Electricity utilization safety self-adaptive prevention and control system of electrical equipment

Abstract

The invention belongs to the technical field of power utilization systems, and particularly relates to a power utilization safety self-adaptive prevention and control system for electrical equipment, which comprises a multi-dimensional sensing layer, a cloud application layer, an edge computing layer, a cloud application layer and a power utilization safety self-adaptive prevention and control layer, wherein the multi-dimensional sensing layer comprises a sensor for acquiring electrical parameters and environmental data of the equipment; and the cloud computing layer is used for receiving the key data and the analysis result uploaded by the edge computing layer, performing long-term trend analysis, historical data storage and global optimization strategy generation, and issuing an optimization strategy to the edge computing layer. According to the method, transmission delay can be reduced to quickly find and process faults, only key data and results are uploaded, the transmission quantity, the bandwidth requirement and the communication cost are greatly reduced, the complex working conditions of industrial equipment are adapted through a self-adaptive adjustment analysis strategy, the fault recognition accuracy is improved, dependence on cloud resources is reduced, the independence and reliability of the system are enhanced, and the method is suitable for popularization and application. And an edge node algorithm is optimized, so that the resource utilization rate is improved, and efficient data processing under resource limitation is realized.

Description

Technical Field

[0001] This invention belongs to the field of power system technology, specifically relating to an adaptive control system for power safety of electrical equipment. Background Art

[0002] Against the backdrop of the accelerated development of the Industrial Internet and intelligent manufacturing, the safe operation of industrial equipment has an increasingly prominent impact on enterprise production efficiency and economic benefits. As a key technology to ensure the normal operation of equipment, the traditional centralized mode of industrial equipment power safety monitoring has problems such as excessive communication load and insufficient real-time response, making it difficult to adapt to the needs of intelligent upgrading of industrial equipment. At the same time, with the large-scale deployment of industrial IoT devices, massive amounts of sensor data need to be processed and analyzed in real time. When dealing with such data, traditional cloud computing architecture faces bottlenecks such as prominent transmission delay, excessive bandwidth consumption, and lack of real-time performance.

[0003] Edge computing, an emerging computing architecture, offers a new technological path for equipment monitoring by collaborating with technologies such as cloud computing and artificial intelligence. It integrates core capabilities of networking, computing, storage, and applications at the network edge, close to physical devices or data sources, to build a distributed open platform with features such as intelligent sensing, real-time analysis, data optimization, intelligent applications, and security and privacy protection. Applying it to electrical safety monitoring of industrial equipment can effectively improve fault diagnosis and handling efficiency, ensuring the safe and stable operation of equipment.

[0004] Currently, research on the application of edge computing in the field of electrical safety monitoring of industrial equipment mainly focuses on the following directions:

[0005] In terms of data acquisition and transmission, traditional monitoring systems suffer from high communication pressure and poor real-time performance during data acquisition and transmission. Edge computing, on the other hand, can reduce data transmission volume by relying on local processing and improve system response efficiency. In the field of fault diagnosis and prediction, edge computing-based solutions can conduct real-time analysis of equipment operating status locally, identify potential faults in a timely manner, and improve the accuracy and efficiency of diagnosis. Furthermore, in terms of energy efficiency optimization, real-time monitoring and analysis of equipment energy consumption using edge computing can achieve energy efficiency optimization and energy saving goals.

[0006] However, existing edge computing applications still have significant shortcomings in this field:

[0007] The multi-dimensional data generated by industrial equipment operation, such as current, temperature, and vibration, needs to be processed and fused efficiently. However, existing edge computing algorithms still have room for improvement in efficiency and accuracy when processing such data. Moreover, the operating conditions of industrial equipment are complex and changeable, and existing edge computing algorithms cannot adjust their analysis strategies in real time to adapt to diverse operating conditions. This results in limitations on the accuracy and response speed of fault identification. Furthermore, edge nodes usually have limited resources, and existing algorithms are lacking in the deployment and optimization of edge nodes, leading to wasted computing resources and low processing efficiency.

[0008] Therefore, there is an urgent need for an edge computing algorithm that can efficiently process multi-dimensional perception layer data, adjust analysis strategies in real time, and adapt to complex working conditions, so as to improve the accuracy and real-time performance of electrical safety monitoring of industrial equipment. Summary of the Invention

[0009] The purpose of this invention is to provide an adaptive control system for electrical safety of electrical equipment, which can reduce transmission delay to quickly detect and handle faults, upload only key data and results, significantly reduce transmission volume, bandwidth requirements and communication costs, adapt to complex operating conditions of industrial equipment through adaptive adjustment analysis strategies, improve fault identification accuracy, reduce dependence on cloud resources, enhance system independence and reliability, and optimize edge node algorithms to improve resource utilization, thereby achieving efficient data processing under resource constraints.

[0010] The specific technical solution adopted by this invention is as follows:

[0011] An adaptive control system for electrical safety of electrical equipment, comprising:

[0012] A: Multi-dimensional perception layer, including sensors for collecting device electrical parameters and environmental data;

[0013] B: Edge computing layer, which is equipped with edge computing nodes, including:

[0014] B1: Data preprocessing module, used to perform noise reduction and feature extraction preprocessing on the data collected by the multi-dimensional perception layer;

[0015] B2: Adaptive algorithm module, used to adjust data analysis strategies in real time according to equipment operating conditions, including adjusting feature extraction methods, model parameters and decision thresholds;

[0016] B3: Fault identification module, used to identify potential equipment faults based on preprocessed data and adaptively adjusted analysis strategies;

[0017] B4: Control Decision Module, used to generate control commands based on fault identification results to control the operating status of the equipment;

[0018] B5: Communication interface module, used for data interaction with the multi-dimensional perception layer;

[0019] C: Cloud application layer, used to receive key data and analysis results uploaded by the edge computing layer, perform long-term trend analysis, historical data storage and global optimization strategy generation, and distribute the optimization strategy to the edge computing layer.

[0020] A control method for an adaptive control system for electrical safety of electrical equipment, the method employing the aforementioned adaptive control system for electrical safety, the control method comprising the following steps:

[0021] S1: Data Acquisition: Real-time acquisition of electrical parameters and environmental data of the equipment through sensors in the multi-dimensional perception layer;

[0022] S2: Data preprocessing: Denoising and feature extraction are performed on the collected data to obtain feature vectors;

[0023] S3: Adaptive Strategy Adjustment: Dynamically adjust the analysis strategy of the edge computing algorithm based on the current operating conditions of the device, including adjusting the feature extraction method, model parameters and decision threshold;

[0024] S4: Fault Identification and Diagnosis: Based on the preprocessed data and adjusted analysis strategies, identify potential equipment faults and determine the fault type and severity.

[0025] S5: Control Decision: Based on the fault identification results, generate corresponding control commands to control the operating status of the equipment, and realize fault early warning and safety prevention and control;

[0026] S6: Data Upload and Optimization: Upload key data and analysis results to the cloud application layer, receive optimization strategies from the cloud, and update the algorithms and parameters of edge computing nodes.

[0027] The technical effects achieved by this invention are as follows:

[0028] This invention performs data processing and analysis locally through edge computing nodes, reducing data transmission latency, timely detection and handling of potential faults, and uploading only key data and analysis results to the cloud, significantly reducing data transmission volume, lowering network bandwidth requirements and communication costs. Furthermore, by adaptively adjusting the analysis strategy, it can better adapt to the complex operating conditions of industrial equipment and improve the accuracy of fault identification.

[0029] Furthermore, since most data processing and analysis are completed on edge computing nodes, the reliance on cloud computing resources is reduced, the system's independence and reliability are improved, and resource utilization efficiency is enhanced by optimizing the algorithms of edge computing nodes. This enables efficient data processing and analysis on resource-constrained edge devices, thereby allowing analysis strategies to be adjusted in real time according to the operating conditions of industrial equipment, adapting to the equipment monitoring needs in harsh environments such as mines and steel plants. Attached Figure Description

[0030] Figure 1 This is a flowchart of the overall architecture of the system of the present invention;

[0031] Figure 2 This is the overall flowchart of the method of the present invention. Detailed Implementation

[0032] To make the objectives and advantages of this invention clearer, the invention will be specifically described below with reference to embodiments. It should be understood that the following text is merely used to describe one or more specific embodiments of the invention and does not strictly limit the scope of protection specifically claimed by the invention.

[0033] Example 1:

[0034] like Figure 1 As shown, an adaptive control system for electrical safety of electrical equipment includes:

[0035] A: Multi-dimensional perception layer, including sensors for collecting device electrical parameters and environmental data;

[0036] The sensors in the multi-dimensional sensing layer in step A include current sensors, voltage sensors, temperature sensors, vibration sensors, humidity sensors, dust sensors, etc., which are used to collect electrical parameters and environmental data of the equipment in real time. The data collected by the sensors is transmitted to the edge computing layer through wired or wireless communication.

[0037] B: Edge computing layer. The edge computing layer is the core of the system and consists of multiple edge computing nodes. Each edge computing node is deployed close to the device. Edge computing nodes include:

[0038] Through this, the sensor in the perception layer communicates with the data collected by the sensor, performs preliminary format conversion and protocol parsing, and then processes the data through the following modules;

[0039] B1: Data preprocessing module, used to perform noise reduction and feature extraction preprocessing on the data collected by the multi-dimensional perception layer, thereby obtaining feature vectors that are easy to analyze. Preprocessing methods include signal processing techniques such as wavelet transform and empirical mode decomposition;

[0040] In step B1, the data preprocessing module employs a multi-dimensional data fusion processing strategy:

[0041] B11: Wavelet transform is used to reduce noise in current and voltage parameters;

[0042] B12: The vibration signal is processed by combining morphological non-subsampled wavelet transform and S-transform;

[0043] B13: Temperature and humidity data are processed using a combination of moving average filtering and median filtering;

[0044] B2: Adaptive algorithm module, used to adjust data analysis strategies in real time according to equipment operating conditions, including adjusting feature extraction methods, model parameters and decision thresholds. This adaptive algorithm module adopts fuzzy PID algorithm, which includes operating condition feature extraction, operating condition classification based on K-means clustering, fuzzy PID parameter adjustment and strategy update.

[0045] B3: Fault Identification Module, used to identify potential equipment faults based on preprocessed data and adaptively adjusted analysis strategies. This fault identification module adopts an improved lightweight neural network model combined with vibration analysis and current fluctuation analysis. The lightweight neural network model is an improved TinyML architecture. The vibration analysis adopts a fast Fourier transform and linear regression model, and the current fluctuation analysis adopts a method combining short-time Fourier transform and wavelet packet analysis. The analysis results are fused through a weighted voting method.

[0046] B4: Control Decision Module, used to generate control commands based on fault identification results to control the operating status of the equipment. The control decision module adopts a combination of rule-based decision tree algorithm and model-based predictive control algorithm to ensure the accuracy and real-time performance of control.

[0047] B5: Communication interface module, used for data interaction with the multi-dimensional perception layer;

[0048] C: Cloud application layer, used to receive key data and analysis results uploaded by the edge computing layer, perform long-term trend analysis, historical data storage and global optimization strategy generation, and distribute the optimization strategy to the edge computing layer.

[0049] Example 2:

[0050] See appendix Figure 2 A control method for an adaptive control system for electrical safety of electrical equipment, the control method employing the aforementioned adaptive control system for electrical safety, and the control method includes the following steps:

[0051] S1: Data Acquisition: Real-time acquisition of electrical parameters and environmental data of the equipment through sensors in the multi-dimensional perception layer;

[0052] S2: Data preprocessing: Denoising and feature extraction are performed on the collected data to obtain feature vectors;

[0053] Data preprocessing in S2 includes:

[0054] S21: Perform wavelet decomposition on the current and voltage signals to obtain coefficients in different frequency bands. Perform threshold processing on the high-frequency coefficients to remove noise components. Perform wavelet reconstruction on the processed coefficients to obtain the denoised signal.

[0055] S22: Perform morphological non-subsampled wavelet transform on the vibration signal to obtain decomposition coefficients at different scales. Perform S-transform on the decomposed coefficients to obtain the time-frequency distribution matrix. Extract the feature parameters of the time-frequency distribution matrix, such as energy distribution and peak frequency.

[0056] S23: Filter the temperature and humidity data to remove outliers. The method of combining moving average filtering and median filtering is used to remove sudden noise and outliers.

[0057] S24: Extract electrical parameter features, vibration features, and temperature features and perform normalization processing;

[0058] Furthermore, the following key feature parameters are extracted from the preprocessed data:

[0059] Electrical parameter characteristics: fundamental residual current, insulation conductance, fundamental residual power, time-period average of full-wave load current, time-period peak shift average, deviation count, etc.

[0060] Vibration characteristics: vibration amplitude, frequency components, kurtosis value, margin index, etc.;

[0061] Temperature characteristics: mean temperature, rate of temperature change, hot spot temperature, etc.

[0062] S3: Adaptive Strategy Adjustment: Dynamically adjust the analysis strategy of the edge computing algorithm based on the current operating conditions of the device, including adjusting the feature extraction method, model parameters and decision threshold;

[0063] The adaptive policy adjustment in S3 includes:

[0064] S31: Extract characteristic parameters reflecting the current operating conditions from the preprocessed data, such as the degree of fluctuation of load current and the energy distribution of vibration signals;

[0065] S32: Use the K-means clustering algorithm to classify the current operating condition into normal operating condition, light load operating condition, heavy load operating condition or abnormal operating condition;

[0066] S33: Based on the fuzzy PID algorithm, the parameters of the edge computing algorithm are adjusted according to the working condition classification results. The input of the fuzzy PID algorithm is the working condition feature parameters, and the output is the adjusted algorithm parameters (such as the learning rate and decision threshold of the neural network).

[0067] S34: Update the analysis strategy based on the adjusted parameters, including feature extraction methods, model parameters, and decision thresholds;

[0068] The mathematical model of the adaptive policy adjustment algorithm is as follows:

[0069] Let the feature vector of the current operating condition be X = [x1, x2, ..., x]. n After classifying the operating conditions, the resulting operating condition category is C. The output of the fuzzy PID algorithm is the adjustment parameter Δθ. Therefore, the formula for calculating the adjusted algorithm parameter θ' is:

[0070] θ' = θ + Δθ = θ + f(X,C)

[0071] Where θ is the decision threshold, and f(X,C) represents the parameter adjustment function based on fuzzy PID, which calculates the adjustment parameters according to different working conditions and feature vectors;

[0072] S4: Fault Identification and Diagnosis: Based on the preprocessed data and adjusted analysis strategies, identify potential equipment faults and determine the fault type and severity.

[0073] Fault identification in S4 includes:

[0074] S41: Select relevant feature parameters using mutual information and recursive feature elimination methods;

[0075] S42: Utilizes an improved TinyML architecture neural network model to analyze feature vectors and identify fault modes, enabling efficient operation on resource-constrained edge devices. The network structure includes an input layer, multiple convolutional layers, pooling layers, and fully connected layers, with the following specific structure:

[0076] Input layer: Receives the preprocessed feature vectors.

[0077] Convolutional layer 1: Uses a 3x3 convolutional kernel to extract local features.

[0078] Pooling layer 1: Use 2x2 max pooling to reduce feature dimension.

[0079] Convolutional layer 2: Uses a 3x3 convolutional kernel to further extract features.

[0080] Pooling layer 2: Use 2x2 average pooling to reduce feature dimension.

[0081] Fully connected layer 1: Maps the feature vectors to a low-dimensional space.

[0082] Fully connected layer 2: Outputs fault type and probability;

[0083] To improve the model's generalization and robustness, transfer learning and data augmentation techniques are employed during training. The pre-trained model is trained on a publicly available industrial equipment fault dataset and then fine-tuned on a dataset specific to industrial equipment.

[0084] S43: Determine the location and severity of the fault by combining the equipment structure and working principle;

[0085] S44: Use a long short-term memory network model to predict fault development trends;

[0086] This method combines Fast Fourier Transform (FFT) and linear regression models to analyze equipment vibration signals and identify mechanical faults such as bearing and gear failures. The specific steps are as follows:

[0087] The vibration signal is subjected to FFT transformation to obtain its spectral characteristics;

[0088] Extract characteristic frequency components and energy distribution from the spectrum;

[0089] Linear regression models are used to analyze features and identify fault types and severity.

[0090] The current fluctuation analysis method combines short-time Fourier transform (STFT) and wavelet packet analysis to analyze the time-frequency characteristics of the current signal and identify electrical faults. The specific steps are as follows:

[0091] Perform STFT transformation on the current signal to obtain the time-frequency distribution matrix;

[0092] Wavelet packet decomposition is performed on the time-frequency distribution matrix to extract energy features in different frequency bands;

[0093] Support Vector Machine (SVM) is used to classify energy characteristics and identify fault types;

[0094] The fault identification algorithm employs a weighted voting method to fuse the results of the neural network model, vibration analysis, and current fluctuation analysis, yielding the final fault identification result. The fusion formula is as follows:

[0095] P = ω1P1 + ω2P2 + ω3P3

[0096] Where P represents the final failure probability, P1, P2, and P3 represent the failure probabilities of the neural network model, vibration analysis, and current fluctuation analysis, respectively, and ω1, ω2, and ω3 represent the corresponding weight coefficients, which are determined according to the performance and reliability of different methods.

[0097] S5: Control Decision: Based on the fault identification results, corresponding control commands are generated to control the operating status of the equipment, realize fault early warning and safety prevention and control. The control decision algorithm adopts a combination of rule-based decision tree and model-based predictive control to ensure the accuracy and real-time performance of the control.

[0098] Control decisions include:

[0099] S51: Determine the decision threshold based on the fault type, severity, and equipment safety requirements;

[0100] S52: Rule-based decision tree generation includes control strategies such as alarms, load reduction, and shutdown;

[0101] The structure of a decision tree is as follows:

[0102] Root node: Determines if a fault exists.

[0103] Intermediate nodes: Determine the type and severity of the fault.

[0104] Leaf node: Generates corresponding control commands, such as alarm, load reduction, and shutdown;

[0105] S53: Model-based predictive control is used to evaluate and select the optimal control strategy. The predictive control model is established based on a combination of the physical model of the equipment and the data-driven model, which can accurately predict the response of the equipment under different control strategies.

[0106] The specific implementation steps of the control decision algorithm are as follows:

[0107] Based on the fault identification results, determine the fault type and severity.

[0108] Based on the fault type and severity, the corresponding control strategy is found from the decision tree.

[0109] Predictive control models are used to evaluate the effectiveness of different control strategies and select the optimal control strategy.

[0110] Generate control commands to control the operating status of industrial equipment;

[0111] S54: Generate control commands and send them to the actuators, monitor the control effect to form a closed-loop control;

[0112] S6: Data Upload and Optimization: Upload key data and analysis results to the cloud application layer, receive optimization strategies from the cloud, and update the algorithms and parameters of edge computing nodes;

[0113] Data uploading and optimization include:

[0114] S61: The edge computing layer uploads device operating status, fault alarm records, and analysis results in real time or on a timed basis according to data importance;

[0115] S62: The cloud application layer uses time series analysis and machine learning algorithms for long-term trend analysis;

[0116] S63: An optimization strategy for generating algorithm parameters, feature selection, and decision thresholds using a genetic algorithm or particle swarm optimization algorithm;

[0117] S64: The cloud will distribute the optimization strategy to the edge computing layer to update the local algorithm and parameters.

[0118] Example 3:

[0119] This embodiment describes the specific implementation steps of Embodiment 2, as follows:

[0120] Data acquisition and preprocessing are fundamental to the system and directly impact the accuracy of subsequent analysis and decision-making. The specific steps are as follows:

[0121] Before system deployment, all sensors are calibrated to ensure the accuracy of measurement data. The calibration method uses a standard signal source input, records the sensor output, and establishes a calibration curve.

[0122] Data collection:

[0123] Based on the characteristics of different parameters, reasonable sampling frequencies are set. The sampling frequency for electrical parameters such as current and voltage is 1kHz, the sampling frequency for vibration signals is 10kHz, and the sampling frequency for slowly changing parameters such as temperature and humidity is 1Hz. A combination of hardware and software synchronization methods is used to ensure that the time synchronization accuracy of multi-sensor data is within 1ms.

[0124] Data preprocessing:

[0125] Low-pass filtering is applied to current, voltage, and other signals to remove high-frequency noise. The filter used is a Butterworth filter, with the cutoff frequency determined based on the signal characteristics. Time-domain features (such as mean, variance, and peak value) and frequency-domain features (such as spectral energy distribution and characteristic frequency amplitude) are extracted from the original signal. The feature extraction method combines wavelet transform and short-time Fourier transform, and the extracted features are normalized to convert features with different dimensions into uniform dimensionless values, facilitating subsequent analysis and comparison.

[0126] The mathematical model for data preprocessing is as follows:

[0127] Let the original signal be x(t) and the filtered signal be y(t), then:

[0128] y(t) = x(t)·h(t)

[0129] Where h(t) is the impulse response function of the filter;

[0130] Feature extraction uses wavelet transform, defined as:

[0131]

[0132] in, Here, is the wavelet basis function, a is the scaling parameter, and b is the translation parameter;

[0133] Adaptive policy adjustment:

[0134] Adaptive strategy adjustment can dynamically adjust the analysis strategy of edge computing algorithms based on the operating conditions of industrial equipment, thereby improving the accuracy and real-time performance of fault identification. The specific steps are as follows:

[0135] Feature parameters reflecting the current operating condition of industrial equipment are extracted from the preprocessed data, such as the fluctuation of load current and the energy distribution of vibration signals. Then, a K-means clustering-based operating condition classification algorithm is used to divide the current operating condition into different categories, such as normal operating condition, light load operating condition, heavy load operating condition, and abnormal operating condition. The number of clusters is determined based on the equipment characteristics and historical data.

[0136] Based on the results of the operating condition classification, the parameters of the edge computing algorithm are adjusted using a fuzzy PID algorithm. The fuzzy PID algorithm takes operating condition characteristic parameters as input and outputs the adjusted algorithm parameters (such as the learning rate and decision threshold of the neural network). Finally, based on the adjusted parameters, the analysis strategy of the edge computing algorithm is updated, covering aspects such as feature extraction methods, model parameters, and decision thresholds.

[0137] The mathematical model for adaptive policy adjustment is as follows:

[0138] Let the feature vector of the current operating condition be:

[0139] The current operating condition feature vector is X = [x1, x2, ..., x...]. n After classifying the operating conditions, the resulting operating condition category is C. The output of the fuzzy PID algorithm is the adjustment parameter Δθ. Therefore, the formula for calculating the adjusted algorithm parameter θ' is:

[0140] θ' = θ + Δθ = θ + f(X,C)

[0141] Fault identification and diagnosis:

[0142] Based on the current operating conditions and equipment characteristics, the most relevant feature parameters are selected as input data for fault identification. The feature selection process integrates mutual information and recursive feature elimination to ensure the selected features possess the strongest discriminative power. Subsequently, a well-trained lightweight neural network model is used to analyze the feature vectors, thereby identifying equipment fault modes. This network employs an improved TinyML architecture, enabling efficient operation on resource-constrained edge devices.

[0143] Based on the fault mode identification results, combined with the equipment structure and working principle, the specific location and severity of the fault are determined. The fault localization process adopts a combination of model-based and rule-based approaches. Furthermore, based on the current fault state and historical data, the development trend and potential impact range of the fault are predicted to support control decisions. The fault prediction uses a Long Short-Term Memory (LSTM) network model, which can effectively capture long-term dependencies in time-series data.

[0144] The mathematical model for fault identification is as follows:

[0145] Let the input feature vector be:

[0146] F = [f1, f2, ..., f n ]

[0147] If the weight matrix of the neural network model is W and the bias vector is b, then the output of the model is:

[0148] y=σ(W 2 ·σ(W 1 ·F+b 1 )+b 2 )

[0149] Where σ is the activation function, using ReLU or Sigmoid, and W 1 and W 2 Let b be the weight matrix. 1 and b 2 It is the bias vector;

[0150] Control Decision-Making and Execution:

[0151] The specific steps for control decision-making and execution are as follows:

[0152] Based on the fault type and severity, and in conjunction with the equipment's safety requirements, appropriate decision thresholds are determined. The determination of these thresholds employs a combination of statistical methods based on historical data and expert experience. Based on the fault identification results and decision thresholds, corresponding control strategies are generated. These control strategies include different levels such as alarm, load reduction, and shutdown. The appropriate control strategy is selected based on the fault severity.

[0153] The control strategy is translated into specific control commands, such as adjusting the output frequency of the frequency converter or disconnecting the circuit breaker. The generation of control commands employs a rule-based decision tree algorithm to ensure the accuracy and real-time performance of the control. These commands are then sent to the actuators to control the operating status of the industrial equipment. Simultaneously, the equipment status after control execution is monitored, forming a closed-loop control system to ensure that the control effect meets expectations.

[0154] The mathematical model for control decision-making is as follows:

[0155] Let the fault identification result be y, and the decision threshold be θ, then the generation rule for the control command u is:

[0156] If y < θ1, then u is running normally;

[0157] elif: θ1≤y<θ2, then u is a warning;

[0158] elif: θ2≤y<θ3, then u is a deload;

[0159] In all other cases, 'u' indicates shutdown;

[0160] Among them, θ1, θ2, and θ3 are decision thresholds at different levels, which are determined based on user presets or device characteristics and security requirements.

[0161] Data Upload and Optimization:

[0162] Through long-term analysis and optimization in the cloud, the performance and reliability of the system are continuously improved. The specific steps are as follows:

[0163] Key data and analysis results are uploaded to the cloud application layer through the edge computing layer. This includes device operating status, fault alarm records, and analysis results. The upload frequency is determined by the importance of the data; key data is uploaded in real time, while general data is uploaded on a scheduled basis. The cloud application layer performs long-term trend analysis on the uploaded data to identify performance change trends and potential problems of the device. The analysis method combines time series analysis and machine learning algorithms. Based on the long-term trend analysis results, optimization strategies are generated, including parameter optimization, feature selection optimization, and decision threshold optimization of the edge computing algorithm. The optimization strategies are generated using optimization algorithms such as genetic algorithms and particle swarm optimization. Subsequently, the cloud application layer distributes the optimization strategies to the edge computing layer, and the edge computing layer updates its local algorithms and parameters according to the optimization strategies to achieve continuous system optimization.

[0164] The mathematical model for data uploading and optimization is as follows:

[0165] Suppose the data uploaded by the edge computing layer is:

[0166] D = [d1, d2, ..., d n ]

[0167] If the long-term trend analysis model for the cloud application layer is M, then the analysis results are:

[0168] R = M(D)

[0169] The optimization strategy θ* is:

[0170] θ*=argmin L(R,θ)

[0171] Where L is the loss function, which represents the difference between the analysis results and the expected results;

[0172] By performing data processing and analysis locally through edge computing nodes, data transmission latency is reduced, potential faults can be detected and handled in a timely manner, and only key data and analysis results are uploaded to the cloud, which significantly reduces the amount of data transmission, lowers network bandwidth requirements and communication costs, and can better adapt to the complex operating conditions of industrial equipment by adaptively adjusting analysis strategies, thereby improving the accuracy of fault identification.

[0173] Furthermore, since most data processing and analysis are completed on edge computing nodes, the reliance on cloud computing resources is reduced, the system's independence and reliability are improved, and resource utilization efficiency is enhanced by optimizing the algorithms of edge computing nodes. This enables efficient data processing and analysis on resource-constrained edge devices, thereby allowing analysis strategies to be adjusted in real time according to the operating conditions of industrial equipment, adapting to the equipment monitoring needs in harsh environments such as mines and steel plants.

[0174] The above description is merely a preferred embodiment of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention. Structures, devices, and operating methods not specifically described or explained in this invention are implemented according to conventional methods in the art unless otherwise specified or limited.

Claims

1. An adaptive control system for electrical safety of electrical equipment, characterized in that, include: A: Multi-dimensional perception layer, including sensors for collecting device electrical parameters and environmental data; B: Edge computing layer, which is equipped with edge computing nodes, including: B1: Data preprocessing module, used to perform noise reduction and feature extraction preprocessing on the data collected by the multi-dimensional perception layer; B2: Adaptive algorithm module, used to adjust data analysis strategies in real time according to equipment operating conditions, including adjusting feature extraction methods, model parameters and decision thresholds; B3: Fault identification module, used to identify potential equipment faults based on preprocessed data and adaptively adjusted analysis strategies; B4: Control Decision Module, used to generate control commands based on fault identification results to control the operating status of the equipment; B5: Communication interface module, used for data interaction with the multi-dimensional perception layer; C: Cloud application layer, used to receive key data and analysis results uploaded by the edge computing layer, perform long-term trend analysis, historical data storage and global optimization strategy generation, and distribute the optimization strategy to the edge computing layer.

2. The adaptive control system for electrical safety according to claim 1, characterized in that: The sensors in the multi-dimensional sensing layer in step A include at least one of a current sensor, a voltage sensor, a temperature sensor, a vibration sensor, a humidity sensor, and a dust sensor.

3. The adaptive control system for electrical safety according to claim 2, characterized in that: In step B1, the data preprocessing module employs a multi-dimensional data fusion processing strategy. B11: Wavelet transform is used to reduce noise in current and voltage parameters; B12: The vibration signal is processed by combining morphological non-subsampled wavelet transform and S-transform; B13: Temperature and humidity data are processed using a combination of moving average filtering and median filtering.

4. The adaptive control system for electrical safety according to claim 3, characterized in that: The adaptive algorithm module in B2 adopts the fuzzy PID algorithm, which includes working condition feature extraction, working condition classification based on K-means clustering, fuzzy PID parameter adjustment and strategy update.

5. The adaptive control system for electrical safety according to claim 4, characterized in that: The fault identification module in B3 uses an improved lightweight neural network model combined with vibration analysis and current fluctuation analysis. The lightweight neural network model is an improved TinyML architecture. The vibration analysis uses a fast Fourier transform and linear regression model, and the current fluctuation analysis uses a combination of short-time Fourier transform and wavelet packet analysis. The analysis results are fused by a weighted voting method.

6. A control method for an adaptive control system for electrical safety of electrical equipment, characterized in that: The prevention and control method employs the adaptive power safety prevention and control system as described in any one of claims 1-5, and the method includes the following steps: S1: Data Acquisition: Real-time acquisition of electrical parameters and environmental data of the equipment through sensors in the multi-dimensional perception layer; S2: Data preprocessing: Denoising and feature extraction are performed on the collected data to obtain feature vectors; S3: Adaptive Strategy Adjustment: Dynamically adjust the analysis strategy of the edge computing algorithm based on the current operating conditions of the device, including adjusting the feature extraction method, model parameters and decision threshold; S4: Fault Identification and Diagnosis: Based on the preprocessed data and adjusted analysis strategies, identify potential equipment faults and determine the fault type and severity. S5: Control Decision: Based on the fault identification results, generate corresponding control commands to control the operating status of the equipment, and realize fault early warning and safety prevention and control; S6: Data Upload and Optimization: Upload key data and analysis results to the cloud application layer, receive optimization strategies from the cloud, and update the algorithms and parameters of edge computing nodes.

7. The prevention and control method according to claim 6, characterized in that: The data preprocessing in S2 includes: S21: Perform wavelet decomposition, high-frequency coefficient thresholding, and wavelet reconstruction on current and voltage signals; S22: Perform morphological non-subsampled wavelet transform, S-transform, and time-frequency feature parameter extraction on the vibration signal; S23: Filter the temperature and humidity data to remove outliers; S24: Extract electrical parameter features, vibration features, and temperature features and perform normalization processing.

8. The prevention and control method according to claim 7, characterized in that: The adaptive strategy adjustment in S3 includes: S31: Extract characteristic parameters reflecting the current operating conditions from the preprocessed data; S32: Use the K-means clustering algorithm to classify the current operating condition into normal operating condition, light load operating condition, heavy load operating condition or abnormal operating condition; S33: Based on the fuzzy PID algorithm, the parameters of the edge calculation algorithm are adjusted according to the working condition classification results; S34: Update the analysis strategy based on the adjusted parameters.

9. The prevention and control method according to claim 8, characterized in that: The fault identification in S4 includes: S41: Select relevant feature parameters using mutual information and recursive feature elimination methods; S42: Utilize an improved TinyML architecture neural network model to analyze feature vectors and identify fault modes; S43: Determine the location and severity of the fault by combining the equipment structure and working principle; S44: Use a long short-term memory network model to predict fault development trends.

Images