Switch cabinet anti-explosion multi-parameter rapid identification and alarm method

Abstract

The invention discloses a switch cabinet anti-explosion multi-parameter rapid identification and alarm method, and relates to the technical field of power equipment safety, and the method comprises the steps: obtaining a known switch cabinet historical fault record, analyzing the attribute preference and occurrence rule of different faults, constructing a multi-modal time sequence fault diagnosis model, and generating a fault knowledge base; operating parameters are monitored in real time, a field programmable gate array is used for solidifying primary judgment logic, and if the current change rate and the arc light signal intensity exceed preset threshold values at the microsecond level at the same time, a primary fault trigger signal is generated; the edge intelligent calculation unit automatically extracts the original data of the multi-source heterogeneous sensor in a set time window before and after the triggering moment, and substitutes the original data into the multi-modal time sequence fault diagnosis model to obtain a diagnosis result and a risk level of a known switch cabinet fault; and in combination with the multi-modal time sequence fault diagnosis model, executing a layered response mechanism. The beneficial effect of the invention is that the fault loss is reduced.

Description

Technical Field

[0001] This invention relates to the field of power equipment safety technology, specifically to a method for rapid identification and alarm of multiple parameters for fire and explosion prevention in switchgear. Background Technology

[0002] While current switchgear fire and explosion prevention technologies have been applied to some extent, several shortcomings remain: most systems rely on a single sensor or single parameter for monitoring, making it difficult to identify potential fire and explosion risks in a timely and accurate manner; monitoring data is mostly discrete or low-frequency sampled, failing to reflect dynamic changes under complex operating conditions, resulting in delayed early warnings; at the same time, existing alarm mechanisms have limited response speed and grading strategies, making it difficult to make precise interventions in high-risk fire and explosion events in a short period of time, thereby affecting equipment safety and operational reliability. Summary of the Invention

[0003] To address the aforementioned technical issues, this paper provides a method for rapid identification and alarm of multiple parameters for fire and explosion prevention in switchgear. This technical solution solves the problem that the existing alarm mechanisms have limited response speed and hierarchical strategies, making it difficult to make accurate interventions for high-risk fire and explosion events in a short period of time.

[0004] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A method for rapid identification and alarm of multiple parameters for fire and explosion protection in switchgear, comprising: Obtain historical fault records of known switchgear, analyze the attribute preferences and occurrence patterns of different faults of known switchgear, construct a multimodal time series fault diagnosis model, and generate a knowledge base of known switchgear faults; Based on integrated multi-source heterogeneous sensors, known switchgear operating parameters are monitored in real time. The primary judgment logic is solidified using FPGA field-programmable gate array. If the current change rate and the arc signal intensity simultaneously exceed the preset threshold within a microsecond time window, a primary fault trigger signal is generated. Based on the generation of the primary fault trigger signal, the edge intelligent computing unit automatically extracts the original data of multi-source heterogeneous sensors within a set time window before and after the trigger time, packages them and substitutes them into the multimodal time series fault diagnosis model for correlation matching, and obtains the diagnosis results and risk level of known switchgear faults. Based on the diagnostic results and risk levels of known switchgear faults, a hierarchical response mechanism is implemented in conjunction with a multimodal time series fault diagnosis model.

[0005] Preferably, historical fault data of known switchgear is obtained, data preprocessing is performed, and each known switchgear fault sample is labeled with fault type and severity level; The acquisition of known historical fault data of switchgear includes: three-phase current waveform sequence, arc intensity sequence, gas concentration sequence, infrared temperature sequence and ultrasonic signal sequence; Using the DTW dynamic time warping algorithm, timestamps are aligned to generate time series of known switchgear fault samples; The distance metric between two sample sequences is calculated using the Euclidean distance formula, and a distance matrix between the two sample sequences is constructed. By using the recursive relationship of dynamic time warping, the cumulative minimum distance of each grid point in the distance matrix is ​​calculated step by step, and the optimal alignment path between the two sequences is determined by backtracking. Based on determining the optimal alignment path between two sequences through backtracking, the known historical fault data of switchgear are uniformly aligned to the same reference timestamp, and a multimodal sample matrix with timestamp alignment is constructed.

[0006] Preferably, based on the time series of each known switchgear fault sample, the mean and variance of the time series of each known switchgear fault sample are calculated to characterize the average level and dispersion of the time series of each known switchgear fault sample, the peak value and skewness of the time series of each known switchgear fault sample are obtained, and the time domain features of each known switchgear fault sample are extracted. Using the time-domain characteristics of each known switchgear fault sample as the input signal, a fast Fourier transform is performed to convert the input signal from the time domain to the frequency domain, and the spectral representation of each known switchgear fault sample is calculated. Based on the spectral representation of each known switchgear fault sample, the amplitude and phase information of each known switchgear fault sample in the frequency domain are obtained, and the frequency domain features of each known switchgear fault sample are extracted. The time-domain and frequency-domain features of each known switchgear fault sample are spliced ​​together and normalized to construct a manual feature vector for each known switchgear fault sample.

[0007] Preferably, a dual-path parallel processing architecture is used, with a timestamp-aligned multimodal sample matrix and the manually generated feature vectors of each known switchgear fault sample as input. Based on the manual feature vectors of known switchgear fault samples, a multilayer perceptron is designed, and linear and nonlinear transformations are used as activation functions to extract the traditional feature paths of each known switchgear fault sample. Based on the timestamp-aligned multimodal sample matrix, the projection of the time series of each known switchgear fault sample into an embedding vector is obtained, thus obtaining the position encoding information of the time series of each known switchgear fault sample. The Transformer encoding layer is designed, which takes the position encoding information of the time series of each known switch cabinet fault sample as input, and uses the self-attention mechanism to capture the dependency relationship between the time series of each known switch cabinet fault sample and extract the self-attention feature path of each known switch cabinet fault sample. The traditional feature path and the self-attention feature path are weighted and summed to obtain the fusion feature vector of each known switchgear fault sample; Using the fused feature vectors of each known switchgear fault sample as input, and the probability distribution of the fault type and the predicted risk level of each known switchgear fault sample as output, a multimodal time series fault diagnosis model is constructed to generate a knowledge base of known switchgear faults.

[0008] Preferably, based on integrated multi-source heterogeneous sensors, known switchgear operating parameters are monitored in real time, known switchgear current signals and arc signals are acquired, converted into digital signals by analog-to-digital converters, and synchronously input to multiple independent processing pipelines inside the FPGA field-programmable gate array for data preprocessing; Calculate the ratio of the difference in current values ​​between adjacent sampling points of a known switchgear to the sampling time interval to obtain the real-time current change rate of the known switchgear; Using a hardware description language, a basic judgment logic is constructed to monitor the current change rate and arc signal intensity in real time; Calculate the ratio of the difference in current values ​​between adjacent sampling points of a known switchgear to the sampling time interval to obtain the real-time current change rate of the known switchgear; If the current change rate and the arc signal intensity are both detected to exceed their respective preset thresholds within a microsecond time window, it is determined that the primary fault triggering condition is met, and a primary fault triggering signal is generated. The preset threshold for the current change rate is 5A per microsecond; the preset threshold for the arc signal intensity is 500mV.

[0009] Preferably, a primary fault trigger signal is generated as a hardware interrupt trigger to wake up the edge intelligent unit and start the data capture module; The data capture module automatically retrieves the original data from the multi-source heterogeneous sensors within a set time window based on the trigger time, including current waveforms, arc intensity, gas concentration, infrared temperature, and ultrasonic signals. Based on the acquisition of raw data from multi-source heterogeneous sensors, data preprocessing is performed to construct multi-modal time-series samples of multi-source heterogeneous sensors; Using multi-modal time-series samples from multi-source heterogeneous sensors as input, the model is substituted into a multi-modal time-series fault diagnosis model. The Sigmoid activation function is used, and the diagnosis results and risk levels of known switchgear faults are used as outputs. The risk level R includes: low risk: 0 ≤ R < 0.4, medium risk: 0.4 ≤ R < 0.7; high risk: 0.7 ≤ R < 1.0.

[0010] Preferably, the risk level formula is as follows: ; in, Given the known fault risk level of the switchgear, For the risk emergence function, Given the current state vector of a known switchgear fault, Given the known risk sharpness parameters for switchgear, The fault categories inferred by the multimodal time series fault diagnosis model The probability, Fault Category Typical pattern vectors, Given the current state vector of the switchgear fault. With the fault prototype The KL divergence between them.

[0011] Preferably, based on the diagnostic results and risk levels of known switchgear faults, a hierarchical response mechanism is implemented in conjunction with a multimodal time series fault diagnosis model; Based on the diagnostic results and risk levels of known switchgear faults, the faults are classified into three emergency levels: low risk, medium risk, and high risk, and corresponding response strategies are implemented.

[0012] Preferably, if the fault is determined to be low-risk, an early warning signal is issued through the remote monitoring system to prompt maintenance personnel to perform planned inspections and maintenance, and to record fault data and update the knowledge base of known switchgear faults in a synchronous manner to support subsequent fault prediction and health management. If the fault is determined to be of medium risk, the on-site audible and visual alarm device will be triggered, and a detailed fault diagnosis report and handling suggestions will be pushed to the mobile terminal of the maintenance personnel to assist them in making accurate on-site intervention and decision support. If a high-risk fault is identified, the protection system will be immediately triggered to perform an emergency trip operation, achieving microsecond-level power cut-off and suppressing the further development of the fault. At the same time, the highest priority alarm will be sent to the central control room, automatically activating the emergency plan and linking the safety control system to start the comprehensive fault isolation and handling process. The hierarchical response mechanism also includes dynamically adjusting response thresholds and strategies based on historical fault handling data and real-time fault conditions to improve the overall fault handling efficiency and operational security of the system.

[0013] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention proposes a rapid multi-parameter identification and alarm method for switchgear fire and explosion protection. By constructing a multimodal time-series fault diagnosis model and combining it with real-time monitoring by multi-source heterogeneous sensors, this invention achieves rapid and accurate identification of the switchgear's operating status. It utilizes FPGA to embed primary judgment logic, enabling microsecond-level fault triggering and improving response speed. Through edge intelligent computing units for data association matching and risk level assessment, it can accurately determine the fault type and severity. Combined with a hierarchical response mechanism, it can implement differentiated handling for different risk levels, thereby significantly improving the switchgear's fire and explosion protection capabilities, enhancing equipment operational safety and reliability, and reducing fault losses. Attached Figure Description

[0014] Figure 1 This is a flowchart of a method for rapid identification and alarm of multiple parameters for fire and explosion prevention in switchgear. Detailed Implementation

[0015] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0016] Reference Figure 1 As shown, a method for rapid identification and alarm of multiple parameters for fire and explosion prevention in switchgear includes: S1. Obtain historical fault records of known switchgear, analyze the attribute preferences and occurrence patterns of different faults of known switchgear, construct a multimodal time series fault diagnosis model, and generate a knowledge base of known switchgear faults; Step S1 includes the following: Obtain historical fault data of known switchgear, perform data preprocessing, and label the fault type and severity level for each known switchgear fault sample; The acquisition of known historical fault data of switchgear includes: three-phase current waveform sequence, arc intensity sequence, gas concentration sequence, infrared temperature sequence and ultrasonic signal sequence; Using the DTW dynamic time warping algorithm, timestamps are aligned to generate time series of known switchgear fault samples; The distance metric between two sample sequences is calculated using the Euclidean distance formula, and a distance matrix between the two sample sequences is constructed. By using the recursive relationship of dynamic time warping, the cumulative minimum distance of each grid point in the distance matrix is ​​calculated step by step, and the optimal alignment path between the two sequences is determined by backtracking. Based on determining the optimal alignment path between two sequences through backtracking, the known historical fault data of switchgear are uniformly aligned to the same reference timestamp, and a multimodal sample matrix with timestamp alignment is constructed.

[0017] Step S1 also includes the following: Based on the time series of known switchgear fault samples, the mean and variance of the time series of known switchgear fault samples are calculated to characterize the average level and dispersion of the time series of known switchgear fault samples, obtain the peak value and skewness of the time series of known switchgear fault samples, and extract the time domain features of each known switchgear fault sample. Using the time-domain characteristics of each known switchgear fault sample as the input signal, a fast Fourier transform is performed to convert the input signal from the time domain to the frequency domain, and the spectral representation of each known switchgear fault sample is calculated. Based on the spectral representation of each known switchgear fault sample, the amplitude and phase information of each known switchgear fault sample in the frequency domain are obtained, and the frequency domain features of each known switchgear fault sample are extracted. The time-domain and frequency-domain features of each known switchgear fault sample are spliced ​​together and normalized to construct a manual feature vector for each known switchgear fault sample.

[0018] Step S1 also includes the following: A dual-path parallel processing architecture is implemented, using a timestamp-aligned multimodal sample matrix and manually generated feature vectors of known switchgear fault samples as inputs. Based on the manual feature vectors of known switchgear fault samples, a multilayer perceptron is designed, and linear and nonlinear transformations are used as activation functions to extract the traditional feature paths of each known switchgear fault sample. Based on the timestamp-aligned multimodal sample matrix, the projection of the time series of each known switchgear fault sample into an embedding vector is obtained, thus obtaining the position encoding information of the time series of each known switchgear fault sample. The Transformer encoding layer is designed, which takes the position encoding information of the time series of each known switch cabinet fault sample as input, and uses the self-attention mechanism to capture the dependency relationship between the time series of each known switch cabinet fault sample and extract the self-attention feature path of each known switch cabinet fault sample. The traditional feature path and the self-attention feature path are weighted and summed to obtain the fusion feature vector of each known switchgear fault sample; Using the fused feature vectors of each known switchgear fault sample as input, and the probability distribution of the fault type and the predicted risk level of each known switchgear fault sample as output, a multimodal time series fault diagnosis model is constructed to generate a knowledge base of known switchgear faults.

[0019] When using it, refer to the content of step S1 above: Existing switchgear fault diagnosis technologies often rely on single-modal data or analyze only traditional time-domain features, neglecting the complex interactions between different modal information. Traditional methods struggle to capture high-dimensional features and temporal relationships in multimodal data, resulting in weak diagnostic accuracy and fault prediction capabilities. Furthermore, existing fault diagnosis models often depend on manual feature extraction, lacking adaptive learning capabilities and struggling to cope with complex and dynamic power equipment fault scenarios. This step integrates multimodal time series data with manual features, and uses the DTW dynamic time warping algorithm to align timestamps, ensuring data consistency and temporal order. Simultaneously, by employing the Transformer self-attention mechanism to extract temporal dependencies, it can automatically learn deep features of fault data, greatly improving the accuracy of fault identification and risk prediction. The dual-channel processing architecture of time-domain and frequency-domain features enhances the model's robustness and generalization ability.

[0020] S2. Based on integrated multi-source heterogeneous sensors, the known operating parameters of the switchgear are monitored in real time. The primary judgment logic is solidified using an FPGA field-programmable gate array. If the current change rate and the arc light signal intensity simultaneously exceed a preset threshold within a microsecond time window, a primary fault trigger signal is generated. The known operating parameters of the switchgear include: current signal and arc light signal. Step S2 includes the following: Based on integrated multi-source heterogeneous sensors, the known operating parameters of the switchgear are monitored in real time, and the known switchgear current signal and arc signal are acquired. The signals are converted into digital signals by analog-to-digital converters and synchronously input to multiple independent processing pipelines inside the FPGA field programmable gate array for data preprocessing. Calculate the ratio of the difference in current values ​​between adjacent sampling points of a known switchgear to the sampling time interval to obtain the real-time current change rate of the known switchgear; Using a hardware description language, a basic judgment logic is constructed to monitor the current change rate and arc signal intensity in real time; Calculate the ratio of the difference in current values ​​between adjacent sampling points of a known switchgear to the sampling time interval to obtain the real-time current change rate of the known switchgear; If the current change rate and the arc signal intensity are both detected to exceed their respective preset thresholds within a microsecond time window, it is determined that the primary fault triggering condition is met, and a primary fault triggering signal is generated. The preset threshold for the current change rate is 5A per microsecond; the preset threshold for the arc signal intensity is 500mV.

[0021] When using it, refer to the content of step S2 above: Current fault detection technologies mostly rely on single signal monitoring, resulting in slow response times, inability to respond promptly to transient faults, and limited processing capabilities. Traditional systems cannot effectively integrate multi-source heterogeneous data, leading to insufficient accurate real-time judgment, especially in high-voltage equipment, which may cause equipment damage. This step utilizes an FPGA (Field-Programmable Gate Array) to process current change rate and arc signal in parallel, making real-time judgments and generating fault trigger signals. Through a microsecond-level time window, it can quickly respond to situations where thresholds are exceeded simultaneously, significantly improving fault response speed and enhancing system safety and reliability.

[0022] S3. Based on the generation of the primary fault trigger signal, the edge intelligent computing unit automatically extracts the original data of multi-source heterogeneous sensors within a set time window before and after the trigger time, packages them and substitutes them into the multimodal time series fault diagnosis model for correlation matching, and obtains the diagnosis results and risk level of known switchgear faults. Step S3 includes the following: The generation of a primary fault trigger signal is used as a hardware interrupt trigger to wake up the edge intelligent unit and start the data capture module. The data capture module automatically retrieves the original data from the multi-source heterogeneous sensors within a set time window based on the trigger time, including current waveforms, arc intensity, gas concentration, infrared temperature, and ultrasonic signals. Based on the acquisition of raw data from multi-source heterogeneous sensors, data preprocessing is performed to construct multi-modal time-series samples of multi-source heterogeneous sensors; Using multi-modal time-series samples from multi-source heterogeneous sensors as input, the model is substituted into a multi-modal time-series fault diagnosis model. The Sigmoid activation function is used, and the diagnosis results and risk levels of known switchgear faults are used as outputs. The risk level R includes: low risk: 0 ≤ R < 0.4, medium risk: 0.4 ≤ R < 0.7; high risk: 0.7 ≤ R < 1.0; The risk level formula is as follows: ; in, Given the known fault risk level of the switchgear, For the risk emergence function, Given the current state vector of a known switchgear fault, Given the known risk sharpness parameters for switchgear, The fault categories inferred by the multimodal time series fault diagnosis model The probability, Fault Category Typical pattern vectors, Given the current state vector of the switchgear fault. With the fault prototype The KL divergence between them.

[0023] When using it, refer to the content of step S3 above: Currently, fault diagnosis technologies largely rely on single sensor data (such as current or temperature), failing to effectively integrate multi-source heterogeneous data. This results in poor fault identification accuracy and slow response, particularly in high-voltage equipment where missed or false diagnoses are common. This step awakens edge units with trigger signals, automatically extracts multi-source heterogeneous sensor data, performs preprocessing, and constructs a multimodal time-series sample input fault diagnosis model. This method effectively improves diagnostic accuracy and real-time performance, integrates multiple signal information, enhances fault identification and risk assessment capabilities, helps better predict and respond to faults, and improves equipment safety.

[0024] S4. Based on the diagnostic results and risk levels of known switchgear faults, and combined with the multimodal time series fault diagnosis model, a hierarchical response mechanism is executed. Step S4 includes the following: Based on the diagnostic results and risk levels of known switchgear faults, and combined with a multimodal time series fault diagnosis model, a hierarchical response mechanism is implemented. Based on the diagnostic results and risk levels of known switchgear faults, the faults are classified into three emergency levels: low risk, medium risk, and high risk, and corresponding response strategies are implemented.

[0025] Step S4 also includes the following: If the fault is determined to be low-risk, an early warning signal is issued through the remote monitoring system to prompt maintenance personnel to perform planned inspections and maintenance, and to record fault data and update the knowledge base of known switchgear faults in a synchronous manner to support subsequent fault prediction and health management. If the fault is determined to be of medium risk, the on-site audible and visual alarm device will be triggered, and a detailed fault diagnosis report and handling suggestions will be pushed to the mobile terminal of the maintenance personnel to assist them in making accurate on-site intervention and decision support. If a high-risk fault is identified, the protection system will be immediately triggered to perform an emergency trip operation, achieving microsecond-level power cut-off and suppressing the further development of the fault. At the same time, the highest priority alarm will be sent to the central control room, automatically activating the emergency plan and linking the safety control system to start the comprehensive fault isolation and handling process. The hierarchical response mechanism also includes dynamically adjusting response thresholds and strategies based on historical fault handling data and real-time fault conditions to improve the overall fault handling efficiency and operational security of the system.

[0026] When using it, refer to the content of step S4 above: Current fault diagnosis systems generally suffer from insufficient prediction accuracy, lack of hierarchical response mechanisms, and limited real-time optimization capabilities, making it difficult to achieve rapid and accurate early warning and handling when facing complex power system faults. This step constructs a multimodal time series fault diagnosis model, integrates data from multiple sensors, and performs efficient time series analysis to achieve more accurate prediction of fault type and occurrence time. Simultaneously, combined with a hierarchical response mechanism, differentiated response strategies are implemented based on the fault risk level (low-risk, medium-risk, and high-risk), and response thresholds and handling schemes can be dynamically adjusted to ensure efficient and safe fault handling. This mechanism not only improves the accuracy and speed of fault prediction but also optimizes the overall fault handling process, significantly enhancing the safety, reliability, and operation and maintenance management capabilities of the power system.

[0027] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A method for rapid identification and alarm of multiple parameters for fire and explosion prevention in switchgear, characterized in that, include: S1. Obtain historical fault records of known switchgear, analyze the attribute preferences and occurrence patterns of different faults of known switchgear, construct a multimodal time series fault diagnosis model, and generate a knowledge base of known switchgear faults; S2. Based on integrated multi-source heterogeneous sensors, the known operating parameters of the switchgear are monitored in real time. The primary judgment logic is solidified using FPGA field programmable gate array. If the current change rate and the arc signal intensity both exceed the preset threshold within a microsecond time window, a primary fault trigger signal is generated. S3. Based on the generation of the primary fault trigger signal, the edge intelligent computing unit automatically extracts the original data of multi-source heterogeneous sensors within a set time window before and after the trigger time, packages them and substitutes them into the multimodal time series fault diagnosis model for correlation matching, and obtains the diagnosis results and risk level of known switchgear faults. S4. Based on the diagnostic results and risk levels of known switchgear faults, and combined with a multimodal time series fault diagnosis model, a hierarchical response mechanism is implemented.

2. The method for rapid identification and alarm of multiple parameters for fire and explosion prevention of switchgear according to claim 1, characterized in that, S1 includes: Obtain historical fault data of known switchgear, perform data preprocessing, and label the fault type and severity level for each known switchgear fault sample; The acquisition of known historical fault data of switchgear includes: three-phase current waveform sequence, arc intensity sequence, gas concentration sequence, infrared temperature sequence and ultrasonic signal sequence; Using the DTW dynamic time warping algorithm, timestamps are aligned to generate time series of known switchgear fault samples; The distance metric between two sample sequences is calculated using the Euclidean distance formula, and a distance matrix between the two sample sequences is constructed. By using the recursive relationship of dynamic time warping, the cumulative minimum distance of each grid point in the distance matrix is ​​calculated step by step, and the optimal alignment path between the two sequences is determined by backtracking. Based on determining the optimal alignment path between two sequences through backtracking, the known historical fault data of switchgear are uniformly aligned to the same reference timestamp, and a multimodal sample matrix with timestamp alignment is constructed.

3. The method for rapid identification and alarm of multiple parameters for fire and explosion prevention of switchgear according to claim 2, characterized in that, S1 further includes: Based on the time series of known switchgear fault samples, the mean and variance of the time series of known switchgear fault samples are calculated to characterize the average level and dispersion of the time series of known switchgear fault samples, obtain the peak value and skewness of the time series of known switchgear fault samples, and extract the time domain features of each known switchgear fault sample. Using the time-domain characteristics of each known switchgear fault sample as the input signal, a fast Fourier transform is performed to convert the input signal from the time domain to the frequency domain, and the spectral representation of each known switchgear fault sample is calculated. Based on the spectral representation of each known switchgear fault sample, the amplitude and phase information of each known switchgear fault sample in the frequency domain are obtained, and the frequency domain features of each known switchgear fault sample are extracted. The time-domain and frequency-domain features of each known switchgear fault sample are spliced ​​together and normalized to construct a manual feature vector for each known switchgear fault sample.

4. The method for rapid identification and alarm of multiple parameters for fire and explosion prevention of switchgear according to claim 3, characterized in that, S1 further includes: A dual-path parallel processing architecture is implemented, using a timestamp-aligned multimodal sample matrix and manually generated feature vectors of known switchgear fault samples as inputs. Based on the manual feature vectors of known switchgear fault samples, a multilayer perceptron is designed, and linear and nonlinear transformations are used as activation functions to extract the traditional feature paths of each known switchgear fault sample. Based on the timestamp-aligned multimodal sample matrix, the projection of the time series of each known switchgear fault sample into an embedding vector is obtained, thus obtaining the position encoding information of the time series of each known switchgear fault sample. The Transformer encoding layer is designed, which takes the position encoding information of the time series of each known switch cabinet fault sample as input, and uses the self-attention mechanism to capture the dependency relationship between the time series of each known switch cabinet fault sample and extract the self-attention feature path of each known switch cabinet fault sample. The traditional feature path and the self-attention feature path are weighted and summed to obtain the fusion feature vector of each known switchgear fault sample; Using the fused feature vectors of each known switchgear fault sample as input, and the probability distribution of the fault type and the predicted risk level of each known switchgear fault sample as output, a multimodal time series fault diagnosis model is constructed to generate a knowledge base of known switchgear faults.

5. The method for rapid identification and alarm of multiple parameters for fire and explosion prevention of switchgear according to claim 1, characterized in that, S2 includes: Based on integrated multi-source heterogeneous sensors, the known operating parameters of the switchgear are monitored in real time, and the known switchgear current signal and arc signal are acquired. The signals are converted into digital signals by analog-to-digital converters and synchronously input to multiple independent processing pipelines inside the FPGA field programmable gate array for data preprocessing. Calculate the ratio of the difference in current values ​​between adjacent sampling points of a known switchgear to the sampling time interval to obtain the real-time current change rate of the known switchgear; Using a hardware description language, a basic judgment logic is constructed to monitor the current change rate and arc signal intensity in real time; Calculate the ratio of the difference in current values ​​between adjacent sampling points of a known switchgear to the sampling time interval to obtain the real-time current change rate of the known switchgear; If the current change rate and the arc signal intensity are both detected to exceed their respective preset thresholds within a microsecond time window, it is determined that the primary fault triggering condition is met, and a primary fault triggering signal is generated. The preset threshold for the current change rate is 5A per microsecond; the preset threshold for the arc signal intensity is 500mV.

6. The method for rapid identification and alarm of multiple parameters for fire and explosion prevention of switchgear according to claim 5, characterized in that, S3 includes: The generation of a primary fault trigger signal is used as a hardware interrupt trigger to wake up the edge intelligent unit and start the data capture module. The data capture module automatically retrieves the original data from the multi-source heterogeneous sensors within a set time window based on the trigger time, including current waveforms, arc intensity, gas concentration, infrared temperature, and ultrasonic signals. Based on the acquisition of raw data from multi-source heterogeneous sensors, data preprocessing is performed to construct multi-modal time-series samples of multi-source heterogeneous sensors; Using multi-modal time-series samples from multi-source heterogeneous sensors as input, the model is substituted into a multi-modal time-series fault diagnosis model. The Sigmoid activation function is used, and the diagnosis results and risk levels of known switchgear faults are used as outputs. The risk level R includes: low risk: 0 ≤ R < 0.4, medium risk: 0.4 ≤ R < 0.7; high risk: 0.7 ≤ R < 1.

0.

7. The method for rapid identification and alarm of multiple parameters for fire and explosion prevention of switchgear according to claim 6, characterized in that, The risk level formula is as follows: ； in, Given the known fault risk level of the switchgear, For the risk emergence function, Given the current state vector of a known switchgear fault, Given the known risk sharpness parameters for switchgear, The fault categories inferred by the multimodal time series fault diagnosis model The probability, Fault Category Typical pattern vectors, Given the current state vector of the switchgear fault. With the fault prototype The KL divergence between them.

8. The method for rapid identification and alarm of multiple parameters for fire and explosion prevention of switchgear according to claim 6, characterized in that, S4 includes: Based on the diagnostic results and risk levels of known switchgear faults, and combined with a multimodal time series fault diagnosis model, a hierarchical response mechanism is implemented. Based on the diagnostic results and risk levels of known switchgear faults, the faults are classified into three emergency levels: low risk, medium risk, and high risk, and corresponding response strategies are implemented.

9. A method for rapid identification and alarm of multiple parameters for fire and explosion prevention in switchgear according to claim 8, characterized in that, S4 further includes: If the fault is determined to be low-risk, an early warning signal is issued through the remote monitoring system to prompt maintenance personnel to perform planned inspections and maintenance, and to record fault data and update the knowledge base of known switchgear faults in a synchronous manner to support subsequent fault prediction and health management. If the fault is determined to be of medium risk, the on-site audible and visual alarm device will be triggered, and a detailed fault diagnosis report and handling suggestions will be pushed to the mobile terminal of the maintenance personnel to assist them in making accurate on-site intervention and decision support. If a high-risk fault is identified, the protection system will be immediately triggered to perform an emergency trip operation, achieving microsecond-level power cut-off and suppressing the further development of the fault. At the same time, the highest priority alarm will be sent to the central control room, automatically activating the emergency plan and linking the safety control system to start the comprehensive fault isolation and handling process. The hierarchical response mechanism also includes dynamically adjusting response thresholds and strategies based on historical fault handling data and real-time fault conditions to improve the overall fault handling efficiency and operational security of the system.

Images