Electric power accurate first-aid repair method based on multi-modal semantic sensing and information integration

Abstract

The invention relates to an electric power accurate first-aid repair method, in particular to an electric power accurate first-aid repair method based on multi-modal semantic sensing information integration. According to the method, the multi-mode semantic communication perception integrated technology is utilized for accurate first-aid repair of the power system for the first time, and safe and stable operation of the power system is protected. And under the condition that a multi-mode semantic communication and sensing integrated technology is used, accurate power emergency repair is realized.

Description

Technical Field

[0001] The present invention relates to the field of communication, and in particular to an electric power precise repair method based on the integration of multi-modal semantic communication and induction. Background Art

[0002] The precise repair technology of power systems requires precise location, weather, displacement, settlement, tilt and other parameter indicators to guide precise repair. However, in traditional repair systems, multiple sensors are often used for independent perception, and the independently perceived parameters and original images, texts, and voices are transmitted to the background analysis system through various communication media, which brings huge pressure on data analysis and fusion to the background analysis system, and also brings huge pressure on data transmission to the communication system; the pressure of calculation and communication brings a lot of delay, thus affecting the precise repair of power systems.

[0003] Obviously, the perception, communication and information processing of the above-mentioned precise repair technology of the power system have brought two problems: First, perception, communication and information processing are independent of each other, and thus cannot meet the needs of supporting precise repair of the power system, that is, the perception accuracy, communication-related indicators and information processing performance cannot achieve precise repair. Secondly, since perception, communication and information processing cannot be jointly optimized, a large amount of perception resources, communication resources and information processing resources are wasted. Therefore, the perception, communication and signal processing of the traditional precise repair of the power system can no longer meet the requirements of precise repair of the power system. Summary of the invention

[0004] In order to study the integrated network of communication, perception and information processing to support the precise emergency repair business of the power system, the present invention discloses a precise emergency repair method of the power system based on the integration of multimodal semantic communication, perception and information.

[0005] The embodiment of the present invention provides the following technical solutions:

[0006] A precise power repair method based on multimodal semantic communication and communication integration, the method comprising:

[0007] Step A: Based on the power repair business, a semantic expression form integrating communication, perception and signal processing is constructed to extract multimodal semantic information from the information source;

[0008] Step B: adopt the semantic transmission coding of Beidou satellite and 5G to fuse multimodal semantic information at the channel level and transmit it;

[0009] Step C, performing semantic-based information processing at the data processing terminal of the edge computing base station;

[0010] Step D, calculating the amount of information transmitted by semantics, the semantic error rate, and the total delay of semantic perception, communication, and information processing;

[0011] Step E: Calculate the perception accuracy and the accuracy of multimodal semantic information processing, and optimize the perception and information processing algorithms according to the accuracy.

[0012] Wherein, step A specifically includes:

[0013] A1, based on the power repair business, construct a semantic expression data set of perception, which is represented by {S Area ,O Shape ,B Shape ,R A-S}, where S Area Represents the precise positioning position coordinates of the target object, which is S Area =(x Area ,y Area ,z Area ,τ), where x Area Represents the horizontal coordinate of the target object, y Area Represents the vertical coordinate of the target object, z Area Represents the height of the target object from the ground, τ is the precise positioning accuracy, and the positioning of this target object is sensed by the Beidou satellite. Shape Represents the shape feature of the target object, which can be obtained through the three-dimensional image of the target object. Shape Represents the shape boundary features of the target object, which is obtained through the target object image. A-S Represents the motion information of the target object, which is R A-S ={v,a,l}, where v represents the object's speed, a represents the object's acceleration, and l represents the target object's trajectory, which is sensed by the Beidou satellite.

[0014] A2, the semantic expression data set defined in step A1, constructs the semantic expression form of communication, perception and signal processing integration as follows:

[0015]

[0016] Among them, S Area Represents the precise positioning coordinates of the target object, n represents the nth channel, a total of N channels, k n represents the channel coefficient, W τ The representative variance is τ 2 The matrix representation of additive white Gaussian noise (AWGN) is: q represents the qth modulation symbol, v represents the speed of the target object, s represents the transmission distance, and f s,l represents the transmission frequency, Δf represents the modulation frequency error, λ represents the phase, j represents the imaginary unit, and τ represents the variance.

[0017] A3, semantic sources generate raw data, such as pictures, videos, voices, and texts. The generated data contains certain semantic information to be shared with the semantic destination, and semantic information is extracted from the information source. The semantic information is defined as I features. Among them, N born represents the semantic information generation factor, n i represents the i-th semantic feature, h qos (n) represents the joint probability distribution. Therefore, the semantic information obtained from the information source can be expressed as:

[0018] Wherein, step B specifically includes:

[0019] B1, using the semantic transmission coding of Beidou satellite and 5G, defines the input data x using the nonlinear encoder function j ε : Mapped into vector u=j s (x) where s represents the channel parameter, R represents the channel matrix, and A H Represents the data format height, B W represents the data format width, C represents the data format height, R D Represents the reconstructed matrix. Set vector u=j s (x) Each value is quantized into Q uniform quantization levels, and the quantized set is: C Y ={y1,y2,...y Q}, through the quantizer: in, represents the quantized matrix. in, Represents plaintext in encryption.

[0020] B2, through a nonlinear decoder function A θ : Pass the estimated u to the input reconstruction.

[0021] B3, channel-level information fusion multimodal semantic information is expressed as R fusi on represents the fused multimodal semantic information, m represents the data of the mth modality, M represents the data of M modalities in total, and H m represents the Rayleigh fading wireless channel, F m represents the encoding matrix, x m Represents the transmission signal, n AWGN represents additive white Gaussian noise.

[0022] Wherein, step C specifically includes:

[0023] C1, cleans, normalizes and segment the input text, removes special characters, punctuation marks, etc., and divides the text into sentences or paragraphs. Use named entity recognition technology to identify key entities involved in the text, such as power equipment, to achieve accurate fault location.

[0024] C2, extract the relationship between entities from the text, such as the association between equipment and fault type. Consider the context information, solve the problems of reference resolution and word sense disambiguation, and ensure the accurate understanding of the meaning of the text. Integrate the existing power field knowledge base with the text and define the similarity T between the texts. SIM , Among them, λ x , α x are the mean and variance of x, respectively. Respectively The mean and variance of b1 and b2 are stability coefficients. Domain expertise is used to explain and supplement the content in the text.

[0025] C3, based on the integrated knowledge and reasoning rules, performs logical reasoning, fault troubleshooting decisions, etc., solves problems or provides suggestions. According to the processing results, it generates corresponding fault troubleshooting plans, reports, visual charts, etc., so that maintenance personnel can understand and implement them.

[0026] Wherein, step D specifically comprises:

[0027] D1, calculates the amount of information transmitted by semantics, Among them, W B represents the channel bandwidth, represents the signal-to-noise ratio at the receiving end, f represents the transmission signal frequency, for The statistical distribution function of .

[0028] D2, calculate the semantic error rate, Among them, ω1, ω2 represent modulation specific parameters, represents the conditional bit error rate, for The statistical distribution function of .

[0029] D3, the total delay of computing semantic perception, communication, and information processing in, represents the perceived delay, v sense represents the perception rate, represents the statistical distribution function of the lth channel, represents the data processing delay, R trans Represents the data transfer rate, represents the communication delay, L data Represents the data size, Wtunnel Represents the channel bandwidth.

[0030] Wherein, step E specifically includes:

[0031] E1, calculates the perception accuracy and uses the minimum mean square error to represent the perception accuracy. Where Z represents the receiving calculation beamforming matrix at the base station, K B represents the transmit calculation beamforming matrix of the bth sensing signal, u B represents the transmit sensing beamforming vector, Z H represents Z and the conjugate transposed matrix, U h represents the channel matrix from the hth transmitter to the base station, V represents the sensing beamforming vector, and δ represents the mean square value of the channel noise;

[0032] E2, calculates the accuracy of information processing and defines the similarity of information to represent the accuracy of information processing. Among them, SSVM represents information similarity, k th Represents the comparative similarity, l th represents the number of rows in the comparison similarity, k max It represents the total number of rows for comparison similarity, and vps represents the similarity coefficient.

[0033] E3, improve the accuracy of perception and multimodal semantic information processing, ensure that the data set is cleaned, denoised and preprocessed to reduce the impact of noise and errors. Use error analysis and feedback mechanisms to iteratively optimize the perception and information processing algorithms to improve their accuracy and reliability.

[0034] E4, multimodal semantic communication enhances interactivity and participation through multiple input methods, saves a lot of communication resources, improves communication efficiency, reduces communication latency, and outputs multimodal perception and information processing models to further improve the overall accuracy of the system.

[0035] Compared with the prior art, the above technical solution has the following advantages:

[0036] The present invention utilizes the integration of perception, communication and information processing to meet the needs of supporting accurate emergency repair of power systems, that is, to achieve accurate emergency repair by achieving perception accuracy, communication-related indicators and information processing performance. Secondly, perception, communication and information processing are jointly optimized to save a large amount of perception resources, communication resources and information processing resources. The overall delay is minimized, while optimizing the amount of unloaded data and cache space. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0038] Figure 1 A flowchart of a method for precise power repair based on multimodal semantic communication and information integration is provided in accordance with an embodiment of the present invention. DETAILED DESCRIPTION

[0039] As described in the background technology section, how to meet the strict requirements of power precision repair technology while reducing the overall information transmission delay and completing multimodal semantic communication, perception, and information resource allocation is an urgent problem to be solved by people in this field.

[0040] The core idea of ​​the present invention is to utilize the integrated technology of perception, communication and information processing in the precise positioning of electric power, and to complete the multimodal semantic communication, perception and information resource allocation based on multimodal semantic communication, so that perception, communication and information processing can be jointly optimized to reduce the information transmission delay, thereby improving the overall performance of the system. Compared with the traditional electric power precise repair technology, the goal of improving overall performance can be achieved.

[0041] See also Figure 1 The embodiment of the present invention provides a precise power repair method based on multi-modal semantic communication and communication integration, including:

[0042] Step A: Based on the power repair business, a semantic expression form integrating communication, perception and signal processing is constructed to extract multimodal semantic information from the information source;

[0043] Step B: adopt the semantic transmission coding of Beidou satellite and 5G to fuse multimodal semantic information at the channel level and transmit it;

[0044] Step C, performing semantic-based information processing at the data processing terminal of the edge computing base station;

[0045] Step D, calculating the amount of information transmitted by semantics, the semantic error rate, and the total delay of semantic perception, communication, and information processing;

[0046] Step E: Calculate the perception accuracy and the accuracy of multimodal semantic information processing, and optimize the perception and information processing algorithms according to the accuracy.

[0047] Wherein, step A specifically includes:

[0048] A1, based on the power repair business, construct a semantic expression data set of perception, which is represented by {S Area ,O Shape ,B Shape ,R A-S}, where S Area Represents the precise positioning position coordinates of the target object, which is S Area =(x Area ,y Area ,z Area ,τ), where x Area Represents the horizontal coordinate of the target object, y Area Represents the vertical coordinate of the target object, z Area Represents the height of the target object from the ground, τ is the precise positioning accuracy, and the positioning of this target object is sensed by the Beidou satellite. Shape Represents the shape feature of the target object, which can be obtained through the three-dimensional image of the target object. Shape Represents the shape boundary features of the target object, which is obtained through the target object image. A-S Represents the motion information of the target object, which is R A-S ={v,a,l}, where v represents the object's speed, a represents the object's acceleration, and l represents the target object's trajectory, which is sensed by the Beidou satellite.

[0049] A2, the semantic expression data set defined in step A1, constructs the semantic expression form of communication, perception and signal processing integration as follows:

[0050]

[0051] Among them, S Area Represents the precise positioning coordinates of the target object, n represents the nth channel, a total of N channels, k n represents the channel coefficient, W τ The representative variance is τ 2 The matrix representation of additive white Gaussian noise (AWGN) is: q represents the qth modulation symbol, v represents the speed of the target object, s represents the transmission distance, and f s,l represents the transmission frequency, Δf represents the modulation frequency error, λ represents the phase, j represents the imaginary unit, and τ represents the variance.

[0052] A3, semantic sources generate raw data, such as pictures, videos, voices, and texts. The generated data contains certain semantic information to be shared with the semantic destination, and semantic information is extracted from the information source. The semantic information is defined as I features. Among them, N born represents the semantic information generation factor, n i represents the i-th semantic feature, h qos(n) represents the joint probability distribution. Therefore, the semantic information obtained from the information source can be expressed as:

[0053] Wherein, step B specifically includes:

[0054] B1, using the semantic transmission coding of Beidou satellite and 5G, defines the input data x using the nonlinear encoder function j ε : Mapped into vector u=j s (x) where s represents the channel parameter, R represents the channel matrix, and A H Represents the data format height, B W represents the data format width, C represents the data format height, R D Represents the reconstructed matrix. Set vector u=j s (x) Each value is quantized into Q uniform quantization levels, and the quantized set is: C Y ={y1,y2,...y Q}, through the quantizer: in, represents the quantized matrix. in, Represents plaintext in encryption.

[0055] B2, through a nonlinear decoder function A θ : Pass the estimated u to the input reconstruction.

[0056] B3, channel-level information fusion multimodal semantic information is expressed as R fusi on represents the fused multimodal semantic information, m represents the data of the mth modality, M represents the data of M modalities in total, and H m represents the Rayleigh fading wireless channel, F m represents the encoding matrix, x m Represents the transmission signal, n AWGN represents additive white Gaussian noise.

[0057] Wherein, step C specifically includes:

[0058] C1, cleans, normalizes and segment the input text, removes special characters, punctuation marks, etc., and divides the text into sentences or paragraphs. Use named entity recognition technology to identify key entities involved in the text, such as power equipment, to achieve accurate fault location.

[0059] C2, extract the relationship between entities from the text, such as the association between equipment and fault type. Consider the context information, solve the problems of reference resolution and word sense disambiguation, and ensure the accurate understanding of the meaning of the text. Integrate the existing power field knowledge base with the text and define the similarity T between the texts. SIM , Among them, λ x , α x are the mean and variance of x, respectively. Respectively The mean and variance of b1 and b2 are stability coefficients. Domain expertise is used to explain and supplement the content in the text.

[0060] C3, based on the integrated knowledge and reasoning rules, performs logical reasoning, fault troubleshooting decisions, etc., solves problems or provides suggestions. According to the processing results, it generates corresponding fault troubleshooting plans, reports, visual charts, etc., so that maintenance personnel can understand and implement them.

[0061] Wherein, step D specifically comprises:

[0062] D1, calculates the amount of information transmitted by semantics, Among them, W B represents the channel bandwidth, represents the signal-to-noise ratio at the receiving end, f represents the transmission signal frequency, for The statistical distribution function of .

[0063] D2, calculate the semantic error rate, Among them, ω1, ω2 represent modulation specific parameters, represents the conditional bit error rate, for The statistical distribution function of .

[0064] D3, the total delay of computing semantic perception, communication, and information processing in, represents the perceived delay, v sense represents the perception rate, represents the statistical distribution function of the lth channel, represents the data processing delay, R trans Represents the data transfer rate, represents the communication delay, L data Represents the data size, W tunnel Represents the channel bandwidth.

[0065] Wherein, step E specifically includes:

[0066] E1, calculates the perception accuracy and uses the minimum mean square error to represent the perception accuracy. Where Z represents the receiving calculation beamforming matrix at the base station, K B represents the transmit calculation beamforming matrix of the bth sensing signal, u B represents the transmit sensing beamforming vector, Z H represents Z and the conjugate transposed matrix, U h represents the channel matrix from the hth transmitter to the base station, V represents the sensing beamforming vector, and δ represents the mean square value of the channel noise;

[0067] E2, calculates the accuracy of information processing and defines the similarity of information to represent the accuracy of information processing. Among them, SSVM represents information similarity, k th Represents the comparative similarity, l th Represents the number of rows in the comparison similarity, k max It represents the total number of rows for the comparison of similarity, and vps represents the similarity coefficient.

[0068] E3, improve the accuracy of perception and multimodal semantic information processing, ensure that the data set is cleaned, denoised and preprocessed to reduce the impact of noise and errors. Use error analysis and feedback mechanisms to iteratively optimize the perception and information processing algorithms to improve their accuracy and reliability.

[0069] E4, multimodal semantic communication enhances interactivity and participation through multiple input methods, saves a lot of communication resources, improves communication efficiency, reduces communication latency, and outputs multimodal perception and information processing models to further improve the overall accuracy of the system.

[0070] Compared with the prior art, the above technical solution has the following advantages:

[0071] This method combines multiple sensors and data sources, such as sound, images, vibrations, etc., to obtain information about power failures from different angles. By comprehensively utilizing data from different modes, the fault situation can be more comprehensively grasped, and the accuracy and efficiency of emergency repairs can be improved. Combining multimodal data with semantic understanding, an integrated emergency repair technology is realized. By integrating multimodal data and semantic information, more accurate emergency repair judgments and guidance can be provided, reducing interference from human factors, and improving emergency repair effects and safety. By comprehensively analyzing multimodal data and semantic information, the location, type, and severity of power failures can be quickly located. A large amount of perception resources, communication resources, and information processing resources are saved, while minimizing overall latency.

[0072] The present invention discloses a precise power repair method, a precise power repair method based on multimodal semantic communication perception integration. The method uses multimodal semantic communication perception integration technology for the first time to perform precise power repair on the power system and protect the safe and stable operation of the power system. By using the multimodal semantic communication perception integration technology, precise power repair is achieved.

[0073] The various parts in this manual are described in a progressive manner, and each part focuses on the differences from other parts. The same or similar parts between the various parts can be referenced to each other.

[0074] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A precise power repair method based on the integration of multimodal semantic communication and communication, characterized in that: The steps include: Step A: Based on the power repair business, a semantic expression form integrating communication, perception and signal processing is constructed to extract multimodal semantic information from the information source; Step B: adopt the semantic transmission coding of Beidou satellite and 5G to fuse multimodal semantic information at the channel level and transmit it; Step C, performing semantic-based information processing at the data processing terminal of the edge computing base station; Step D, calculating the amount of information transmitted by semantics, the semantic error rate, and the total delay of semantic perception, communication, and information processing; Step E, calculating the perception accuracy and the accuracy of multimodal semantic information processing, and optimizing the perception and information processing algorithms according to the accuracy to improve the accuracy of the power repair service; Wherein, step A specifically includes: A1, based on the precise repair business, construct a set of perceived semantic expression data, which is represented by {S Area ,O Shape ,B Shape ,R A-S }, where S Area Represents the precise positioning coordinates of the target object, which is S Area =(x Area ,y Area ,z Area ,τ), where x Area Represents the horizontal coordinate of the target object, y Area Represents the vertical coordinate of the target object, z Area represents the height of the target object from the ground, τ is the precise positioning accuracy, and the positioning of this target object is sensed by the Beidou satellite; Shape Represents the shape feature of the target object, which can be obtained through the three-dimensional image of the target object; B Shape Represents the shape boundary features of the target object, which is obtained through the target object image; R A-S Represents the motion information of the target object, which is R A-S ={v,a,l}, where v represents the object's velocity, a represents the object's acceleration, and l represents the target object's trajectory, which is sensed by BeiDou satellites; A2, the semantic expression data set defined in step A1, constructs the semantic expression form of communication, perception and signal processing integration as follows: Among them, S Area Represents the precise positioning coordinates of the target object, n represents the nth channel, a total of N channels, k n represents the channel coefficient, W τ The representative variance is τ 2 The matrix representation of additive white Gaussian noise (AWGN) is: q represents the qth modulation symbol, v represents the speed of the target object, s represents the transmission distance, and f s,l represents the transmission frequency, Δf represents the modulation frequency error, λ represents the phase, j represents the imaginary unit, and τ represents the variance; A3, semantic sources generate raw data, such as pictures, videos, voices, and texts. The generated data contains certain semantic information to be shared with the semantic destination. Semantic information is extracted from the information source and defined as the semantic information containing I features. Among them, N born represents the semantic information generation factor, n i represents the i-th semantic feature, h qos (n) represents the joint probability distribution, so the semantic information obtained from the information source can be expressed as: Wherein, step B specifically includes: B1, using the semantic transmission coding of Beidou satellite and 5G, defines the input data x using the nonlinear encoder function j ε : Mapped into vector u=j s (x) where s represents the channel parameter, R represents the channel matrix, and A H Represents the data format height, B W represents the data format width, C represents the data format height, R D Represents the reconstructed matrix, and the vector u=j s (x) Each value is quantized into Q uniform quantization levels, and the quantized set is: C Y ={y1,y2,...y Q }, through the quantizer: in, Represents the quantized matrix, let in, Represents the plaintext in encryption; B2, through a nonlinear decoder function A θ : Pass the estimated u to the input Reconstruction of B3, channel-level information fusion multimodal semantic information is expressed as R fusion represents the fused multimodal semantic information, m represents the data of the mth modality, M represents the data of M modalities in total, and H m represents the Rayleigh fading wireless channel, F m represents the encoding matrix, x m Represents the transmission signal, n AWGN represents additive white Gaussian noise; Wherein, step C specifically includes: C1, cleans, normalizes and segment the input text, removes special characters and punctuation marks, and divides the text into sentences or paragraphs. It uses named entity recognition technology to identify key entities involved in the text, such as power equipment, to accurately locate faults. C2, extract the relationship between entities from the text, such as the association between equipment and fault type, consider context information, solve problems such as reference resolution and word sense disambiguation, ensure accurate understanding of the meaning of the text, integrate the existing power field knowledge base with the text, and define the similarity T between the texts. SIM , Among them, λ x , α x are the mean and variance of x, respectively. Respectively The mean and variance of b1 and b2 are stability coefficients. Use domain expertise to explain and supplement the content in the text. C3, based on the integrated knowledge and reasoning rules, conducts logical reasoning and troubleshooting decisions, solves problems or provides suggestions, and generates corresponding troubleshooting plans, reports, visualization charts, etc. according to the processing results, so that maintenance personnel can understand and implement them; Wherein, step D specifically comprises: D1, calculates the amount of information transmitted by semantics, Among them, W B represents the channel bandwidth, represents the signal-to-noise ratio at the receiving end, f represents the transmission signal frequency, for The statistical distribution function of D2, calculate the semantic error rate, Among them, ω1, ω2 represent modulation specific parameters, represents the conditional bit error rate, for The statistical distribution function of D3, the total delay of computing semantic perception, communication, and information processing in, represents the perceived delay, v sense represents the perception rate, represents the statistical distribution function of the lth channel, represents the data processing delay, R trans Represents the data transfer rate, represents the communication delay, L data Represents the data size, W tunnel represents the channel bandwidth; Wherein, step E specifically includes: E1, calculates the perception accuracy and uses the minimum mean square error to represent the perception accuracy. Where Z represents the receiving calculation beamforming matrix at the base station, K B represents the transmit calculation beamforming matrix of the bth sensing signal, u B represents the transmit sensing beamforming vector, Z H represents Z and the conjugate transposed matrix, U h represents the channel matrix from the hth transmitter to the base station, V represents the sensing beamforming vector, and δ represents the mean square value of the channel noise; E2, calculates the accuracy of information processing and defines the similarity of information to represent the accuracy of information processing. Among them, SSVM represents information similarity, k th Represents the comparative similarity, l th Represents the number of rows in the comparison similarity, k max It represents the total number of rows compared with similarity, and vps represents the similarity coefficient; E3, improve the accuracy of perception and multimodal semantic information processing, ensure that the data set is cleaned, denoised and preprocessed to reduce the impact of noise and errors, and use error analysis and feedback mechanisms to iteratively optimize the perception and information processing algorithms to improve their accuracy and reliability; E4, multimodal semantic communication enhances interactivity and participation through multiple input methods, saves a lot of communication resources, improves communication efficiency, reduces communication latency, and outputs multimodal perception and information processing models to further improve the overall accuracy of the system.

Images