A non-contact electromagnetic power supply and energy storage monitoring system for high-voltage transmission lines

Abstract

The present invention discloses a non-contact electromagnetic power supply and energy storage supervision system for high-voltage transmission lines, including a data acquisition module, a feature extraction module, a preliminary clustering module, a local neighborhood support fusion module, a secondary clustering module, a loss function construction module and an energy storage supervision module. The present invention belongs to the field of energy storage supervision, and specifically refers to a non-contact electromagnetic power supply and energy storage supervision system for high-voltage transmission lines. This scheme introduces a dynamic fuzzy factor; based on the dynamic allocation of signal feature weights to reduce the impact of environmental noise, flexibly divide electromagnetic signal boundary samples, increase neighborhood density constraints, and improve the accuracy of abnormal signal support, thereby reducing the risk of false alarms and missed alarms, and improving the accuracy of energy storage supervision; by introducing an adaptive smoothing factor, constructing a time-weighted contrast loss, reducing the interference of non-critical time periods on the results, and introducing an abnormal penalty mechanism to improve the ability to identify abnormal signals, thereby improving the effect of energy storage supervision.

Description

Technical Field

[0001] The present invention relates to the field of energy storage supervision, and in particular to a non-contact electromagnetic power supply and energy storage supervision system for high-voltage transmission lines. Background Art

[0002] The electromagnetic power supply and energy storage supervision system is a comprehensive monitoring system that combines non-contact power collection, energy storage equipment management and signal status analysis. Its core purpose is to use the electromagnetic field around the high-voltage transmission line to achieve power supply through electromagnetic induction, while monitoring the operating status of the energy storage equipment in real time, detecting anomalies, and providing operation optimization and fault warnings. However, the general electromagnetic power supply and energy storage supervision system has problems such as the complex distribution of electromagnetic signals in the transmission line and improper handling of the differences in signal distribution between different states, which leads to false alarms and missed reports, and poor supervision effect; the general electromagnetic power supply and energy storage supervision system has problems such as improper division of abnormal states, ignoring the differences in data characteristics in different time periods, resulting in poor accuracy in identifying abnormal states, and thus poor energy storage supervision effect. Summary of the invention

[0003] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides a non-contact electromagnetic power supply and energy storage supervision system for high-voltage transmission lines. In view of the problem that the general electromagnetic power supply and energy storage supervision system has the problem of complex electromagnetic signal distribution of transmission lines and improper handling of signal distribution differences between different states, which leads to false alarms and missed alarms, and poor supervision effect, this scheme is based on the introduction of dynamic fuzzy factors to effectively distinguish abnormal signal samples from normal signal samples; based on the dynamic allocation of signal feature weights to reduce the impact of environmental noise, the electromagnetic signal boundary samples are flexibly divided by adaptively adjusting the threshold of the boundary samples, and the dynamic adjustment coefficient is introduced to adjust the direction of the restricted neighborhood range, increase the constraints of neighborhood density, improve the accuracy of abnormal signal support, thereby reducing the risk of false alarms and missed alarms, and improving the accuracy of energy storage supervision; for the problem that the general electromagnetic power supply and energy storage supervision system has improper division of abnormal states and ignores the differences in data characteristics in different time periods, resulting in poor accuracy in identifying abnormal states, and then leading to poor energy storage supervision effects, this scheme introduces an adaptive smoothing factor to finely divide the state of electromagnetic signal data, and reduces the interference of non-critical time periods on the results by constructing a time-weighted contrast loss. In addition, an abnormal penalty mechanism is introduced to improve the ability to identify abnormal signals, thereby improving the energy storage supervision effect.

[0004] The technical solution adopted by the present invention is as follows: A non-contact electromagnetic power supply and energy storage supervision system for high-voltage transmission lines provided by the present invention includes a data acquisition module, a feature extraction module, a preliminary clustering module, a local neighborhood support fusion module, a secondary clustering module, a loss function construction module and an energy storage supervision module;

[0005] The data acquisition module acquires historical electromagnetic signal data;

[0006] The feature extraction module uses a deep autoencoder to reduce the dimension of high-dimensional electromagnetic signal data and extract key features that reflect the operating status of the transmission line;

[0007] The preliminary clustering module classifies the extracted feature data by using a fuzzy clustering algorithm to preliminarily identify electromagnetic signal samples in normal, abnormal and uncertain states;

[0008] The local neighborhood support fusion module constructs a direction-restricted neighborhood with the boundary sample as the center, and calculates the membership of the boundary sample by determining the support information of the sample in the neighborhood;

[0009] The secondary clustering module jointly optimizes the membership matrix after the initial clustering and the local neighborhood support fusion, recalculates the cluster center and sample membership, and realizes the final classification;

[0010] The loss function construction module constructs an optimization objective function that integrates reconstruction, contrast, and clustering losses;

[0011] The energy storage supervision module dynamically monitors the power extraction efficiency and abnormal status of the energy storage system based on the operating status recognition and clustering results of the real-time electromagnetic signals, and issues an early warning.

[0012] Furthermore, in the data acquisition module, the historical electromagnetic signal data includes transmission line signal data, environmental data, time series tags and operating status; the operating status includes normal operation and abnormal operation; the operating status is used as a data label; and the data label does not participate in clustering operations.

[0013] Furthermore, the feature extraction module is specifically: using a deep autoencoder to reduce the dimension of the collected high-dimensional electromagnetic signal to extract key features; the encoder feature extraction is expressed as: ; The decoder reconstructs the electromagnetic signal It is expressed as: ; Reconstruction error The calculation is expressed as: ; Wherein, Z is the feature extraction result of the encoder, reflecting the core features of the electromagnetic signal; F(·) is the nonlinear mapping function of the encoder; G(·) is the nonlinear mapping function of the decoder; GELU(·) is the GELU activation function; X is the collected high-dimensional electromagnetic signal data matrix; W and b are the weight matrix and bias vector of the encoder respectively; is the square of the L2 norm.

[0014] Furthermore, the preliminary clustering module is specifically as follows: taking the extracted features as input, calculating the membership of each sample to each cluster center through the fuzzy clustering algorithm; dividing the electromagnetic signal samples into three categories according to the membership matrix: normal state, abnormal state and uncertain state; introducing the adaptive fuzzy factor , the membership matrix update is expressed as: ; ; ; Initial normal state set It is expressed as: ; Initial uncertain state set It is expressed as: ; Preliminary abnormal state set It is expressed as: ;Introducing dynamic threshold and dynamic adjustment coefficient , the labeled sample is represented as: ; ; ;in, is the i-th electromagnetic signal sample For the jth cluster center The degree of membership; is the kth cluster center; K1 is the total number of clusters, k is the total number of clusters index, and j is the total number of clusters index different from k; is a labeled sample; is the membership degree of the i-th electromagnetic signal sample to the k-th cluster center; It is a global data center; is the weighted Euclidean distance; D is the data dimension; and are the values ​​of the dth dimension of the sample and cluster center respectively; is the feature weight; is the importance prior score; exp(·) is the exponential function; is the basic threshold; is the local density; std(·) is the standard deviation; is the global mean of all sample memberships; is the control factor; is the global standard deviation of all sample memberships; is the value of the dth dimension of the kth cluster center; It will be When the maximum value is taken, the corresponding j is assigned to j.

[0015] Furthermore, the local neighborhood support fusion module is specifically as follows: taking the boundary sample as the center, constructing a direction-restricted neighborhood based on its distance to the cluster center, selecting neighboring samples with consistent directions and satisfying distance conditions, and using the membership information of the determined samples in the neighborhood to calculate the final membership of the boundary samples, where the boundary samples are samples of the preliminary uncertain state set; introducing dynamic adjustment parameters , the distance constraint constructed by the hemisphere is expressed as: ; ; Direction constraints are expressed as: ; The basic probability distribution of local neighborhood support fusion is expressed as: ; Introducing neighborhood density constraints, neighborhood fusion is expressed as: ; ; The membership degree of the boundary sample is expressed as: ;in, is a sample of neighbors; is the direction angle, which constrains the direction consistency between neighbor samples and cluster centers; and are the vectors from boundary samples to neighbor samples and cluster centers respectively; is the cluster center of the neighbor sample pair The support degree reflects the matching degree between the electromagnetic signal characteristics and the operating status; is the complete set of neighbor sample pairs support level; is the membership of the neighbor samples to the cluster center; is the global support of the boundary samples to the cluster center, which is used to judge the running status; L is the number of neighborhood samples, and r is the neighborhood sample index; is the support of samples in the neighborhood to the cluster center; K is the normalization factor; A and B are different operating states; and is the basic probability distribution value of the sample to the state; is the final membership of the boundary samples to the cluster center; Yes The standard deviation of all feature dimensions is calculated; d(·) is the Euclidean distance.

[0016] Furthermore, the secondary clustering module specifically includes: substituting the fused membership matrix into the fuzzy clustering algorithm to recalculate the cluster center and sample membership; introducing an adaptive smoothing factor , the final membership matrix It is expressed as: ; ; ; Final normal state set It is expressed as: ; Final abnormal state set It is expressed as: ;in, is the membership matrix of the samples obtained by preliminary clustering; is the membership matrix of the boundary samples obtained by fusion of local neighborhood supports; dens(·) is the density weight; is the number of preliminary clustering samples; is to take the maximum value.

[0017] Furthermore, the loss function construction module is specifically: constructing a time-weighted contrast loss , expressed as: ; Introduce abnormal penalty and construct clustering loss , expressed as: ; The overall loss function LL is expressed as: ;in, is the true label of the sample; is the sample classification label; N is the total number of samples; is the weighted Euclidean distance; is the regularization coefficient; and is the loss weight; is the sample timestamp; It is the global time center; is the L2 norm; is the anomaly coefficient; is the abnormal threshold; Yes Take the natural logarithm; and optimize the weights of the autoencoder through backpropagation and gradient descent until the loss function converges.

[0018] Furthermore, the energy storage supervision module is an autoencoder completed based on loss convergence; electromagnetic signal data is collected in real time, and energy storage supervision is performed based on the operating status of the electromagnetic signal data after feature extraction, preliminary clustering, local neighborhood support fusion and secondary clustering.

[0019] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0020] (1) In view of the problem that the general electromagnetic power supply and energy storage supervision system has the problem of false alarms and missed alarms due to the complex distribution of electromagnetic signals in the transmission lines and improper handling of the differences in signal distribution between different states, resulting in poor supervision effect, this scheme introduces a dynamic fuzzy factor to effectively distinguish abnormal signal samples from normal signal samples; dynamically allocates the feature weights of the signal to reduce the impact of environmental noise, flexibly divides the electromagnetic signal boundary samples by adaptively adjusting the threshold of the boundary samples, introduces a dynamic adjustment coefficient to adjust the direction of the restricted neighborhood, increases the constraint of the neighborhood density, and improves the accuracy of the abnormal signal support, thereby reducing the risk of false alarms and missed alarms and improving the accuracy of energy storage supervision.

[0021] (2) In view of the problem that the general electromagnetic power supply and energy storage supervision system has improper classification of abnormal states and ignores the differences in data characteristics in different time periods, resulting in poor accuracy in identifying abnormal states and poor energy storage supervision effects, this scheme introduces an adaptive smoothing factor to finely divide the states of electromagnetic signal data, constructs a time-weighted contrast loss to reduce the interference of non-critical time periods on the results, and introduces an abnormal penalty mechanism to improve the ability to identify abnormal signals, thereby improving the energy storage supervision effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 A schematic diagram of the process flow of a non-contact electromagnetic power supply and energy storage monitoring system for high-voltage transmission lines provided by the present invention.

[0023] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0024] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0025] In the description of the present invention, it should be understood that terms such as “upper”, “lower”, “front”, “back”, “left”, “right”, “top”, “bottom”, “inside” and “outside” indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore should not be understood as limiting the present invention.

[0026] Example 1, see Figure 1 The present invention provides a non-contact electromagnetic power supply and energy storage supervision system for high-voltage transmission lines, including a data acquisition module, a feature extraction module, a preliminary clustering module, a local neighborhood support fusion module, a secondary clustering module, a loss function construction module and an energy storage supervision module;

[0027] The data acquisition module collects historical electromagnetic signal data; and sends the data to the feature extraction module;

[0028] The feature extraction module uses a deep autoencoder to reduce the dimension of high-dimensional electromagnetic signal data, extract key features that reflect the operating status of the transmission line, and send the data to the preliminary clustering module;

[0029] The preliminary clustering module classifies the extracted feature data through a fuzzy clustering algorithm, preliminarily identifies electromagnetic signal samples in normal, abnormal and uncertain states; and sends the data to the local neighborhood support fusion module;

[0030] The local neighborhood support fusion module constructs a direction-restricted neighborhood with the boundary sample as the center, calculates the membership of the boundary sample by determining the support information of the sample in the neighborhood, and sends the data to the secondary clustering module;

[0031] The secondary clustering module jointly optimizes the membership matrix after the initial clustering and the local neighborhood support fusion, recalculates the cluster center and sample membership, and realizes the final classification; and sends the data to the loss function construction module;

[0032] The loss function construction module constructs an optimization objective function of comprehensive reconstruction, comparison and clustering losses and sends the data to the energy storage supervision module;

[0033] The energy storage supervision module dynamically monitors the power extraction efficiency and abnormal status of the energy storage system based on the operating status recognition and clustering results of the real-time electromagnetic signals, and issues an early warning.

[0034] Example 2, see Figure 1 This embodiment is based on the above embodiment. In the data acquisition module, the historical electromagnetic signal data includes transmission line signal data, environmental data, time series tags and operating status; the operating status includes normal operation and abnormal operation; the operating status is used as a data label; the data label does not participate in the clustering operation; the core dimensions of the signal data include power frequency electromagnetic signals, harmonic characteristics, electromagnetic field fluctuations and environmental noise signals; the specific operation is to deploy electromagnetic coupling coils around the high-voltage transmission line, sense the power frequency electric field of the transmission line, convert the induced voltage into direct current, and collect the induced signal through the electromagnetic coupling coil.

[0035] Example 3, see Figure 1 , this embodiment is based on the above embodiment, and the feature extraction module specifically uses a deep autoencoder to reduce the dimension of the collected high-dimensional electromagnetic signal and extract key features; the encoder feature extraction is expressed as: ; The decoder reconstructs the electromagnetic signal It is expressed as: ; Reconstruction error The calculation is expressed as: ; Wherein, Z is the feature extraction result of the encoder, reflecting the core features of the electromagnetic signal; F(·) is the nonlinear mapping function of the encoder; G(·) is the nonlinear mapping function of the decoder; GELU(·) is the GELU activation function; X is the collected high-dimensional electromagnetic signal data matrix; W and b are the weight matrix and bias vector of the encoder respectively; is the square of the L2 norm; the electromagnetic environment around the transmission line is complex, and the high-dimensional electromagnetic signal contains both normal power frequency signals and abnormal signals of line faults and load fluctuations. It is necessary to extract the core features that reflect the operating status of the transmission line from the complex signal.

[0036] Example 4, see Figure 1 This embodiment is based on the above embodiment. Specifically, the preliminary clustering module takes the extracted features as input and calculates the membership of each sample to each cluster center through the fuzzy clustering algorithm; according to the membership matrix, the electromagnetic signal samples are divided into three categories: normal state, abnormal state and uncertain state; an adaptive fuzzy factor is introduced , the membership matrix update is expressed as: ; ; ; Initial normal state set It is expressed as: ; Initial uncertain state set It is expressed as: ; Preliminary abnormal state set It is expressed as: ;Introducing dynamic threshold and dynamic adjustment coefficient , the labeled sample is represented as: ; ; ;in, is the i-th electromagnetic signal sample For the jth cluster center The degree of membership; is the kth cluster center; K1 is the total number of clusters, k is the total number of clusters index, and j is the total number of clusters index different from k; is a labeled sample; is the membership degree of the i-th electromagnetic signal sample to the k-th cluster center; It is a global data center; is the weighted Euclidean distance; D is the data dimension; and are the values ​​of the dth dimension of the sample and cluster center respectively; is the feature weight; is the importance prior score; exp(·) is the exponential function; is the basic threshold; is the local density; std(·) is the standard deviation; is the global mean of all sample memberships; is the control factor; is the global standard deviation of all sample memberships; is the value of the dth dimension of the kth cluster center; It will be The corresponding j value when the maximum value is taken is assigned to j; the normal operating status and abnormal status of the transmission line are identified through preliminary clustering, and the samples whose status cannot be determined are isolated to avoid false alarms and missed alarms.

[0037] By performing the above operations, in order to solve the problem that the general electromagnetic power supply and energy storage supervision system has complex electromagnetic signal distribution of transmission lines and improper handling of signal distribution differences between different states, which leads to false alarms and missed alarms and poor supervision effect, this scheme introduces dynamic fuzzy factors to effectively distinguish abnormal signal samples from normal signal samples; reduces the influence of environmental noise based on the dynamic allocation of signal feature weights, flexibly divides electromagnetic signal boundary samples by adaptively adjusting the threshold of boundary samples, introduces dynamic adjustment coefficients to adjust the direction of the restricted neighborhood, increases the constraints of neighborhood density, and improves the accuracy of abnormal signal support, thereby reducing the risk of false alarms and missed alarms and improving the accuracy of energy storage supervision.

[0038] Example 5, see Figure 1 , this embodiment is based on the above embodiment. The local neighborhood support fusion module specifically takes the boundary sample as the center, builds a direction-restricted neighborhood based on its distance to the cluster center, selects neighboring samples with consistent directions and satisfying distance conditions, and uses the membership information of the determined samples in the neighborhood to calculate the final membership of the boundary samples. The boundary samples are samples of the preliminary uncertain state set; introduces dynamic adjustment parameters , the distance constraint constructed by the hemisphere is expressed as: ; ; Direction constraints are expressed as: ; The basic probability distribution of local neighborhood support fusion is expressed as: ; Introducing neighborhood density constraints, neighborhood fusion is expressed as: ; ; The membership degree of the boundary sample is expressed as: ;in, is a sample of neighbors; is the direction angle, which constrains the direction consistency between neighbor samples and cluster centers; and are the vectors from boundary samples to neighbor samples and cluster centers respectively; is the cluster center of the neighbor sample pair The support degree reflects the matching degree between the electromagnetic signal characteristics and the operating status; is the complete set of neighbor sample pairs support level; is the membership of the neighbor samples to the cluster center; is the global support of the boundary samples to the cluster center, which is used to judge the running status; L is the number of neighborhood samples; is the support of samples in the neighborhood to the cluster center; K is the normalization factor; A and B are different operating states; and is the basic probability distribution value of the sample to the state; is the final membership of the boundary samples to the cluster center; Yes The standard deviation of all feature dimensions is calculated; d(·) is the Euclidean distance; in the operation of high-voltage transmission lines, samples of boundary states contain potential signals of abnormal states; directly ignoring the existence of noise samples will lead to false alarms; local neighborhood support fusion is used to perform refined analysis of boundary samples and improve the ability to accurately identify abnormal states.

[0039] Example 6, see Figure 1 This embodiment is based on the above embodiment. The secondary clustering module specifically substitutes the fused membership matrix into the fuzzy clustering algorithm to recalculate the cluster center and sample membership; introduces an adaptive smoothing factor , the final membership matrix It is expressed as: ; ; ; Final normal state set It is expressed as: ; Final abnormal state set It is expressed as: ;in, is the membership matrix of the samples obtained by preliminary clustering; is the membership matrix of the boundary samples obtained by fusion of local neighborhood supports; dens(·) is the density weight; is the number of preliminary clustering samples; The maximum value is taken; after the state of the boundary samples is redistributed through the local neighborhood support fusion, it is necessary to re-cluster them together with the determined samples to optimize the final classification effect.

[0040] Embodiment 7, see Figure 1 Based on the above embodiment, this embodiment has a loss function construction module that specifically includes: constructing a time-weighted contrast loss , expressed as: ; Introduce abnormal penalty and construct clustering loss , expressed as: ; The overall loss function LL is expressed as: ;in, is the true label of the sample; is the sample classification label; N is the total number of samples; is the weighted Euclidean distance; is the regularization coefficient; and is the loss weight; is the sample timestamp; It is the global time center; is the L2 norm; is the anomaly coefficient; is the abnormal threshold; Yes Take the natural logarithm; and optimize the weights of the autoencoder through back propagation and gradient descent until the loss function converges; used to optimize feature extraction, electromagnetic signal reconstruction and anomaly classification; improve the ability to detect abnormal signals of power extraction efficiency abnormalities and energy storage overload.

[0041] By performing the above operations, the general electromagnetic power storage supervision system has the problem of improper division of abnormal states and ignoring the differences in data characteristics in different time periods, resulting in poor accuracy in identifying abnormal states, and then leading to poor energy storage supervision effects. This scheme introduces an adaptive smoothing factor to finely divide the states of electromagnetic signal data, constructs a time-weighted contrast loss to reduce the interference of non-critical time periods on the results, and introduces an abnormal penalty mechanism to improve the ability to identify abnormal signals, thereby improving the energy storage supervision effect.

[0042] Embodiment 8, see Figure 1 This embodiment is based on the above embodiment. The energy storage supervision module is an autoencoder completed based on loss convergence. Electromagnetic signal data is collected in real time. After feature extraction, preliminary clustering, local neighborhood support fusion and secondary clustering, energy storage supervision is performed based on the operating status of the electromagnetic signal data. If the operating status of the electromagnetic signal data is abnormal, an early warning is issued to the management personnel.

[0043] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0044] While the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that many changes, modifications, substitutions and variations can be made to the embodiments without departing from the principles and spirit of the invention.

[0045] The present invention and its embodiments are described above, and such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and embodiment similar to the technical solution without creativity, which should belong to the protection scope of the present invention.

Claims

1. A non-contact electromagnetic power supply and energy storage monitoring system for high-voltage transmission lines, characterized by: The system includes data acquisition module, feature extraction module, preliminary clustering module, local neighborhood support fusion module, secondary clustering module, loss function construction module and energy storage supervision module; The data acquisition module acquires historical electromagnetic signal data; The feature extraction module uses a deep autoencoder to reduce the dimension of high-dimensional electromagnetic signal data and extract key features that reflect the operating status of the transmission line; The preliminary clustering module classifies the extracted feature data by using a fuzzy clustering algorithm to preliminarily identify electromagnetic signal samples in normal, abnormal and uncertain states; The local neighborhood support fusion module takes the boundary sample as the center, constructs a direction-restricted neighborhood based on its distance to the cluster center, selects neighboring samples with consistent directions and satisfying distance conditions, and uses the membership information of the determined samples in the neighborhood to calculate the final membership of the boundary samples; the boundary samples are samples of the preliminary uncertain state set; the determined samples are samples of the normal state set or the abnormal state set; The secondary clustering module jointly optimizes the membership matrix after the initial clustering and the local neighborhood support fusion, recalculates the cluster center and sample membership, and realizes the final classification; The loss function construction module constructs an optimization objective function that integrates reconstruction, contrast, and clustering losses; The energy storage supervision module dynamically monitors the power extraction efficiency and abnormal status of the energy storage system based on the operating status recognition and clustering results of the real-time electromagnetic signals, and issues an early warning.

2. According to claim 1, a non-contact electromagnetic power supply and energy storage monitoring system for high-voltage transmission lines is characterized by: The feature extraction module specifically comprises: using a deep autoencoder to reduce the dimension of the collected high-dimensional electromagnetic signal and extract key features; the encoder feature extraction is expressed as: ; Decoder reconstructs electromagnetic signal It is expressed as: ; Reconstruction error The calculation is expressed as: ; Wherein, Z is the feature extraction result of the encoder, reflecting the core features of the electromagnetic signal; F(·) is the nonlinear mapping function of the encoder; G(·) is the nonlinear mapping function of the decoder; GELU(·) is the GELU activation function; X is the collected high-dimensional electromagnetic signal data matrix; W and b are the weight matrix and bias vector of the encoder respectively; is the square of the L2 norm.

3. A non-contact electromagnetic power supply and energy storage monitoring system for high-voltage transmission lines according to claim 2, characterized in that: The local neighborhood support fusion module specifically includes: introducing a dynamic adjustment parameter , the distance constraint constructed by the hemisphere is expressed as: ; ; The direction constraint is expressed as: ; The basic probability distribution of local neighborhood support fusion is expressed as: ; Introducing neighborhood density constraints, neighborhood fusion is expressed as: ; ; The membership degree of the boundary sample is expressed as: ;in, is a sample of neighbors; is the direction angle, which constrains the direction consistency between neighbor samples and cluster centers; and are the vectors from boundary samples to neighbor samples and cluster centers respectively; is the cluster center of the neighbor sample pair The support degree reflects the matching degree between the electromagnetic signal characteristics and the operating status; is the complete set of neighbor sample pairs support level; is the membership of the neighbor samples to the cluster center; is the global support of the boundary samples to the cluster center, which is used to judge the running status; L is the number of neighborhood samples; is the support of samples in the neighborhood to the cluster center; K is the normalization factor; A and B are different operating states; and is the basic probability distribution value of the sample to the state; is the final membership of the boundary samples to the cluster center; Yes The standard deviation of all feature dimensions is calculated; d(·) is the Euclidean distance.

4. A non-contact electromagnetic power supply and energy storage monitoring system for high-voltage transmission lines according to claim 3, characterized in that: The secondary clustering module specifically includes: substituting the fused membership matrix into the fuzzy clustering algorithm to recalculate the cluster center and sample membership; introducing an adaptive smoothing factor , the final membership matrix It is expressed as: ; ; ; Final normal state set It is expressed as: ; Final abnormal state set It is expressed as: ;in, is the membership matrix of the samples obtained by preliminary clustering; is the membership matrix of the boundary samples obtained by fusion of local neighborhood supports; dens(·) is the density weight; is the number of preliminary clustering samples; is to take the maximum value.

5. A non-contact electromagnetic power supply and energy storage monitoring system for high-voltage transmission lines according to claim 4, characterized in that: The loss function construction module is specifically: constructing time-weighted contrast loss , expressed as: ; Introduce abnormal penalty and construct clustering loss , expressed as: ; Overall loss function It is expressed as: ;in, is the true label of the sample; is the sample classification label; N is the total number of samples; is the weighted Euclidean distance; is the regularization coefficient; and is the loss weight; is the sample timestamp; It is the global time center; is the L2 norm; is the anomaly coefficient; is the abnormal threshold; Yes Take the natural logarithm; and optimize the weights of the autoencoder through backpropagation and gradient descent until the loss function converges.

6. A non-contact electromagnetic power supply and energy storage monitoring system for high-voltage transmission lines according to claim 5, characterized in that: The preliminary clustering module specifically takes the extracted features as input, calculates the membership of each sample to each cluster center through the fuzzy clustering algorithm; divides the electromagnetic signal samples into three categories according to the membership matrix: normal state, abnormal state and uncertain state; introduces the adaptive fuzzy factor , the membership matrix update is expressed as: ; ; ; Initial normal state set It is expressed as: ; Initial uncertain state set It is expressed as: ; Preliminary abnormal state set It is expressed as: ;Introducing dynamic threshold and dynamic adjustment coefficient , the labeled sample is represented as: ; ; ;in, is the i-th electromagnetic signal sample For the jth cluster center The degree of membership; is the kth cluster center; K1 is the total number of clusters, k is the total number of clusters index, and j is the total number of clusters index different from k; is a labeled sample; is the membership degree of the i-th electromagnetic signal sample to the k-th cluster center; It is a global data center; is the weighted Euclidean distance; D is the data dimension; and are the values ​​of the dth dimension of the sample and cluster center respectively; is the feature weight; is the importance prior score; exp(·) is the exponential function; is the basic threshold; is the local density; std(·) is the standard deviation; is the global mean of all sample memberships; is the control factor; is the global standard deviation of all sample memberships; is the value of the dth dimension of the kth cluster center; It will be When the maximum value is taken, the corresponding j is assigned to j.

7. A non-contact electromagnetic power supply and energy storage monitoring system for high-voltage transmission lines according to claim 6, characterized in that: In the data acquisition module, the historical electromagnetic signal data includes transmission line signal data, environmental data, time series tags and operating status; the operating status includes normal operation and abnormal operation; the operating status is used as a data label; the data label does not participate in clustering operations.

8. A non-contact electromagnetic power supply and energy storage monitoring system for high-voltage transmission lines according to claim 7, characterized in that: The energy storage supervision module is an autoencoder completed based on loss convergence; it collects electromagnetic signal data in real time, and performs energy storage supervision based on the operating status of the electromagnetic signal data after feature extraction, preliminary clustering, local neighborhood support fusion and secondary clustering.

Images