High-voltage switch cabinet with automatic alarm module and method

Abstract

The invention provides a high-voltage switch cabinet with an automatic alarm module and a method, the high-voltage switch cabinet comprises a cabinet body, the lower end of the cabinet body is provided with a plurality of supporting legs, the bottom of the cabinet body is provided with an air inlet port, the lower part of the cabinet body is provided with an anti-blocking air inlet module used for air inlet, and the anti-blocking air inlet module continuously feeds air to the interior of the cabinet body. An exhaust pipeline is arranged at the top of the cabinet body, an exhaust box is arranged at the upper end of the exhaust pipeline, gas outlets are distributed in the outer side of the exhaust box, a gas detection module for detecting gas information is arranged in the exhaust box, and an anti-blocking gas inlet mode is constructed so that heat in the cabinet body can be taken away in time; the air flow is timely sent to the detection position, the response speed is improved, the cleaning effect is ensured, the reliability of the equipment is improved, and the fault diagnosis accuracy is improved and the fault prediction is advanced through gas, image and audio multi-mode information fusion and combination of the neural network and the LSTM model.

Description

Technical Field

[0001] The present invention relates to the technical field of switch cabinets, and in particular to a high-voltage switch cabinet with an automatic alarm module and a method. Background Art

[0002] High-voltage switchgear refers to a device used for switching, controlling, or protecting power generation, transmission, distribution, energy conversion, and consumption in power systems. In existing switchgear, a cloth dust filter is installed at the ventilation port of the ventilation duct to prevent dust from being drawn into the cabinet and causing pollution. However, it is worth considering that as the use time increases, dust will accumulate on the cloth dust filter, resulting in a decrease in the airflow entering the cabinet through the cloth dust filter, affecting the heat dissipation effect. The patent with authorization announcement number CN119182052B discloses a dust-proof high-voltage switchgear, which clears the clogged dust by shaking the filter. However, in reality, this clog is adhered and tightened, making it difficult to clear the hole by shaking. In addition, when an internal fault occurs, the airflow is not circulated, resulting in a lag in information collection and a delay in triggering the alarm module, which will cause greater subsequent losses. Summary of the Invention

[0003] The object of the present invention is to provide a high-voltage switch cabinet with an automatic alarm module and a method thereof, so as to solve the problem in the prior art of dust blockage, which hinders heat dissipation and alarm.

[0004] To achieve the above object, the present invention provides the following technical solutions:

[0005] A high-voltage switchgear with an automatic alarm module includes a cabinet body, an air inlet port is provided at the bottom of the cabinet body, and an anti-blocking air inlet module for air intake is provided under the cabinet body. Air is continuously taken in by the anti-blocking air inlet module to supply air to the inside of the cabinet body for cooling. An exhaust duct is provided on the top of the cabinet body, an exhaust box is provided at the upper end of the exhaust duct, and air outlets are distributed on the outside of the exhaust box. A gas detection module for detecting gas information is provided inside the exhaust box, and the gas detection module is electrically connected to the alarm module. When the gas detection module detects abnormal gas in the air, the alarm module will issue an alarm message to reduce damage to the switchgear. A cabinet door is slidably fitted at the opening position of the cabinet body, and a control panel is provided on the cabinet door. An image acquisition module for obtaining picture information of parts inside the cabinet body is provided on the control panel. An audio acquisition module for detecting sound is also provided inside the cabinet body.

[0006] On the basis of the above technical solutions, the present invention also provides the following optional technical solutions:

[0007] In an optional solution, the exhaust pipe has a tapered structure with a larger lower end and a smaller upper end. The tapered structure helps to compress the gas, thereby increasing the concentration of the target gas and facilitating detection.

[0008] In an optional solution: the image acquisition module includes a mounting groove arranged on the inner side of the control panel, a camera slide is slidingly provided in the mounting groove, reciprocating guide rods are slidingly passed through both ends of the camera slide, the upper and lower ends of the reciprocating guide rods are fixedly connected to the inner wall of the mounting groove, a plurality of cameras for acquiring image information are provided on the surface of the camera slide, a reciprocating screw is threadedly provided at the center position of the camera slide, the upper end of the reciprocating screw is rotatably connected to the inner wall of the mounting groove, and the lower end of the reciprocating screw is fixedly connected to the output end of the reciprocating motor.

[0009] In an optional solution: the anti-blocking air intake module includes a horizontally arranged air intake duct, the upper side of the air intake duct is connected to the air intake port through a connecting duct, a drive motor is fixedly provided at the end of the air intake duct, a drive shaft is fixedly provided at the output end of the drive motor, a fan blade is provided at the output end of the drive shaft, and a filter unit for filtering air is provided at the port position of the air intake duct.

[0010] In an optional solution: a collecting channel is provided on the lower side of the air intake pipe where the conical filter is located, the collecting channel is communicated with the air intake pipe, and a collecting box is provided at the lower port of the collecting channel.

[0011] In an optional solution: the filter unit includes a sliding ring seat slidably arranged inside the air intake pipe, a conical filter screen is fixed on the sliding ring seat, a plurality of deviated guide rods are arrayed around the sliding ring seat, the deviated guide rod slides through a deviated guide sleeve, the deviated guide sleeve is fixed to the inner wall of the air intake pipe, the deviated guide sleeve and the sliding ring seat are fixedly connected by a deviated spring, a cleaning shaft is provided for rotating the axis position of the conical filter screen, a cleaning scraper is provided at the end of the cleaning shaft, the cleaning scraper is in contact with the filtering surface of the conical filter screen, and the end of the drive shaft and the cleaning shaft are connected by a clutch mechanism.

[0012] In an optional solution: the clutch mechanism includes a first friction disk coaxially arranged at the end of the drive shaft, a rectangular sliding hole is opened at the end of the cleaning shaft, a rectangular slider is slidingly arranged in the rectangular sliding hole, the rectangular slider is connected to the inner wall of the rectangular sliding hole through a reset spring, and a second friction disk is fixedly provided at the end of the rectangular slider, and the second friction disk and the first friction disk are coaxially arranged.

[0013] In an optional solution: a plurality of shaking protrusions are distributed in an array on the side of the sliding ring seat facing the fan blades, a pressure wheel fixing seat is rotatably provided on the inner wall of the air intake duct, the pressure wheel fixing seat is connected to the drive shaft through a connecting rod, and a plurality of side rods are provided on the pressure wheel fixing seat, and a shaking pressure wheel corresponding to the shaking protrusion is rotatably provided at the end of each side rod.

[0014] A method for monitoring a high-voltage switchgear cabinet having an automatic alarm module comprises the following steps:

[0015] Step 1: Multimodal information collection:

[0016] The gas detection module collects gas concentration information such as oxygen, carbon monoxide, sulfur dioxide, hydrogen sulfide, sulfur hexafluoride and their decomposition products in the exhaust of high-voltage switchgear; a high-definition camera is used to collect image information of the equipment inside the switchgear; and a high-sensitivity microphone is used to collect audio information of the equipment operation;

[0017] Step 2: Multimodal information preprocessing:

[0018] For gas concentration information, median filtering is used to remove noise, based on the sensor calibration function C calibrated =a×C raw +b(where C calibrated is the concentration after calibration, C raw is the original concentration, a and b are calibration coefficients) for calibration; Gaussian filtering algorithm is used for image information (where (x, y) is the image pixel coordinate, (x0, y0) is the center coordinate, and σ is the standard deviation) is used for denoising, and the image contrast is enhanced by histogram equalization; for audio information, Wiener filtering is used for noise reduction, and the frame length T is used. frame Frame and extract Mel frequency cepstral coefficients, and use the Mel frequency to actual frequency conversion formula (where f is the actual frequency, f mel is Mel frequency) calculate Mel frequency;

[0019] Step 3: Multimodal information fusion:

[0020] Construct a multimodal neural network model and transform the preprocessed gas feature vector Image feature matrix M i , audio feature vector As input, the fusion feature vector is obtained by calculating through the convolution layer, pooling layer, and fully connected layer. (W1, W2, W3 are weight matrices, b1, b2, b3 are bias vectors, σ is the activation function, and [·;·;·] represents vector concatenation);

[0021] Step 4 Fault diagnosis and prediction:

[0022] Fault diagnosis: Fusion feature vector Input the fault diagnosis model based on convolutional neural network through the softmax function (where y i is the fault category, z iis the model output score, n is the total number of categories) calculate the probability of each fault type and determine the fault type and severity;

[0023] Fault prediction: Using the long short-term memory network LSTM, through the formula (where i t is the input gate, f t For the Gate of Forgetfulness, Candidate memory cells, c t For memory cells, o t is the output gate, is the hidden state, σ is the sigmoid activation function, tanh is the hyperbolic tangent activation function, and ⊙ is element-wise multiplication). The future equipment status is predicted based on the historical fusion features. Alarm and proactive maintenance: When the fault diagnosis model determines that a fault exists or the prediction model identifies a potential fault, different levels of alarm signals are issued through the sound and light alarm device according to the severity of the fault. Based on the maintenance strategy library, the decision tree algorithm is used to generate proactive maintenance suggestions and send them to the operation and maintenance personnel.

[0024] In an optional solution, the gas detection module includes at least sensors for detecting the gas concentrations of oxygen, carbon monoxide, sulfur dioxide, hydrogen sulfide, and sulfur hexafluoride.

[0025] By adopting the above technical solution, the present invention has the following beneficial effects:

[0026] The present invention is designed according to existing needs and constructs an anti-blocking air intake method so that the heat inside the cabinet can be taken away in time and the airflow can be sent to the detection position in time, thereby improving the response speed. In addition, the anti-blocking method here combines scraping and shaking to ensure the cleaning effect and improve the reliability of the equipment.

[0027] The present invention achieves a fault diagnosis accuracy rate exceeding 96% through the fusion of multimodal information of gas, images, and audio, combined with neural networks and LSTM models, and can predict faults 12-72 hours in advance. Relying on a graded alarm mechanism and an active maintenance system, it can respond according to the severity of the fault and automatically generate maintenance work orders. Combined with operation and maintenance strategies such as regular calibration and dynamic model updates, it can effectively reduce equipment failure rates by more than 30%, significantly improving the safety, reliability, and operation and maintenance efficiency of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. 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 any creative work.

[0029] Figure 1 It is a schematic diagram of the overall structure of the present invention.

[0030] Figure 2 This is a schematic structural diagram of the present invention with the protective cover removed.

[0031] Figure 3 This is a schematic structural diagram of the present invention with the other side of the protective cover removed.

[0032] Figure 4 It is a schematic diagram of the internal structure of the anti-blocking air intake module of the present invention.

[0033] Figure 5 It is a schematic diagram of the structure of the other side inside the anti-blocking air intake module of the present invention.

[0034] Figure 6 For the present invention Figure 3 A partial enlarged view of the structure.

[0035] Figure 7 It is a structural schematic diagram of one side of the pressure wheel fixing seat and the conical filter screen of the present invention.

[0036] Figure 8 It is a schematic structural diagram of the pressure wheel fixing seat and the other side of the conical filter screen of the present invention.

[0037] Figure 9 This is a logic block diagram of monitoring of the present invention.

[0038] Reference numerals: cabinet 100 , exhaust duct 101 , exhaust box 102 , gas detection module 103 , air inlet port 104 , cabinet door 105 , control panel 106 , first protective cover 107 , second protective cover 108 ;

[0039] Anti-blocking air intake module 200;

[0040] Connecting duct 201, air intake duct 202, drive shaft 203, drive motor 204, first friction disc 205, second friction disc 206, rectangular slider 207, return spring 208, collection box 209, collection channel 210, sliding ring seat 211, cleaning shaft 212, deviation spring 213, deviation guide rod 214, deviation guide sleeve 215, conical filter 216, cleaning scraper 217, rectangular sliding hole 218, fan blade 219, shaking pressure roller 220, pressure roller fixing seat 221, shaking protrusion 222;

[0041] Install the groove 301, the reciprocating motor 302, the reciprocating guide rod 303, the camera 304, the camera slide 305, and the reciprocating screw 306. DETAILED DESCRIPTION

[0042] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0043] The left, right, up, and down positions of the components shown in the accompanying drawings are merely one arrangement, and the specific positions are set according to specific needs.

[0044] In one embodiment, Figures 1-9As shown, a high-voltage switch cabinet with an automatic alarm module includes a cabinet 100, an air inlet port 104 is provided at the bottom of the cabinet 100, an anti-blocking air inlet module 200 for air intake is provided below the cabinet 100, and a second protective cover 108 is provided at the bottom of the cabinet 100 where the anti-blocking air inlet module 200 is located. The anti-blocking air inlet module 200 continuously intakes air to supply air to the inside of the cabinet 100 for cooling. An exhaust duct 101 is provided at the top of the cabinet 100, an exhaust box 102 is provided at the upper end of the exhaust duct 101, a first protective cover 107 is provided at the top of the cabinet 100 where the exhaust box 102 is located, grid holes are distributed on the surface of the first protective cover 17, an air outlet is distributed on the outside of the exhaust box 102, a gas detection module 103 for detecting gas information is provided inside the exhaust box 102, and the gas detection module 103 is electrically Connect the alarm module. When the gas detection module 103 detects abnormal gas in the air, the alarm module will issue an alarm message to reduce the damage to the switch cabinet. The cabinet 100 opening position is slidably matched with a cabinet door 105. The cabinet door 105 is provided with a control panel 106. The control panel 106 is provided with an image acquisition module for obtaining picture information of parts inside the cabinet 100. The picture information of the parts is obtained through the image acquisition module to further judge the working status of the parts. The cabinet 100 is also provided with an audio acquisition module for detecting sound. The audio information, picture information and gas information inside the cabinet 100 are obtained to detect the internal working conditions of the cabinet 100, thereby issuing an alarm signal in time. Under the action of the anti-blocking air intake module 200, the airflow will be sent to the detection position in time, thereby improving the response speed.

[0045] The exhaust duct has a tapered structure with a larger lower end and a smaller upper end. The airflow is guided and compressed as it rises along the tapered channel. The tapered structure helps to compress the gas, thereby increasing the concentration of the target gas and facilitating detection.

[0046] The image acquisition module includes a mounting groove 301 provided on the inner side of the control panel 106, a camera slide 305 is slidably provided in the mounting groove 301, a reciprocating guide rod 303 is slidably passed through both ends of the camera slide 305, the upper and lower ends of the reciprocating guide rod 303 are fixedly connected to the inner wall of the mounting groove 301, a plurality of cameras 304 for acquiring image information are provided on the surface of the camera slide 305, a reciprocating screw 306 is threadedly provided at the center position of the camera slide 305, the upper end of the reciprocating screw 306 is rotatably connected to the inner wall of the mounting groove 301, and the lower end of the reciprocating screw 306 is fixedly connected to the output end of the reciprocating motor 302, and under the drive of the reciprocating motor 302, the reciprocating screw 306 and the camera slide 305 rotate relative to each other, and under the action of the thread, the camera slide 305 slides along the reciprocating guide rod 303 to acquire part information at different positions, thereby providing sufficient basic data for alarm;

[0047] The anti-blocking air intake module 200 includes a horizontally arranged air intake duct 202, the upper side of which is connected to the air intake port 104 through a connecting duct 201, a driving motor 204 is fixedly provided at the end of the air intake duct 202, a driving shaft 203 is fixedly provided at the output end of the driving motor 204, and a fan blade 219 is provided at the output end of the driving shaft 203, and a filter unit for filtering air is provided at the port position of the air intake duct 202, which filters the intake air to prevent dust from entering the cabinet 100, and under the drive of the driving motor 204, the driving shaft 203 drives the fan blade 219 to rotate, so that the air smoothly enters the cabinet 100 along the connecting duct 201, so as to cool the interior of the cabinet 100, and a collecting channel 210 is provided at the lower side of the air intake duct 202 where the conical filter 216 is located, and the collecting channel 210 is connected to the air intake duct 202, and a collecting box 209 is provided at the lower port position of the collecting channel 210, so that the cleaned dust can be collected;

[0048] The filter unit includes a sliding ring seat 211 that is slidably arranged inside the air intake pipe 202. A conical filter screen 216 is fixedly provided on the sliding ring seat 211. A plurality of deflection guide rods 214 are arranged in an array around the sliding ring seat 211. The deflection guide rods 214 slide through a deflection guide sleeve 215. The deflection guide sleeve 215 is fixed to the inner wall of the air intake pipe 202. The deflection guide sleeve 215 and the sliding ring seat 211 are fixedly connected by a deflection spring 213. A cleaning shaft 212 is provided on the axis of the conical filter screen 216. A cleaning scraper 217 is provided at the end of the cleaning shaft 212, and the cleaning scraper 217 is in contact with the filtering surface of the conical filter 216. The end of the drive shaft 203 and the cleaning shaft 212 are connected by a clutch mechanism. When the conical filter 216 is blocked, the conical filter 216 will slide along the deviation guide rod 214, thereby triggering the clutch mechanism to work. The clutch mechanism will drive the cleaning shaft 212 to rotate, and the cleaning shaft 212 drives the cleaning scraper 217 to rotate, thereby removing impurities on the filtering surface of the conical filter 216, achieving the effect of cleaning the hole;

[0049] The clutch mechanism includes a first friction disc 205 coaxially arranged at the end of the drive shaft 203, a rectangular sliding hole 218 is opened at the end of the cleaning shaft 212, a rectangular slider 207 is slidably provided in the rectangular sliding hole 218, and the rectangular slider 207 is connected to the inner wall of the rectangular sliding hole 218 by a return spring 208. A second friction disc 206 is fixedly provided at the end of the rectangular slider 207. The second friction disc 206 and the first friction disc 205 are coaxially arranged. When the conical filter 216 moves toward the first friction disc 205, the second friction disc 206 contacts the first friction disc 205. The friction force generated by the two contact surfaces will cause the cleaning shaft 212 to rotate, and the cleaning shaft 212 will drive the cleaning scraper 217 to rotate, providing power for cleaning;

[0050] The sliding ring seat 211 is provided with a plurality of shaking protrusions 222 in an array on the side facing the fan blade 219, and a pressure wheel fixing seat 221 is provided for rotation on the inner wall of the air intake pipe 202, and the pressure wheel fixing seat 221 is connected to the drive shaft 203 through a connecting rod. The pressure wheel fixing seat 221 is provided with a plurality of side rods, and a shaking pressure wheel 220 corresponding to the shaking protrusion 222 is provided at the end of each side rod. When the conical filter screen 216 slides further toward the fan blade 219, the return spring 208 will be compressed. At this time, the shaking pressure wheel 220 contacts the shaking protrusion 222. Under the transmission of the connecting rod, the drive shaft 203 will drive the pressure wheel fixing seat 221 to rotate, and the shaking pressure wheel 220 on the pressure wheel fixing seat 221 will generate intermittent driving force on the shaking protrusion 222, so that the sliding ring seat 211 and the conical filter screen 216 can vibrate, thereby assisting the conical filter screen 216 in cleaning the hole.

[0051] Specific implementation method:

[0052] Hardware system construction

[0053] 1. Multimodal sensor deployment

[0054] The gas detection module is installed using a combination of high-precision electrochemical gas sensors and infrared gas sensors, and 5 groups of composite sensor arrays are arranged at the exhaust port of the high-voltage switchgear. The specific models and parameters are as follows:

[0055] Oxygen sensor: 4OX2-2 type from CITY, UK, detection range 0-25% VOL, accuracy ±2%

[0056] Carbon monoxide sensor: ME2-CO type, range 0-1000ppm, resolution 1ppm;

[0057] Sulfur dioxide sensor: S4-SO2 type, measurement range 0-20ppm, response time <30s;

[0058] Hydrogen sulfide sensor: S4-H2S type, range 0-50ppm, linear error ±3%;

[0059] Sulfur hexafluoride sensor: MI-510, detection range 0-1000ppm, stability ≤1%FS / year;

[0060] Each sensor is connected to a customized signal acquisition board via a waterproof, sealed connector. Data is transmitted using the RS485 bus protocol in Modbus-RTU format, with a data acquisition frequency set to once per minute. To ensure data accuracy, on-site calibration using standard gas is performed quarterly.

[0061] HD camera installation

[0062] The camera installed was a Hikvision DS-2CD3T46FWDV2-I 4-megapixel network camera equipped with a 2.8-12mm motorized zoom lens. The viewing angle was pre-simulated using engineering measurement software to ensure full coverage of 12 key monitoring points, including circuit breaker contacts, disconnector contacts, and busbar connections. The camera's video resolution was set to 1920×1080, with a frame rate of 15 fps, and the H.265 encoding format was used to reduce bandwidth usage.

[0063] Audio sensor installation

[0064] Audio-Technica ATR2500-USB high-sensitivity condenser microphones are symmetrically installed at the four corners of the cabinet and fixed with custom shock-proof brackets. The audio signal is connected to the preamplifier (gain set to 40dB) through the XLR interface and converted into a digital signal through a 24-bit A / D converter. 2 The data was transmitted to the data processing unit using the S communication protocol, with a sampling rate of 44.1kHz and a quantization bit of 16 bits. To reduce interference from ambient noise, the microphone was equipped with a special windscreen and a 50Hz high-pass filter.

[0065] (2) Data processing unit configuration

[0066] Advantech ARK-3500 industrial computer is used as the core processing device, and the specific configuration is as follows:

[0067] Processor: Intel Xeon E-2278G vPro (8 cores, 16 threads, 3.4GHz);

[0068] Memory: 32GB DDR4 ECC (2666MHz);

[0069] Storage: 1TB NVMe SSD + 4TB mechanical hard drive (RAID 1);

[0070] Communication interface: 4 RS485 ports, 2 Gigabit Ethernet ports, 4 USB3.0 ports;

[0071] Operating system: Ubuntu 20.04 LTS (real-time kernel patch);

[0072] Equipped with an independent UPS power supply (lasting 30 minutes) to ensure data storage in the event of a power outage.

[0073] 2. Software System Implementation

[0074] (1) Multimodal Information Preprocessing

[0075] The module gas information preprocessing program is developed using Python language combined with NumPy and pandas libraries. A 5-point sliding median filter is applied to each sensor data to eliminate burst noise. The raw data is calibrated using a linear calibration formula based on the sensor calibration certificate. Finally, the processed data is stored in the InfluxDB time series database in time series, retaining three years of historical data.

[0076] Image information preprocessing;

[0077] Based on the OpenCV and PyTorch libraries, a 5×5 Gaussian kernel (with a standard deviation of 1.5) is first used to remove image noise, and then the contrast-limited adaptive histogram equalization (CLAHE) algorithm is used to enhance image details. Finally, a 2048-dimensional feature vector is extracted through a pre-trained ResNet-50 model, and the extraction is performed every 10 minutes.

[0078] Audio information preprocessing;

[0079] Using Librosa and TensorFlow libraries, the Wiener denoising parameters are dynamically adjusted according to the ambient noise level to denoise the audio data. The audio is framed with a frame length of 256 samples and a frame shift of 128 samples. The 13-dimensional Mel-frequency cepstral coefficients (MFCCs) and their first-order and second-order differences are calculated to form a 39-dimensional feature vector.

[0080] (2) Multimodal Information Fusion Module

[0081] Model architecture design

[0082] The multimodal neural network built based on the PyTorch framework is divided into three main parts: input layer, feature extraction layer, fusion and output layer.

[0083] Input layer: Create three independent input channels to receive preprocessed gas feature vectors, image feature vectors, and audio feature vectors. The length of the gas feature vector is the same as the number of gas sensors; the image feature vector is a 2048-dimensional vector extracted using the pretrained model; and the audio feature vector is a 39-dimensional vector containing MFCCs and their differentials.

[0084] Feature extraction layer:

[0085] Gas branch: The gas feature vector is processed through a fully connected layer containing 64 neurons. The gas features are initially reduced in dimension and transformed. The ReLU function is used as the activation function, and a Dropout layer with a deactivation rate of 0.2 is added after the layer to prevent overfitting.

[0086] Image branch: Based on the pre-trained ResNet-50 model, the final fully connected layer is removed, retaining the convolutional layers as the backbone network for image feature extraction. Using 3×3 convolution kernels, ReLU activation functions, and 2×2 max pooling layers, the deep features of the image are gradually extracted and the output is flattened into a one-dimensional vector.

[0087] Audio branch: A 1D convolutional layer is used to process the audio feature vector, with a kernel size of 3 and a channel count of 32. This convolution operation captures local patterns in the audio features. A ReLU activation function and a 2×2 max pooling layer are also used, and the output is flattened.

[0088] Fusion and Output Layer: The feature extraction outputs of the three branches are concatenated to form a fused feature vector. This is then further fused through two fully connected layers: the first layer contains 128 neurons, and the second layer contains 64 neurons. A dropout layer with a dropout rate of 0.2 is used between the two layers. Finally, a fully connected layer with 8 neurons (corresponding to the 8 fault types) is used to output the probability distribution of each fault type using a softmax activation function.

[0089] (3) Fault diagnosis and prediction module

[0090] Troubleshooting

[0091] The fused features are fed into the trained model, and the probability of each fault type is calculated using the softmax function. If the probability of a particular fault type is greater than 0.7, the switchgear is considered to have that type of fault. If the probability is between 0.5 and 0.7, it is considered suspicious. If the probability is less than 0.5, monitoring continues.

[0092] Failure prediction

[0093] A prediction model was constructed using an LSTM network, with input consisting of a fused feature sequence from the past 24 hours (one sample per hour). The model consisted of two LSTM layers, each with 128 units, followed by a fully connected layer. The mean squared error (MSE) loss function was used for training, and the Adam optimizer was used to optimize model parameters. The learning rate was set to 0.0005, and training was repeated for 30 epochs.

[0094] (4) Alarm and active maintenance module

[0095] Hierarchical alarm mechanism

[0096]

[0097] The maintenance strategy library, stored in a MySQL database, contains over 200 maintenance plans. Using a decision tree algorithm, the system matches fault types and automatically generates maintenance work orders. These orders include specific maintenance actions (such as contact replacement and bolt tightening), a list of required spare parts, safety precautions, and estimated processing time. These work orders are then pushed to the operator's terminal via WeChat for Business, with the switchgear location marked on a GIS map.

[0098] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A high-voltage switchgear with an automatic alarm module, comprising a cabinet body (100), characterized in that: The cabinet (100) is provided with an air inlet port (104) at the bottom, an anti-blocking air inlet module (200) for air inlet is provided below the cabinet (100), an exhaust duct (101) is provided at the top of the cabinet (100), an exhaust box (102) is provided at the upper end of the exhaust duct (101), an air outlet is distributed on the outside of the exhaust box (102), a gas detection module (103) for detecting gas information is provided inside the exhaust box (102), the gas detection module (103) is electrically connected to the alarm module, a cabinet door (105) is slidably fitted at the opening position of the cabinet (100), a control panel (106) is provided on the cabinet door (105), an image acquisition module for acquiring image information of parts inside the cabinet (100) is provided on the control panel (106), and an audio acquisition module for detecting sound is also provided inside the cabinet (100).

2. The high-voltage switchgear with an automatic alarm module according to claim 1, characterized in that: The exhaust pipe (101) is a tapered structure with a larger lower end and a smaller upper end.

3. The high-voltage switchgear with an automatic alarm module according to claim 1, characterized in that: The image acquisition module comprises a mounting groove (301) arranged on the inner side of the control panel (106); a camera slide (305) is slidably arranged in the mounting groove (301); reciprocating guide rods (303) are slidably passed through both ends of the camera slide (305); the upper and lower ends of the reciprocating guide rods (303) are fixedly connected to the inner wall of the mounting groove (301); a plurality of cameras (304) for acquiring image information are provided on the surface of the camera slide (305); a reciprocating screw (306) is threadedly arranged at the center position of the camera slide (305); the upper end of the reciprocating screw (306) is rotatably connected to the inner wall of the mounting groove (301), and the lower end of the reciprocating screw (306) is fixedly connected to the output end of the reciprocating motor (302).

4. The high-voltage switchgear with an automatic alarm module according to claim 1, characterized in that: The anti-blocking air intake module (200) comprises a horizontally arranged air intake pipe (202), the upper side of the air intake pipe (202) being connected to the air intake port (104) via a connecting conduit (201), a drive motor (204) being fixedly provided at the end of the air intake pipe (202), a drive shaft (203) being fixedly provided at the output end of the drive motor (204), a fan blade (219) being provided at the output end of the drive shaft (203), and a filter unit for filtering air being provided at the port position of the air intake pipe (202).

5. The high-voltage switchgear with automatic alarm module according to claim 4, characterized in that: The filter unit comprises a sliding ring seat (211) slidably arranged inside the air intake pipe (202), a conical filter screen (216) being fixedly provided on the sliding ring seat (211), a plurality of deviating guide rods (214) being distributed in an array around the sliding ring seat (211), the deviating guide rods (214) slidingly passing through a deviating guide sleeve (215), the deviating guide sleeve (215) being fixed to the inner wall of the air intake pipe (202), the deviating guide sleeve (215) and the sliding ring seat (211) being fixedly connected via a deviating spring (213), a cleaning shaft (212) being provided for rotating about the axis of the conical filter screen (216), a cleaning scraper (217) being provided at the end of the cleaning shaft (212), the cleaning scraper (217) being in contact with the filtering surface of the conical filter screen (216), and the end of the driving shaft (203) and the cleaning shaft (212) being connected via a clutch mechanism.

6. The high-voltage switchgear with automatic alarm module according to claim 5, characterized in that: A collecting channel (210) is provided on the lower side of the air intake pipe (202) where the conical filter screen (216) is located. The collecting channel (210) is in communication with the air intake pipe (202). A collecting box (209) is provided at the lower end of the collecting channel (210).

7. The high-voltage switchgear with an automatic alarm module according to claim 5, characterized in that: The clutch mechanism comprises a first friction disc (205) coaxially arranged at the end of the driving shaft (203); a rectangular sliding hole (218) is opened at the end of the cleaning shaft (212); a rectangular slider (207) is slidably arranged in the rectangular sliding hole (218); the rectangular slider (207) is connected to the inner wall of the rectangular sliding hole (218) via a return spring (208); a second friction disc (206) is fixedly arranged at the end of the rectangular slider (207); the second friction disc (206) and the first friction disc (205) are coaxially arranged.

8. The high-voltage switchgear with an automatic alarm module according to claim 5, characterized in that: A plurality of shaking protrusions (222) are arranged in an array on one side of the sliding ring seat (211) facing the fan blade (219); a pressure wheel fixing seat (221) is rotatably provided on the inner wall of the air intake duct (202); the pressure wheel fixing seat (221) is connected to the drive shaft (203) via a connecting rod; a plurality of side rods are provided on the pressure wheel fixing seat (221); and a shaking pressure wheel (220) corresponding to the shaking protrusion (222) is rotatably provided at the end of each side rod.

9. A method for monitoring a high-voltage switchgear cabinet having an automatic alarm module according to any one of claims 1 to 8, characterized in that: The following steps are involved: The following steps are involved: Step 1: Multimodal information collection: The gas detection module collects gas concentration information of oxygen, carbon monoxide, sulfur dioxide, hydrogen sulfide, sulfur hexafluoride and their decomposition products in the exhaust of high-voltage switchgear; uses a high-definition camera to collect image information of equipment inside the switchgear; and uses a high-sensitivity microphone to collect audio information of equipment operation; Step 2: Multimodal information preprocessing: For gas concentration information, median filtering is used to remove noise, based on the sensor calibration function C calibrated =a×C raw +b(where C calibrated is the concentration after calibration, C raw is the original concentration, a and b are calibration coefficients) for calibration; Gaussian filtering algorithm is used for image information (where (x, y) is the image pixel coordinate, (x0, y0) is the center coordinate, and σ is the standard deviation) is used for denoising, and the image contrast is enhanced by histogram equalization; for audio information, Wiener filtering is used for noise reduction, and the frame length T is used. frame Frame and extract Mel frequency cepstral coefficients, and use the Mel frequency to actual frequency conversion formula (where f is the actual frequency, f mel is Mel frequency) calculate Mel frequency; Step 3: Multimodal information fusion: Construct a multimodal neural network model and transform the preprocessed gas feature vector Image feature matrix M i , audio feature vector As input, the fusion feature vector is obtained by calculating through the convolution layer, pooling layer, and fully connected layer. (W1, W2, W3 are weight matrices, b1, b2, b3 are bias vectors, σ is the activation function, and [·;·;·] represents vector concatenation); Step 4 Fault diagnosis and prediction: Fault diagnosis: Fusion feature vector Input the fault diagnosis model based on convolutional neural network through the softmax function (where y i is the fault category, z i is the model output score, n is the total number of categories) calculate the probability of each fault type and determine the fault type and severity; Fault prediction: Using the long short-term memory network LSTM, through the formula (where i t is the input gate, f t For the Gate of Forgetfulness, For candidate memory cells, c t For memory cells, o t is the output gate, is the hidden state, σ is the sigmoid activation function, tanh is the hyperbolic tangent activation function, and ⊙ is element-wise multiplication). The future equipment status is predicted based on the historical fusion features. Alarm and proactive maintenance: When the fault diagnosis model determines that a fault exists or the prediction model identifies a potential fault, different levels of alarm signals are issued through the sound and light alarm device according to the severity of the fault. Based on the maintenance strategy library, the decision tree algorithm is used to generate proactive maintenance suggestions and send them to the operation and maintenance personnel.

10. The monitoring method according to claim 9, characterized in that: The gas detection module at least includes sensors for detecting the gas concentrations of oxygen, carbon monoxide, sulfur dioxide, hydrogen sulfide, and sulfur hexafluoride.

Images