A method and system for intelligent monitoring and control of high-voltage distribution cabinet status

Abstract

The present invention discloses a method and system for intelligent monitoring and control of the status of a high-voltage distribution cabinet, which relates to the technical field of power equipment control. The partial discharge signal of the target device is collected in real time and denoised to obtain a corrected signal; the partial discharge signal includes a transient ground voltage signal and an ultrasonic signal; the corrected signal is used as the input of a pre-trained state evaluation model to obtain the state type; and control decisions are made based on the state type. By eliminating the interference of noise on the partial discharge signal, potential fault information of the electrical equipment can be more accurately captured and identified, thereby making accurate control decisions, avoiding the expansion of the fault, and ensuring the safe operation of the power system.

Description

Technical Field

[0001] The present invention relates to the technical field of power equipment control, and in particular to a method and system for intelligently monitoring and controlling the status of a high-voltage distribution cabinet. Background Art

[0002] In high-voltage power distribution systems, distribution cabinets, as key electrical equipment, undertake the crucial task of transmitting and distributing electrical energy. Their stable operation is directly related to the safety and reliability of the power system. Insulation is an integral component of high-voltage electrical equipment, and even minor damage can lead to latent defects. As insulation gradually deteriorates, defects can expand until equipment failure occurs, seriously impacting the normal operation of the power system.

[0003] Partial discharge (PD) is considered the root cause of insulation breakdown. Because PD signals from different types of insulation defects differ, the cause of the PD can be inferred from the PD signal, effectively detecting the type of defect present in electrical equipment. With the development of intelligent technology, PD signals have gained increasing attention as a key precursor to faults. Monitoring PD signals can promptly capture potential electrical faults in distribution cabinets, providing a crucial basis for fault early warning. However, PD signals collected in the field are often coupled with noise or interference, such as periodic narrowband interference and white noise. Severe noise can overwhelm the PD signal, weakening its expressive power and leading to inaccurate status recognition, which can cause protective devices to malfunction or fail to operate. Summary of the Invention

[0004] The purpose of the present invention is to solve the problem of misoperation or refusal of protection devices mentioned in the above background technology, and to propose a method and system for intelligent monitoring and control of the status of a high-voltage distribution cabinet.

[0005] A first aspect of the present invention provides a method for intelligently monitoring and controlling the status of a high-voltage distribution cabinet, the method comprising:

[0006] Real-time acquisition of partial discharge signals of target devices; the partial discharge signals include transient voltage signals and ultrasonic signals;

[0007] performing denoising processing on the partial discharge signal to obtain a corrected signal;

[0008] Using the correction signal as input to a pre-trained state evaluation model to obtain a state type;

[0009] A control decision is made according to the state type; the control decision includes load adjustment, operation mode adjustment and fault isolation.

[0010] Optionally, the performing denoising on the partial discharge signal to obtain a corrected signal includes:

[0011] performing variational modal decomposition on a first target signal to obtain a plurality of modal components; the first target signal is any one of a transient voltage signal and an ultrasonic signal;

[0012] Calculating the kurtosis of each modal component, including the modal components with kurtosis values greater than a preset threshold into the first component set, and including the modal components with kurtosis values not greater than the preset threshold into the second component set;

[0013] Using a wavelet threshold denoising method to filter the first component set, and using a triple standard deviation criterion to reversely suppress the second component set to obtain a denoised modal component set;

[0014] Signal reconstruction is performed based on the filtered modal component set to obtain a second target signal as a correction signal of the first target signal.

[0015] Optionally, performing variational modal decomposition on the first target signal to obtain multiple modal components includes:

[0016] Performing empirical mode decomposition on the first target signal to determine a decomposition layer number K as the number of layers of variational mode decomposition;

[0017] A weighted fusion of envelope entropy and reconstruction error is used as the objective function, and the penalty factor α of the variational mode decomposition algorithm is optimized with the goal of minimizing the objective function to obtain the penalty factor α;

[0018] According to the decomposition layer number K and the penalty factor α, variational modal decomposition is performed on the first target signal to obtain multiple modal components.

[0019] Optionally, filtering the first component set by using a wavelet threshold denoising method includes:

[0020] Performing a wavelet transform on the target modal component to obtain wavelet coefficients; the wavelet coefficients include approximate coefficients and detail coefficients; the target modal component is any modal component in the first component set;

[0021] An improved threshold function is used to perform threshold processing on the detail coefficients; specifically:

[0022] in, is the wavelet threshold; median() means taking the median; is the detail coefficient of the j-th level scale; is the length of the detail coefficient at the j-th level; is the detail coefficient after threshold processing; sgn() is the sign function; n is a constant;

[0023] The approximate coefficients and the detail coefficients after threshold processing are subjected to inverse wavelet transform to obtain denoised modal components.

[0024] Optionally, the state evaluation model consists of a first feature extraction branch, a second feature extraction branch, an attention module, and a classifier; wherein:

[0025] The first feature extraction branch is configured to extract features from the correction signal of the transient ground voltage signal using a gated recurrent unit to obtain a first feature;

[0026] The second feature extraction branch is used to extract features of the corrected signal of the ultrasonic signal using a gated recurrent unit to obtain a second feature;

[0027] The attention module is used to fuse and enhance the combined features of the first feature and the second feature using a multi-head self-attention mechanism to obtain a third feature;

[0028] The classifier is used to process the third feature using a fully connected network and output a state type.

[0029] A second aspect of the present invention provides an intelligent monitoring and control system for a high-voltage distribution cabinet state, the system comprising:

[0030] A sensor monitoring module is used to collect partial discharge signals of target equipment in real time; the partial discharge signals include transient voltage signals and ultrasonic signals;

[0031] A signal denoising module, configured to perform denoising processing on the partial discharge signal to obtain a corrected signal;

[0032] an abnormality identification module, configured to use the correction signal as input to a pre-trained state evaluation model to obtain a state type;

[0033] The action generation module is used to make control decisions according to the state type; the control decisions include load adjustment, operation mode adjustment and fault isolation.

[0034] Optionally, the signal denoising module includes:

[0035] a signal decomposition module, configured to perform variational modal decomposition on a first target signal to obtain a plurality of modal components; the first target signal is any one of a transient voltage signal and an ultrasonic signal;

[0036] A leading classification module, configured to calculate the kurtosis of each modal component, include modal components with kurtosis values greater than a preset threshold into a first component set, and include modal components with kurtosis values not greater than the preset threshold into a second component set;

[0037] an item-by-item denoising module, configured to filter the first component set using a wavelet threshold denoising method, and perform reverse suppression on the second component set using a triple standard deviation criterion, to obtain a denoised modal component set;

[0038] The signal reconstruction module is used to reconstruct the signal according to the filtered modal component set to obtain a second target signal as a correction signal of the first target signal.

[0039] Optionally, the signal decomposition module includes:

[0040] a layer number determination module, configured to perform empirical mode decomposition on the first target signal and determine the number of decomposition layers K as the number of layers of variational mode decomposition;

[0041] a penalty factor determination module, configured to optimize the penalty factor α of the variational mode decomposition algorithm by adopting a weighted fusion of envelope entropy and reconstruction error as an objective function and minimizing the objective function to obtain the penalty factor α;

[0042] A decomposition execution module is used to perform variational modal decomposition on the first target signal according to the decomposition layer number K and the penalty factor α to obtain multiple modal components.

[0043] Optionally, the item-by-item denoising module includes:

[0044] A wavelet transform module is used to perform wavelet transform on the target modal component to obtain wavelet coefficients; the wavelet coefficients include approximate coefficients and detail coefficients; the target modal component is any modal component in the first component set;

[0045] The threshold processing module is used to perform threshold processing on the detail coefficients using an improved threshold function; specifically:

[0046] in, is the wavelet threshold; median() means taking the median; is the detail coefficient of the j-th level scale; is the length of the detail coefficient at the j-th level; is the detail coefficient after threshold processing; sgn() is the sign function; n is a constant;

[0047] The inverse wavelet transform module is used to perform inverse wavelet transform on the approximate coefficients and the detail coefficients after threshold processing to obtain denoised modal components.

[0048] Optionally, the state evaluation model consists of a first feature extraction branch, a second feature extraction branch, an attention module, and a classifier; wherein:

[0049] The first feature extraction branch is configured to extract features from the correction signal of the transient ground voltage signal using a gated recurrent unit to obtain a first feature;

[0050] The second feature extraction branch is used to extract features of the corrected signal of the ultrasonic signal using a gated recurrent unit to obtain a second feature;

[0051] The attention module is used to fuse and enhance the combined features of the first feature and the second feature using a multi-head self-attention mechanism to obtain a third feature;

[0052] The classifier is used to process the third feature using a fully connected network and output a state type.

[0053] Beneficial effects of the present invention:

[0054] The present invention proposes a method for intelligent monitoring and control of the state of a high-voltage distribution cabinet, the method comprising: real-time acquisition of partial discharge signals of target equipment; the partial discharge signals include transient ground voltage signals and ultrasonic signals;

[0055] De-noising the partial discharge signal to obtain a corrected signal;

[0056] The correction signal is used as the input of the pre-trained state evaluation model to obtain the state type;

[0057] Based on the status type, control decisions are made; control decisions include load adjustment, operation mode adjustment, and fault isolation.

[0058] By collecting partial discharge signals in real time and performing denoising processing to eliminate the interference of noise on partial discharge signals, it is possible to more accurately capture and identify potential fault information of electrical equipment, thereby making accurate control decisions, avoiding the expansion of faults, and ensuring the safe operation of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 A flow chart of a method for intelligent monitoring and control of the status of a high-voltage distribution cabinet is provided for an embodiment of the present invention;

[0060] Figure 2 A schematic diagram of the structure of a state evaluation model is provided for an embodiment of the present invention;

[0061] Figure 3 The present invention provides an architecture diagram of an intelligent monitoring and control system for a high-voltage distribution cabinet state. DETAILED DESCRIPTION

[0062] 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. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0063] The embodiment of the present invention provides a method for intelligent monitoring and control of the status of a high-voltage power distribution cabinet. Figure 1 , Figure 1 A flowchart of a method for intelligent monitoring and control of the status of a high-voltage distribution cabinet provided by an embodiment of the present invention. The method comprises the following steps:

[0064] S101, collecting partial discharge signals of a target device in real time.

[0065] S102: De-noising the partial discharge signal to obtain a corrected signal.

[0066] S103: Using the corrected signal as input to a pre-trained state evaluation model to obtain a state type.

[0067] S104: Make a control decision based on the state type.

[0068] Among them, the partial discharge signal includes the transient earth voltage signal TEV and the ultrasonic signal US; the control decision includes load adjustment, operation mode adjustment and fault isolation.

[0069] An intelligent monitoring and control method for the status of a high-voltage distribution cabinet provided in an embodiment of the present invention collects partial discharge signals in real time and performs denoising processing to eliminate noise interference on the partial discharge signals. This method can more accurately capture and identify potential fault information of electrical equipment, thereby making accurate control decisions, avoiding the expansion of faults, and ensuring the safe operation of the power system.

[0070] In one embodiment, step S102 includes:

[0071] Step 1: Perform variational modal decomposition on the first target signal to obtain multiple modal components.

[0072] Step 2: Calculate the kurtosis of each modal component, include the modal components with kurtosis values greater than a preset threshold into the first component set, and include the modal components with kurtosis values not greater than the preset threshold into the second component set.

[0073] Step three: use the wavelet threshold denoising method to filter the first component set, and use the triple standard deviation criterion to perform reverse suppression on the second component set to obtain the denoised modal component set.

[0074] Step 4: Reconstruct the signal based on the filtered modal component set to obtain a second target signal as a correction signal of the first target signal.

[0075] The first target signal is either a transient voltage signal or an ultrasonic signal.

[0076] In one embodiment, variational modal decomposition is performed on the first target signal to obtain multiple modal components including:

[0077] By using algorithms such as particle swarm optimization or sparrow search algorithm, the number of decomposition layers K and penalty factor α are jointly optimized to obtain the optimal combination, and variational mode decomposition is performed on the first target signal based on the optimal combination. Specifically, in various optimization algorithms, the fitness function can adopt a weighted fusion of envelope entropy and reconstruction error: .in, is the fitness function; H is the normalized envelope entropy; E is the normalized reconstruction error; and is the weight coefficient, .

[0078] In one implementation, considering the computational efficiency of real-time monitoring, the first target signal can be subjected to empirical mode decomposition first to determine the number of decomposition layers K, which will be used as the number of layers for variational mode decomposition. After the number of layers is determined, the penalty factor α of the variational mode decomposition algorithm is optimized with the goal of minimizing the above-mentioned fitness function to obtain the penalty factor α. Finally, based on the number of decomposition layers K and the penalty factor α, the first target signal is subjected to variational mode decomposition to obtain multiple modal components. Empirical mode decomposition (EMD) is an adaptive signal processing method that can automatically select the number of decomposition layers based on the characteristics of the signal without the need for pre-setting. Determining the number of decomposition layers through EMD can effectively avoid the problems of over-decomposition due to too many layers or under-decomposition due to too few layers, thereby enabling the subsequent variational mode decomposition (VMD) algorithm to better capture the key features of the signal.

[0079] In one embodiment, the preset threshold can be set to 3. The kurtosis of the normal distribution is about 3. The PD signal interval deviates significantly from the normal distribution, and its kurtosis value is much greater than 3. The white noise and narrowband interference signal intervals obey the normal distribution, and their kurtosis values are less than or equal to 3, so 3 is used as the dividing line, and the components with kurtosis values greater than 3 are regarded as modal components dominated by partial discharge signals, and the components with kurtosis values not greater than 3 are regarded as modal components dominated by noise. The two modal components are included in different sets and filtered using different denoising methods. Specifically, for the first component set, that is, the modal components dominated by partial discharge signals, the wavelet threshold denoising method is used for processing, which can effectively extract the partial discharge characteristic signals in the components. For the second component set, that is, the modal components dominated by noise signals, the triple standard deviation criterion is used for reverse suppression, which can effectively remove the noise characteristic signals in the components. This classification processing method can retain the partial discharge characteristics while effectively removing noise, thereby improving the purity and feature fidelity of the signal.

[0080] In one embodiment, filtering the first component set using a wavelet threshold denoising method includes:

[0081] Step 1: Perform wavelet transform on the target modal component to obtain wavelet coefficients; the wavelet coefficients include approximate coefficients and detail coefficients.

[0082] Step 2: Use the improved threshold function to perform threshold processing on the detail coefficients; specifically:

[0083] in, is the wavelet threshold; median() means taking the median; is the detail coefficient of the j-th level scale; is the length of the detail coefficient at the j-th level; is the detail coefficient after threshold processing; sgn() is the sign function; n is a constant.

[0084] Step 3: Perform inverse wavelet transform on the approximate coefficients and detail coefficients after threshold processing to obtain the denoised modal components.

[0085] The target modal component is any modal component in the first component set.

[0086] In one implementation, conventional threshold functions include soft threshold functions and hard threshold functions. The threshold operation of the hard threshold is discontinuous, which can cause the denoised signal to appear abrupt or uneven. In particular, at transient changes in the signal, this may cause large fluctuations, affecting the smoothness of the signal. The soft threshold function can achieve smooth noise reduction, but it may introduce amplitude scaling, thereby distorting the original signal. In this embodiment, the parameter n is set to 2, which can maintain continuity and smoothness at the threshold point while minimizing the amplitude distortion introduced by the soft threshold function.

[0087] In one embodiment, the triple standard deviation criterion is used to reversely suppress the second component set. Specifically, the following steps are performed: Calculate the mean and standard deviation of the second target modal component, subtract the mean from the second modal component, take the absolute value, set points within triple the standard deviation to zero, and retain the remaining points. This allows the partial discharge characteristic signal to be extracted from the modal component dominated by the noise signal.

[0088] In one embodiment, the state assessment model may be a machine learning model, such as a support vector machine.

[0089] In one embodiment, the state evaluation model may be a neural network model. For details, see Figure 2 , Figure 2 This is a schematic diagram of the structure of a state evaluation model provided by an embodiment of the present invention. The state evaluation model consists of a first feature extraction branch FFE, a second feature extraction branch SFE, an attention module AM, and a classifier Cls; wherein:

[0090] The first feature extraction branch is used to extract features of the correction signal of the transient ground voltage signal using a gated recurrent unit (GRU) to obtain a first feature.

[0091] The second feature extraction branch is used to extract features from the corrected signal of the ultrasonic signal using a gated recurrent unit (GRU) to obtain a second feature. Specifically, the second feature has the same dimension as the first feature.

[0092] The attention module uses a multi-head self-attention mechanism to fuse and enhance the joint features of the first and second features to obtain the third feature. Specifically, the first and second features are concatenated to form a joint feature matrix. The multi-head attention module (MHA) calculates attention weights based on the joint feature matrix and outputs the weighted results. The output matrix is fed into the subsequent feedforward network (FFN) for further processing to obtain the third feature.

[0093] The classifier is used to process the third feature using a fully connected network and output the state type.

[0094] By combining multimodal signals, the model can adapt to more complex equipment monitoring environments and provide more accurate diagnostic support.

[0095] In one embodiment, status types can be categorized as normal, slightly abnormal, moderately abnormal, and severely abnormal. For normal status, no adjustments are required. For any abnormality, abnormality information is uploaded to an early warning system. Specifically, for slightly abnormal conditions, load adjustments can be made to reduce the frequency of partial discharges. For moderate abnormal conditions, the equipment can be switched to low-power or standby mode to prevent further development of partial discharges. For severe abnormal conditions, system fault isolation can be implemented, transferring the load to a backup system.

[0096] The embodiment of the present invention provides a high-voltage distribution cabinet state intelligent monitoring and control system. Figure 3 , Figure 3 This is an architecture diagram of an intelligent monitoring and control system for high-voltage distribution cabinet status provided by an embodiment of the present invention. The system includes:

[0097] The sensor monitoring module is used to collect the partial discharge signal of the target device in real time.

[0098] The signal denoising module is used to denoise the partial discharge signal to obtain a corrected signal.

[0099] The anomaly recognition module is used to take the correction signal as the input of the pre-trained state evaluation model to obtain the state type.

[0100] The action generation module is used to make control decisions based on the state type.

[0101] Among them, partial discharge signals include transient ground voltage signals and ultrasonic signals; control decisions include load adjustment, operation mode adjustment and fault isolation.

[0102] An intelligent monitoring and control method for the status of a high-voltage distribution cabinet provided in an embodiment of the present invention collects partial discharge signals in real time and performs denoising processing to eliminate noise interference on the partial discharge signals. This method can more accurately capture and identify potential fault information of electrical equipment, thereby making accurate control decisions, avoiding the expansion of faults, and ensuring the safe operation of the power system.

[0103] In one embodiment, the signal denoising module includes:

[0104] The signal decomposition module is used to perform variational modal decomposition on a first target signal to obtain multiple modal components. The first target signal is either a transient voltage signal or an ultrasonic signal.

[0105] The dominant classification module is used to calculate the kurtosis of each modal component, include the modal components with kurtosis values greater than a preset threshold into the first component set, and include the modal components with kurtosis values not greater than the preset threshold into the second component set.

[0106] The item-by-item denoising module is used to filter the first component set using a wavelet threshold denoising method, and to reversely suppress the second component set using a triple standard deviation criterion to obtain a denoised modal component set.

[0107] The signal reconstruction module is used to reconstruct the signal according to the filtered modal component set to obtain a second target signal as a correction signal of the first target signal.

[0108] In one embodiment, the signal decomposition module includes:

[0109] The layer number determination module is used to perform empirical mode decomposition on the first target signal and determine the decomposition layer number K as the layer number of variational mode decomposition.

[0110] The penalty factor determination module is used to adopt the weighted fusion of envelope entropy and reconstruction error as the objective function, minimize the objective function, optimize the penalty factor α of the variational mode decomposition algorithm, and obtain the penalty factor α.

[0111] The decomposition execution module is used to perform variational modal decomposition on the first target signal according to the decomposition layer number K and the penalty factor α to obtain multiple modal components.

[0112] In one embodiment, the item-by-item denoising module includes:

[0113] The wavelet transform module is used to perform wavelet transform on the target modal component to obtain wavelet coefficients. The wavelet coefficients include approximate coefficients and detail coefficients. The target modal component is any modal component in the first component set.

[0114] The threshold processing module is used to perform threshold processing on the detail coefficients using an improved soft threshold function; specifically:

[0115] in, is the wavelet threshold; median() means taking the median; is the detail coefficient of the j-th level scale; is the length of the detail coefficient at the j-th level; is the detail coefficient after threshold processing; sgn() is the sign function; n is a constant.

[0116] The inverse wavelet transform module is used to perform inverse wavelet transform on the approximate coefficients and the detail coefficients after threshold processing to obtain the denoised modal components.

[0117] It should be noted that, in this document, terms such as "comprises", "includes" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or apparatus that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements that are inherent to such process, method, article or apparatus.

[0118] The embodiments of the present invention are described in detail above, but the contents described are only preferred embodiments of the present invention and should not be considered to limit the scope of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. A method for intelligent monitoring and control of high-voltage distribution cabinet status, characterized in that: The method comprises: Real-time acquisition of partial discharge signals of target devices; the partial discharge signals include transient voltage signals and ultrasonic signals; performing denoising processing on the partial discharge signal to obtain a corrected signal; Using the correction signal as input to a pre-trained state evaluation model to obtain a state type; Making control decisions based on the state type; the control decisions include load adjustment, operation mode adjustment, and fault isolation; The performing denoising on the partial discharge signal to obtain a corrected signal includes: performing variational modal decomposition on a first target signal to obtain a plurality of modal components; the first target signal is any one of a transient voltage signal and an ultrasonic signal; Calculating the kurtosis of each modal component, including the modal components with kurtosis values greater than a preset threshold into the first component set, and including the modal components with kurtosis values not greater than the preset threshold into the second component set; Using a wavelet threshold denoising method to filter the first component set, and using a triple standard deviation criterion to reversely suppress the second component set to obtain a denoised modal component set; Reconstructing a signal based on the filtered modal component set to obtain a second target signal as a correction signal of the first target signal; The performing variational modal decomposition on the first target signal to obtain multiple modal components includes: Performing empirical mode decomposition on the first target signal to determine a decomposition layer number K as the number of layers of variational mode decomposition; A weighted fusion of envelope entropy and reconstruction error is used as the objective function, and the penalty factor α of the variational mode decomposition algorithm is optimized with the goal of minimizing the objective function to obtain the penalty factor α; performing variational modal decomposition on the first target signal according to the decomposition layer number K and the penalty factor α to obtain a plurality of modal components; The state evaluation model consists of a first feature extraction branch, a second feature extraction branch, an attention module and a classifier; wherein: A first feature extraction branch is used to extract features of the correction signal of the transient ground voltage signal using a gated recurrent unit to obtain a first feature; The second feature extraction branch is used to extract features of the corrected signal of the ultrasonic signal using a gated recurrent unit to obtain a second feature; specifically, the second feature and the first feature have the same dimension; The attention module is used to fuse and enhance the joint features of the first and second features using a multi-head self-attention mechanism to obtain a third feature. Specifically, the first and second features are concatenated to obtain a joint feature matrix. The multi-head attention module is used to calculate the attention weights based on the joint feature matrix and output the weighted matrix. The output matrix is sent to the subsequent feedforward network for further processing to obtain the third feature. The classifier is used to process the third feature using a fully connected network and output the state type.

2. The method for intelligent monitoring and control of high-voltage distribution cabinet status according to claim 1, characterized in that: The filtering of the first component set by using the wavelet threshold denoising method includes: Performing a wavelet transform on the target modal component to obtain wavelet coefficients; the wavelet coefficients include approximate coefficients and detail coefficients; the target modal component is any modal component in the first component set; An improved threshold function is used to perform threshold processing on the detail coefficients; specifically: ;in, is the wavelet threshold; median() means taking the median; is the detail coefficient of the j-th level scale; is the length of the detail coefficient at the j-th level; is the detail coefficient after threshold processing; sgn() is the sign function; n is a constant; The approximate coefficients and the detail coefficients after threshold processing are subjected to inverse wavelet transform to obtain denoised modal components.

3. A high-voltage distribution cabinet state intelligent monitoring and control system, characterized in that: The system comprises: A sensor monitoring module is used to collect partial discharge signals of target equipment in real time; the partial discharge signals include transient voltage signals and ultrasonic signals; A signal denoising module, configured to perform denoising processing on the partial discharge signal to obtain a corrected signal; an abnormality identification module, configured to use the correction signal as input to a pre-trained state evaluation model to obtain a state type; An action generation module, configured to make control decisions based on the state type; the control decisions include load adjustment, operation mode adjustment, and fault isolation; The signal denoising module includes: a signal decomposition module, configured to perform variational modal decomposition on a first target signal to obtain a plurality of modal components; the first target signal is any one of a transient voltage signal and an ultrasonic signal; A leading classification module, configured to calculate the kurtosis of each modal component, include modal components with kurtosis values greater than a preset threshold into a first component set, and include modal components with kurtosis values not greater than the preset threshold into a second component set; an item-by-item denoising module, configured to filter the first component set using a wavelet threshold denoising method, and perform reverse suppression on the second component set using a triple standard deviation criterion, to obtain a denoised modal component set; a signal reconstruction module, configured to reconstruct a signal based on the filtered modal component set to obtain a second target signal as a correction signal of the first target signal; The signal decomposition module includes: a layer number determination module, configured to perform empirical mode decomposition on the first target signal and determine the number of decomposition layers K as the number of layers of variational mode decomposition; a penalty factor determination module, configured to optimize the penalty factor α of the variational mode decomposition algorithm by adopting a weighted fusion of envelope entropy and reconstruction error as an objective function and minimizing the objective function to obtain the penalty factor α; a decomposition execution module, configured to perform variational modal decomposition on the first target signal according to the decomposition layer number K and the penalty factor α to obtain a plurality of modal components; The state evaluation model consists of a first feature extraction branch, a second feature extraction branch, an attention module and a classifier; wherein: A first feature extraction branch is used to extract features of the correction signal of the transient ground voltage signal using a gated recurrent unit to obtain a first feature; The second feature extraction branch is used to extract features of the corrected signal of the ultrasonic signal using a gated recurrent unit to obtain a second feature; specifically, the second feature and the first feature have the same dimension; The attention module is used to fuse and enhance the joint features of the first and second features using a multi-head self-attention mechanism to obtain a third feature. Specifically, the first and second features are concatenated to obtain a joint feature matrix. The multi-head attention module is used to calculate the attention weights based on the joint feature matrix and output the weighted matrix. The output matrix is sent to the subsequent feedforward network for further processing to obtain the third feature. The classifier is used to process the third feature using a fully connected network and output the state type.

4. The intelligent monitoring and control system for high-voltage distribution cabinet status according to claim 3 is characterized in that: The item-by-item denoising module includes: A wavelet transform module is used to perform wavelet transform on the target modal component to obtain wavelet coefficients; the wavelet coefficients include approximate coefficients and detail coefficients; the target modal component is any modal component in the first component set; The threshold processing module is used to perform threshold processing on the detail coefficients using an improved threshold function; specifically: in, is the wavelet threshold; median() means taking the median; is the detail coefficient of the j-th level scale; is the length of the detail coefficient at the j-th level; is the detail coefficient after threshold processing; sgn() is the sign function; n is a constant; The inverse wavelet transform module is used to perform inverse wavelet transform on the approximate coefficients and the detail coefficients after threshold processing to obtain denoised modal components.

Images