandwidth priority index is constructed, a total order relationship determined by the bandwidth priority index and the compressed data amount is established in the same edge collection device, a cache scheduling and bandwidth scheduling is performed, and an actual upload branch set in the current upload scheduling period is selected;

[0078] For the actual upload branch set, a structured upload data record containing the edge collection device identifier, the downstream power branch identifier, the number of downstream power branches, the start time and end time of the current sampling time window, the number of reserved sampling time points in the current sampling time window, and the compressed and encoded coefficient vector obtained by orthogonal transformation is generated and uploaded.

[0079] By deploying edge collection devices inside the distribution box and establishing accurate number mapping relationship with each downstream power branch, the scheme first completes the collection of the initial voltage signal of each branch, providing a standardized and traceable original data basis for subsequent dynamic decision-making and data processing; then it uses historical sampling data to build a voltage data set and calculates the voltage mean and instantaneous signal and noise power estimation to obtain the signal-to-noise ratio of each branch in real time, solving the problem that the previous communication system cannot judge the channel quality in a noisy environment, and ensuring the accuracy of subsequent sampling and transmission decision; then by constructing a multi-level sampling frequency and signal-to-noise ratio threshold set, the signal-to-noise ratio is mapped to a multi-grade sampling period, realizing the adaptive adjustment of automatically increasing the sampling density when the channel condition is good and automatically reducing the sampling frequency when the channel is poor, avoiding the waste of redundant data and resources caused by excessive sampling; after the sampling period is determined, the scheme sets a sampling time window at each statistical and sampling decision-making moment, and generates uniform sampling time points within the window according to the sampling period, ensuring the balance and continuity of the sampling timing; then, according to the difference between the voltage of each sampling point and the voltage at the decision-making moment, a pre-filtering energy threshold is constructed to discard abnormal or severely noisy sampling points, and only keep the data that significantly contributes to signal characteristics, solving the problem of false sampling caused by noise mutation or interference; after pre-filtering, a fixed-length signal vector is constructed for each branch voltage sequence, and a linear transformation is performed using an orthogonal normalization basis to generate compressed encoding coefficients, effectively compressing the data volume while preserving the main information of the signal; finally, according to the compressed data volume, signal-to-noise ratio and available uplink bandwidth of each branch, a bandwidth priority index is constructed and buffer scheduling is performed to select the optimal upload branch set in the current scheduling period, ensuring that critical branches are uploaded first and bandwidth is fully utilized; structured upload data records are generated for the selected branches and packaged for transmission, realizing a complete closed loop from local collection to cloud transmission.

[0080] The deployment edge collection device establishes a number relationship with the downstream power branch, sets a sampling initial time and a fixed sampling period, and performs initial voltage sampling of each downstream power branch, specifically including:

[0081] Install one edge collection device in each distribution box, assign a unique number to each edge collection device, form an edge collection device index set, and record the total number of edge collection devices;

[0082] Connect a number of downstream power branches to each edge collection device, generate a unique branch identifier for each downstream power branch according to the combination rule of edge collection device number and corresponding edge collection device downstream power branch sequence number, and record the number of downstream power branches for each edge collection device;

[0083] A voltage signal acquisition module is installed on each downstream power branch, the system starting time is set as the initial sampling time, and a fixed sampling period is set in the system starting stage as the uniform sampling step on the continuous time axis;

[0084] An initial sampling frequency is set in the system starting stage, and the initial sampling frequency and the sampling period are reciprocal of each other;

[0085] After each module is enabled, the voltage sampling of all downstream power branches is completed at the first sampling time point, the original voltage value at the system starting time is obtained, and the initial voltage sampling record is stored in the edge collection device.

[0086] An edge collection device is deployed in each distribution box, numbered as ; wherein, is the unique identifier of the th edge collection device; is the index number of the edge collection device; is the total number of edge collection devices deployed in the system;

[0087] Each edge device is connected with downstream power branches, each branch numbered as ; wherein, is the serial number of the power branch under the edge device ; is the number of downstream power branches connected to the edge device ; is the identifier of the th edge device and the power branch numbered as under the device;

[0088] On each branch , a voltage signal acquisition module is installed, the initial sampling clock is set to , and the sampling period in the system starting stage is set to ; wherein, is the continuous time independent variable; is the specific time point at the system starting time; is the fixed sampling period used in the system starting stage;

[0089] The initial sampling frequency in the system starting stage is set to ;

[0090] After all modules are enabled, the original voltage value of each branch is collected at the first time step , denoted as: ; wherein, is the branch on the continuous time a voltage signal value corresponding to the moment; a branch an initial voltage sampling value at a system start moment.

[0091] The historical voltage sampling data set is constructed based on historical sampling data of each downstream power branch, and a voltage mean value, an instantaneous signal power estimate, and a noise power estimate are calculated to obtain an instantaneous signal-to-noise ratio, specifically including:

[0092] At the statistical and sampling decision moment, the number of effective historical samples participating in statistics is obtained according to the number of decisions completed since the system starts and a maximum historical sample number target value; when the actual sample number is less than the maximum historical sample number target value, the actual sample number is used; when the actual sample number is not less than the maximum historical sample number target value, the maximum historical sample number target value is used.

[0093] Among the plurality of historical decision moments corresponding to the number of effective historical samples, the historical decision moment set is constructed by indexing each moment in the order of decision from the statistical and sampling decision moment, and arranged in chronological order.

[0094] The voltage sampling value corresponding to each moment is read from the historical decision moment set, and the voltage mean value corresponding to the statistical and sampling decision moment is obtained by performing an arithmetic average operation on all voltage sampling values.

[0095] Based on the voltage sampling values in the historical decision moment set, the arithmetic average operation is performed on the square of each voltage sampling value to obtain the instantaneous signal power estimate value at the statistical and sampling decision moment.

[0096] Based on the voltage mean value, the difference between each historical voltage sampling value and the voltage mean value is squared and arithmetically averaged to obtain the noise power estimate value at the statistical and sampling decision moment.

[0097] When the noise power estimate value is greater than zero, the ratio of the instantaneous signal power estimate value to the noise power estimate value is taken as the instantaneous signal-to-noise ratio, and is truncated according to a preset maximum effective upper limit of the instantaneous signal-to-noise ratio; when the ratio exceeds the maximum effective upper limit of the instantaneous signal-to-noise ratio, the instantaneous signal-to-noise ratio is limited to the maximum effective upper limit of the instantaneous signal-to-noise ratio.

[0098] When the noise power estimate value is equal to zero and the instantaneous signal power estimate value is greater than zero, the instantaneous signal-to-noise ratio is set to the maximum effective upper limit of the instantaneous signal-to-noise ratio; when the noise power estimate value and the instantaneous signal power estimate value are both equal to zero, the instantaneous signal-to-noise ratio is set to zero.

[0099] At the th sampling moment , the number of effective historical samples of the branch is:

[0100] ; wherein, is a discrete index of the statistical and sampling decision; is a discrete index corresponds to the statistical and sampling decision moment; is a maximum historical sample number target value required for calculating the statistic; is the decision moment at which the statistic is calculated; corresponds to the branch effective historical sample number;

[0101] Let the corresponding set of historical statistical and sampling decision moments be:

[0102] ; wherein, is the offset relative to the current decision index in the set of historical statistical and sampling decision moments; is the statistical and sampling decision moment in the decision sequence that is offset by decision intervals relative to the current index ;

[0103] Calculate the signal mean: ; wherein, is the voltage mean of the branch calculated according to the sampling values of the most recent decision moments at the statistical and sampling decision moment ;

[0104] Calculate the instantaneous signal power: ; wherein, is the instantaneous signal power estimate of the branch at the statistical and sampling decision moment ;

[0105] Calculate the noise power:

[0106] ; wherein, is the noise power estimate of the branch at the statistical and sampling decision moment ;

[0107] Perform steps S201 to S203 to calculate the instantaneous signal-to-noise ratio:

[0108] S201, when , let ; wherein, is the instantaneous signal-to-noise ratio of the branch at the statistical and sampling decision moment ; The maximum effective upper bound of the instantaneous signal-to-noise ratio;

[0109] S202, when , ;

[0110] S203, when , .

[0111] The set of sampling frequency levels and the set of signal-to-noise ratio thresholds are selected according to the interval of the instantaneous signal-to-noise ratio, and the matching sampling period is set, specifically including:

[0112] The set of preset sampling frequency levels is arranged in ascending order of numerical value to form a sampling frequency level index;

[0113] According to the maximum effective upper bound of the instantaneous signal-to-noise ratio, the signal-to-noise ratio range from zero to the maximum effective upper bound of the instantaneous signal-to-noise ratio is divided into several continuous subintervals according to equal intervals, and the signal-to-noise ratio threshold value is set for each subinterval, and a one-to-one correspondence is established with the sampling frequency level, so that the subinterval with lower signal-to-noise ratio corresponds to lower sampling frequency level, and the subinterval with higher signal-to-noise ratio corresponds to higher sampling frequency level;

[0114] At each statistical and sampling decision moment, the current instantaneous signal-to-noise ratio is compared with the signal-to-noise ratio threshold value one by one; when the instantaneous signal-to-noise ratio falls into any subinterval, the sampling frequency level corresponding to the falling subinterval is selected as the sampling frequency of the current downstream power branch; when the instantaneous signal-to-noise ratio reaches or exceeds the highest threshold value, the highest sampling frequency is selected from the set of sampling frequency levels;

[0115] According to the selected sampling frequency level, the sampling period is set, and the sampling period and the sampling frequency are reciprocal of each other. In the subsequent sampling time window, the voltage sampling is performed according to the selected sampling period.

[0116] The set of sampling frequency levels available for the system to select is , and ; wherein, is the th sampling frequency value in the set of sampling frequency levels ; is the sampling frequency level index, and the value range is ; is the total number of sampling frequency levels;

[0117] The set of signal-to-noise ratio thresholds for hierarchical selection of sampling frequency is , specifically:

[0118] , , ; wherein, is the first signal-to-noise ratio threshold value; is the second signal-to-noise ratio threshold value;

[0119] is the first signal-to-noise ratio threshold value; , is the second signal-to-noise ratio threshold value; is the lower bound of the signal-to-noise ratio threshold; is the upper bound of the signal-to-noise ratio threshold;

[0120] In step S301 and step S302, the current signal-to-noise ratio is matched with the threshold set, specifically:

[0121] S301, when , , the sampling frequency of the branch is set as: ; wherein, is the statistical and sampling decision time is the sampling frequency selected by the branch ;

[0122] S302, when , set ;

[0123] The corresponding sampling period is set as: ; wherein, is the statistical and sampling decision time , the sampling period used by the branch in the subsequent sampling window.

[0124] The statistical and sampling decision time is set as the current sampling time window, and a segmented and uniform sampling time point sequence is generated in the current sampling time window according to the sampling period, and a corresponding voltage sampling sequence is obtained, specifically including:

[0125] The sampling time window length is set for each downstream power branch, and at the statistical and sampling decision time, the time interval from the statistical and sampling decision time to the statistical and sampling decision time plus the sampling time window length is set as the current sampling time window;

[0126] In the current sampling time window, according to the sampling period obtained at the statistical and sampling decision time, the sampling time point sequence corresponding to the downstream power branch is generated in a recursive manner, wherein the first sampling time point is consistent with the statistical and sampling decision time, and each subsequent sampling time point is increased by one sampling period based on the previous sampling time point;

[0127] The sampling time window length is added to the statistical and sampling decision time to obtain the next statistical and sampling decision time, and the sampling and decision in the next sampling time window are started at the next statistical and sampling decision time.

[0128] Within this sampling time window, all sampling time points no later than the end of the sampling time window are selected to form a set of sampling time points, which are then arranged in chronological order. The corresponding voltage sampling values ​​are read at each sampling time point to form a voltage sampling sequence that corresponds one-to-one with the set of sampling time points.

[0129] Set the sampling time window length to The window time interval is ;

[0130] Within this time interval, based on time... Determined sampling period The sampling time of this branch is calculated recursively, specifically as follows:

[0131] , And set the next statistical and sampling decision time as: ;in, In the first Within each window, the branch The Each sampling time; The sequence number of the sampling point within the window;

[0132] The conditions will be met. All The set of sampling time points is composed of:

[0133] The corresponding sampling voltage sequence is:

[0134] ;in, branch road In the window The set of all sampling time points within; branch road At sampling time The voltage signal at that time.

[0135] Within the current sampling time window, the local deviation is calculated based on the deviation between the voltage at each sampling time point and the voltage at the statistical and sampling decision time points of the current sampling time window. A pre-filtering energy threshold is then constructed and pre-filtering is performed to form a set of retained sampling time points and a retained voltage sequence. Specifically, this includes:

[0136] Within each sampling time window, for each sampling time point in the set of sampling time points, calculate the difference between the voltage at each sampling time point and the voltage at the statistical sampling decision time of this sampling time window, and obtain the absolute value of the difference as the local deviation.

[0137] The pre-filtering energy threshold is constructed based on the noise power estimation value in the sampling time window, and the noise power estimation value is used as a parameter of the pre-filtering energy threshold;

[0138] For each sampling time point in the set of sampling time points, the local deviation is compared with the pre-filtering energy threshold; when the local deviation is greater than or equal to the pre-filtering energy threshold, the corresponding sampling time point is retained; when the local deviation is less than the pre-filtering energy threshold, the corresponding sampling time point is discarded;

[0139] The sampling time points retained under the pre-filtering rule are grouped into a set of retained sampling time points, and the voltage sampling values corresponding to the retained sampling time points are grouped into a retained voltage sequence in time sequence.

[0140] For each sampling point time in the set , the local difference energy is calculated, specifically: ; wherein, is the absolute value of the local voltage deviation of the branch at the sampling time ;

[0141] The energy threshold is constructed: ; wherein, is the pre-filtering threshold value of the branch at the statistical and sampling decision time ;

[0142] The pre-filtering decision is performed for each sampling point:

[0143] When , the sampling point is retained;

[0144] When , the sampling point is discarded;

[0145] The set of retained sampling points is constructed as:

[0146] , and the retained voltage values corresponding to are arranged in time increasing order to form a sequence:

[0147] ; wherein, is the set of sampling time points judged to be retained under the pre-filtering rule; is the sequence of retained voltage values corresponding to in one-to-one correspondence and in time increasing order.

[0148] The reserved voltage sequence of each downstream power branch in the current sampling time window is used to construct a fixed-length signal vector, an orthogonal normalized basis is selected, a linear orthogonal transformation is performed, and a compressed coding coefficient vector is generated, which specifically includes:

[0149] The length of the coding vector and the number of coding coefficients are set, and the number of coding coefficients is not greater than the length of the coding vector;

[0150] A basis vector set with a length equal to the length of the coding vector is constructed, in which each basis vector is only one in one component position and zero in the remaining component positions, and the inner product of each basis vector is zero and the inner product of each basis vector is one, forming an orthogonal normalized basis;

[0151] The number of elements of the reserved sampling time point set in the current sampling time window is counted as the number of reserved sampling time points, and the reserved voltage sequence is rearranged in chronological order to obtain a time-ordered reserved voltage sequence;

[0152] According to the relationship between the number of reserved sampling time points and the length of the coding vector, a signal vector participating in transformation is constructed: when the number of reserved sampling time points is greater than or equal to the length of the coding vector, the first several reserved voltage values in time order are sequentially filled into each component of the signal vector; when the number of reserved sampling time points is greater than zero and less than the length of the coding vector, all reserved voltage values are sequentially filled into each component in the front part of the signal vector and the remaining components are filled with zero; when the number of reserved sampling time points is equal to zero, all components of the signal vector are filled with zero;

[0153] For each basis vector in the orthogonal normalized basis, a linear transformation operation is performed, the components of the signal vector are multiplied by the corresponding components of the current basis vector and summed to obtain the corresponding coding coefficient, and all coding coefficients are combined into a compressed coding coefficient vector as the compressed coding coefficient vector of the corresponding downstream power branch in the current sampling time window.

[0154] The length of the coding vector is set to , the number of coding coefficients is , and ;

[0155] The basis vector is constructed as: , , ; wherein, is the th basis vector; is the value of the basis vector in the th component; is the basis vector number index; is the vector component position; For dimension The real vector space;

[0156] From this, we can derive the orthogonality normality relation: ;in, Number the basis vectors by index;

[0157] The number of retained sampling points is ;in, branch road The number of sampling points retained after pre-filtering within the current window;

[0158] In chronological order Rearranged into a sequence:

[0159] ;in, To preserve the position index in the voltage sequence Element; To preserve the position numbers in the sequence;

[0160] Construction length is vector Its components are determined according to the following rules:

[0161] S601, when season , ;in, The length used for orthogonal transformation is The signal vector, corresponding to the branch The pre-filtered sampled information within the current window; For vectors In the The values ​​of each component;

[0162] S602, when season:

[0163] ;

[0164] S603, when season , ;

[0165] Based on basis vector pairs Perform a linear transformation to obtain the compressed encoding result:

[0166] , ;in, branch road The corresponding number One coding coefficient;

[0167] all the transform coefficients are arranged into a vector: ; wherein, is a branch compressed encoding coefficient vector under the current window.

[0168] The bandwidth priority index is constructed according to the compressed data amount, the instantaneous signal-to-noise ratio and the uplink communication bandwidth of each downstream power branch, a total order relationship is established in the same edge collection device according to the bandwidth priority index and the compressed data amount, buffer scheduling and bandwidth scheduling are performed, an actual upload branch set in the current upload scheduling period is selected, and the method specifically comprises the following steps:

[0169] At the beginning of each upload scheduling period, the uplink communication bandwidth available to each edge collection device in the current upload scheduling period is obtained, the number of bits occupied by a single encoding coefficient after quantization of the compressed encoding coefficient is obtained, and the time length of the current upload scheduling period is set;

[0170] For each downstream power branch, the compressed data amount required to be uploaded by each downstream power branch in the current upload scheduling period is calculated according to the product of the number of encoding coefficients required to be uploaded by each downstream power branch in the current upload scheduling period and the number of bits occupied by a single encoding coefficient;

[0171] At the time of statistics and sampling decision, the bandwidth priority index is constructed according to the compressed data amount and the instantaneous signal-to-noise ratio of each downstream power branch, so that the downstream power branch with larger compressed data amount and lower instantaneous signal-to-noise ratio obtains a larger bandwidth priority index, and the downstream power branch with smaller compressed data amount and higher instantaneous signal-to-noise ratio obtains a smaller bandwidth priority index, and a total order relationship is established in the same edge collection device according to the sorting order determined by the bandwidth priority index and the compressed data amount;

[0172] For all downstream power branches of the same edge collection device, the downstream power branches are arranged in order according to the total order relationship determined by the bandwidth priority index and the compressed data amount from the downstream power branch with smaller bandwidth priority index; in the case of the same bandwidth priority index, the downstream power branch with smaller compressed data amount is arranged in front, and an ordered downstream power branch sequence is obtained;

[0173] For each edge acquisition device, starting from the beginning of the ordered downstream power branch sequence, the corresponding compressed data volume is accumulated sequentially. Accumulation continues as long as the accumulated result does not exceed the product of the available uplink bandwidth and the duration of the current upload scheduling cycle. Accumulation stops when the volume exceeds the product of the available uplink bandwidth and the duration of the current upload scheduling cycle after adding the next downstream power branch, and the number of downstream power branches participating in the accumulation is recorded as the number of downstream power branches that can be scheduled for upload in the current upload scheduling cycle. When the compressed data volume of any single downstream power branch is greater than the product of the available uplink bandwidth and the duration of the current upload scheduling cycle, the number of downstream power branches that can be scheduled for upload is set to zero.

[0174] When the number of downstream power branches that can be scheduled for upload is greater than zero, the first few downstream power branches in the ordered downstream power branch sequence that match the number of downstream power branches that can be scheduled for upload will be grouped into the actual upload branch set for this upload scheduling cycle; when the number of downstream power branches that can be scheduled for upload is equal to zero, no data upload will be performed in this upload scheduling cycle.

[0175] Get the The uplink communication bandwidth available to each edge device in the current scheduling period is denoted as . The number of bits occupied by the coding coefficients after quantization is denoted as ;

[0176] Set the upload scheduling period length to... ;

[0177] Calculate branches The amount of compressed data that needs to be uploaded within a scheduling cycle for:

[0178] ;

[0179] The bandwidth priority function is constructed as follows: ;in, To make statistical and sampling decisions Location, side road The bandwidth priority value;

[0180] For the An edge device indexes all branches. The sequence is obtained by rearranging the elements according to the following total order relation. Specifically:

[0181] or Thus satisfying:

[0182] ;in, For the sorted position at the th Branch index of the position; Numbering the position in the ordered sequence; , Any two branch indexes selected in the ordering comparison process; Total order relation defined on the binary tuple ;

[0183] Set integer To be the maximum non-negative integer satisfying the following formula: ; Wherein, The number of branches that the th edge device can choose to upload data in the current scheduling period;

[0184] And when , it also satisfies ;

[0185] If for all , , let ;

[0186] When , the branch set selected for uploading in the current scheduling period is constructed as:

[0187] ; Wherein, The branch set actually selected for the th edge device to perform data upload in the current scheduling period;

[0188] When , let be an empty set, and no data upload is performed in the current period.

[0189] For the actual upload branch set, generate a structured upload data record containing the edge collection device identifier, downstream power branch identifier, downstream power branch quantity, this sampling time window start time and end time, the number of reserved sampling time points in this sampling time window, the compressed encoding coefficient vector obtained by orthogonal transformation, and upload, specifically including:

[0190] For each edge collection device, traverse the downstream power branch in the actual upload branch set of the current upload scheduling period, and construct a structured upload data record for each downstream power branch. The structured upload data record includes at least the following fields: edge collection device identifier, downstream power branch identifier, downstream power branch quantity, this sampling time window start time, this sampling time window end time, the number of reserved sampling time points in this sampling time window, the compressed encoding coefficient vector obtained by orthogonal transformation;

[0191] The same edge collection device sends the structured upload data corresponding to all actual upload branches in the current scheduling period to the data center on the uplink.

[0192] For each branch , generate a structured upload data tuple:

[0193] ; wherein, is a structured data record that branch needs to upload in the current scheduling period;

[0194] Upload the set to the data center in the form of a data message; wherein, is the structured data tuple set corresponding to all selected upload branches of the th edge device in the current scheduling period.

[0195] It should be noted that in this paper, relational terms such as first and second are used only to distinguish one entity or operation from another, and do not necessarily require or imply that there is any such actual relationship or order between these entities or operations. Moreover, the terms "include", "contain" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device that includes a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such a process, method, article or device.

[0196] The above is only the preferred embodiment of the present application, it should be noted that for those skilled in the art, without departing from the technical principles of the present application, can make a number of improvements and refinements, these improvements and refinements should also be considered as the protection scope of the present application.​

Claims

1. A method for power line communication optimization collection with fusion edge computing, characterized in that, The method comprises the following steps: Deploying an edge collection device, establishing a numbering relationship with downstream power branches, setting a sampling initial time and a fixed sampling period, and performing initial voltage sampling of each downstream power branch; Based on the historical sampling data of each downstream power branch, a historical voltage sampling data set is constructed, the mean voltage, instantaneous signal power estimation and noise power estimation are calculated, and the instantaneous signal-to-noise ratio is obtained; A set of sampling frequency levels and a set of signal-to-noise ratio thresholds are constructed, and the sampling frequency level is selected according to the interval of the instantaneous signal-to-noise ratio, and the matching sampling period is set; Set the sampling time window at the statistical and sampling decision time, generate a segmented and uniform sampling time point sequence according to the sampling period in the sampling time window, and obtain the corresponding voltage sampling sequence; In the sampling time window, calculate the local deviation according to the voltage at each sampling time point and the voltage at the statistical and sampling decision time of the sampling time window, construct a pre-filtering energy threshold and perform pre-filtering to form a reserved sampling time point set and a reserved voltage sequence; For each downstream power branch in the sampling time window, a fixed-length signal vector is constructed, an orthogonal normalization basis is selected, a linear orthogonal transformation is performed, and a compressed coding coefficient vector is generated; According to the compressed data amount, the instantaneous signal-to-noise ratio and the uplink communication bandwidth of each downstream power branch, a bandwidth priority index is constructed, a total order relationship is established in the same edge collection device according to the bandwidth priority index and the compressed data amount, a buffer scheduling and bandwidth scheduling are performed, and an actual uploading branch set in the current uploading scheduling period is selected; For the actual uploading branch set, a structured uploading data record containing the edge collection device identifier, the downstream power branch identifier, the number of downstream power branches, the start time and end time of the sampling time window, the number of reserved sampling time points in the sampling time window, and the compressed coding coefficient vector obtained by orthogonal transformation is generated and uploaded. 2.The method of claim 1, wherein, The method comprises the following steps: Deploying an edge collection device, establishing a numbering relationship with downstream power branches, setting a sampling initial time and a fixed sampling period, and performing initial voltage sampling of each downstream power branch, specifically comprising: Install an edge collection device in each distribution box, assign a unique number to each edge collection device, form an edge collection device index set, and record the total number of edge collection devices; Connect a plurality of downstream power branches to each edge collection device, generate a unique branch identifier for each downstream power branch according to the combination rule of edge collection device number and downstream power branch sequence number of the corresponding edge collection device, and record the number of downstream power branches of each edge collection device; Install a voltage signal collection module on each downstream power branch, set the system startup time as the sampling initial time, and set the fixed sampling period as the uniform sampling step on the continuous time axis during the system startup stage; Set the initial sampling frequency during the system startup stage, and the initial sampling frequency and the sampling period are inversely proportional to each other; After the modules are enabled, complete the voltage sampling of all downstream power branches at the first sampling time point, obtain the original voltage value at the system startup time, and store it as an initial voltage sampling record in the edge collection device. 3.The method of claim 2, wherein, The historical voltage sampling data set is constructed based on historical sampling data of each downstream power branch, and voltage mean value, instantaneous signal power estimation and noise power estimation are calculated to obtain instantaneous signal-to-noise ratio, specifically comprising: At the statistical and sampling decision moment, the number of effective historical samples participating in statistics is obtained according to the number of decisions completed since the system starts and the maximum historical sample number target value; when the actual sample number is less than the maximum historical sample number target value, the actual sample number is used; when the actual sample number is not less than the maximum historical sample number target value, the maximum historical sample number target value is used; Among the plurality of historical decision moments corresponding to the number of effective historical samples, each is indexed in order from the statistical and sampling decision moment, a historical decision moment set is constructed, and arranged in chronological order; The voltage sampling value corresponding to each moment is read from the historical decision moment set, and the voltage mean value corresponding to the statistical and sampling decision moment is obtained by performing arithmetic average operation on all voltage sampling values; Based on the voltage sampling values in the historical decision moment set, the arithmetic average operation is performed on the square of each voltage sampling value to obtain the instantaneous signal power estimation value at the statistical and sampling decision moment; Based on the voltage mean value, the square operation is performed on the difference between each historical voltage sampling value and the voltage mean value, and the arithmetic average is performed to obtain the noise power estimation value at the statistical and sampling decision moment; When the noise power estimation value is greater than zero, the ratio of the instantaneous signal power estimation value to the noise power estimation value is taken as the instantaneous signal-to-noise ratio, and the instantaneous signal-to-noise ratio is truncated according to the preset maximum effective upper limit of the instantaneous signal-to-noise ratio; when the ratio exceeds the maximum effective upper limit of the instantaneous signal-to-noise ratio, the instantaneous signal-to-noise ratio is limited to the maximum effective upper limit of the instantaneous signal-to-noise ratio; When the noise power estimation value is equal to zero and the instantaneous signal power estimation value is greater than zero, the instantaneous signal-to-noise ratio is set to the maximum effective upper limit of the instantaneous signal-to-noise ratio; when the noise power estimation value and the instantaneous signal power estimation value are both equal to zero, the instantaneous signal-to-noise ratio is set to zero.

4. The method of claim 3, wherein the method further comprises: The sampling frequency level set and the signal-to-noise ratio threshold set are constructed, the sampling frequency level is selected according to the interval where the instantaneous signal-to-noise ratio is located, and the matching sampling period is set, specifically comprising: A set of preset sampling frequency levels is arranged, each sampling frequency level is arranged in ascending order of value to form a sampling frequency level index; According to the maximum effective upper limit of the instantaneous signal-to-noise ratio, the signal-to-noise ratio range from zero to the maximum effective upper limit of the instantaneous signal-to-noise ratio is divided into several continuous subintervals according to equal intervals, a signal-to-noise ratio threshold value is set for each subinterval, and a one-to-one correspondence is established between the signal-to-noise ratio threshold value and the sampling frequency level, so that the subinterval with lower signal-to-noise ratio corresponds to lower sampling frequency level, and the subinterval with higher signal-to-noise ratio corresponds to higher sampling frequency level; At each statistical and sampling decision moment, the current instantaneous signal-to-noise ratio is compared with the signal-to-noise ratio threshold value one by one; when the instantaneous signal-to-noise ratio falls into any subinterval, the sampling frequency level corresponding to the subinterval where the instantaneous signal-to-noise ratio falls into is selected as the sampling frequency of the current downstream power branch; when the instantaneous signal-to-noise ratio reaches or exceeds the highest threshold value, the highest sampling frequency is selected from the sampling frequency level set; A sampling period is set according to the selected sampling frequency level, and the sampling period and the sampling frequency are reciprocal of each other, and voltage sampling is performed in the subsequent sampling time window according to the selected sampling period.

5. The method of claim 4, wherein the method further comprises: The sampling time window is set at the statistical and sampling decision moment, a segmented and uniform sampling time point sequence is generated in the sampling time window according to the sampling period, and a corresponding voltage sampling sequence is obtained, and specifically comprising: The sampling time window length is set for each downstream power branch, and at the statistical and sampling decision moment, a time interval from the statistical and sampling decision moment to the statistical and sampling decision moment plus the sampling time window length is set as the current sampling time window; In the current sampling time window, the sampling time point sequence corresponding to the downstream power branch is generated in a recursive manner according to the sampling period obtained at the statistical and sampling decision moment, wherein the first sampling time point is consistent with the statistical and sampling decision moment, and each subsequent sampling time point is increased by one sampling period based on the previous sampling time point; The sampling time window length is added to the statistical and sampling decision moment to obtain the next statistical and sampling decision moment, and the sampling and decision in the next sampling time window are started at the next statistical and sampling decision moment; In the current sampling time window, all sampling time points no later than the end time of the sampling time window are screened to form a sampling time point set and arranged in chronological order, and the corresponding voltage sampling value is read at each sampling time point to form a voltage sampling sequence corresponding to the sampling time point set.

6. The method of claim 5, wherein the method further comprises: In the current sampling time window, the local deviation is calculated according to the deviation of the voltage at each sampling time point from the voltage at the statistical and sampling decision moment of the current sampling time window, the pre-filtering energy threshold is constructed and pre-filtering is performed to form a reserved sampling time point set and a reserved voltage sequence, and specifically comprising: In each sampling time window, the difference between the voltage at each sampling time point in the sampling time point set and the voltage at the statistical and sampling decision moment of the current sampling time window is calculated, and the absolute value of the difference is obtained as the local deviation; The pre-filtering energy threshold is constructed based on the noise power estimate value of the current sampling time window, and the noise power estimate value is used as the parameter of the pre-filtering energy threshold; For each sampling time point in the sampling time point set, the local deviation is compared with the pre-filtering energy threshold; when the local deviation is greater than or equal to the pre-filtering energy threshold, the corresponding sampling time point is retained, and when the local deviation is less than the pre-filtering energy threshold, the corresponding sampling time point is discarded; The sampling time points retained under the pre-filtering rule are combined to form a reserved sampling time point set, and the voltage sampling values corresponding to the reserved sampling time point set are combined to form a reserved voltage sequence.

7. The method of claim 6, wherein the method further comprises: For each downstream power branch, the reserved voltage sequence in the current sampling time window is used to construct a fixed-length signal vector, an orthogonal normalization basis is selected, a linear orthogonal transformation is performed, and a compressed coding coefficient vector is generated, and specifically comprising: The coding vector length and the number of coding coefficients are set, and the number of coding coefficients is not greater than the coding vector length. Construct a basis vector set with a length equal to the length of the encoding vector, in which each basis vector only takes a value of one at one component position and zero at the rest of the component positions, each basis vector has a zero inner product with another basis vector and a one inner product with itself, forming an orthogonal normalized basis; Count the number of elements of the reserved sampling time point set in the current sampling time window as the number of reserved sampling time points, and rearrange the reserved voltage sequence according to the time sequence of the reserved sampling time points to obtain a time-ordered reserved voltage sequence; According to the relationship between the number of reserved sampling time points and the length of the encoding vector, a signal vector participating in the transformation is constructed: when the number of reserved sampling time points is greater than or equal to the length of the encoding vector, the first several time-ordered reserved voltage values are sequentially filled into each component of the signal vector; when the number of reserved sampling time points is greater than zero and less than the length of the encoding vector, all reserved voltage values are sequentially filled into each component of the front part of the signal vector and the remaining components are filled with zero values; when the number of reserved sampling time points is equal to zero, all components of the signal vector are filled with zero values; For each basis vector in the orthogonal normalized basis, a linear transformation operation is performed to multiply each component of the signal vector with the corresponding component of the current basis vector and sum them up to obtain the corresponding encoding coefficient, and all encoding coefficients are combined into a compressed encoding coefficient vector as the compressed encoding coefficient vector of the corresponding downstream power branch in the current sampling time window. 8.The method of claim 7, wherein, The bandwidth priority index is constructed according to the compressed data amount, the instantaneous signal-to-noise ratio and the uplink communication bandwidth of each downstream power branch, and a total order relationship is established in the same edge collection device according to the bandwidth priority index and the compressed data amount, and the buffer scheduling and bandwidth scheduling are performed to select the actual upload branch set in the current upload scheduling period, which specifically includes: At the beginning of each upload scheduling period, the available uplink communication bandwidth of each edge collection device in the current upload scheduling period is obtained, the number of bits occupied by a single encoding coefficient after quantization of the compressed encoding coefficient is obtained, and the time length of the current upload scheduling period is set; For each downstream power branch, the compressed data amount of each downstream power branch that needs to be uploaded in the current upload scheduling period is calculated according to the product of the number of encoding coefficients that each downstream power branch needs to upload in the current upload scheduling period and the number of bits occupied by a single encoding coefficient; At the time of statistics and sampling decision, the bandwidth priority index is constructed according to the compressed data amount and the instantaneous signal-to-noise ratio of each downstream power branch, so that the downstream power branch with larger compressed data amount and lower instantaneous signal-to-noise ratio obtains a larger bandwidth priority index, and the downstream power branch with smaller compressed data amount and higher instantaneous signal-to-noise ratio obtains a smaller bandwidth priority index, and a total order relationship is established in the same edge collection device according to the bandwidth priority index and the compressed data amount to determine the order; For all downstream power branches of the same edge collection device, the total order relationship determined by the bandwidth priority index and the compressed data amount is used to arrange them in order from the downstream power branch with the smallest bandwidth priority index to the downstream power branch with the largest bandwidth priority index; in the case of the same bandwidth priority index, the downstream power branch with smaller compressed data amount is arranged in front, and an ordered downstream power branch sequence is obtained. For each edge collection device, the corresponding compressed data amount is sequentially accumulated from the head of the ordered downstream power branch sequence; when the accumulation result does not exceed the product of the available uplink bandwidth and the time length of the current upload scheduling period, the accumulation continues; when the next downstream power branch is added, the product of the available uplink bandwidth and the time length of the current upload scheduling period will be exceeded, and the accumulation is stopped; the number of downstream power branches participating in the accumulation is recorded as the number of downstream power branches that can be arranged for upload in the current upload scheduling period; when the compressed data amount of any single downstream power branch is greater than the product of the available uplink bandwidth and the time length of the current upload scheduling period, the number of downstream power branches that can be arranged for upload is set to zero; When the number of downstream power branches that can be arranged for upload is greater than zero, the first several downstream power branches in the ordered downstream power branch sequence are grouped into the actual upload branch set of the current upload scheduling period; when the number of downstream power branches that can be arranged for upload is equal to zero, no data upload is performed in the current upload scheduling period. 9.The method of claim 8, wherein, For the actual upload branch set, a structured upload data record containing edge collection device identification, downstream power branch identification, downstream power branch number, this sampling time window start time and end time, the number of reserved sampling time points in this sampling time window, and compressed encoding coefficient vector obtained by orthogonal transformation is generated and uploaded, specifically including: For each edge collection device, the downstream power branches in the actual upload branch set of the current upload scheduling period are traversed, and a structured upload data record is constructed for each downstream power branch. The structured upload data record includes at least the following fields: edge collection device identification, downstream power branch identification, downstream power branch number, this sampling time window start time, this sampling time window end time, the number of reserved sampling time points in this sampling time window, and compressed encoding coefficient vector obtained by orthogonal transformation; The structured upload data corresponding to all actual upload branches of the same edge collection device in the current upload scheduling period is sent to the data center on the uplink.

Images