Anti-explosion electrical cabinet warning method and device

Abstract

The invention discloses an explosion-proof electrical cabinet warning method and device, and relates to the technical field of electrical control, and the method comprises the steps: reconstructing and extracting pressure features based on the pressure data in an electrical cabinet through adaptive compressed sensing and an iterative hard threshold algorithm, and constructing a dynamic pressure model; evaluating the pressure state in the cabinet in real time through a dynamic pressure model in combination with a deep reinforcement learning algorithm and a fuzzy cognitive map theory, and generating an intelligent control strategy; and according to the real-time evaluation result and the intelligent control strategy, generating a graded warning signal by using a multi-stage audio-visual tactile feedback mechanism. According to the invention, accurate modeling and prediction of pressure change are realized, reliable data support is provided for pressure monitoring, abnormal warning and control strategy formulation of the explosion-proof electrical cabinet by analyzing the nonlinear and non-stationary characteristics of pressure data, and the safety and reliability of the explosion-proof electrical cabinet are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of electrical control technology, and more specifically, to a method and device for warning of explosion-proof electrical cabinets. Background Technology

[0002] Intelligentization and digitalization have become major trends in the development of industrial equipment. In high-risk industries such as petroleum, chemical, and mining, the safety and reliability of explosion-proof electrical equipment are paramount. As key safety equipment, explosion-proof electrical cabinets not only need to meet stringent explosion-proof requirements but also require intelligent and visual features to adapt to the needs of modern industry. Visualized intelligent explosion-proof electrical cabinets, as a new generation of explosion-proof equipment, integrate advanced sensing technology, data analysis algorithms, and human-machine interfaces. They can monitor, analyze, and display various parameters within the cabinet in real time, such as pressure, temperature, and humidity, thereby improving the safety, reliability, and maintainability of the equipment. This intelligent upgrade not only improves work efficiency but also significantly reduces the probability of safety accidents, providing strong technical support for safe production in high-risk industries.

[0003] Currently, research on visualized intelligent explosion-proof electrical cabinets mainly focuses on several aspects: First, the application of sensor technology, using various high-precision sensors installed inside the cabinet to achieve real-time monitoring of key parameters such as pressure, temperature, and humidity; second, data acquisition and transmission technology, utilizing industrial IoT technology to transmit collected data to the control center for processing and analysis in real time; third, data visualization technology, intuitively displaying equipment operating status and various parameters through a graphical interface; in addition, some research also involves the application of artificial intelligence and machine learning algorithms for equipment status prediction and fault diagnosis. The combination of these technologies endows explosion-proof electrical cabinets with a certain degree of intelligence and visualization capabilities, providing operators with more intuitive and comprehensive equipment operating information, helping to promptly identify potential problems and take corresponding measures.

[0004] However, existing visual intelligent explosion-proof electrical cabinet technologies still have some significant limitations. First, most existing technologies focus only on data collection and simple display, lacking in-depth analysis of the data's inherent structure and dynamic characteristics, resulting in an inability to accurately capture and predict complex pressure change patterns. Second, existing visualization methods often display parameters in isolation, ignoring the interrelationships between parameters and the dynamic behavior of the overall system, making it difficult for operators to fully grasp the equipment's operating status. Furthermore, current intelligent functions are mostly limited to simple threshold alarms and basic trend analysis, lacking the ability to understand and predict long-term dynamic behavior, making it difficult to cope with sudden anomalies and dynamic changes in the system. These problems severely restrict the predictive accuracy and alarm decision support capabilities of visual intelligent explosion-proof electrical cabinets, affecting their application effectiveness in actual industrial environments.

[0005] No effective solutions have yet been proposed to address the problems in the relevant technologies. Summary of the Invention

[0006] To address the problems in related technologies, this invention proposes an alarm method and device for explosion-proof electrical cabinets. This method enables accurate modeling and prediction of pressure changes, handling not only the nonlinear and non-stationary characteristics of pressure data but also adapting to dynamic pressure changes within the cabinet. It provides reliable data support and decision-making basis for pressure monitoring, anomaly alarms, and control strategy formulation for explosion-proof electrical cabinets, effectively improving the safety and reliability of these cabinets. Furthermore, it solves the problems of existing technologies that focus solely on data acquisition and simple display, lacking in-depth analysis of the data's internal structure and dynamic characteristics, displaying parameters in isolation, ignoring the interrelationships between parameters and the dynamic behavior of the overall system. This makes it difficult for operators to fully grasp the equipment's operating status, lacking understanding and predictive ability of long-term dynamic behavior, and struggling to cope with sudden anomalies and dynamic changes in the system.

[0007] Therefore, the specific technical solution adopted by the present invention is as follows: According to one aspect of the present invention, a method for warning of explosion-proof electrical cabinets is provided, the method comprising the following steps: S1. Based on the pressure data inside the electrical cabinet, the pressure features are reconstructed and extracted using adaptive compressed sensing and iterative hard thresholding algorithm to build a dynamic pressure model. S2. By using a dynamic pressure model, combined with deep reinforcement learning algorithms and fuzzy cognitive graph theory, the pressure state inside the cabinet is evaluated in real time, and an intelligent control strategy is generated. S3. Based on real-time evaluation results and intelligent control strategies, generate graded warning signals using a multi-level audiovisual and tactile feedback mechanism; Based on pressure data inside the electrical cabinet, a dynamic pressure model is constructed by reconstructing and extracting pressure features using adaptive compressed sensing and iterative hard thresholding algorithms, including the following steps: S11. Calculate the self-similarity index of the stress data and construct an adaptive compressed sensing matrix to generate reconstructed stress data; S12. Based on the reconstructed pressure data, pressure features are extracted nonlinearly and reduced in dimensionality using topological pressure manifold and continuous tomography algorithm. The pressure features include topological features and geometric features. S13. Based on the pressure characteristics, construct a dynamic pressure model using a nonlinear autoregressive exogenous model.

[0008] Furthermore, the self-similarity index of the stress data is calculated, and an adaptive compressed sensing matrix is ​​constructed to generate reconstructed stress data, including the following steps: S111. Decompose the collected pressure data at multiple scales to obtain pressure fractal components at different scales, and calculate the corresponding pressure data self-similarity index. The calculation expression is as follows: ; In the formula, H i The pressure data self-similarity index represents the pressure data for each pressure fractal component. R This represents the range of the pressure fractal component, that is, the difference between the maximum and minimum values; S This indicates that at the same time scale t Internal, pressure fractal component F i(t) Standard deviation; t Indicates a time scale; F i(t) Indicates at any point in time t No. i Pressure fractal components at various scales; m Represents pressure fractal components F i(t) In time scale t The average value within; n Representing time scale t The number of sampling points within; S112. Based on the self-similarity index of pressure data, construct an adaptive compressed sensing matrix and adaptively sample the pressure fractal components at different scales. S113. Based on the adaptive compressed sensing matrix, the pressure fractal components at different scales are reconstructed using an iterative hard thresholding algorithm, and then merged to obtain the reconstructed pressure data.

[0009] Furthermore, based on the reconstructed pressure data, pressure features are nonlinearly extracted and dimensionality reduced using a topological pressure manifold and a persistent tomography algorithm, including the following steps: S121. Construct a high-dimensional manifold representation of the pressure data and apply discrete Morse theory to analyze the critical point structure of the pressure manifold; S122. Construct a complex sequence of pressure data points and identify topological features in the pressure data; S123. Reconstruct the local structure of the pressure manifold using tomography and extract its geometric features.

[0010] Furthermore, based on the stress characteristics and a nonlinear autoregressive exogenous model, a dynamic stress model is constructed, including the following steps: S131. Based on topological and geometric features, construct a feature set and design the structure of the dynamic pressure model to obtain the nonlinear function; S132. Using deep neural networks to learn nonlinear functions; S133. With the goal of minimizing the prediction error, a dynamic stress model is trained using the backpropagation algorithm, and the hyperparameters of the dynamic stress model are optimized using cross-validation.

[0011] Furthermore, by using a dynamic pressure model, combined with deep reinforcement learning algorithms and fuzzy cognitive graph theory, the pressure state inside the cabinet is assessed in real time, and an intelligent control strategy is generated, including the following steps: S21. Based on the dynamic pressure model, the deep Q-network algorithm is used to evaluate the pressure status inside the cabinet; S22. Based on the stress state assessment results, use fuzzy cognitive map theory to predict potential risks inside the cabinet; S23. Based on the assessment results of the pressure state and the prediction results of potential risks, a fuzzy logic controller is used to generate an intelligent control strategy.

[0012] Furthermore, based on the dynamic pressure model, the deep Q-network algorithm is used to evaluate the pressure state inside the cabinet, including the following steps: S211. Define the state space and action space based on the actual operating parameters of the explosion-proof electrical cabinet; S212. Based on the dynamic pressure model and combined with actual operating parameters, design a reward function; S213. Using a deep neural network, construct a Q-function approximator and optimize the network parameters through empirical replay and target network methods.

[0013] Furthermore, based on the stress state assessment results, and using fuzzy cognitive mapping theory, the prediction of potential risks within the cabinet includes the following steps: S221. Based on the risk factors of explosion-proof electrical cabinets, define the set of concept nodes in a fuzzy cognitive graph; S222. Establish a weight matrix based on the influence between concept nodes; S223. Using iterative calculation and activation functions, update the concept node values ​​and predict potential risks by analyzing steady-state values.

[0014] Furthermore, based on the assessment results of the stress state and the prediction results of potential risks, a fuzzy logic controller is used to generate an intelligent control strategy, including the following steps: S231. Based on the pressure control requirements, define the input and output linguistic variables of the fuzzy controller; S232. Design a fuzzy rule base and construct an intelligent control strategy; S233. Based on intelligent control strategies, the control quantity is determined using fuzzy reasoning and defuzzification methods.

[0015] Furthermore, based on real-time evaluation results and intelligent control strategies, a multi-level audiovisual and tactile feedback mechanism is used to generate graded warning signals, including the following steps: S31. Based on the stress status assessment results and potential risk prediction results, construct a multi-level warning mechanism; S32. Design multi-layered alarms using visual, auditory, and tactile feedback; S33. Generate fault diagnosis information and handling suggestions based on real-time pressure values ​​and trends.

[0016] According to another aspect of the present invention, a visual intelligent explosion-proof electrical cabinet warning device is also provided, the visual intelligent explosion-proof electrical cabinet warning device comprising: The feature extraction and model building module is used to reconstruct and extract pressure features based on pressure data inside the electrical cabinet using adaptive compressed sensing and iterative hard thresholding algorithms, and to build a dynamic pressure model. The state assessment and strategy generation module is used to assess the pressure state inside the cabinet in real time and generate intelligent control strategies by using a dynamic stress model, combined with deep reinforcement learning algorithms and fuzzy cognitive graph theory. The intelligent control and hierarchical alarm module is used to generate hierarchical alarm signals based on real-time evaluation results and intelligent control strategies, using a multi-level audiovisual and tactile feedback mechanism. The feature extraction and model building module is connected to the state assessment and strategy generation module, and the state assessment and strategy generation module is connected to the intelligent control and hierarchical alarm module.

[0017] The beneficial effects of this invention are as follows: (1) This invention constructs a model that can accurately capture the dynamic characteristics of pressure changes inside explosion-proof electrical cabinets by integrating the topological features, geometric features, and temporal characteristics of pressure data. This invention uses time delay embedding technology to map pressure data to a high-dimensional space, and applies discrete Morse theory and continuous homology analysis to extract the topological features of the data. It reconstructs the local structure of the data and extracts geometric features through manifold learning and tomography. At the same time, based on the extracted pressure features, this invention designs an improved nonlinear autoregressive exogenous model structure and uses deep neural network to learn the nonlinear function in the model to achieve accurate modeling and prediction of pressure changes. It can not only handle the nonlinear and non-stationary characteristics of pressure data, but also adapt to the dynamic changes of pressure inside the cabinet. It provides reliable data support and decision-making basis for pressure monitoring, abnormal alarm and control strategy formulation of explosion-proof electrical cabinets, and effectively improves the safety and reliability of explosion-proof electrical cabinets.

[0018] (2) This invention has significant advantages in pressure feature extraction. By mapping pressure data to a high-dimensional space and applying topological data analysis techniques, it can capture data structures and patterns that are difficult to identify using traditional time series analysis methods. In addition, the application of discrete Morse theory enables this invention to identify key points in pressure data, including local extrema and saddle points. These points often correspond to important state transitions of the pressure system. Continuous cohomology analysis provides a multi-scale perspective and can identify stable topological features in pressure data. These features reflect the essential patterns of pressure changes. At the same time, through manifold learning and tomography techniques, this invention can also reconstruct the local geometric structure of pressure data and extract geometric features such as curvature and geodesic distance. These features can reflect the local and global characteristics of pressure changes. Through this multi-dimensional and multi-scale feature extraction method, the internal structure and dynamic characteristics of pressure data can be fully and deeply understood.

[0019] (3) This invention designs a dynamic pressure model through a nonlinear autoregressive exogenous model structure and uses a deep neural network to learn the nonlinear functions in it, thereby achieving accurate capture of the complex dynamic characteristics of the pressure system. It not only considers the influence of historical pressure values ​​and external inputs, but also incorporates the aforementioned extracted topological and geometric features, so that the dynamic pressure model can comprehensively consider various influencing factors of pressure changes. At the same time, the application of deep neural networks gives the model powerful nonlinear modeling capabilities and adaptability, and can automatically learn complex patterns and long-term and short-term dependencies in pressure data. Through backpropagation algorithm and cross-validation, the model parameters and hyperparameters are optimized, which can achieve high-precision prediction of pressure changes while ensuring the model's generalization ability. It can not only adapt to the nonlinear and non-stationary characteristics of the pressure system in the explosion-proof electrical cabinet, but also dynamically adjust the model parameters according to real-time data. It can still maintain high prediction accuracy when facing sudden changes or abnormal situations in the system. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a flowchart illustrating an explosion-proof electrical cabinet alarm method according to an embodiment of the present invention; Figure 2 This is a schematic block diagram of an explosion-proof electrical cabinet alarm device according to an embodiment of the present invention. Detailed Implementation

[0022] To further illustrate the various embodiments, the present invention provides accompanying drawings, which are part of the disclosure of the present invention. These drawings are mainly used to illustrate the embodiments and can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. With reference to these drawings, those skilled in the art should be able to understand other possible implementation methods and the advantages of the present invention. The components in the drawings are not drawn to scale, and similar component symbols are generally used to represent similar components.

[0023] According to an embodiment of the present invention, an explosion-proof electrical cabinet warning method and device are provided.

[0024] The present invention will now be further described in conjunction with the accompanying drawings and specific embodiments, such as... Figure 1 As shown, according to an embodiment of the present invention, an explosion-proof electrical cabinet alarm method is provided, which includes the following steps: S1. Based on the pressure data inside the electrical cabinet, the pressure data is reconstructed using adaptive compressed sensing and iterative hard thresholding algorithm, and pressure features are extracted to construct a dynamic pressure model. S2. By using a dynamic pressure model, combined with deep reinforcement learning algorithms and fuzzy cognitive graph theory, the pressure state inside the cabinet is evaluated in real time, and an intelligent control strategy is generated. S3. Based on real-time evaluation results and intelligent control strategies, generate graded warning signals using a multi-level audiovisual and tactile feedback mechanism; Based on the pressure data inside the electrical cabinet, the pressure data is reconstructed using adaptive compressed sensing and iterative hard thresholding algorithms, and pressure features are extracted to construct a dynamic pressure model, including the following steps: S11. Calculate the self-similarity index of the stress data and construct an adaptive compressed sensing matrix to generate reconstructed stress data; S12. Based on the reconstructed pressure data, pressure features are extracted nonlinearly and reduced in dimensionality using topological pressure manifold and continuous tomography algorithm. The pressure features include topological features and geometric features. S13. Based on the pressure characteristics, construct a dynamic pressure model using a nonlinear autoregressive exogenous model.

[0025] In one embodiment, calculating the stress data self-similarity index and constructing an adaptive compressed sensing matrix to generate reconstructed stress data includes the following steps: S111. Decompose the collected pressure data at multiple scales to obtain pressure fractal components at different scales, and calculate the corresponding pressure data self-similarity index. Specifically, firstly, wavelet fractal decomposition is used to process the original pressure data. x ( t Decomposed into N scale components, the expression is: x ( t )=Σ Fi(t) , i =1,2,..., N In the formula, x ( t ) represents any point in time. t Pressure value, F i(t) For any point in time t No. i The pressure data is then divided into fractal components at each scale; subsequently, the pressure data self-similarity index for each pressure fractal component is calculated. H i : ; In the formula, H i The pressure data self-similarity index represents the pressure data for each pressure fractal component. R This represents the range of the pressure fractal component, that is, the difference between the maximum and minimum values; S This indicates that at the same time scale t Internal, pressure fractal component F i(t) Standard deviation; t Indicates a time scale; F i(t) Indicates at any point in time t No. i Pressure fractal components at various scales; m Represents pressure fractal components F i(t) In time scale t The average value within; n Representing time scale t The number of sampling points within.

[0026] S112. Based on the self-similarity index of pressure data, construct an adaptive compressed sensing matrix and adaptively sample the pressure fractal components at different scales. Specifically, firstly, based on the pressure data self-similarity index of each pressure fractal component... H i Establish an adaptive compressed sensing matrix F i The expression is F i = R · H i ( H i · I +(1- H i )· R In the formula, I It is the identity matrix.R The matrix is ​​random; then, adaptive compression sampling is performed on each pressure fractal component, expressed as follows: y i = F i ·F i(t) In the formula, y i This represents the adaptive compression sampling result.

[0027] S113. Based on the adaptive compressed sensing matrix, the pressure fractal components at different scales are reconstructed using an iterative hard thresholding algorithm, and then merged to obtain the reconstructed pressure data.

[0028] Specifically, each component is reconstructed using the Iterative Hard Thresholding (IHT) algorithm, expressed as follows: F' i =IHT( y i , F i , H i In the formula, y i These are compression measurements obtained through compression sensing, representing partial information about the original pressure components. F i It is a compressed sensing matrix, which describes how compressed measurements are obtained from the raw signal; H i It is a pressure data self-similarity index, which reflects the characteristics of this component and is used to guide the reconstruction process; the goal of the IHT algorithm used in this invention is to find a reconstructed signal. F' i This makes it conform to compression measurement y i It maintains appropriate sparsity.

[0029] Specifically, first, ① initialize the reconstructed signal F'i; ② repeat the following steps: a) using the current F' i Calculate compression measurements, compared with actual measurements. y i Compare; b) Update based on comparison results F' i c) Apply threshold operation to retain F' i The most important part; ③ The choice of threshold is dynamic, depending on each component. H i Adjust the value accordingly; H iThe larger the value, the smoother the signal, and the higher the threshold can be set; H i The smaller the value, the more drastic the signal change, and the lower the threshold should be set.

[0030] Specifically, finally, the complete signal is reconstructed through inverse fractal transformation. a) Perform inverse wavelet transform on each reconstructed component, expressed as follows: x '( t )=Σ F' i(t) a) Restore the original signal to its original scale (this process uses the same wavelet basis functions as during decomposition to ensure consistency in reconstruction); b) Add all the restored signal components to obtain the final reconstructed complete pressure data; c) Compare the original pressure data and the reconstructed pressure data, calculate the error between them, and if the error exceeds a preset threshold, adjust the reconstruction parameters and repeat the above process until satisfactory reconstruction quality is achieved.

[0031] In one embodiment, based on reconstructed pressure data, nonlinear extraction and dimensionality reduction of pressure features using a topological pressure manifold and a persistent tomography algorithm includes the following steps: S121. Construct a high-dimensional manifold representation of the pressure data and apply discrete Morse theory to analyze the critical point structure of the pressure manifold; Specifically, the reconstructed pressure data is first mapped to a high-dimensional space to form a pressure manifold. Then, the time series data is converted into point cloud data using the sliding window method, with the expression P(t)=[p(t),p(t+τ0),...,p(t+(n-1)τ0)], where P(t) is a point in the n-dimensional phase space, p(t) is the pressure value at time t, τ0 is the time delay, and n is the embedding dimension.

[0032] Specifically, the discrete Morse theory is then applied to identify critical points on the pressure manifold, including: a) constructing a discrete vector field and calculating the gradient of each data point in each dimension; b) defining a discrete Morse function based on the density of the data points; c) identifying critical points, including local minima (pressure troughs, i.e., the function values ​​of all adjacent points are greater than that point), local maxima (pressure peaks, i.e., the function values ​​of all adjacent points are less than that point), and saddle points (inflection points of pressure change, adjacent points whose function values ​​are both greater than and less than that point in different directions); d) constructing a Morse-Smail complex, i.e., forming the topological skeleton of pressure change by connecting critical points; and e) analyzing the complex structure, i.e., determining the overall topological structure of pressure change by calculating the connection relationships between critical points.

[0033] S122. Construct a complex sequence of pressure data points and identify topological features in the pressure data; Specifically, a complex sequence of pressure data points is constructed. In this embodiment, the Lipsian complex sequence is used. Then, the persistent homology groups of different dimensions are calculated, and a persistent graph PD={(bi,di)} is constructed, where bi is the birth time of the topological feature and di is the extinction time of the topological feature. In this embodiment, features with long durations correspond to stable structures in the pressure data, including persistent pressure fluctuation patterns and stable pressure distribution regions.

[0034] Specifically, by analyzing the persistence graph, topological features in the pressure data are identified, including stable connected components (representing stable pressure regions in this embodiment), persistent toroidal structures (representing periodic pressure fluctuations in this embodiment), and cavity structures (representing stable gaps and abnormal regions in the pressure distribution in this embodiment), and the length of the persistence interval corresponding to these topological features is recorded.

[0035] S123. Reconstruct the local structure of the pressure manifold using tomography and extract its geometric features.

[0036] Specifically, a manifold learning algorithm (Isomap algorithm is used in this embodiment) is applied to project the high-dimensional pressure manifold into a low-dimensional space. Then, tomography is used to reconstruct the local structure of the manifold, which is achieved through Radon transform. The expression is Rf(s,θ)=∫Lf(x,y)dL, where Rf(s,θ) is the result of Radon transform, representing the line integral value along a specific direction; s is the distance between the integral line and the origin; θ is the angle between the integral line and the positive x-axis; ∫ represents the integration operation; L is the integration path, i.e., the straight line passing through the point (scosθ,ssinθ) and perpendicular to the direction (cosθ,sinθ); f(x,y) is a function on the pressure manifold, representing the pressure value at the point (x,y); and dL represents the small line element along the integration path L.

[0037] Specifically, the two-dimensional pressure distribution function f(x,y) is transformed into its line integrals in different directions using the Radon transform, which helps to capture the directional characteristics and local structure of the pressure distribution. By calculating the Rf(s,θ) values ​​for different s and θ, the integral characteristics of the pressure manifold in various directions and locations can be obtained, providing a basis for subsequent local structure reconstruction.

[0038] Specifically, the local structure is reconstructed using a back-projection algorithm, and geometric features are calculated, including principal curvature (reflecting the severity of pressure changes in this embodiment), Gaussian curvature (representing the overall shape of the pressure distribution in this embodiment), and geodesic distance (used to measure the similarity between different pressure states in this embodiment). These features combine the topological and geometric information of the pressure data, providing a comprehensive feature representation for subsequent pressure state assessment and anomaly detection.

[0039] In one embodiment, constructing a dynamic stress model based on a nonlinear autoregressive exogenous model, according to stress characteristics, includes the following steps: S131. Based on topological and geometric features, construct a feature set and design the structure of the dynamic pressure model to obtain the nonlinear function; It should be noted that the nonlinear autoregressive exogenous model (NARX model) is a nonlinear autoregressive exogenous model suitable for modeling dynamic systems. It can effectively capture the nonlinear characteristics and time dependence of pressure changes in explosion-proof electrical cabinets, while integrating the influence of external environmental factors. The model has a flexible structure and can be combined with deep learning technology to provide powerful predictive capabilities. At the same time, the NARX model can handle long-term dependencies and system noise while maintaining certain physical meaning, and has high computational efficiency. These characteristics make it an ideal choice for dynamic pressure modeling of explosion-proof electrical cabinets in this application, providing a reliable foundation for achieving high-precision pressure state assessment and control.

[0040] Specifically, first, a feature set F = {f1, f2, ..., fn} is constructed, where f1, f2, and f3 are the lengths of the principal persistent intervals of 0-dimensional, 1-dimensional, and 2-dimensional persistent homology, respectively, corresponding to stable connected components, circular ring structures, and cavity structures in the topological features; f4 is the persistent entropy, which quantifies the complexity of the pressure data topological structure; f5 and f6 are the mean and variance of the principal curvatures in the geometric features, respectively, representing the severity of pressure changes; f7 is the average value of the Gaussian curvature in the geometric features, reflecting the overall shape of the pressure distribution; and f8 is the maximum value of the geodesic distance in the geometric features, representing the maximum similarity difference between different pressure states.

[0041] Specifically, a nonlinear autoregressive exogenous model is then used as the basic framework to design a dynamic pressure model structure and establish a nonlinear function, expressed as y(t)=f(y(t-1),y(t-2),y(t-3),u(t-1),u(t-2),F)+ε(t), where y(t) is the pressure value at time t, u(t) is the external input (temperature in this embodiment), F is the aforementioned feature set, and ε(t) is the error term.

[0042] S132. Using deep neural networks to learn nonlinear functions; Specifically, a) a deep neural network structure is designed as the implementation of the nonlinear function f in the dynamic stress model, including an input layer (in this embodiment, it is used to receive historical stress values ​​P(t-1),...,P(t-np), external input u(t-1),...,u(t-nu), and feature set F), a hidden layer (in this embodiment, it is used to use a recurrent neural network LSTM to capture the nonlinear dynamic characteristics of stress changes), and an output layer (in this embodiment, it is used to generate the predicted future stress value P(t)); b) network parameters are determined, including the number of layers and the number of neurons per layer, to meet the needs of dynamic stress modeling; c) an appropriate activation function is selected (in this embodiment, the ReLU function is used) to enhance the model's ability to express nonlinear relationships; d) the forward propagation algorithm is implemented so that the deep neural network can generate stress prediction values ​​based on the input.

[0043] S133. With the goal of minimizing the prediction error, a dynamic stress model is trained using the backpropagation algorithm, and the hyperparameters of the dynamic stress model are optimized using cross-validation.

[0044] Specifically, a) Define the loss function, using the mean squared error (MSE) as the loss function, expressed as MSE = (1 / N) Σ(y true -y pred )^2, where N is the sample size, y true y represents the actual pressure value. pred a) Model prediction values; b) Algorithm optimization using the Adam optimizer with an initial learning rate of 0.001; c) Training process: First, the dataset is divided into a training set (70%), a validation set (15%), and a test set (15%). Then, mini-batch gradient descent is used, with a batch size of 64 in this embodiment. Finally, the number of training epochs is set to 200, and an early stopping strategy is used, i.e., training stops if the validation set loss does not improve within 20 epochs; d) Cross-validation to optimize hyperparameters using 5-fold cross-validation. Optimized hyperparameters include the number of hidden layers (as described in the above embodiment). The parameters are: [3,4,5], number of neurons per layer ([16,32,64,128] in the above embodiment), learning rate ([0.1,0.01,0.001,0.0001] in the above embodiment), and batch size ([32,64,128] in the above embodiment). Then, a grid search method is used to traverse all hyperparameter combinations and select the combination with the best performance on the validation set. e) Model evaluation: Root mean square error (RMSE) and mean absolute percentage error (MAPE) are calculated on the test set, and a comparison chart of predicted and actual values ​​is drawn to intuitively evaluate the model performance.

[0045] In one embodiment, a dynamic pressure model, combined with deep reinforcement learning algorithms and fuzzy cognitive graph theory, is used to assess the pressure state inside the cabinet in real time and generate an intelligent control strategy, including the following steps: S21. Based on the dynamic pressure model, the deep Q-network algorithm is used to evaluate the pressure status inside the cabinet; S22. Based on the stress state assessment results, use fuzzy cognitive map theory to predict potential risks inside the cabinet; S23. Based on the assessment results of the pressure state and the prediction results of potential risks, a fuzzy logic controller is used to generate an intelligent control strategy.

[0046] In one embodiment, evaluating the pressure state inside the cabinet based on a dynamic pressure model and using a deep Q-network algorithm includes the following steps: S211. Define the state space and action space based on the actual operating parameters of the explosion-proof electrical cabinet; Specifically, the state space is defined as S={s1,s2,s3,s4}, where s1 represents the pressure value inside the cabinet, s2 represents the temperature inside the cabinet, s3 represents the sealing strip pressure, and s4 represents the cabinet vibration intensity.

[0047] Specifically, the action space A={a1,a2,a3} is defined, where a1 represents adjusting the sealing strip pressure, a2 represents starting the auxiliary pressurization system, and a3 represents triggering the alarm.

[0048] S212. Based on the dynamic pressure model and combined with actual operating parameters, design a reward function; Specifically, a reward function R(s,a) is designed to reflect the effect of the action on the positive pressure maintenance effect of the explosion-proof electrical cabinet: R(s,a)=w1 (PP tg )²+w2 E+w3 T, where P is the current pressure, P tg Let E be the target pressure, T be the energy consumption, and w1, w2, and w3 be the weighting coefficients corresponding to the pressure deviation, energy consumption, and temperature change, respectively. In this embodiment, w1 = -0.6, w2 = -0.3, and w3 = -0.1.

[0049] S213. Using a deep neural network, construct a Q-function approximator and optimize the network parameters through empirical replay and target network methods.

[0050] Specifically, construct a Q-network: Q(S,A;θ)=f(S,A;θ), where θ is the network parameter and f is the deep neural network function.

[0051] Specifically, the Q-network is optimized using experience replay and the target network, and the constructed loss function is L(θ)=E[(r+γmax_A'Q(S',A';θ-)-Q(S,A;θ))²], where E represents the expected value, which is approximated in practice by the average of multiple samples (64 samples in this embodiment); r is the immediate reward, which is equal to the function value of the reward function R(s,a); γ is the discount factor (0.95 in this embodiment); max_A' represents the selection of the Q-value with the largest value among all possible next actions A'; S' is the next state; A' is the possible action of the next state; and θ- is the target network parameter (100 in this embodiment).

[0052] In one embodiment, predicting potential risks within the cabinet based on stress state assessment results and using fuzzy cognitive mapping theory includes the following steps: S221. Based on the risk factors of explosion-proof electrical cabinets, define the set of concept nodes in a fuzzy cognitive graph; Specifically, based on the theory of fuzzy cognitive graph (FCM), the FCM concept node set C={C1,C2,C3,C4} is defined, where C1 represents the degree of pressure abnormality, C2 represents the degree of temperature abnormality, C3 represents the degree of aging of the sealing strip, and C4 represents the degree of external environmental interference. These nodes are directly related to the output of the dynamic pressure model in step S1 and the critical state of the explosion-proof electrical cabinet.

[0053] S222. Establish a weight matrix based on the influence between concept nodes; Specifically, a weight matrix W is constructed. In this embodiment, W = [0 0.5 0.3 0.2; 0.40 0.2 0.1; 0.6 0.30 0.2; 0.3 0.2 0.40]. ij Representing concept C i For C j The intensity and direction of the influence.

[0054] Specifically, in the above embodiments, w 13 =0.3 indicates that the degree of pressure abnormality (C1) has a positive impact of 0.3 on the degree of aging of the sealing strip (C3), that is, pressure abnormality may accelerate the aging of the sealing strip; the weight value ranges from [-1,1], with positive values ​​indicating positive effects, negative values ​​indicating negative effects, and 0 indicating no direct effect.

[0055] S223. Using iterative calculation and activation functions, update the concept node values ​​and predict potential risks by analyzing steady-state values.

[0056] Specifically, the activation function f(x) is designed. In this embodiment, the hyperbolic tangent function f(x) = (e^xe^(-x)) / (e^x+e^(-x)) is used.

[0057] Specifically, iteratively update the concept node values: A i(t+1) =f(A i(t) +Σw ji A j(t) In the formula, Ai(t+1) represents the concept C at time t+1. i The activation value, A i(t) The concept of time t, C i The activation value, w ji Representing concept C j For C i Influence weight, A j(t) The concept of time t, C j The activation value is ∑, which represents the summation over all concept nodes j ≠ i. This update process uses the output of the dynamic stress model in step S1 as the initial value to simulate the interaction between various risk factors. Specifically, analyze the steady-state value and assess the degree of potential risk: Risk index RI = v1 A 1(∞) +v2 A 2(∞) +v3 A 3(∞) +v4 A 4(∞) In the formula, v i As the weight of each concept node, A i(∞) As steady-state values, in this embodiment, v1=0.4, v2=0.3, v3=0.2, and v4=0.1, respectively reflecting the contribution of the risk factors of each conceptual node (C1 pressure anomaly degree, C2 temperature anomaly degree, C3 sealing strip aging degree, and C4 external environmental interference degree) to the overall risk. In the above embodiment, the pressure anomaly degree (v1=0.4) is considered the most important risk factor because it is directly related to the core function of the explosion-proof electrical cabinet; the temperature anomaly degree (v2=0.3) is the second most important because temperature changes may affect the performance and lifespan of electrical equipment; the sealing strip aging degree (v3=0.2) ranks third because it affects the sealing performance of the cabinet; the external environmental interference degree (v4=0.1), although important, has a relatively small impact, so it is assigned the lowest weight.

[0058] Specifically, this invention uses the output of the dynamic pressure model in step S1 as input to the FCM through the above process. By simulating the complex interactions between various risk factors, it achieves the prediction of potential risks to explosion-proof electrical cabinets. The iterative process of FCM considers the dynamic changes of multiple factors such as pressure, temperature, sealing strip condition, and external environment, providing a comprehensive risk assessment perspective.

[0059] In one embodiment, generating an intelligent control strategy based on the assessment results of the stress state and the prediction results of potential risks, using a fuzzy logic controller, includes the following steps: S231. Based on the pressure control requirements, define the input and output linguistic variables of the fuzzy controller; Specifically, the input language variables are defined, including ① pressure deviation E: {negative large, negative medium, negative small, zero, positive small, positive medium, positive large}, ② pressure change rate EC: {negative large, negative medium, negative small, zero, positive small, positive medium, positive large}.

[0060] Specifically, define output language variables, including ① auxiliary pressurization system power U1: {off, low power, medium power, high power, full power}, ② alarm level U2: {no alarm, minor warning, moderate warning, severe warning, emergency alarm}.

[0061] S232. Design a fuzzy rule base and construct an intelligent control strategy; Specifically, a fuzzy rule base is designed, containing the following rules (example of intelligent control strategy): First control rule R1: If (pressure deviation E is negative and large) and (pressure change rate EC is negative and large), then (auxiliary pressurization system power U1 is full power) and (alarm level U2 is emergency alarm). Second control rule R2: If (pressure deviation E is zero) and (pressure change rate EC is zero), then (auxiliary pressurization system power U1 is off) and (alarm level U2 is no alarm). Third control rule R3: If (pressure deviation E is positive) and (pressure change rate EC is positive), then (auxiliary pressurization system power U1 is off) and (alarm level U2 is moderate warning); Fourth control rule R4: If (pressure deviation E is negative medium) and (pressure change rate EC is negative small), then (auxiliary pressurization system power U1 is medium power) and (alarm level U2 is minor warning); Fifth control rule R5: If (pressure deviation E is positive and small) and (pressure change rate EC is negative and medium), then (auxiliary pressurization system power U1 is low power) and (alarm level U2 is minor warning).

[0062] S233. Based on intelligent control strategies, the control quantity is determined using fuzzy reasoning and defuzzification methods.

[0063] Specifically, in this embodiment, the Mamdani inference method is used for fuzzy inference. The activation intensity of each rule is calculated as follows: μi=min(μE(Ei),μEC(ECi)), where μE(Ei) and μEC(ECi) are the membership degrees of the inputs E and EC on their respective fuzzy sets, respectively.

[0064] Specifically, the centroid method is used for defuzzification to obtain the precise control quantity: U1=(Σiμi u1i) / (Σiμi); U2=(Σiμi) u2i) / (Σiμi); where μi is the activation intensity of the i-th rule, and u1i and u2i are the output values ​​of the auxiliary pressurization system power and alarm level for the corresponding rule, respectively.

[0065] Specifically, the control strategy includes a) adjusting the power output of the auxiliary pressurization system based on the U1 value obtained from defuzzification; and b) triggering an alarm signal of the corresponding level based on the U2 value obtained from defuzzification.

[0066] Specifically, in this embodiment, the control cycle is set to 1 second, and a control decision is made every second based on the current pressure status and trend. At the same time, upper and lower pressure thresholds are set, and when the pressure exceeds the safe range, emergency protection measures are immediately triggered, including cutting off the power supply and starting emergency exhaust.

[0067] In one embodiment, generating graded warning signals using a multi-level audiovisual and tactile feedback mechanism based on real-time evaluation results and intelligent control strategies includes the following steps: S31. Based on the stress status assessment results and potential risk prediction results, construct a multi-level warning mechanism; Specifically, based on the stress state assessment and risk prediction results in step S2, a multi-level warning mechanism (five-level warning mechanism) is designed, including: Under normal conditions (risk index RI < 0.2), no alarm is required. A minor warning (0.2 ≤ Risk Index RI < 0.4) is issued, followed by a low-level visual cue. Moderate warning (0.4 ≤ Risk Index RI < 0.6), at which point moderate visual and auditory cues are provided; A severe warning (0.6 ≤ Risk Index RI < 0.8) is issued, at which point high-level visual and auditory cues are provided; In an emergency (risk index RI ≥ 0.8), emergency visual, auditory, and tactile feedback should be provided.

[0068] S32. Design multi-layered alarms using visual, auditory, and tactile feedback; Specifically, based on the multi-level alarm mechanism, multiple alarm layers are designed, including normal state alarm, minor warning alarm, moderate warning alarm, severe warning alarm and emergency state alarm; Normal status alarm: At this time, the electrical cabinet displays a green indicator light, indicating that the system is operating normally; A minor warning alarm will sound, at which point the electrical cabinet will display a yellow indicator light accompanied by a soft beeping sound to alert the operator. A moderate warning alarm will be triggered, at which point the electrical cabinet will display an orange indicator light accompanied by a loud beeping sound, prompting the operator to check the system. A serious warning alarm will sound, at which point the electrical cabinet will display a red indicator light, accompanied by a continuous beeping sound and an error code displayed on the screen; Emergency alarm: The electrical cabinet activates a red rotating warning light, emits a high-decibel alarm sound, and alerts the operator via mobile phone vibration. At the same time, the display screen shows emergency fault information and operational instructions.

[0069] S33. Generate fault diagnosis information and handling suggestions based on real-time pressure values ​​and trends.

[0070] Specifically, firstly, pressure values ​​and trend graphs are collected and displayed in real time so that operators can intuitively understand the current status; when an abnormality is detected, the cause of the fault is automatically analyzed, and possible causes of the fault and corresponding handling suggestions are displayed on the screen.

[0071] Specifically, the fault diagnosis information includes, but is not limited to: the cause of the abnormal pressure (including poor sealing, temperature changes), possible impacts (including equipment damage, safety hazards, etc.), and recommended troubleshooting steps (including checking door seals and adjusting temperature settings); the handling suggestions are presented in a combination of text and graphics to provide clear guidance and help operators solve problems quickly and accurately.

[0072] Specifically, through the above steps, operators can clearly receive real-time status information of the electrical cabinet and take corresponding measures after receiving an alarm to ensure the safe and stable operation of the electrical cabinet.

[0073] like Figure 2 As shown, according to another embodiment of the present invention, a visual intelligent explosion-proof electrical cabinet warning device is also provided, the visual intelligent explosion-proof electrical cabinet warning device comprising: Feature extraction and model building module 1: Based on the pressure data collected inside the electrical cabinet, the pressure data is reconstructed using adaptive compressed sensing and iterative hard thresholding algorithm, and pressure features are extracted to build a dynamic pressure model; The state assessment and strategy generation module 2 is used to assess the pressure state inside the cabinet in real time and generate intelligent control strategies by using a dynamic pressure model, combined with deep reinforcement learning algorithms and fuzzy cognitive graph theory. The intelligent control and hierarchical alarm module 3 is used to generate hierarchical alarm signals based on real-time evaluation results and intelligent control strategies using a multi-level audiovisual and tactile feedback mechanism. The feature extraction and model building module 1 is connected to the state evaluation and strategy generation module 2, and the state evaluation and strategy generation module 2 is connected to the intelligent control and hierarchical warning module 3.

[0074] In summary, by utilizing the technical solutions described above, this invention constructs a model capable of accurately capturing the dynamic characteristics of pressure changes within explosion-proof electrical cabinets by integrating the topological, geometric, and temporal features of pressure data. This invention maps pressure data to a high-dimensional space using time-delay embedding technology and extracts topological features of the data using discrete Morse theory and continuous cohomology analysis. It reconstructs the local structure of the data and extracts geometric features through manifold learning and tomography techniques. Based on the extracted pressure features, this invention designs an improved nonlinear autoregressive exogenous model structure and utilizes deep neural networks to learn the nonlinear functions in the model, achieving accurate modeling and prediction of pressure changes. This not only handles pressure... The nonlinear and non-stationary characteristics of the data allow it to adapt to dynamic changes in pressure within the cabinet, providing reliable data support and decision-making basis for pressure monitoring, anomaly warning, and control strategy formulation in explosion-proof electrical cabinets, effectively improving the safety and reliability of these cabinets. Simultaneously, this invention has significant advantages in pressure feature extraction. By mapping pressure data to a high-dimensional space and applying topological data analysis techniques, it can capture data structures and patterns that are difficult to identify using traditional time-series analysis methods. The application of discrete Morse theory enables this invention to identify key points in the pressure data, including local extrema and saddle points, which often correspond to important state transitions in the pressure system. Continuous homology analysis provides a multi-scale... This invention, through a multi-dimensional, multi-scale feature extraction method, can identify stable topological features in pressure data, reflecting the essential patterns of pressure changes. Furthermore, by employing manifold learning and tomography techniques, it can reconstruct the local geometry of pressure data and extract geometric features such as curvature and geodesic distance. These features reflect both local and global characteristics of pressure changes. This comprehensive and in-depth understanding of the intrinsic structure and dynamic characteristics of pressure data is achieved through this multi-dimensional, multi-scale feature extraction method. In addition, this invention designs a dynamic pressure model using a nonlinear autoregressive exogenous model structure and utilizes a deep neural network to learn the nonlinear functions within it, achieving accurate capture of the complex dynamic characteristics of the pressure system. This not only considers historical pressure values ​​and external... The influence of the input is also incorporated into the aforementioned extracted topological and geometric features, enabling the dynamic pressure model to comprehensively consider various influencing factors of pressure changes. The application of deep neural networks endows the model with powerful nonlinear modeling capabilities and adaptability, enabling it to automatically learn complex patterns and long-term and short-term dependencies in pressure data. Through backpropagation algorithms and cross-validation, the model parameters and hyperparameters are optimized, achieving high-precision prediction of pressure changes while ensuring the model's generalization ability. It can not only adapt to the nonlinear and non-stationary characteristics of pressure systems in explosion-proof electrical cabinets, but also dynamically adjust model parameters according to real-time data, maintaining high prediction accuracy even when facing sudden changes or abnormal situations in the system.

[0075] 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 within the protection scope of the present invention.

Claims

1. A method for issuing an alarm for an explosion-proof electrical cabinet, characterized in that, include: S1. Based on the pressure data inside the electrical cabinet, pressure features are reconstructed and extracted using adaptive compressed sensing and iterative hard thresholding algorithms to construct a dynamic pressure model; specifically including: S11. Calculate the self-similarity index of the stress data and construct an adaptive compressed sensing matrix to generate reconstructed stress data; S12. Based on the reconstructed pressure data, pressure features are extracted nonlinearly and reduced in dimensionality using topological pressure manifold and continuous tomography algorithm. The pressure features include topological features and geometric features. S13. Based on the pressure characteristics, construct a dynamic pressure model using a nonlinear autoregressive exogenous model. S2. By using a dynamic pressure model, combined with deep reinforcement learning algorithms and fuzzy cognitive graph theory, the pressure state inside the cabinet is evaluated in real time, and an intelligent control strategy is generated. S3. Based on real-time evaluation results and intelligent control strategies, generate graded alarm signals using a multi-level audiovisual and tactile feedback mechanism.

2. The method for warning of an explosion-proof electrical cabinet according to claim 1, characterized in that, The calculation of the stress data self-similarity index and the construction of an adaptive compressed sensing matrix to generate reconstructed stress data include: S111. Decompose the collected pressure data at multiple scales to obtain pressure fractal components at different scales, and calculate the corresponding pressure data self-similarity index. The calculation expression is as follows: ； In the formula, H i The pressure data self-similarity index represents the pressure data for each pressure fractal component. R This represents the range of the pressure fractal component, that is, the difference between the maximum and minimum values; S This indicates that at the same time scale τ Internal, pressure fractal component F i(t) Standard deviation; τ Indicates a time scale; F i(t) Indicates at any point in time t No. i Pressure fractal components at various scales; μ Represents pressure fractal components F i(t) In time scale τ The average value within; n Representing time scale τ The number of sampling points within; S112. Based on the self-similarity index of pressure data, construct an adaptive compressed sensing matrix and adaptively sample the pressure fractal components at different scales. S113. Based on the adaptive compressed sensing matrix, the pressure fractal components at different scales are reconstructed using an iterative hard thresholding algorithm, and then merged to obtain the reconstructed pressure data.

3. The alarm method for an explosion-proof electrical cabinet according to claim 1, characterized in that, The reconstructed pressure data, using topological pressure manifolds and persistent tomography algorithms, nonlinearly extracts pressure features and reduces dimensionality, including: S121. Construct a high-dimensional manifold representation of the pressure data and apply discrete Morse theory to analyze the critical point structure of the pressure manifold; S122. Construct a complex sequence of pressure data points and identify topological features in the pressure data; S123. Reconstruct the local structure of the pressure manifold using tomography and extract its geometric features.

4. The method for warning of an explosion-proof electrical cabinet according to claim 1, characterized in that, The construction of a dynamic stress model based on stress characteristics and a nonlinear autoregressive exogenous model includes: S131. Based on topological and geometric features, construct a feature set and design the structure of the dynamic pressure model to obtain the nonlinear function; S132. Using deep neural networks to learn nonlinear functions; S133. With the goal of minimizing the prediction error, a dynamic stress model is trained using the backpropagation algorithm, and the hyperparameters of the dynamic stress model are optimized using cross-validation.

5. The method for warning of an explosion-proof electrical cabinet according to claim 1, characterized in that, The method of using a dynamic pressure model, combined with deep reinforcement learning algorithms and fuzzy cognitive graph theory, to assess the pressure state inside the cabinet in real time and generate intelligent control strategies includes: S21. Based on the dynamic pressure model, the deep Q-network algorithm is used to evaluate the pressure status inside the cabinet; S22. Based on the stress state assessment results, use fuzzy cognitive map theory to predict potential risks inside the cabinet; S23. Based on the assessment results of the pressure state and the prediction results of potential risks, a fuzzy logic controller is used to generate an intelligent control strategy.

6. The method for warning of an explosion-proof electrical cabinet according to claim 5, characterized in that, The assessment of the pressure state inside the cabinet based on the dynamic pressure model and using the deep Q-network algorithm includes: S211. Define the state space and action space based on the actual operating parameters of the explosion-proof electrical cabinet; S212. Based on the dynamic pressure model and combined with actual operating parameters, design a reward function; S213. Using a deep neural network, construct a Q-function approximator and optimize the network parameters through empirical replay and target network methods.

7. The method for warning of an explosion-proof electrical cabinet according to claim 5, characterized in that, The prediction of potential risks within the cabinet based on the pressure state assessment results and using fuzzy cognitive map theory includes: S221. Based on the risk factors of explosion-proof electrical cabinets, define the set of concept nodes in a fuzzy cognitive graph; S222. Establish a weight matrix based on the influence between concept nodes; S223. Using iterative calculation and activation functions, update the concept node values ​​and predict potential risks by analyzing steady-state values.

8. The method for warning of an explosion-proof electrical cabinet according to claim 5, characterized in that, The intelligent control strategy generated using a fuzzy logic controller, based on the assessment results of the stress state and the prediction results of potential risks, includes the following steps: S231. Based on the pressure control requirements, define the input and output linguistic variables of the fuzzy controller; S232. Design a fuzzy rule base and construct an intelligent control strategy; the intelligent control strategy includes a first control rule, a second control rule, a third control rule, a fourth control rule, and a fifth control rule; S233. Based on intelligent control strategies, the control quantity is determined using fuzzy reasoning and defuzzification methods.

9. The alarm method for an explosion-proof electrical cabinet according to claim 1, characterized in that, The generation of graded warning signals based on real-time evaluation results and intelligent control strategies, utilizing a multi-level audiovisual and tactile feedback mechanism, includes: S31. Based on the stress status assessment results and potential risk prediction results, a multi-level warning mechanism is constructed; the multi-level warning mechanism includes normal state, minor warning, moderate warning, severe warning and emergency state; S32. Design a multi-layered alarm system using visual, auditory, and tactile feedback; the multi-layered alarm system includes a normal state alarm, a minor warning alarm, a moderate warning alarm, a severe warning alarm, and an emergency state alarm. S33. Generate fault diagnosis information and handling suggestions based on real-time pressure values ​​and trends.

10. An explosion-proof electrical cabinet alarm device, used to implement the explosion-proof electrical cabinet alarm method according to any one of claims 1-9, characterized in that, The explosion-proof electrical cabinet warning device includes: The feature extraction and model building module is used to reconstruct and extract pressure features based on pressure data inside the electrical cabinet using adaptive compressed sensing and iterative hard thresholding algorithms, and to build a dynamic pressure model; specifically, it includes: Calculate the self-similarity index of stress data and construct an adaptive compressed sensing matrix to generate reconstructed stress data; Based on the reconstructed pressure data, pressure features are extracted nonlinearly and reduced in dimensionality using topological pressure manifold and persistent tomography algorithm. The pressure features include topological and geometric features. Based on the stress characteristics, a dynamic stress model is constructed using a nonlinear autoregressive exogenous model. The state assessment and strategy generation module is used to assess the pressure state inside the cabinet in real time and generate intelligent control strategies by using a dynamic stress model, combined with deep reinforcement learning algorithms and fuzzy cognitive graph theory. The intelligent control and hierarchical alarm module is used to generate hierarchical alarm signals based on real-time evaluation results and intelligent control strategies, using a multi-level audiovisual and tactile feedback mechanism. The feature extraction and model building module is connected to the state evaluation and strategy generation module, and the state evaluation and strategy generation module is connected to the intelligent control and hierarchical alarm module.

Images