A method and system for constructing a power line communication noise library

Abstract

The present invention discloses a method and system for constructing a power line communication noise library, comprising: using each edge server, respectively utilizing all first neighbor entity sets and all second neighbor entity sets in a cloud-side noise time series knowledge graph, combining a noise data feature set and a multi-mode mapping relationship model, to complete the first main entity set and the first guest entity set in the edge quadruple of the edge noise time series knowledge graph, thereby obtaining a second main entity set and a second guest entity set; receiving the completed edge noise time series knowledge graph uploaded by each edge server through a cloud server, aggregating all completed edge quadruple groups, and then completing the cloud-side noise time series knowledge graph based on the aggregation result, so as to update the cloud-side noise library and use the updated cloud-side noise library as the power line communication noise library. The present invention utilizes cloud-edge collaboration between a cloud server and multiple edge servers to improve the construction accuracy of the power line communication noise library.

Description

Technical Field

[0001] The present invention relates to the field of power line communication noise library modeling, and in particular to a power line communication noise library construction method and system. Background Art

[0002] Power line communication (PLC) is subject to interference from various loads on branch lines. This, coupled with the relatively harsh communication environment, results in noise on power lines characterized by a wide spectrum, high burstiness, and high intensity. This PLC noise interference causes bit errors during signal transmission over the power line channel, significantly reducing data transmission accuracy and severely impacting communication quality. To minimize power line noise interference and improve communication reliability, accurate noise library modeling and noise feature analysis are required. A time-series knowledge graph abstracts real-world knowledge into a complex graph network consisting of billions of quadruples. Leveraging this graph can effectively improve the accuracy of noise library construction and accurately describe the characteristics of PLC noise.

[0003] However, the traditional method of constructing a power line communication noise library does not consider the impact of the missing time series knowledge graph on the construction of the noise library, and cannot fully utilize the resources of the cloud server. At the same time, it does not consider constructing the power line communication noise library from multiple dimensions, resulting in low modeling accuracy of the power line communication noise library. Summary of the Invention

[0004] The present invention provides a method and system for constructing a power line communication noise library, which utilizes the cloud-edge collaboration between a cloud server and multiple edge servers to optimize the scalability of the noise time series knowledge graph, thereby improving the construction accuracy of the power line communication noise library.

[0005] In order to solve the above technical problems, an embodiment of the present invention provides a method for constructing a power line communication noise library, comprising:

[0006] Obtain a noise data feature set, and through multiple edge servers, obtain a multimodal mapping relationship model of the cloud-side noise database, including a set of first neighbor entities corresponding to several similar main entities, a set of second neighbor entities corresponding to several similar guest entities, and the cloud-side noise database from the cloud-side noise time series knowledge graph in the cloud-side noise database.

[0007] Through each of the edge servers, respectively using all of the first neighbor entity sets and all of the second neighbor entity sets, combined with the noise data feature set and the multimodal mapping relationship model of the cloud-side noise library, the first main entity set and the first guest entity set in the side quadruple of the side noise time series knowledge graph are completed to obtain the corresponding second main entity set and second guest entity set, so as to complete the completion of the side quadruple and the side noise time series knowledge graph;

[0008] Receiving, through the cloud server, the completed edge-side noise time series knowledge graph uploaded by each edge server, aggregating all the completed edge-side quadruples, and then completing the cloud-side noise time series knowledge graph according to the aggregation result to update the cloud-side noise library, and then using the updated cloud-side noise library as the power line communication noise library;

[0009] Among them, the side quadruple includes the first main entity set, the first guest entity set, the first entity association set and the first timestamp, the first main entity set includes several side noise main entities, the first guest entity set includes several side noise guest entities, the first entity association set includes the association between each of the side noise main entities and each of the side noise guest entities, and the first timestamp includes the establishment time of the association between each of the side noise main entities and each of the side noise guest entities.

[0010] In the implementation of the embodiment of the present invention, the cloud-side noise time series knowledge graph of the cloud server is used to complete the edge quadruple in the edge noise time series knowledge graph of multiple edge servers. Based on the aggregation results of the completed edge noise time series knowledge graphs of multiple edge servers, the cloud-side noise time series knowledge graph of the cloud server is completed. This can collect effective information from the edge noise time series knowledge graph of each edge server, integrate the knowledge graphs, and achieve cloud-edge collaboration between the cloud server and multiple edge servers, thereby solving the problems of poor scalability and small scale of the cloud-side noise time series knowledge graph and the edge noise time series knowledge graph. It also fully utilizes the computing resources of the cloud server and the edge server to achieve high-precision construction of the noise time series knowledge graph and the power line communication noise library. At the same time, since the noise data received by each edge server is different, the edge noise time series knowledge graph of each edge server is also different. By completing the cloud-side time series knowledge graph by multiple edge servers, information complementarity between edge servers can be achieved, edge server information can be fully utilized, and the accuracy of cloud-side noise time series knowledge graph completion can be improved. In addition, most of the calculations are completed through local edge servers, reducing the load on cloud servers and improving processing efficiency.

[0011] As a preferred solution, each of the edge servers utilizes all the first neighbor entity sets and all the second neighbor entity sets, respectively, in combination with the noise data feature set and the multimodal mapping relationship model of the cloud-side noise library, to perform entity completion on the first main entity set and the first guest entity set in the edge quadruple of the edge noise time series knowledge graph, and obtain the corresponding second main entity set and second guest entity set to complete the completion of the edge quadruple and the edge noise time series knowledge graph, specifically including:

[0012] Through each of the edge servers, according to a preset main entity completion function, combined with the weight coefficients of each of the similar main entities in the cloud-side noise time series knowledge graph, the first neighbor entity set corresponding to each of the similar main entities, the noise data feature set, the weight parameters of the noise data feature set, and the multimodal mapping relationship model of the cloud-side noise library, entity completion is performed on the first main entity set to obtain the second main entity set corresponding to the first main entity set, so as to form a main entity completion result corresponding to each of the edge servers, and complete the main entity completion of the edge quadruple in the edge noise time series knowledge graph;

[0013] Through each of the edge servers, in accordance with a preset guest entity completion function, combined with the weight coefficients of each of the similar guest entities in the cloud-side noise time series knowledge graph, the second neighbor entity set corresponding to each of the similar guest entities, the noise data feature set, the weight parameters of the noise data feature set, and the multimodal mapping relationship model of the cloud-side noise library, the first guest entity set is completed, and the second guest entity set corresponding to the first guest entity set is obtained to form the guest entity completion results corresponding to each of the edge servers, thereby completing the guest entity completion of the edge quadruple in the edge noise time series knowledge graph.

[0014] According to a preferred embodiment of the present invention, based on the cloud-side noise time series knowledge graph of the cloud server, the multi-modal mapping relationship model of the cloud-side noise library, and the completion results of multiple edge servers, the integrated utilization of noise resources is realized through weighted aggregation. The first main entity set and the first guest entity set in the constructed edge noise time series knowledge graph are aggregated and completed respectively to solve the problem of missing entities in the edge noise time series knowledge graph, thereby improving the construction accuracy of the power line communication noise library.

[0015] As a preferred solution, the main entity completion function and the guest entity completion function are specifically:

[0016]

[0017] Where, Represents the second main entity set corresponding to the first main entity set of the side quadruple in the side noise time series knowledge graph, represents the second guest entity set corresponding to the first guest entity set of the edge quadruple in the edge noise time series knowledge graph, f(g) represents the nonlinear entity completion function, N represents the number of similar main entities in the cloud side noise time series knowledge graph, L n ci represents the first neighbor entity set corresponding to the nth similar main entity in the cloud-side noise time series knowledge graph, L n cjrepresents the set of second neighbor entities corresponding to the nth similar guest entity in the cloud-side noise time series knowledge graph, α ci n represents the weight coefficient of the nth similar main entity in the cloud-side noise time series knowledge graph, α cj n represents the weight coefficient of the nth similar customer entity in the cloud-side noise time series knowledge graph, Z represents the noise data feature set, χ represents the weight parameter of the noise data feature set, V represents the multimodal mapping relationship model of the cloud-side noise library, V = {k1, k2}, k1 represents the mapping relationship between the concept model and the internal model, k2 represents the mapping relationship between the concept model and the external model, and δ represents the weight parameter of the multimodal mapping relationship model of the cloud-side noise library.

[0018] A preferred solution for implementing the embodiment of the present invention is to use a nonlinear entity completion function to weightedly aggregate the first neighbor entity set corresponding to each similar main entity, the noise data feature set, and the mapping relationship between the concept model and the internal model and the mapping relationship between the concept model and the external model in the multimodal mapping relationship model of the cloud-side noise library to complete the first main entity set of the side quadruple in the side noise time series knowledge graph. Similarly, a nonlinear entity completion function is used to weightedly aggregate the second neighbor entity set corresponding to each similar guest entity, the noise data feature set, and the mapping relationship between the concept model and the internal model and the mapping relationship between the concept model and the external model in the multimodal mapping relationship model of the cloud-side noise library to complete the first guest entity set of the side quadruple in the side noise time series knowledge graph. This can describe the cloud-side noise library from multiple dimensions and further improve the construction accuracy of the power line communication noise library.

[0019] As a preferred solution, all the completed edge quadruple groups are aggregated, and then the cloud-side noise time series knowledge graph is completed according to the aggregation results, specifically:

[0020] Aggregating all completed edge quadruple groups through the cloud server, and completing the cloud-side noise time series knowledge graph in accordance with a preset cloud-side quadruple completion function and the aggregation results; wherein the cloud-side quadruple completion function is specifically:

[0021]

[0022] Where G c represents the cloud-side quadruple in the cloud-side noise time series knowledge graph, represents the edge quadruple of the completed edge noise temporal knowledge graph uploaded by the i-th edge server to the cloud server, T i erepresents the first timestamp of the edge quadruple of the completed edge noise time series knowledge graph uploaded by the i-th edge server to the cloud server, and g(g) represents the cloud-side noise time series knowledge graph completion function.

[0023] A preferred solution for implementing the embodiment of the present invention is to add the aggregation results of all completed edge-side quadruple groups to the cloud-side quadruple group of the original cloud-side noise time series knowledge graph to integrate the effective information of the original cloud-side quadruple group and all completed edge-side quadruple groups, thereby improving the accuracy of the cloud-side noise time series knowledge graph.

[0024] As a preferred solution, the method for constructing a power line communication noise library further includes:

[0025] Through each of the edge servers, according to all the completed edge noise time series knowledge graphs, searching for corresponding entity names in the cloud-side noise library of the cloud server to update the edge-cloud mapping relationship of the cloud-side noise library;

[0026] Updating, by the cloud server, a mapping relationship in a multimodal mapping relationship model of the cloud-side noise library according to the edge-cloud mapping relationship;

[0027] The edge-cloud mapping relationship refers to the mapping relationship between the edge-side noise of the edge server and the cloud-side noise of the cloud server.

[0028] A preferred solution of an embodiment of the present invention is implemented, and the mapping relationship between the edge-side noise of the edge server and the cloud-side noise of the cloud server contained in the cloud-side noise library is updated according to all the completed edge-side noise time series knowledge graphs, so as to enhance the connection between the various modes in the cloud-side noise library, thereby improving the accuracy of the power line communication noise library.

[0029] As a preferred solution, the noise data feature set is obtained as follows:

[0030] The power terminal collects carrier data of power line communication in real time, normalizes the carrier data, and then filters the normalized carrier data using a bandpass filter to obtain corresponding noise data, and uploads the noise data to each edge server;

[0031] The noise data is subjected to feature extraction by each of the edge servers to obtain the noise data feature set.

[0032] A preferred solution for implementing an embodiment of the present invention is to normalize the carrier data collected in real time by the power terminal to remove the pollution of the power line carrier communication noise caused by non-stationary noise sources such as the distribution network environmental noise. Then, based on the frequency band characteristics of the power line carrier communication noise, a bandpass filter is used to filter the normalized carrier data, which can filter out the remaining noise in the carrier data, thereby improving the accuracy of the noise data feature set.

[0033] In order to solve the same technical problem, an embodiment of the present invention further provides a power line communication noise library construction system, comprising:

[0034] The data acquisition module is used to obtain a noise data feature set and, through multiple edge servers, obtain a set of first neighbor entities corresponding to several similar main entities, a set of second neighbor entities corresponding to several similar guest entities, and a multimodal mapping relationship model of the cloud-side noise library from the cloud-side noise time series knowledge graph in the cloud-side noise library.

[0035] A graph completion module is configured to, through each of the edge servers, respectively utilize all the first neighbor entity sets and all the second neighbor entity sets, combine the noise data feature set and the multimodal mapping relationship model of the cloud-side noise library, and perform entity completion on the first main entity set and the first guest entity set in the side quadruple of the side noise time series knowledge graph to obtain the corresponding second main entity set and second guest entity set, so as to complete the completion of the side quadruple and the side noise time series knowledge graph; wherein the side quadruple includes the first main entity set, the first guest entity set, a first entity association set and a first timestamp, the first main entity set includes a plurality of side noise main entities, the first guest entity set includes a plurality of side noise guest entities, the first entity association set includes the association between each of the side noise main entities and each of the side noise guest entities, and the first timestamp includes the establishment time of the association between each of the side noise main entities and each of the side noise guest entities;

[0036] The noise library construction module is used to receive the completed edge-side noise time series knowledge graph uploaded by each edge server through the cloud server, aggregate all the completed edge-side quadruples, and then complete the cloud-side noise time series knowledge graph according to the aggregation result to update the cloud-side noise library, and then use the updated cloud-side noise library as the power line communication noise library.

[0037] As a preferred solution, the map completion module specifically includes:

[0038] A first graph completion unit is configured to perform entity completion on the first main entity set through each edge server according to a preset main entity completion function, in combination with the weight coefficient of each similar main entity in the cloud-side noise time series knowledge graph, the first neighbor entity set corresponding to each similar main entity, the noise data feature set, the weight parameter of the noise data feature set, and the multimodal mapping relationship model of the cloud-side noise library, to obtain the second main entity set corresponding to the first main entity set, so as to form a main entity completion result corresponding to each edge server, and complete the main entity completion of the edge quadruple in the edge noise time series knowledge graph;

[0039] The second graph completion unit is used to perform entity completion on the first guest entity set through each of the edge servers in accordance with a preset guest entity completion function, combined with the weight coefficients of each of the similar guest entities in the cloud-side noise time series knowledge graph, the second neighbor entity set corresponding to each of the similar guest entities, the noise data feature set, the weight parameters of the noise data feature set, and the multimodal mapping relationship model of the cloud-side noise library, to obtain the second guest entity set corresponding to the first guest entity set, so as to form the guest entity completion results corresponding to each of the edge servers, and complete the guest entity completion of the edge quadruple in the edge noise time series knowledge graph.

[0040] As a preferred solution, the main entity completion function and the guest entity completion function are specifically:

[0041]

[0042] Where, Represents the second main entity set corresponding to the first main entity set of the side quadruple in the side noise time series knowledge graph, represents the second guest entity set corresponding to the first guest entity set of the edge quadruple in the edge noise time series knowledge graph, f(g) represents the nonlinear entity completion function, N represents the number of similar main entities in the cloud side noise time series knowledge graph, L n ci represents the first neighbor entity set corresponding to the nth similar main entity in the cloud-side noise time series knowledge graph, L n cj represents the set of second neighbor entities corresponding to the nth similar guest entity in the cloud-side noise time series knowledge graph, α ci n represents the weight coefficient of the nth similar main entity in the cloud-side noise time series knowledge graph, α cj nrepresents the weight coefficient of the nth similar customer entity in the cloud-side noise time series knowledge graph, Z represents the noise data feature set, χ represents the weight parameter of the noise data feature set, V represents the multimodal mapping relationship model of the cloud-side noise library, V = {k1, k2}, k1 represents the mapping relationship between the concept model and the internal model, k2 represents the mapping relationship between the concept model and the external model, and δ represents the weight parameter of the multimodal mapping relationship model of the cloud-side noise library.

[0043] As a preferred solution, the power line communication noise library construction system further includes:

[0044] A mapping relationship update module is used to search for corresponding entity names in the cloud-side noise library of the cloud server through each of the edge servers according to all the completed edge noise time series knowledge graphs to update the edge-cloud mapping relationship of the cloud-side noise library; through the cloud server, according to the edge-cloud mapping relationship, update the mapping relationship in the multimodal mapping relationship model of the cloud-side noise library; wherein the edge-cloud mapping relationship refers to the mapping relationship between the edge noise of the edge server and the cloud noise of the cloud server. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 : A schematic flow chart of a method for constructing a power line communication noise library provided in the first embodiment of the present invention;

[0046] Figure 2 : A structural diagram of a power line communication noise library construction system provided in Example 1 of the present invention. DETAILED DESCRIPTION

[0047] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0048] Embodiment one:

[0049] Please refer to Figure 1 , a method for constructing a power line communication noise library provided by an embodiment of the present invention, the method includes steps S1 to S3, each step is specifically as follows:

[0050] Step S1: Obtain a noise data feature set, and through multiple edge servers, obtain a first neighbor entity set corresponding to several similar main entities, a second neighbor entity set corresponding to several similar guest entities, and a multimodal mapping relationship model of the cloud-side noise library from the cloud-side noise time series knowledge graph in the cloud-side noise library.

[0051] As a preferred solution, step S1 includes steps S11 to S14, and each step is specifically as follows:

[0052] In step S11, the carrier data of the power line communication is collected in real time through the power terminal, and the carrier data is normalized. The pollution of the power line carrier communication noise caused by non-stationary noise sources such as distribution network environmental noise is removed through time domain normalization.

[0053] Step S12: Based on the frequency band characteristics of the power line carrier communication noise, a bandpass filter is used to filter the normalized carrier data to obtain corresponding noise data, and the noise data is uploaded to each edge server.

[0054] Step S13: extract features from the noise data through each edge server to obtain a noise data feature set Z.

[0055] Step S14: Obtain the first neighbor entity set L corresponding to several similar main entities from the cloud-side noise time series knowledge graph in the cloud-side noise library through multiple edge servers. n ci , the second neighbor entity set L corresponding to several similar guest entities n cj and the multi-modal mapping relationship model of the cloud-side noise library.

[0056] Step S2: Each edge server uses all first neighbor entity sets L n ci and the set of all second neighbor entities L n cj , combined with the noise data feature set Z and the multi-modal mapping relationship model of the cloud-side noise library, the edge quadruple G of the edge noise time series knowledge graph e The first main entity set H in ei and the first customer entity set H ej Perform entity completion to obtain the corresponding second main entity set and second guest entity set to complete the side quadruple G e And the completion of the side noise time series knowledge graph.

[0057] Among them, the edge quadruple G e Including the first main entity set H ei , the first customer entity set H ej , the first entity association set R e and the first timestamp T e , the edge quadruple is represented by G e ={H ei ,R e ,H ej ,Te}; First main entity set H ei Including several side noise main entities; the first guest entity set H ej Including several side noise guest entities; the first entity association set R e Including the association between each side noise main entity and each side noise guest entity; the first timestamp T e Including the establishment time of the association between each side noise main entity and each side noise guest entity.

[0058] As a preferred solution, step S2 includes steps S21 to S22, and the details of each step are as follows:

[0059] Step S21: Through each edge server, according to the preset main entity completion function, combined with the weight coefficient α of each similar main entity in the cloud side noise time series knowledge graph ci n , the first neighbor entity set L corresponding to each similar main entity n ci , noise data feature set Z, the weight parameter χ of the noise data feature set Z, and the multi-mode mapping relationship model V of the cloud-side noise library, for the first main entity set H ei Perform entity completion to obtain the first main entity set H ei The corresponding second main entity set And the second main entity set The side quadruple G filled into the side noise time series knowledge graph e In order to form the main entity completion results corresponding to each edge server, the edge quadruple G in the edge noise time series knowledge graph is completed. e The main entity of the completion.

[0060] Among them, the cloud side noise time series knowledge graph is composed of the cloud side quadruple G c The cloud side quadruple G c Including the third main entity set H ci , the third guest entity set H cj , the second entity association set R c and the second timestamp T c , cloud-side quadruple G c Represented as G c ={H ci ,R c ,H cj ,T c}; The third main entity set H ci Including several cloud side noise main entities; the third guest entity set H cj Including several cloud-side noise guest entities; the second entity association set R cIncluding the association between each cloud side noise main entity and each cloud side noise guest entity; the second timestamp T c Including the establishment time of the association between each cloud-side noise master entity and each cloud-side noise guest entity.

[0061] It should be noted that the first main entity set H in each edge server ei The main entity of the edge noise to be completed in is recorded as the main entity to be completed. The similar main entity refers to the cloud side quadruple G c The third main entity set H ci A cloud-side noise main entity that satisfies the main entity similarity condition; wherein the main entity similarity condition means that the similarity between the name of the cloud-side noise main entity and the name of the main entity to be completed reaches a preset value and the corresponding association thereof is the same as the association corresponding to the main entity to be completed.

[0062] Step S22: Through each edge server, according to the preset guest entity completion function, combined with the weight coefficient α of each similar guest entity in the cloud side noise time series knowledge graph cj n , the second neighbor entity set L corresponding to each similar guest entity n cj , noise data feature set Z, the weight parameter χ of the noise data feature set Z, and the multimodal mapping relationship model V of the cloud-side noise library, for the first guest entity set H ej Perform entity completion to obtain the first customer entity set H ej Corresponding second customer entity set And the second customer entity set The side quadruple G filled into the side noise time series knowledge graph e In order to form the customer entity completion results corresponding to each edge server, the edge quadruple G in the edge noise time series knowledge graph is completed. e The object entity completion of .

[0063] It should be noted that the side noise main entity H to be completed in the first guest entity set in each edge server ej , recorded as the guest entity to be completed. The similar guest entity refers to the cloud-side quadruple G c The third guest entity set H cj The cloud-side noise object entity that meets the object entity similarity condition; wherein the object entity similarity condition means that the similarity between the name of the cloud-side noise object entity and the name of the object entity to be completed reaches a preset value and the corresponding association is the same as the association corresponding to the object entity to be completed.

[0064] As a preferred solution, the main entity completion function in step S21 and the guest entity completion function in step S22 are specifically described in formula (1).

[0065]

[0066] Where, Represents the second main entity set corresponding to the first main entity set of the side quadruple in the side noise time series knowledge graph; represents the second guest entity set corresponding to the first guest entity set of the edge quadruple in the edge noise time series knowledge graph; f(g) represents the nonlinear entity completion function; N represents the number of similar main entities in the cloud side noise time series knowledge graph; L n ci represents the first neighbor entity set corresponding to the nth similar main entity in the cloud-side noise time series knowledge graph; L n cj represents the set of second neighbor entities corresponding to the nth similar guest entity in the cloud-side noise time series knowledge graph; α ci n represents the weight coefficient of the nth similar main entity in the cloud-side noise time series knowledge graph; α cj n represents the weight coefficient of the nth similar guest entity in the cloud-side noise time series knowledge graph; Z represents the noise data feature set; χ represents the weight parameter of the noise data feature set; V represents the multimodal mapping relationship model of the cloud-side noise library, V = {k1, k2}, k1 represents the mapping relationship between the concept model and the internal model, k2 represents the mapping relationship between the concept model and the external model; δ represents the weight parameter of the multimodal mapping relationship model of the cloud-side noise library.

[0067] In this embodiment, after executing steps SS21 to S22, each edge server will obtain a completed edge noise time series knowledge graph, and then each edge server uploads its completed edge noise time series knowledge graph to the cloud server.

[0068] Step S3: receive the completed edge noise time series knowledge graph uploaded by each edge server through the cloud server, and convert all completed edge quadruple Aggregation is performed, and then the cloud-side noise time series knowledge graph is completed based on the aggregation results to update the cloud-side noise library, and then the updated cloud-side noise library is used as the power line communication noise library.

[0069] As a preferred solution, step S3 includes step S31 to step S32, and each step is specifically as follows:

[0070] Step S31: Receive the completed edge noise time series knowledge graph uploaded by each edge server through the cloud server.

[0071] Step S32: All completed edge quads are sent to the cloud server. Aggregation is performed, and according to the preset cloud-side four-tuple completion function, the cloud-side noise time series knowledge graph is completed in combination with the aggregation results to update the cloud-side noise library, and then the updated cloud-side noise library is used as the power line communication noise library; wherein, the cloud-side four-tuple completion function, please refer to formula (2) for details.

[0072]

[0073] Where G c represents the cloud-side quadruple in the cloud-side noise time series knowledge graph, represents the edge quadruple of the completed edge noise temporal knowledge graph uploaded by the i-th edge server to the cloud server, T i e represents the first timestamp of the edge quadruple of the completed edge noise time series knowledge graph uploaded by the i-th edge server to the cloud server, and g(g) represents the cloud-side noise time series knowledge graph completion function.

[0074] As a preferred solution, an embodiment of the present invention provides a method for constructing a power line communication noise library, further comprising a mapping relationship update process, which includes steps S4 to S5, each of which is specifically as follows:

[0075] In step S4, each edge server searches for the corresponding entity name in the cloud-side noise library of the cloud server according to all completed edge-side noise time series knowledge graphs to update the edge-cloud mapping relationship of the cloud-side noise library.

[0076] Among them, the edge-cloud mapping relationship refers to the mapping relationship between the edge-side noise of the edge server and the cloud-side noise of the cloud server.

[0077] In step S5, the cloud server updates the mapping relationship in the multimodal mapping relationship model of the cloud-side noise library based on the edge-cloud mapping relationship. For details on the mapping relationship in the multimodal mapping relationship model of the cloud-side noise library, please refer to equations (3) and (4).

[0078] k1:A={a1,a2,...,a m ,...}→B={b1,b2,...,b p ,...}; (3)

[0079] k2:A={a1,a2,...,a m ,...}→C={c1,c2,...,c q ,...}; (4)

[0080] In the formula, A represents the conceptual model set; B represents the internal model set; C represents the external model set; a mrepresents the mth noise element concept in the concept pattern set A; b p represents the p-th noise physical structure in the internal mode set B; c q Represents the qth noise data in the external model set C; k1 represents the mapping relationship from the conceptual model to the internal model, that is, mapping the noise element concept to the noise physical structure; k2 represents the mapping relationship from the conceptual model to the internal model, that is, mapping the noise element concept to the noise physical data.

[0081] Please refer to Figure 2 , is a schematic diagram of the structure of a power line communication noise library construction system provided by an embodiment of the present invention. The system includes a data acquisition module M1, a spectrum completion module M2, and a noise library construction module M3. The details of each module are as follows:

[0082] The data acquisition module M1 is used to obtain a noise data feature set and, through multiple edge servers, obtain a set of first neighbor entities corresponding to several similar main entities, a set of second neighbor entities corresponding to several similar guest entities, and a multimodal mapping relationship model of the cloud-side noise library from the cloud-side noise time series knowledge graph in the cloud-side noise library.

[0083] The graph completion module M2 is used to complete the first main entity set and the first guest entity set in the side quadruple of the side noise time series knowledge graph by using all first neighbor entity sets and all second neighbor entity sets respectively, combined with the noise data feature set and the multimodal mapping relationship model of the cloud-side noise library through each edge server, and obtain the corresponding second main entity set and second guest entity set to complete the completion of the side quadruple and the side noise time series knowledge graph; wherein the side quadruple includes the first main entity set, the first guest entity set, the first entity association set and the first timestamp, the first main entity set includes several side noise main entities, the first guest entity set includes several side noise guest entities, the first entity association set includes the association between each side noise main entity and each side noise guest entity, and the first timestamp includes the establishment time of the association between each side noise main entity and each side noise guest entity;

[0084] The noise library construction module M3 is used to receive the completed edge noise time series knowledge graph uploaded by each edge server through the cloud server, aggregate all the completed edge quadruple groups, and then complete the cloud-side noise time series knowledge graph based on the aggregation results to update the cloud-side noise library, and then use the updated cloud-side noise library as the power line communication noise library.

[0085] As a preferred solution, the atlas completion module M2 specifically includes a first atlas completion unit 21 and a second atlas completion unit 22, and the details of each unit are as follows:

[0086] The first graph completion unit 21 is used to perform entity completion on the first main entity set through each edge server according to a preset main entity completion function, combined with the weight coefficient of each similar main entity in the cloud-side noise time series knowledge graph, the first neighbor entity set corresponding to each similar main entity, the noise data feature set, the weight parameter of the noise data feature set, and the multimodal mapping relationship model of the cloud-side noise library, to obtain the second main entity set corresponding to the first main entity set, so as to form the main entity completion result corresponding to each edge server, and complete the main entity completion of the side quadruple in the side noise time series knowledge graph;

[0087] The second graph completion unit 22 is used to complete the first guest entity set through each edge server in accordance with a preset guest entity completion function, combined with the weight coefficients of each similar guest entity in the cloud-side noise time series knowledge graph, the second neighbor entity set corresponding to each similar guest entity, the noise data feature set, the weight parameters of the noise data feature set, and the multimodal mapping relationship model of the cloud-side noise library, to obtain the second guest entity set corresponding to the first guest entity set, so as to form the guest entity completion results corresponding to each edge server, and complete the guest entity completion of the edge quadruple in the edge noise time series knowledge graph.

[0088] As a preferred solution, the main entity completion function mentioned in the first graph completion unit 21 and the guest entity completion function mentioned in the second graph completion unit 22 are specifically:

[0089]

[0090] Where, Represents the second main entity set corresponding to the first main entity set of the side quadruple in the side noise time series knowledge graph, represents the second guest entity set corresponding to the first guest entity set of the edge quadruple in the edge noise time series knowledge graph, f(g) represents the nonlinear entity completion function, N represents the number of similar main entities in the cloud side noise time series knowledge graph, L n ci represents the first neighbor entity set corresponding to the nth similar main entity in the cloud-side noise time series knowledge graph, L n cj represents the set of second neighbor entities corresponding to the nth similar guest entity in the cloud-side noise time series knowledge graph, α ci n represents the weight coefficient of the nth similar main entity in the cloud-side noise time series knowledge graph, α cj nrepresents the weight coefficient of the nth similar customer entity in the cloud-side noise time series knowledge graph, Z represents the noise data feature set, χ represents the weight parameter of the noise data feature set, V represents the multimodal mapping relationship model of the cloud-side noise library, V = {k1, k2}, k1 represents the mapping relationship between the concept model and the internal model, k2 represents the mapping relationship between the concept model and the external model, and δ represents the weight parameter of the multimodal mapping relationship model of the cloud-side noise library.

[0091] As a preferred solution, the power line communication noise library construction system provided by the embodiment of the present invention further includes a mapping relationship updating module M4, which is specifically as follows:

[0092] The mapping relationship update module M4 is used to search for the corresponding entity name in the cloud-side noise library of the cloud server through each edge server according to all the completed edge-side noise time series knowledge graphs to update the edge-cloud mapping relationship of the cloud-side noise library; through the cloud server, according to the edge-cloud mapping relationship, update the mapping relationship in the multimodal mapping relationship model of the cloud-side noise library; wherein, the edge-cloud mapping relationship refers to the mapping relationship between the edge-side noise of the edge server and the cloud-side noise of the cloud server.

[0093] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working process of the system described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0094] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0095] The present invention provides a method and system for constructing a power line communication noise library. The method utilizes a cloud server's cloud-side noise time series knowledge graph to complete edge quadruples in the edge noise time series knowledge graphs of multiple edge servers. The method then completes the cloud server's cloud-side noise time series knowledge graph based on the aggregated results of the completed edge noise time series knowledge graphs of the multiple edge servers. This method collects effective information from the edge noise time series knowledge graphs of each edge server, integrates the knowledge graphs, and achieves cloud-edge collaboration between the cloud server and multiple edge servers. This addresses the issues of poor scalability and small scale of the cloud-side noise time series knowledge graph and the edge noise time series knowledge graph, while fully utilizing the computing resources of the cloud server and edge servers to achieve high-precision construction of the noise time series knowledge graph and the power line communication noise library. Furthermore, since the noise data received by each edge server differs, the edge noise time series knowledge graphs of each edge server also differ. Completing the cloud-side time series knowledge graph by multiple edge servers enables information complementarity between edge servers, fully utilizing edge server information and improving the accuracy of the cloud-side noise time series knowledge graph completion.

[0096] Furthermore, by normalizing the carrier data collected in real time by the power terminal, the pollution of the power line carrier communication noise caused by non-stationary noise sources such as the distribution network environmental noise is removed. Then, based on the frequency band characteristics of the power line carrier communication noise, the normalized carrier data is filtered using a bandpass filter, which can filter out the remaining noise in the carrier data and thus improve the accuracy of the noise data feature set.

[0097] The specific embodiments described above further illustrate the objectives, technical solutions, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for constructing a power line communication noise library, characterized in that: include: Obtain a noise data feature set, and through multiple edge servers, obtain a multimodal mapping relationship model of the cloud-side noise database, including a set of first neighbor entities corresponding to several similar main entities, a set of second neighbor entities corresponding to several similar guest entities, and the cloud-side noise database from the cloud-side noise time series knowledge graph in the cloud-side noise database. Through each of the edge servers, respectively using all of the first neighbor entity sets and all of the second neighbor entity sets, combined with the noise data feature set and the multimodal mapping relationship model of the cloud-side noise library, the first main entity set and the first guest entity set in the side quadruple of the side noise time series knowledge graph are completed to obtain the corresponding second main entity set and second guest entity set, so as to complete the completion of the side quadruple and the side noise time series knowledge graph; Receiving, through the cloud server, the completed edge-side noise time series knowledge graph uploaded by each edge server, aggregating all the completed edge-side quadruples, and then completing the cloud-side noise time series knowledge graph according to the aggregation result to update the cloud-side noise library, and then using the updated cloud-side noise library as the power line communication noise library; The side quadruple includes the first main entity set, the first guest entity set, a first entity association set, and a first timestamp, the first main entity set includes a plurality of side noise main entities, the first guest entity set includes a plurality of side noise guest entities, the first entity association set includes an association between each of the side noise main entities and each of the side noise guest entities, and the first timestamp includes an establishment time of an association between each of the side noise main entities and each of the side noise guest entities; The method aggregates all completed edge quadruples, and then completes the cloud-side noise time series knowledge graph based on the aggregation results, specifically: Aggregating all completed edge quadruple groups through the cloud server, and completing the cloud-side noise time series knowledge graph in accordance with a preset cloud-side quadruple completion function and the aggregation results; wherein the cloud-side quadruple completion function is specifically: Where G c represents the cloud-side quadruple in the cloud-side noise time series knowledge graph, represents the edge quadruple of the completed edge noise temporal knowledge graph uploaded by the i-th edge server to the cloud server, T i e represents the first timestamp of the edge quadruple of the completed edge noise time series knowledge graph uploaded by the i-th edge server to the cloud server, and g(g) represents the cloud-side noise time series knowledge graph completion function.

2. The method for constructing a power line communication noise library according to claim 1, wherein: The edge servers respectively utilize all the first neighbor entity sets and all the second neighbor entity sets, combine the noise data feature set and the multimodal mapping relationship model of the cloud-side noise library, and perform entity completion on the first main entity set and the first guest entity set in the edge quadruple of the edge noise time series knowledge graph to obtain the corresponding second main entity set and the second guest entity set, so as to complete the completion of the edge quadruple and the edge noise time series knowledge graph, specifically including: Through each of the edge servers, according to a preset main entity completion function, combined with the weight coefficients of each of the similar main entities in the cloud-side noise time series knowledge graph, the first neighbor entity set corresponding to each of the similar main entities, the noise data feature set, the weight parameters of the noise data feature set, and the multimodal mapping relationship model of the cloud-side noise library, entity completion is performed on the first main entity set to obtain the second main entity set corresponding to the first main entity set, so as to form a main entity completion result corresponding to each of the edge servers, and complete the main entity completion of the edge quadruple in the edge noise time series knowledge graph; Through each of the edge servers, in accordance with a preset guest entity completion function, combined with the weight coefficients of each of the similar guest entities in the cloud-side noise time series knowledge graph, the second neighbor entity set corresponding to each of the similar guest entities, the noise data feature set, the weight parameters of the noise data feature set, and the multimodal mapping relationship model of the cloud-side noise library, the first guest entity set is completed, and the second guest entity set corresponding to the first guest entity set is obtained to form the guest entity completion results corresponding to each of the edge servers, thereby completing the guest entity completion of the edge quadruple in the edge noise time series knowledge graph.

3. The method for constructing a power line communication noise library according to claim 2, wherein: The main entity completion function and the guest entity completion function are specifically: Where, Represents the second main entity set corresponding to the first main entity set of the side quadruple in the side noise time series knowledge graph, represents the second guest entity set corresponding to the first guest entity set of the edge quadruple in the edge noise time series knowledge graph, f(g) represents the nonlinear entity completion function, N represents the number of similar main entities in the cloud side noise time series knowledge graph, L n ci represents the first neighbor entity set corresponding to the nth similar main entity in the cloud-side noise time series knowledge graph, L n cj represents the set of second neighbor entities corresponding to the nth similar guest entity in the cloud-side noise time series knowledge graph, α ci n represents the weight coefficient of the nth similar main entity in the cloud-side noise time series knowledge graph, α cj n represents the weight coefficient of the nth similar customer entity in the cloud-side noise time series knowledge graph, Z represents the noise data feature set, χ represents the weight parameter of the noise data feature set, V represents the multimodal mapping relationship model of the cloud-side noise library, V = {k1, k2}, k1 represents the mapping relationship between the concept model and the internal model, k2 represents the mapping relationship between the concept model and the external model, and δ represents the weight parameter of the multimodal mapping relationship model of the cloud-side noise library.

4. The method for constructing a power line communication noise library according to claim 1, wherein: Also includes: Through each of the edge servers, according to all the completed edge noise time series knowledge graphs, searching for corresponding entity names in the cloud-side noise library of the cloud server to update the edge-cloud mapping relationship of the cloud-side noise library; Updating, by the cloud server, a mapping relationship in a multimodal mapping relationship model of the cloud-side noise library according to the edge-cloud mapping relationship; The edge-cloud mapping relationship refers to the mapping relationship between the edge-side noise of the edge server and the cloud-side noise of the cloud server.

5. The method for constructing a power line communication noise library according to claim 1, wherein: The acquisition of the noise data feature set is specifically as follows: The power terminal collects carrier data of power line communication in real time, normalizes the carrier data, and then filters the normalized carrier data using a bandpass filter to obtain corresponding noise data, and uploads the noise data to each edge server; The noise data is subjected to feature extraction by each of the edge servers to obtain the noise data feature set.

6. A power line communication noise library construction system, characterized in that: include: The data acquisition module is used to obtain a noise data feature set and, through multiple edge servers, obtain a set of first neighbor entities corresponding to several similar main entities, a set of second neighbor entities corresponding to several similar guest entities, and a multimodal mapping relationship model of the cloud-side noise library from the cloud-side noise time series knowledge graph in the cloud-side noise library. A graph completion module is configured to, through each of the edge servers, respectively utilize all the first neighbor entity sets and all the second neighbor entity sets, combine the noise data feature set and the multimodal mapping relationship model of the cloud-side noise library, and perform entity completion on the first main entity set and the first guest entity set in the side quadruple of the side noise time series knowledge graph to obtain the corresponding second main entity set and second guest entity set, so as to complete the completion of the side quadruple and the side noise time series knowledge graph; wherein the side quadruple includes the first main entity set, the first guest entity set, a first entity association set and a first timestamp, the first main entity set includes a plurality of side noise main entities, the first guest entity set includes a plurality of side noise guest entities, the first entity association set includes the association between each of the side noise main entities and each of the side noise guest entities, and the first timestamp includes the establishment time of the association between each of the side noise main entities and each of the side noise guest entities; A noise library construction module is configured to receive, through a cloud server, the completed edge-side noise time series knowledge graph uploaded by each edge server, aggregate all completed edge-side quadruples, and then complete the cloud-side noise time series knowledge graph based on the aggregation result to update the cloud-side noise library, and then use the updated cloud-side noise library as the power line communication noise library; The method aggregates all completed edge quadruples, and then completes the cloud-side noise time series knowledge graph based on the aggregation results, specifically: Aggregating all completed edge quadruple groups through the cloud server, and completing the cloud-side noise time series knowledge graph in accordance with a preset cloud-side quadruple completion function and the aggregation results; wherein the cloud-side quadruple completion function is specifically: Where G c represents the cloud-side quadruple in the cloud-side noise time series knowledge graph, represents the edge quadruple of the completed edge noise temporal knowledge graph uploaded by the i-th edge server to the cloud server, T i e represents the first timestamp of the edge quadruple of the completed edge noise time series knowledge graph uploaded by the i-th edge server to the cloud server, and g(g) represents the cloud-side noise time series knowledge graph completion function.

7. A power line communication noise library construction system according to claim 6, characterized in that: The graph completion module specifically includes: A first graph completion unit is configured to perform entity completion on the first main entity set through each edge server according to a preset main entity completion function, in combination with the weight coefficient of each similar main entity in the cloud-side noise time series knowledge graph, the first neighbor entity set corresponding to each similar main entity, the noise data feature set, the weight parameter of the noise data feature set, and the multimodal mapping relationship model of the cloud-side noise library, to obtain the second main entity set corresponding to the first main entity set, so as to form a main entity completion result corresponding to each edge server, and complete the main entity completion of the edge quadruple in the edge noise time series knowledge graph; The second graph completion unit is used to perform entity completion on the first guest entity set through each of the edge servers in accordance with a preset guest entity completion function, combined with the weight coefficients of each of the similar guest entities in the cloud-side noise time series knowledge graph, the second neighbor entity set corresponding to each of the similar guest entities, the noise data feature set, the weight parameters of the noise data feature set, and the multimodal mapping relationship model of the cloud-side noise library, to obtain the second guest entity set corresponding to the first guest entity set, so as to form the guest entity completion results corresponding to each of the edge servers, and complete the guest entity completion of the edge quadruple in the edge noise time series knowledge graph.

8. A power line communication noise library construction system according to claim 7, characterized in that: The main entity completion function and the guest entity completion function are specifically: Where, Represents the second main entity set corresponding to the first main entity set of the side quadruple in the side noise time series knowledge graph, represents the second guest entity set corresponding to the first guest entity set of the edge quadruple in the edge noise time series knowledge graph, f(g) represents the nonlinear entity completion function, N represents the number of similar main entities in the cloud side noise time series knowledge graph, L n ci represents the first neighbor entity set corresponding to the nth similar main entity in the cloud-side noise time series knowledge graph, L n cj represents the set of second neighbor entities corresponding to the nth similar guest entity in the cloud-side noise time series knowledge graph, α ci n represents the weight coefficient of the nth similar main entity in the cloud-side noise time series knowledge graph, α cj n represents the weight coefficient of the nth similar customer entity in the cloud-side noise time series knowledge graph, Z represents the noise data feature set, χ represents the weight parameter of the noise data feature set, V represents the multimodal mapping relationship model of the cloud-side noise library, V = {k1, k2}, k1 represents the mapping relationship between the concept model and the internal model, k2 represents the mapping relationship between the concept model and the external model, and δ represents the weight parameter of the multimodal mapping relationship model of the cloud-side noise library.

9. A power line communication noise library construction system according to claim 6, characterized in that: Also includes: A mapping relationship updating module is configured to search for corresponding entity names in the cloud-side noise library of the cloud server through each of the edge servers according to all completed edge-side noise time series knowledge graphs, so as to update the edge-cloud mapping relationship of the cloud-side noise library; Through the cloud server, the mapping relationship in the multimodal mapping relationship model of the cloud-side noise library is updated according to the edge-cloud mapping relationship; wherein the edge-cloud mapping relationship refers to the mapping relationship between the edge-side noise of the edge server and the cloud-side noise of the cloud server.

Images