SF6 gas state intelligent diagnosis and early warning system based on Internet of Things

Abstract

The invention relates to the technical field of power equipment state monitoring and the Internet of Things, in particular to an SF6 gas state intelligent diagnosis and early warning system based on the Internet of Things, which comprises a sensor array module, an edge computing module, an anti-interference communication module, a cloud platform intelligent analysis module, an early warning and linkage control module, an energy management module and a self-checking and fault-tolerant module. The sensor array collects gas state and equipment environment data in real time, and the data are stably transmitted to the cloud platform through the anti-interference communication module after being preprocessed through edge computing. The cloud platform uses a multi-modal data fusion and deep learning algorithm to realize gas leakage trend analysis, leakage source positioning and risk level evaluation; and the early warning and linkage control module triggers graded early warning according to the diagnosis result and is linked with related equipment for emergency response. According to the invention, omnibearing intelligent monitoring and management of the SF6 gas equipment are realized, the monitoring accuracy and the system reliability can be effectively improved, and the equipment fault risk is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of power equipment condition monitoring and Internet of Things technology, and specifically to an intelligent diagnosis and early warning system for SF6 gas state based on the Internet of Things. Background Technique

[0002] Due to its excellent insulation and arc extinguishing performance, SF6 gas is widely used in power equipment such as circuit breakers and GIS combined electrical appliances. However, problems such as SF6 gas leakage and excessive micro water content can lead to serious consequences such as a decline in equipment insulation performance, faults, and even explosions. At the same time, SF6 is a strong greenhouse gas, and leakage will cause environmental hazards.

[0003] At present, traditional SF6 gas monitoring mostly uses single-point sensors, which have problems such as limited monitoring range and single data, and it is difficult to comprehensively reflect the operating state of the equipment. Moreover, the communication of the monitoring system is vulnerable to strong electromagnetic interference in the substation, resulting in unstable data transmission and high packet loss rate. In terms of data analysis, most rely on manual experience judgment, lacking intelligent diagnosis and prediction capabilities, and unable to detect potential hidden dangers in advance.

[0004] With the development of technologies such as the Internet of Things and artificial intelligence, intelligent monitoring of power equipment has become a trend. However, existing technologies still have deficiencies in multi-source data fusion analysis, reliable communication in complex environments, and system self-maintenance, and it is difficult to meet the requirements of the power system for efficient, safe, and intelligent monitoring of SF6 gas equipment. Therefore, an intelligent diagnosis and early warning system for SF6 gas state based on the Internet of Things is proposed to address the above problems. Summary of the Invention

[0005] The purpose of the present invention is to provide an intelligent diagnosis and early warning system for SF6 gas state based on the Internet of Things to solve the problems raised in the above background technique.

[0006] To achieve the above purpose, the present invention provides the following technical solutions:

[0007] An intelligent diagnosis and early warning system for SF6 gas state based on the Internet of Things, including:

[0008] Sensor array module: Deployed at key monitoring points of SF6 gas equipment, including SF6 gas concentration sensors, temperature and humidity sensors, pressure sensors, micro water content sensors, and vibration sensors, to collect gas state parameters and equipment operating environment data in real time;

[0009] Edge computing module: Integrated into the local terminal, used to preprocess gas state parameters and equipment operating environment data, and generate a preliminary diagnosis result based on the dynamic baseline model;

[0010] Anti-interference communication module: Adopts LoRa wireless communication technology, integrating forward error correction coding FEC and orthogonal frequency division multiple access OFDMA modulation technology;

[0011] Cloud platform intelligent analysis module: receives data uploaded by the edge computing module, analyzes gas leakage trends, equipment aging status and environmental coupling effects through multimodal data fusion algorithms, and predicts the location of leakage sources and risk levels based on deep learning models;

[0012] Early warning and linkage control module: triggers graded early warning signals according to the location of the leakage source and the risk level, and links the fan, air supply device and access control system to execute emergency response;

[0013] Energy management module: includes solar power supply unit and intelligent power distribution circuit, providing multi-mode power supply guarantee for the system;

[0014] Self-check and fault-tolerance module: periodically checks sensor accuracy, communication link integrity, and power status, and performs fault isolation and redundant switching.

[0015] As a preferred solution, the dynamic baseline model is constructed in the following way:

[0016] Establish dynamic threshold ranges for SF6 gas concentration, pressure and trace water content based on historical data and equipment operating parameters;

[0017] Introducing environmental coupling factors, combining equipment vibration data with local meteorological information, and dynamically adjusting the warning sensitivity of the baseline model;

[0018] The improved hybrid time series prediction algorithm is adopted, and its modeling method is as follows:

[0019] The prediction results of the ARIMA model are weighted and fused with the seasonal term and trend term components of the Prophet algorithm. The value range of the weight coefficient α is 0.6 to 0.8, and the environmental variable correction term is superimposed. The coupling coefficient β of the correction term is dynamically calculated by the ridge regression algorithm.

[0020] Establish a double residual feedback mechanism:

[0021] The first-order residual between the real-time monitoring value and the model prediction value is calculated, the second-order residual is decomposed by wavelet transform, and the second-order residual is injected back into the model input layer for parameter fine-tuning.

[0022] As a preferred solution, the anti-interference communication module includes:

[0023] The LoRa channel encoder based on convolutional code has the following encoding process: the information bit polynomial is modulo-operated with the preset generator polynomial to generate the encoded data with redundant check bits;

[0024] Adaptive frequency hopping mechanism, the frequency selection method is: allocate available frequency band resources according to the real-time signal strength RSSI ratio of each frequency band, and select the communication frequency band with the least interference;

[0025] Encryption transmission protocol, using the AES-256 algorithm to perform end-to-end encryption on sensor data and control instructions.

[0026] As a preferred solution, the deep learning model includes:

[0027] Leak source localization sub-model: Using an improved multi-scale attention convolutional neural network, dynamically calculating the attention weight matrix of each convolutional kernel through a gated recurrent unit, and weighted fusing convolutional feature maps of different scales;

[0028] Risk prediction sub-model: Using a hybrid architecture of LSTM and temporal convolutional network, and the temporal convolutional network captures multi-scale temporal features through an exponentially increasing dilation factor;

[0029] Federated learning optimization mechanism: Weighted aggregation of local model parameters at each site, and the weight value is inversely proportional to the privacy protection strength of the site data distribution.

[0030] As a preferred solution, the early warning and linkage control module includes:

[0031] Adaptive early warning strategy: Dynamically adjust the early warning level according to the leakage rate, equipment importance level, and personnel presence status;

[0032] Air replenishment control logic: Using a PID-fuzzy composite controller, and its parameter tuning method is:

[0033] The proportional coefficient Kp is dynamically adjusted according to the differential value of the pressure deviation, and the integral coefficient Ki adaptively increases according to the absolute value of the pressure deviation;

[0034] Emergency ventilation mechanism: Based on the dynamic weight path planning algorithm, the path cost consists of the actual movement cost, the estimated remaining cost, and the gas concentration penalty term, where the weight of the concentration penalty term decays exponentially with time.

[0035] As a preferred solution, the energy management module includes:

[0036] Photovoltaic power supply unit: Using the maximum power point tracking algorithm to optimize the output voltage of the photovoltaic array to minimize the equivalent series resistance loss;

[0037] Intelligent power distribution circuit: Allocate the total electric energy proportionally according to the equipment priority weight, and the equipment with a higher priority obtains the power supply resources first.

[0038] As a preferred solution, the self-check and fault tolerance module includes:

[0039] Multi-index fusion diagnosis algorithm: Activate the difference between the sensor readings and the calibration coefficient and the failure threshold through a rectified linear unit, and then output the fault probability distribution after weighted by dynamic weights;

[0040] Fault recovery mechanism: When the communication is interrupted, it automatically switches to the local cache mode, and the amount of stored data does not exceed the maximum capacity limit of the edge device.

[0041] As can be seen from the technical solutions provided by the present invention above, the intelligent SF6 gas status diagnosis and early warning system based on the Internet of Things provided by the present invention has the following beneficial effects:

[0042] Comprehensive and accurate monitoring to ensure equipment safety: The sensor array module deploys a variety of high-precision sensors to collect key parameters such as SF6 gas concentration, temperature and humidity, and pressure in real time, comprehensively covering the equipment operation status information; the edge computing module effectively filters interference through a dynamic baseline model and data preprocessing, and accurately judges the equipment operation status; the cloud platform intelligent analysis module uses multi-modal data fusion and deep learning technologies to realize in-depth analysis of gas leakage trends and equipment aging status, discovers potential fault hazards in advance, and builds a solid defense line for the safe operation of equipment, avoiding safety accidents caused by equipment failures;

[0043] Efficient and reliable communication to ensure data transmission: The anti-interference communication module uses LoRa wireless communication technology, combined with technologies such as forward error correction coding and orthogonal frequency division multiple access modulation, to ensure the accuracy and stability of data transmission in the strong electromagnetic interference environment of the substation; the adaptive frequency hopping mechanism and encryption transmission protocol further improve communication efficiency and data security, ensure the real-time and reliable transmission of monitoring data and control instructions, and provide solid data support for the accurate diagnosis and timely response of the system;

[0044] Intelligent early warning and linkage to improve emergency response capabilities: The early warning and linkage control module implements an adaptive early warning strategy based on the analysis results of the cloud platform, and conveys equipment risk information to relevant personnel in a timely manner through various methods such as audible and visual alarms and mobile terminal push; at the same time, it links the fan, gas replenishment device and access control system, and automatically executes emergency response operations according to different risk levels, such as quickly dispersing leaked gas, replenishing gas pressure, and controlling personnel access, effectively reducing the harm of faults, improving the system's emergency handling ability, and reducing accident losses;

[0045] Scientific energy management to enhance system endurance: The energy management module uses a solar power supply unit to convert solar energy into electrical energy, providing green and renewable energy for the system, reducing dependence on traditional commercial power, and saving energy costs; the intelligent power distribution circuit dynamically distributes electrical energy based on load priority, ensures priority power supply for key modules, and at the same time realizes intelligent switching and management of multiple power sources, ensuring the continuous and stable operation of the system under various environmental conditions, enhancing the system's endurance, and reducing operation and maintenance costs;

[0046] Self-detection and fault tolerance to improve system robustness: The self-check and fault tolerance module periodically detects the accuracy of system sensors, the integrity of communication links, and the power status. Through a multi-index fusion diagnosis algorithm, it can promptly detect and locate faults. Once a fault is detected, it quickly executes fault isolation and redundant switching operations to ensure that the core functions of the system are not affected. The local storage and forwarding mode saves data during communication interruptions, ensuring data integrity, significantly improving the robustness and reliability of the system, reducing system downtime, and enhancing the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 FIG. is a schematic diagram of the overall structure of the SF6 gas status intelligent diagnosis and warning system based on the Internet of Things according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0048] In order to make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0049] To better understand the above technical solutions, the above technical solutions will be described in detail below in conjunction with the accompanying drawings of the specification and the specific embodiments.

[0050] As Figure 1 shown, the embodiment of the present invention provides an SF6 gas status intelligent diagnosis and warning system based on the Internet of Things, including:

[0051] Sensor array module: Deployed at key monitoring points of SF6 gas equipment, including SF6 gas concentration sensors, temperature and humidity sensors, pressure sensors, micro water content sensors, and vibration sensors, which can collect gas state parameters and equipment operation environment data in real time;

[0052] Edge computing module: Integrated in the local terminal, used to preprocess gas state parameters and equipment operation environment data, and generate a preliminary diagnosis result based on a dynamic baseline model. The dynamic baseline model is constructed in the following ways:

[0053] Establish dynamic threshold intervals for SF6 gas concentration, pressure, and micro water content based on historical data and equipment operation parameters;

[0054] Introduce an environmental coupling factor, combine equipment vibration data with local meteorological information, and dynamically adjust the warning sensitivity of the baseline model;

[0055] Adopt an improved hybrid time series prediction algorithm, and its modeling method is:

[0056] Fuse the prediction results of the ARIMA model with the seasonal and trend components of the Prophet algorithm, where the value range of the weight coefficient α is from 0.6 to 0.8, and superimpose the environmental variable correction term. The coupling coefficient β of the correction term is dynamically calculated through the ridge regression algorithm;

[0057] Establish a dual residual feedback mechanism:

[0058] Calculate the first-level residual between the real-time monitoring value and the model prediction value, decompose the second-level residual through wavelet transform, and inject the second-level residual back into the model input layer for parameter fine-tuning;

[0059] Anti-interference communication module: Adopt LoRa wireless communication technology, integrating forward error correction coding FEC and orthogonal frequency division multiple access OFDMA modulation technology;

[0060] Cloud platform intelligent analysis module: Receive the data uploaded by the edge computing module, analyze the gas leakage trend, equipment aging status and environmental coupling impact through the multi-modal data fusion algorithm, and predict the leakage source location and risk level based on the deep learning model. The deep learning model includes:

[0061] Leakage source localization sub-model: Adopt an improved multi-scale attention convolutional neural network, dynamically calculate the attention weight matrix of each convolutional kernel through a gated recurrent unit, and fuse the convolutional feature maps of different scales through weighting;

[0062] Risk prediction sub-model: Adopt a hybrid architecture of LSTM and temporal convolutional network. The temporal convolutional network captures multi-scale temporal features through an exponentially growing dilation factor;

[0063] Federated learning optimization mechanism: Weightedly aggregate the local model parameters of each site, and the weight value is inversely proportional to the privacy protection intensity of the site data distribution;

[0064] Early warning and linkage control module: According to the leakage source location and risk level, trigger a hierarchical early warning signal, and link the fan, air supply device and access control system to execute emergency responses;

[0065] Energy management module: Include a solar power supply unit and an intelligent power distribution circuit to provide multi-mode power supply guarantee for the system;

[0066] Self-check and fault tolerance module: Periodically detect the sensor accuracy, communication link integrity and power supply status, and perform fault isolation and redundant switching.

[0067] In this embodiment, the sensor array module is the core component for accurate data acquisition in the SF6 gas status intelligent diagnosis and early warning system based on the Internet of Things. It is like the "perceiving antenna" of the system, and its performance directly affects the monitoring and diagnostic accuracy of the operating status of SF6 gas equipment. In complex electromagnetic environments such as substations, the operating status of SF6 gas equipment is vulnerable to various factors. The sensor array module, through scientific layout and the collaborative work of multiple sensors, can collect gas status and equipment operating environment data in real time and accurately, providing a reliable basis for subsequent analysis and diagnosis, specifically as follows:

[0068] I. Overall function overview:

[0069] This module deploys various types of sensors at key monitoring points of SF6 gas equipment to build an all-round data acquisition network. For key parameters such as SF6 gas concentration, temperature and humidity, pressure, micro water content, and equipment vibration, advanced sensing technologies are used for real-time monitoring. Combined with the data transmission link, the collected raw data is stably transmitted to the edge computing module, providing basic data support for subsequent data processing, diagnostic analysis, and early warning control of the system;

[0070] II. Composition and functions of sub-modules:

[0071] (1) SF6 gas concentration sensing unit:

[0072] Sensor selection and deployment: Select SF6 gas concentration sensors based on high-precision infrared absorption or electrochemical principles. According to the structural characteristics of SF6 gas equipment, they are densely deployed at key positions prone to gas leakage, such as circuit breaker gas chambers, GIS pipeline connection parts, and transformer gas chambers. The detection accuracy of this sensor can reach the ppm level, and it can quickly respond to small changes in gas concentration;

[0073] Data acquisition and transmission: Continuously and real-time collect SF6 gas concentration data. After converting the analog signal into a digital signal through the built-in analog-to-digital conversion module, it is stably transmitted to the edge computing module through a dedicated data transmission line, providing core data for the system to judge gas leakage;

[0074] (2) Temperature and humidity sensing unit:

[0075] Sensor characteristics: Adopt a digital temperature and humidity composite sensor, with wide measurement range and high-precision characteristics. The temperature measurement range is -40°C - 85°C, and the accuracy can reach ±0.3°C; the humidity measurement range is 0% - 100%RH, and the accuracy is ±2%RH. The sensor is built-in with an intelligent calibration algorithm, which can automatically compensate for the influence of environmental factors on the measurement results;

[0076] Data acquisition and function: Real-time monitoring of the temperature and humidity data of the equipment operating environment, packing the data and transmitting it to the edge computing module; the temperature and humidity data can be used to correct the measurement errors of other sensors caused by environmental factors, and at the same time provide an environmental reference for analyzing the changes in the physical and chemical properties of SF6 gas;

[0077] (III) Pressure sensing unit:

[0078] Sensor principle and parameters: A piezoresistive pressure sensor is adopted, with a measuring range covering 0 - 1.0 MPa and an accuracy of 0.1% FS; based on the piezoresistive effect principle, when the internal pressure of the SF6 gas equipment changes, the resistance value of the piezoresistive resistor inside the sensor changes, and the resistance change is converted into a voltage signal through a measuring circuit, and then the pressure value is obtained;

[0079] Data acquisition and application: Real-time monitoring of the internal SF6 gas pressure of the equipment, and real-time transmission of the pressure data to the edge computing module; the pressure data is one of the key bases for judging whether there are faults such as gas leakage and seal failure in the equipment;

[0080] (IV)Micro water content sensing unit:

[0081] Sensing technology and accuracy: A micro water content sensor based on the principle of capacitance or dew point method can accurately measure trace moisture as low as 10 ppm level in SF6 gas; the capacitance sensor determines the micro water content by detecting the capacitance change caused by moisture in the gas; the dew point method sensor calculates the micro water content indirectly by measuring the dew point temperature of the gas;

[0082] Data acquisition and significance: Continuously collect the micro water content data of SF6 gas and transmit it to the edge computing module; too high micro water content will cause the decline of the internal insulation performance of the equipment, and this data is of great significance for evaluating the insulation state of the equipment and preventing insulation faults;

[0083] (V)Vibration sensing unit:

[0084] Sensor type and performance: An acceleration-type vibration sensor is selected, which can detect vibration acceleration in the range of 0.1 m / s² - 100 m / s², and the frequency response range is 0.1 Hz - 10 kHz; this sensor can sensitively capture the vibration signal changes caused by mechanical faults such as loose parts and bearing wear during the operation of the equipment;

[0085] Data acquisition and analysis: Real-time collect the vibration data of the equipment and transmit the data to the edge computing module; through the analysis and processing of the vibration data, the operating state of the internal mechanical structure of the equipment can be judged, providing important clues for predicting equipment faults;

[0086] (VI)Data transmission and interface module:

[0087] Transmission line design: Shielded cables or optical fibers are used as data transmission lines to effectively resist strong electromagnetic interference in substations and ensure the stable transmission of data collected by sensors. For the data transmission requirements of different types of sensors, an adapted data interface protocol is designed to ensure the accuracy and consistency of data transmission.

[0088] Data aggregation and preprocessing: Aggregate the data collected by each sensor and perform simple preprocessing operations such as data format conversion and verification before transmitting it to the edge computing module to improve data transmission efficiency and reliability.

[0089] III. Key technical principles:

[0090] (I) SF6 gas concentration sensing principle:

[0091] The infrared absorption type SF6 gas concentration sensor is based on the absorption characteristics of SF6 gas for infrared light of a specific wavelength. When infrared light passes through the gas chamber containing SF6 gas, part of the infrared light is absorbed by the SF6 gas. By detecting the attenuation degree of the infrared light intensity, the gas concentration is calculated according to the Lambert-Beer law. The electrochemical sensor utilizes the relationship that the magnitude of the current generated by the electrochemical reaction of SF6 gas on the electrode is proportional to the gas concentration to achieve the measurement of the gas concentration.

[0092] (II) Temperature and humidity sensing principle:

[0093] In a digital temperature and humidity composite sensor, the temperature sensing part usually uses a thermistor or a thermocouple. By detecting the resistance or voltage change caused by the temperature change, it is converted into a digital temperature signal through a signal processing circuit. The humidity sensing part generally uses a polymer humidity-sensitive capacitor, whose capacitance value changes with the ambient humidity. The humidity data is obtained by measuring the capacitance value change.

[0094] (III) Pressure sensing principle:

[0095] The piezoresistive pressure sensor inside has a piezoresistor that deforms under pressure, resulting in a change in its resistance value. The resistance change is converted into a voltage signal through a Wheatstone bridge circuit, and then through signal processing circuits such as amplification and filtering, an electrical signal proportional to the pressure is finally output.

[0096] (IV) Micro water content sensing principle:

[0097] In a capacitive micro water content sensor, after the humidity-sensitive film of the sensor adsorbs moisture in the gas, its dielectric constant changes, resulting in a change in the capacitance value. The micro water content of the gas is determined by measuring the capacitance value change. The dew point method micro water content sensor cools the gas to its dew point temperature. When the water vapor in the gas begins to condense into dew, by detecting the dew point temperature and combining parameters such as gas pressure, the micro water content of the gas is calculated.

[0098] (V) Vibration sensing principle:

[0099] The acceleration-type vibration sensor is based on the piezoelectric effect or piezoresistive effect. When the sensor is subjected to vibration, the internal piezoelectric material or piezoresistive element generates a change in charge or resistance. The weak signal is amplified and converted into a voltage signal through a charge amplifier or a signal conditioning circuit, and then an electrical signal output related to the vibration acceleration is obtained.

[0100] IV. Working process of the module:

[0101] (I) Initialization stage:

[0102] After the system starts, the sensor array module completes the self-check operation of each sensor, detecting whether the sensor power supply is normal, whether the communication link is unobstructed, and whether the performance of the sensor itself meets the standards.

[0103] Initialize the working parameters of the sensor, such as the sampling frequency, measurement range, data transmission protocol, etc., and establish a data connection channel with the edge computing module.

[0104] (II) Data acquisition stage:

[0105] Each type of sensor continuously and real-time collects the corresponding parameter data according to the set sampling frequency. For example, the SF6 gas concentration sensor collects gas concentration data once per second, and the temperature and humidity sensor collects temperature and humidity data once every 5 seconds, etc.

[0106] The sensor performs preliminary processing on the collected raw data, such as filtering and denoising, range conversion, etc., and converts the analog signal into a digital signal.

[0107] (III) Data transmission stage:

[0108] The data transmission and interface module aggregates the data processed by each sensor and packages and encapsulates it according to a specific data format.

[0109] Through a shielded cable or fiber optic transmission line, the encapsulated data is stably transmitted to the edge computing module, and a verification mechanism is adopted during the transmission process to ensure the accuracy and integrity of the data.

[0110] (IV) Data update and maintenance stage:

[0111] The sensor array module continuously collects data and updates the data transmitted to the edge computing module in real time.

[0112] Regularly calibrate and maintain the sensor, calibrate the accuracy of the sensor according to the usage duration and environmental conditions, and replace the aging or damaged sensor to ensure the long-term stable operation of the module.

[0113] (V) End stage:

[0114] When the system stops running or receives a stop instruction, the sensor array module stops data acquisition and transmission operations, shuts down the power supply of each sensor, saves the working parameters and status information of the sensors, and waits for the next system startup.

[0115] In this embodiment, the edge computing module, as the "local brain" of the SF6 gas status intelligent diagnosis and early warning system based on the Internet of Things, undertakes the key tasks of data preprocessing and preliminary diagnosis; in the complex electromagnetic environment of the substation, the sensor array module will generate a large amount of raw data. The edge computing module reduces the data transmission pressure by processing data locally, and at the same time quickly outputs the preliminary diagnosis results, providing a strong guarantee for the efficient operation of the system. Specifically, it is reflected in:

[0116] I. Overall function overview:

[0117] This module is integrated into the local terminal, receives the SF6 gas status parameters and equipment operation environment data transmitted by the sensor array module, purifies the quality of the raw data through preprocessing means such as noise filtering, data normalization, and outlier removal; at the same time, analyzes the preprocessed data based on the dynamic baseline model, generates preliminary diagnosis results, and uploads the key data and diagnosis information to the intelligent analysis module of the cloud platform to achieve hierarchical processing and efficient transmission of data;

[0118] II. Composition and functions of sub-modules:

[0119] (I) Data preprocessing unit:

[0120] Noise filtering: An adaptive filtering algorithm is adopted to dynamically adjust the filtering parameters according to the time-domain and frequency-domain characteristics of the data for the electromagnetic interference noise, environmental fluctuation noise, etc. existing in the data collected by the sensors; for example, the Kalman filtering algorithm is used to effectively filter out random noise and retain the real signal characteristics by combining the historical values and predicted values of the sensor data;

[0121] Data normalization: The data collected by different types of sensors with different dimensions and value ranges is converted into a unified numerical interval, such as [0,1] or a standard distribution with a mean of 0 and a standard deviation of 1, through the min-max normalization or Z-score normalization method, which is convenient for subsequent data processing and analysis;

[0122] Outlier removal: An outlier detection mechanism combining the 3σ principle based on statistics and the isolation forest algorithm is established; for data points that are significantly deviated from the mean by 3 times the standard deviation, they are initially determined as outliers; then the isolation forest algorithm is used to deeply mine the data, identify the isolated points and outlier data, and remove them to ensure the reliability of the input data;

[0123] (II) Dynamic baseline model processing unit:

[0124] Model construction and update: Based on historical data and equipment operation parameters (equipment model, commissioning time, environmental temperature and humidity, etc.), construct dynamic threshold intervals for SF6 gas concentration, pressure, and micro water content; introduce an environmental coupling factor, combine equipment vibration data with local meteorological information (atmospheric pressure, wind speed), and dynamically adjust the warning sensitivity of the baseline model; adopt an improved hybrid time series prediction algorithm, such as the formula (where is the predicted value of the gas state at time is the weight coefficient of the ARIMA model ; ARIMA is the autoregressive integrated moving average model, is the order of the autoregressive term, is the order of differencing, is the order of the moving average term; is the Prophet algorithm model, is the seasonal component of the Prophet algorithm, is the trend component of the Prophet algorithm, is the environmental variable correction term (including parameters such as temperature and vibration); is the environmental coupling coefficient (dynamically calculated through ridge regression)), and through a double residual feedback mechanism (where is the first-level residual, is the true value of the gas state at time is the predicted value of the gas state at time), (where is the second-level residual, is the result of wavelet transform on the first-level residual ), realize dynamic fine-tuning and optimization of model parameters;

[0125] Preliminary diagnostic analysis: Input the preprocessed data into the dynamic baseline model, compare the real-time data with the dynamic threshold interval, and analyze the change trend and deviation degree of the data; if the data exceeds the threshold range or shows an abnormal change trend, combine the prediction results of the model to generate a preliminary diagnostic conclusion, such as judging whether there are potential faults such as SF6 gas leakage and abnormal pressure fluctuations in the equipment;

[0126] (III) Data transmission and interface unit:

[0127] Data Screening and Packing: According to the requirements of the intelligent analysis module of the cloud platform, screen the preprocessed data and preliminary diagnosis results, and extract key data and diagnostic information; pack and encapsulate the screened data in a specific data format, and add metadata such as data identifiers and timestamps to facilitate data parsing and processing by the cloud platform;

[0128] Communication Protocol Adaptation: Adopt the LoRa wireless communication technology of the anti-interference communication module, adapt its forward error correction coding (FEC) and orthogonal frequency division multiple access (OFDMA) modulation technologies to ensure reliable data transmission in the strong electromagnetic interference environment of the substation; establish a stable data transmission link, monitor the data transmission status in real time, and automatically perform retransmission or error correction when transmission interruptions or errors occur;

[0129] (4) Local Storage and Backup Unit:

[0130] Data Storage Management: Set up local storage devices to locally store the original sensor data, preprocessed data, and preliminary diagnosis results; use a database management system to classify and index the data for quick query and call; set different storage periods according to the importance and usage frequency of the data, and regularly clean up expired data to release storage resources;

[0131] Data Backup Mechanism: Establish a data backup strategy to regularly back up key data to external storage devices or cloud storage platforms; when the system fails or data is lost, be able to quickly restore data from the backup to ensure the continuity of the system and the integrity of the data;

[0132] III. Key Technical Principles:

[0133] (1) Principle of Data Preprocessing Technology:

[0134] The adaptive filtering algorithm realizes the dynamic suppression of noise by establishing a state space model of the data, using the state estimation of the previous moment and the observed data of the current moment to continuously update the filtering parameters; the min-max normalization maps the data to a specified interval through a linear transformation, and the formula is (where, is the original data, and are the minimum and maximum values of the data respectively, is the normalized data; and are the lower and upper limits of the target interval); the Z-score normalization is based on the mean and standard deviation of the data for standardization, and the formula is (where, is the data mean, (where σ is the standard deviation of the data); the 3σ principle is based on the assumption of a normal distribution and holds that the probability of data falling outside the range of the mean ± 3 times the standard deviation is extremely small, and such data points can be determined as outliers; the Isolation Forest algorithm constructs random binary trees and calculates the path lengths of data points, and the data points with shorter paths are more likely to be outliers;

[0135] (2) Principle of the dynamic baseline model:

[0136] The dynamic baseline model combines traditional statistical models and machine learning algorithms, and uses historical data and equipment operation parameters to establish a basic threshold interval; by introducing an environmental coupling factor and considering the influence of environmental factors on gas state parameters, the model can adapt to changes under different working conditions; the improved hybrid time series prediction algorithm integrates the ability of the ARIMA model to capture time series trends and seasonality, and the advantages of the Prophet algorithm in dealing with complex trends and outliers, and further improves the prediction accuracy through an environmental variable correction term; the dual residual feedback mechanism uses wavelet transform to perform multi-scale analysis on the residuals, extracts the detailed information in the residuals, and reversely optimizes the model parameters to improve the fitting accuracy of the model;

[0137] (3) Principle of data transmission and storage:

[0138] In terms of data transmission, the LoRa wireless communication technology uses spread spectrum modulation technology, combines forward error correction coding and orthogonal frequency division multiple access technology to enhance the signal anti-interference ability and transmission efficiency; in terms of data storage, the database management system realizes the efficient storage, fast query and secure management of data through data file storage, index structure establishment and transaction processing mechanisms; the backup technology ensures the security and recoverability of data through data replication, incremental backup, etc.;

[0139] IV. Working process of the module:

[0140] (1) Initialization stage:

[0141] After the edge computing module is started, it completes the self-check of hardware devices, including the status detection of the processor, memory, storage device, communication interface, etc., to ensure the normal operation of the device;

[0142] Load the data preprocessing algorithm, initial parameters and configuration files of the dynamic baseline model, establish communication connections with the sensor array module and the intelligent analysis module of the cloud platform, and initialize the local storage database;

[0143] (2) Data reception stage:

[0144] Real-time monitor the data transmission channel of the sensor array module and receive the original data such as SF6 gas concentration, temperature and humidity, pressure, micro water content and vibration;

[0145] Perform integrity and accuracy verification on the received data. If data errors or omissions are found, request the sensor array module to retransmit in a timely manner;

[0146] (3) Data preprocessing stage:

[0147] Send the original data to the noise filtering, data normalization, and outlier removal units in sequence, and process it according to the preset algorithms and parameters to purify the data quality;

[0148] Perform quality assessment on the preprocessed data. If the data quality does not meet the requirements, return for reprocessing or mark it as abnormal data;

[0149] (4) Preliminary diagnosis stage:

[0150] Input the preprocessed data into the dynamic baseline model processing unit, use the model for data analysis and prediction, and generate preliminary diagnosis results;

[0151] Review and confirm the preliminary diagnosis results, and add diagnostic basis and relevant explanatory information;

[0152] (5) Data transmission stage:

[0153] According to the requirements of the cloud platform intelligent analysis module, screen the key data and diagnosis results, and perform packaging and encapsulation;

[0154] Through the anti-interference communication module, transmit the encapsulated data to the cloud platform intelligent analysis module, and monitor the data transmission status in real time to ensure that the data is delivered accurately and error-free;

[0155] (6) Data storage and backup stage:

[0156] Store the original data, preprocessed data, and preliminary diagnosis results in the local database, and classify and store them according to data types and time;

[0157] According to the set backup strategy, regularly back up the key data, and record the backup time and status information;

[0158] (7) End stage:

[0159] When the system stops running or receives a stop instruction, the edge computing module stops data reception, processing, and transmission operations, closes the communication connections with the sensor array module and the cloud platform intelligent analysis module, saves the current working status and parameters, and releases system resources.

[0160] In this embodiment, the anti-interference communication module includes:

[0161] A LoRa channel encoder based on convolutional codes. The encoding process is: perform modulo operation on the information bit polynomial and the preset generating polynomial to generate encoded data with redundant check bits;

[0162] An adaptive frequency hopping mechanism, and its frequency point selection method is as follows: allocate available frequency band resources according to the real-time signal strength RSSI ratio of each frequency band, and select the communication frequency band with the least interference;

[0163] An encryption transmission protocol, which uses the AES-256 algorithm to perform end-to-end encryption on sensor data and control instructions;

[0164] Furthermore, the anti-interference communication module is the key hub to ensure reliable data transmission in the SF6 gas status intelligent diagnosis and early warning system based on the Internet of Things; in the complex environment of strong electromagnetic interference in the substation, this module integrates a variety of advanced communication technologies to ensure the stable and efficient transmission of sensor data and control instructions between the edge computing module and the cloud platform, and builds a solid data transmission defense line for the normal operation of the system, specifically manifested as:

[0165] I. Overall function overview:

[0166] This module uses LoRa wireless communication technology as the basis, integrates forward error correction coding (FEC), orthogonal frequency division multiple access (OFDMA) modulation technology, adaptive frequency hopping mechanism and encryption transmission protocol to build a complete anti-interference communication system; it is responsible for uploading the monitoring data of SF6 gas equipment processed by the edge computing module to the cloud platform, and at the same time receiving the control instructions issued by the cloud platform and transmitting them to the early warning and linkage control module to ensure the accuracy, integrity and security of data transmission in the complex electromagnetic environment;

[0167] II. Composition and functions of sub-modules:

[0168] (I) LoRa channel coding unit:

[0169] Convolutional code encoder: The LoRa channel encoder based on convolutional code uses the formula (where is the encoded polynomial for transmission in the channel; is the information bit, that is, the encoded representation of the original data to be transmitted: is the number of information bits; is the generating polynomial, which is the core of the pre-set coding rule and determines the way of adding redundant information after encoding. Through this formula calculation, redundant check information is added to the original information bits according to specific rules to generate error correction coding, so that when data transmission errors occur, the receiving end can perform error detection and correction according to the coding rule); before data transmission, perform convolutional coding processing on the data output by the edge computing module to increase the redundancy of the data and improve the anti-interference ability and error correction ability of the data during transmission;

[0170] Coding parameter configuration: According to the actual communication environment and data transmission requirements, parameters such as the coding rate and constraint length of the convolutional code can be flexibly configured; a higher coding rate can improve data transmission efficiency, but the error correction ability is relatively reduced; a larger constraint length enhances the error correction ability, but increases the coding complexity and transmission delay; by reasonably adjusting the parameters, the balance between transmission efficiency and anti-interference performance can be achieved in different scenarios;

[0171] (2) Adaptive frequency hopping unit:

[0172] Frequency point selection algorithm: Use the formula (where, is the frequency point selected for the th frequency band, that is, the frequency finally used for data transmission; is the base frequency point, which is the reference frequency for the entire frequency point selection; is the signal strength of the th frequency band, reflecting the strength of the signal received in this frequency band currently; is the total number of available frequency bands, clarifying the number of selectable frequency ranges; is the sum of the signal strengths of all available frequency bands, used to calculate the relative proportion of the signal strength of each frequency band; is the frequency point interval, determining the frequency difference between adjacent selectable frequency points; is the floor function, ensuring that the calculated frequency point is an integer); this algorithm monitors the signal strength (RSSI) of each frequency band in real time, and dynamically selects the frequency band with the least interference and the best signal quality for data transmission according to the relative size of the signal strength; when a certain frequency band is affected by strong electromagnetic interference and the signal strength drops, the system quickly switches to other high-quality frequency bands to avoid data transmission interruption or error code;

[0173] Frequency hopping period and strategy: Set a reasonable frequency hopping period, which can not only avoid continuous interference frequency bands in time, but also not increase the communication overhead due to too frequent frequency hopping; at the same time, formulate multiple frequency hopping strategies, such as random frequency hopping, sequential frequency hopping, etc., and flexibly select according to different electromagnetic interference characteristics and communication requirements to further enhance the anti-interference ability;

[0174] (3) Modulation and demodulation unit:

[0175] OFDMA modulation technology: Use orthogonal frequency division multiple access (OFDMA) modulation technology to divide the total bandwidth into multiple orthogonal subcarriers, and each subcarrier can transmit data independently; at the sending end, the data to be transmitted is allocated to different subcarriers for parallel transmission, improving the spectrum utilization rate and data transmission rate; at the same time, through the orthogonality between subcarriers, multi-path fading and inter-symbol interference are effectively suppressed, ensuring the reliable transmission of data in a complex electromagnetic environment;

[0176] Demodulation and Signal Recovery: At the receiving end, corresponding demodulation algorithms are used to process the received signals, separate the original data from each subcarrier, and restore complete and accurate information through signal processing techniques; during the demodulation process, combined with the error correction ability of LoRa channel coding, correct the error codes generated during the transmission process to ensure the accuracy of the data;

[0177] (IV) Encryption Transmission Unit:

[0178] AES-256 Encryption Algorithm: The Advanced Encryption Standard AES-256 algorithm is used to perform end-to-end encryption on sensor data and control instructions; the AES-256 algorithm uses a 256-bit key to encrypt data in groups, and through multiple rounds of byte substitution, row shift, column confusion, and round key addition operations, convert the original data into ciphertext; before data transmission, use the pre-negotiated key to encrypt the data, and only the receiving end with the correct key can decrypt and restore the data to ensure the security of the data during transmission and prevent the data from being stolen or tampered with;

[0179] Key Management: Establish a perfect key management mechanism, including key generation, distribution, update, and storage; use a secure key generation algorithm to generate high-strength keys, and distribute the keys to both communication parties through a secure channel; update the keys regularly to reduce the risk of the keys being cracked; at the same time, encrypt and store the keys to ensure the security of the keys;

[0180] (V) Communication Link Monitoring Unit:

[0181] Status Monitoring: Real-time monitor various parameters of the communication link, such as signal strength, bit error rate, data transmission rate, etc.; through the analysis of these parameters, judge the operating status of the communication link, and timely detect link failures or performance degradation;

[0182] Fault Handling: When a communication link failure or abnormality is detected, automatically start the fault handling mechanism; for example, if the bit error rate is found to be too high, trigger the retransmission mechanism to retransmit the error data; if the communication link is interrupted, try to re-establish the connection to ensure the continuity of data transmission; at the same time, report the fault information to the system management platform for operation and maintenance personnel to conduct troubleshooting and repair in a timely manner;

[0183] III. Key Technical Principles:

[0184] (I) Convolutional Code Error Correction Principle:

[0185] Convolutional code is an error control code with memory. It introduces redundant information during the encoding process by performing a convolution operation on the input information sequence and the generating polynomial. At the receiving end, using this redundant information and a specific decoding algorithm (such as the Viterbi decoding algorithm), according to the maximum likelihood criterion, the original correct information sequence is recovered from the possibly erroneous received sequence. By increasing the coding constraint length and reasonably designing the generating polynomial, convolutional code can effectively improve the reliability of data transmission without significantly increasing the transmission bandwidth.

[0186] (2) Principle of adaptive frequency hopping:

[0187] Adaptive frequency hopping technology is based on the real-time perception of the channel environment. By continuously monitoring the signal strength and interference situation of each frequency band, it dynamically adjusts the operating frequency according to a preset frequency point selection algorithm. When a certain frequency band is interfered, the system quickly switches to other idle or less interfered frequency bands for communication, thus avoiding the interference source and achieving stable transmission of the communication link. Its core lies in the fast and accurate channel state perception and efficient frequency point switching mechanism to ensure continuous communication in a complex and changing electromagnetic environment.

[0188] (3) Principle of OFDMA modulation:

[0189] OFDMA technology divides high-speed data services into multiple low-speed sub-data streams, which are respectively modulated onto multiple mutually orthogonal subcarriers for parallel transmission. Since the subcarriers are mutually orthogonal, at the receiving end, the data on each subcarrier can be accurately separated through orthogonal demodulation technology, effectively overcoming the influence of multipath fading and inter-symbol interference. At the same time, OFDMA can flexibly allocate subcarriers to different users or services, realizing multi-user sharing of spectrum resources and improving the spectrum utilization rate and system capacity.

[0190] (4) Principle of AES-256 encryption:

[0191] The AES-256 encryption algorithm is based on the principle of symmetric encryption. The same 256-bit key is used for both encryption and decryption. Through multiple rounds of complex transformations, including byte substitution (SubBytes), row shift (ShiftRows), column mixing (MixColumns), and round key addition (AddRoundKey) and other operations, the original data is scrambled and mixed with the key to generate ciphertext. Only by performing reverse operations with the same key can the ciphertext be restored to the original data. The high-strength encryption characteristics and complex transformation process of AES-256 enable it to resist various cryptographic attacks and ensure the confidentiality and integrity of data.

[0192] In this embodiment, the intelligent analysis module of the cloud platform is the "intelligent center" of the SF6 gas status intelligent diagnosis and early warning system based on the Internet of Things, undertaking the core tasks of deeply mining massive monitoring data and making intelligent decisions. When facing multi-source heterogeneous data uploaded by the edge computing module, this module uses advanced data fusion algorithms and deep learning models to accurately analyze the SF6 gas leakage trend, equipment aging status, and environmental coupling effects, providing key support for the system to achieve intelligent early warning and scientific decision-making, specifically manifested as follows:

[0193] I. Overall function overview:

[0194] This module receives the preprocessed SF6 gas equipment monitoring data uploaded by the edge computing module, covering multi-dimensional information such as gas concentration, pressure, micro water content, temperature and humidity, vibration, etc. It integrates the correlation relationships between different types of data through multi-modal data fusion algorithms, and combines deep learning models to deeply analyze the data, realizing functions such as gas leakage trend prediction, leakage source location, equipment aging status assessment, and risk level classification. At the same time, it provides services such as data query and analysis result display for other modules of the system, and generates visual reports to assist operation and maintenance personnel in making decisions;

[0195] II. Composition and functions of sub-modules:

[0196] (I) Multi-modal data fusion unit:

[0197] Data integration: Align and integrate different types of sensor data uploaded by the edge computing module according to the time series, and construct a data set containing multi-dimensional information. In view of the characteristics of heterogeneous data, standardize the data to unify the data format and dimension, laying a foundation for subsequent fusion analysis;

[0198] Fusion algorithm: Adopt an algorithm combining feature-level fusion and decision-level fusion. In terms of feature-level fusion, use dimensionality reduction algorithms such as principal component analysis (PCA) and independent component analysis (ICA) to extract the key features of each type of data and fuse them. In terms of decision-level fusion, use methods such as evidence theory and Bayesian network to comprehensively make decisions on the preliminary diagnosis results obtained by different sensors based on their respective data, improving the accuracy and reliability of data fusion;

[0199] (II) Deep learning analysis unit:

[0200] Leakage source location sub-model: Adopt an improved multi-scale attention CNN model, and its feature map generation formula is (where is the finally generated feature map for leakage source location analysis; is the number of convolutional kernels, which determines the dimension of the features extracted by the model; Through GRU dynamic calculation, is the attention weight matrix corresponding to the th convolution kernel, which is used to highlight the key features related to the leakage source; is the element-wise multiplication operation to achieve the fusion of attention weights and convolution features; Conv2D is to use the th convolution kernel to perform the result of two-dimensional convolution operation on the input feature map

[0201] Risk prediction sub-model: Using the LSTM-TCN hybrid architecture, the dilation convolution formula of the temporal convolutional network (TCN) is (where is the output at time, which is used for risk prediction; is the length of the convolution kernel: is the weight of the th convolution kernel: is the input at time; is the exponentially growing dilation factor, and is the number of layers of the dilation convolution. By increasing the receptive field of the network through the dilation convolution, the long-term dependence relationship of the data is captured); This sub-model combines the processing ability of LSTM for time series data and the dilation convolution characteristics of TCN to analyze the operation data of SF6 gas equipment, predict the future risk level of the equipment, and discover potential fault hazards in advance;

[0202] Model training and optimization: Using a large amount of historical monitoring data and actual fault cases to train the deep learning model; Adopting the federated learning optimization mechanism (where is the global model parameter; is the number of clients participating in federated learning; is the model parameter of the th client; is the privacy protection strength coefficient; is the th client data distribution and the average distribution of all client data between Divergence, which is used to balance the impact of data differences among clients on model training), integrates multi-party data to optimize model parameters and improve the generalization ability and prediction accuracy of the model while protecting data privacy;

[0203] (3) Data Analysis and Decision-making Unit:

[0204] Trend Analysis: Conduct time-series analysis on the fused data, draw trend charts of parameters such as SF6 gas concentration, pressure, and micro water content, and analyze the change trends of the parameters in combination with historical data and equipment operation rules to determine whether there are abnormal development trends in the equipment;

[0205] Risk Assessment: Quantitatively evaluate the operation risk of SF6 gas equipment according to the prediction results of the deep learning model and preset risk assessment indicators, and divide the risk levels (such as low risk, medium risk, high risk) to provide a decision-making basis for the warning and linkage control module;

[0206] Root Cause Analysis: Use a root cause analysis tool driven by a knowledge graph, and its correlation calculation formula is (where is the correlation between entity and entity for analyzing the cause of the fault is the number of feature dimensions: is the feature encoding of entity in the -th dimension; is the feature encoding of entity in the -th dimension); By constructing a knowledge graph of equipment operation status, analyze the correlation between faults and various influencing factors, quickly locate the root cause of the faults, and provide guidance for equipment maintenance;

[0207] (4) Data Management and Service Unit:

[0208] Data Storage: Establish a distributed database to classify and store the data uploaded by the edge computing module, model training data, analysis result data, etc.; Adopt data compression and indexing technologies to improve data storage efficiency and query speed, and regularly back up the data to ensure data security and integrity;

[0209] Data Query and Display: Provide a data query interface for system users, and users can query relevant data according to conditions such as time, equipment type, parameter type, etc.; Display the analysis results in the form of visual charts (such as line charts, bar charts, heat maps, etc.) and reports to facilitate users to intuitively understand the equipment operation status and analysis results;

[0210] Compliance Report Generation: By using a compliance report generator, it automatically matches the standard terms of the power industry, conducts compliance checks on equipment operation data and analysis results, and generates audit logs and compliance reports to meet the regulatory requirements of the power industry;

[0211] III. Key Technical Principles:

[0212] (I) Multi-modal Data Fusion Principle:

[0213] Multi-modal data fusion is a technology that organically combines multi-source heterogeneous data from different sensors to obtain more comprehensive and accurate information; Feature-level fusion reduces data redundancy by extracting the features of each data and merging similar features; Decision-level fusion is based on the independent decision results of each sensor, and through comprehensive judgment, it obtains the final decision, improving the reliability and robustness of the decision; For example, by combining the data of gas concentration sensors and vibration sensors, through fusion analysis, it can more accurately judge whether there are leaks and mechanical failures in the equipment;

[0214] (II) Deep Learning Model Principle:

[0215] The improved multi-scale attention CNN model enables the model to automatically learn and focus on the key features related to the leakage source location by introducing the attention mechanism, improving the location accuracy; In the LSTM-TCN hybrid architecture, LSTM can effectively handle the long-term dependencies in time series data, and the dilated convolution of TCN further expands the network's perception range of historical data. The combination of the two realizes the accurate prediction of equipment risks; The federated learning optimization mechanism integrates multi-party data to update model parameters without sharing the original data, protecting data privacy while improving model performance;

[0216] (III) Root Cause Analysis Principle of Knowledge Graph:

[0217] The knowledge graph represents various factors (such as equipment parameters, environmental conditions, fault phenomena, etc.) during the operation of the equipment in a structured form by constructing a relationship network between entities and entities; By using the correlation degree calculation formula to analyze the correlation degree between different entities, when the equipment fails, the knowledge graph quickly searches for the entities and relationships related to the fault to trace the root cause of the fault, providing a basis for fault diagnosis and handling.

[0218] In this embodiment, the warning and linkage control module is the core execution unit for risk response in the SF6 gas status intelligent diagnosis and warning system based on the Internet of Things. Like the "action center" of the system, it quickly responds and takes measures after the cloud platform intelligent analysis module outputs the diagnosis result; Through hierarchical warning and intelligent linkage, this module effectively reduces the risks brought by SF6 gas equipment failures and ensures the safe and stable operation of the power system. Specifically, it is reflected as:

[0219] 1.Overall Function Overview:

[0220] This module receives the SF6 gas equipment operating status diagnosis results and risk level assessment information output by the cloud platform intelligent analysis module, triggers the corresponding level of early warning signals based on preset rules and algorithms, and notifies relevant personnel in a variety of forms; at the same time, according to the risk situation, it links the fan, gas replenishment device, access control system and other equipment to automatically perform emergency response operations, forming a closed-loop management from risk identification to disposal;

[0221] 2. Sub-module composition and functions:

[0222] 1. Adaptive warning strategy unit:

[0223] Warning level determination: Based on the gas leakage rate output by the cloud platform, the importance level of the equipment (such as the criticality of the equipment in the power grid, the scope of fault impact, etc.) and the presence of personnel (obtained through the access control system or personnel positioning equipment), a multi-factor comprehensive evaluation algorithm is used to dynamically adjust the warning level; the warning level is divided into three levels: general warning, severe warning and emergency warning, and different levels correspond to different risk levels;

[0224] Warning signal generation: for different warning levels, corresponding sound and light alarm signals, mobile terminal push information and remote monitoring platform pop-up window prompts are generated; general warnings use low-frequency sound and light alarms, and mobile terminals push brief risk warnings; severe warnings increase the frequency of sound and light alarms, and push detailed risk analysis and response suggestions; emergency warnings trigger high-intensity sound and light alarms, and push emergency warning information in a variety of ways such as SMS and phone calls, and pop-up window warnings are displayed in prominent locations on the remote monitoring platform;

[0225] (II) Gas supply control logic unit:

[0226] Pressure monitoring and deviation calculation: Real-time acquisition of internal pressure sensor data of SF6 gas equipment, comparison with the normal operating pressure range of the equipment, and calculation of pressure deviation ; When the pressure is lower than the lower limit of the normal range, the air replenishment control process is started;

[0227] Improved PID-fuzzy composite control: using formula (in, for The proportionality factor of the moment, is the initial proportionality factor, is the adaptive coefficient, is the rate of change of pressure deviation) and for The integration coefficient at time, is the initial integration coefficient, is the adaptive coefficient, is the absolute value of the pressure deviation Dynamically adjust the parameters of the PID controller; combine fuzzy control rules, and according to the pressure deviation and its rate of change, optimize the air replenishment rate and air replenishment volume of the air replenishment device in real time to ensure that the equipment pressure quickly returns to the normal range, and at the same time avoid excessive pressure fluctuations during the air replenishment process;

[0228] (3) Emergency ventilation mechanism unit:

[0229] Path planning: Plan the ventilation path of the fan based on the dynamic weight A* algorithm, and the cost function is (where, is the cost function value of node , is the actual cost from the starting node to node , is the time decay factor, is the decay coefficient, is the time. As time goes by, the path planning is more inclined to reach the target quickly; is the heuristic estimated cost from node to the target node, is the gas concentration influence coefficient, is the gas concentration at node , is the maximum allowable gas concentration); Considering the ventilation path length, time factor and gas concentration distribution comprehensively, calculate the optimal ventilation path to enable the fan to efficiently disperse the leaked SF6 gas and reduce the gas concentration in the environment;

[0230] Fan control: Control the start, stop, speed and wind direction of the fan according to the planned path and gas concentration situation; when a high-concentration SF6 gas leak is detected, start multiple fans to work together and ventilate according to the planned path to quickly reduce the gas concentration in the dangerous area and ensure the safety of personnel and the normal operation of equipment;

[0231] (4) Access control linkage control unit:

[0232] Personnel control: When the system issues a serious or emergency warning, link the access control system to restrict non-essential personnel from entering the area where SF6 gas equipment is located to prevent personnel from being poisoned or other dangers due to contact with the leaked gas; at the same time, record the personnel entry and exit information through the access control system for subsequent safety management and accident tracing;

[0233] Emergency passage management: While restricting personnel entry, ensure the smoothness of the emergency rescue passage to provide guarantee for professional maintenance personnel and rescue teams to quickly enter the scene; through the permission setting and automatic control of the access control system, realize the flexible opening and closing of the emergency passage;

[0234] (V) Information Interaction and Feedback Unit:

[0235] Information Sending: Real-time send early warning information, equipment linkage status, etc. to the mobile terminals of operation and maintenance personnel, remote monitoring platforms, and systems of relevant management departments to ensure that all parties can timely understand the equipment risk situation and the progress of emergency disposal;

[0236] Feedback Receiving and Processing: Receive operation instructions, feedback information, and on-site situation reports sent by operation and maintenance personnel through mobile terminals or monitoring platforms, parse and process the instructions, and adjust the early warning and linkage control strategies; for example, when the operation and maintenance personnel confirm the on-site situation, they can manually lift the early warning or adjust the equipment linkage parameters.

[0237] In this embodiment, the energy management module is the "power heart" of the SF6 gas status intelligent diagnosis and early warning system based on the Internet of Things, undertaking the important mission of providing reliable energy guarantee for the stable operation of the system; in complex application scenarios such as substations, this module realizes multi-mode power supply through the coordinated work of the solar power supply unit and the intelligent power distribution circuit, effectively solving the energy supply problem during the system operation process, and ensuring the continuous and stable operation of the system under various environmental conditions, specifically manifested as:

[0238] I. Overall Function Overview:

[0239] The energy management module mainly consists of two parts: a solar power supply unit and an intelligent power distribution circuit; the solar power supply unit uses photovoltaic power generation technology to convert solar energy into electrical energy to provide green and renewable energy for the system; the intelligent power distribution circuit dynamically distributes electrical energy based on the load priority according to the power consumption requirements of each module of the system, and at the same time has functions of power supply status monitoring, switching, and management to ensure the power supply of the system under different working conditions, and improve the energy utilization efficiency and reliability of the system;

[0240] II. Composition and Functions of Sub-Modules:

[0241] (I) Solar Power Supply Unit:

[0242] Photovoltaic Modules: Select high-efficiency monocrystalline silicon photovoltaic modules, whose photoelectric conversion efficiency can reach more than 22%, with characteristics such as high conversion efficiency, strong stability, and long service life; reasonably plan the installation quantity and layout of photovoltaic modules according to the system power consumption requirements and installation environment to ensure that under good lighting conditions, the maximum amount of solar energy can be absorbed and converted into electrical energy;

[0243] Maximum Power Point Tracking (MPPT) Module: Adopt an advanced MPPT algorithm, and its optimization formula is (where is the optimal operating voltage of the photovoltaic cell, which is the voltage when the output power of the photovoltaic cell reaches the maximum value; is the output current of the photovoltaic cell; is the terminal voltage of the photovoltaic cell; is the equivalent series resistance, representing the resistance loss inside the photovoltaic cell: represents when the expression in the brackets reaches the maximum value value); this module monitors the output voltage and current of the photovoltaic module in real time, and by dynamically adjusting the operating point, makes the photovoltaic module always operate at the maximum power output state, improving the utilization efficiency of solar energy; when environmental factors such as light intensity and temperature change, the MPPT module can respond quickly to ensure that the photovoltaic module stably outputs the maximum power;

[0244] Charge and discharge management module: Responsible for controlling and managing the charging and discharging of the storage battery; when the electric energy output by the photovoltaic module is sufficient, the excess electric energy is stored in the storage battery, and intelligent charging strategies are adopted, such as constant current charging, constant voltage charging, floating charging and other staged charging methods, to ensure the safe and efficient charging of the storage battery and extend the service life of the storage battery; when the power demand of the system is greater than the output of the photovoltaic module, the charge and discharge management module controls the storage battery to supply power to the system, and at the same time monitors the power state of the storage battery in real time. When the power is lower than the set threshold, a low power warning is issued to avoid over-discharging and damaging the storage battery;

[0245] (2) Intelligent power distribution circuit:

[0246] Load priority division: According to the urgency and importance of the power supply requirements of each module in the system, the loads are divided into different priorities; for example, the sensor array module, edge computing module, etc. that undertake key functions such as data acquisition and preliminary processing are set as high-priority loads; while some auxiliary function modules, such as some display devices, are set as low-priority loads; through clear priority division, ensure that when the power supply is insufficient, the normal operation of key modules is preferentially guaranteed;

[0247] Dynamic power distribution: Dynamically distribute electric energy based on load priority, using the formula (where is the electric energy allocated to device ; is the priority weight of device , the higher the weight, the higher the priority of the device, and the larger the proportion obtained in the power distribution: is the power demand of device ; is the sum of the priority weights of all devices; is the total electric energy); the intelligent power distribution circuit monitors the total electric energy of the system and the power demands of each load in real time, calculates and distributes electric energy according to the priority weights, realizes the rational use of electric energy, and avoids insufficient power supply to other key modules due to excessive power consumption of a certain module;

[0248] Power Switching and Management: It has multiple power input interfaces and can be connected to various power sources such as solar power supply and mains power supply. When the solar power supply is insufficient or fails, the intelligent power distribution circuit automatically detects and switches to the backup power supply (such as mains power) to ensure the uninterrupted operation of the system. At the same time, it monitors the status of each power source in real time, including parameters such as voltage, current, and frequency. Once a power abnormality is detected, corresponding measures are taken in a timely manner, such as alarm prompts and automatic power switching, and the power status information is recorded for subsequent maintenance and fault troubleshooting.

[0249] In this embodiment, the self-check and fault tolerance module is the "security guard" for the stable operation of the SF6 gas status intelligent diagnosis and early warning system based on the Internet of Things. It ensures the continuous and reliable operation of the system in a complex environment through real-time monitoring and fault handling of key parts of the system. The following will elaborate on this module in detail from the overall to the details:

[0250] I. Overall Function Overview:

[0251] The self-check and fault tolerance module is mainly responsible for periodically detecting the sensor accuracy, communication link integrity, and power status in the system, and promptly discovering potential faults or abnormal situations. Once a fault is detected, it quickly performs fault isolation and redundant switching operations to ensure that the core functions of the system are not affected. At the same time, through data recording and analysis, it provides a basis for system maintenance and optimization, improving the overall robustness and reliability of the system.

[0252] II. Sub-module Composition and Functions:

[0253] (I) Multi-index Fusion Diagnosis Unit:

[0254] Data Acquisition and Preprocessing: Real-time collect the output data of the sensor array module, the signal status data of the anti-interference communication module, and the power parameter data of the energy management module, etc. Perform preprocessing operations such as filtering and noise reduction on the collected data to remove noise and interference in the data, ensuring the accuracy and effectiveness of the input data.

[0255] Multi-index Fusion Diagnosis Algorithm: Use the formula (where is the diagnosis result vector, and each element corresponds to the probability of a fault type; is the number of monitoring indicators; is the weight of the th indicator, reflecting the importance of this indicator for fault diagnosis; is the rectified linear unit activation function used to introduce non-linear relationships; is the coefficient of the th indicator; is the actual measured value of the th indicator; For the threshold of the

[0256] nth metric), multi-metric fusion diagnosis is performed; through this algorithm, the correlation between multiple metrics is comprehensively analyzed to determine whether there is a fault in the system and the type of fault; for example, when the sensor data is abnormal and the signal strength of the communication link drops, the fault diagnosis result is calculated through the algorithm to determine that there may be a sensor fault or a communication link fault; Weight update mechanism: In order to enable the diagnostic algorithm to adapt to the changes during the system operation, a weight update formula is adopted, where is the update amount of the weight of the nth metric; is the learning rate, which controls the step size of weight update; is the partial derivative of the loss function L with respect to the weight of the nth metric, which is used to measure the impact of the weight on the diagnostic result; is the weight coefficient of entropy, which is used to balance the uncertainty of the diagnostic result; is the entropy of the diagnostic result vector

[0257] (II) Fault recovery mechanism unit:

[0258] Fault detection and location: When the multi-metric fusion diagnosis unit detects a fault in the system, the fault is further analyzed and located in detail; by deeply investigating information such as sensor data, communication link status, and power supply parameters, the specific location and cause of the fault are determined; for example, when abnormal sensor data is detected, by comparing the data of multiple sensors with historical data, it is determined whether it is a single sensor fault or a sensor network fault;

[0259] Fault isolation: Once the fault location is determined, the fault isolation operation is immediately executed; for a sensor fault, the connection between the faulty sensor and the system is cut off to prevent the data of the faulty sensor from affecting the normal operation of the system; for a communication link fault, the faulty section of the link is isolated to avoid the spread of the fault to the entire communication network; for a power supply fault, the faulty power supply module is isolated to prevent damage to other modules;

[0260] Redundant Switching: While isolating the fault, start the redundant switching mechanism; for critical components such as sensors and communication modules, the system is equipped with redundant devices; when the primary device fails, it automatically switches to the redundant device to ensure that the core functions of the system are not affected; for example, when the primary sensor fails, the backup sensor automatically takes over the data acquisition task to ensure that the system can continuously obtain accurate monitoring data;

[0261] Local Storage-Forwarding Mode: When a communication interruption is detected, automatically switch to the local storage-forwarding mode and use the formula (where, is the amount of data already stored in the local cache; is the upper limit of the edge storage capacity; is the moment of communication interruption; is the current moment; is the amount of data generated at time is from the moment of communication interruption to the current moment the integral of the total amount of data generated); during the communication interruption, temporarily store the data collected by the sensors in the local cache, and when the communication resumes, forward the data in the cache to the intelligent analysis module of the cloud platform in the order of data generation time to ensure the integrity and continuity of the data;

[0262] (III) System Status Monitoring and Recording Unit:

[0263] Real-time Status Monitoring: Continuously and real-time monitor the operating status of each module of the system, including the working status of sensors, the connection status of communication links, the output status of power supplies, etc.; through the real-time monitoring of these status information, promptly detect abnormal situations during the operation of the system and provide timely data support for fault diagnosis and handling;

[0264] Data Recording and Storage: Record in detail the key data during the operation of the system, such as sensor acquisition data, fault diagnosis results, fault occurrence time and location, etc., and store them in the local database; these data can not only be used for fault analysis and traceability, but also provide data basis for the optimization and improvement of the system;

[0265] Status Report Generation: Regularly generate system status reports to summarize the operation of the system within a certain period of time, including information such as the number of faults occurred, fault types, and fault handling situations; send the status reports to system administrators and maintenance personnel to facilitate their understanding of the system operation status and formulate corresponding maintenance plans and optimization strategies.

[0266] Although embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. The SF6 gas status intelligent diagnosis and early warning system based on the Internet of Things is characterized by: include: Sensor array module: deployed at key monitoring points of SF6 gas equipment, including SF6 gas concentration sensor, temperature and humidity sensor, pressure sensor, micro-water content sensor and vibration sensor, to collect gas state parameters and equipment operating environment data in real time; Edge computing module: integrated in the local terminal, used to pre-process gas state parameters and equipment operating environment data, and generate preliminary diagnosis results based on the dynamic baseline model; Anti-interference communication module: Adopts LoRa wireless communication technology, integrating forward error correction coding FEC and orthogonal frequency division multiple access OFDMA modulation technology; Cloud platform intelligent analysis module: receives data uploaded by the edge computing module, analyzes gas leakage trends, equipment aging status and environmental coupling effects through multimodal data fusion algorithms, and predicts the location of leakage sources and risk levels based on deep learning models; Early warning and linkage control module: triggers graded early warning signals according to the location of the leakage source and the risk level, and links the fan, air supply device and access control system to execute emergency response; Energy management module: includes solar power supply unit and intelligent power distribution circuit, providing multi-mode power supply guarantee for the system; Self-check and fault-tolerance module: periodically checks sensor accuracy, communication link integrity, and power status, and performs fault isolation and redundant switching.

2. The SF6 gas state intelligent diagnosis and early warning system based on the Internet of Things according to claim 1 is characterized in that: The dynamic baseline model is constructed in the following way: Establish dynamic threshold ranges for SF6 gas concentration, pressure and trace water content based on historical data and equipment operating parameters; Introducing environmental coupling factors, combining equipment vibration data with local meteorological information, and dynamically adjusting the warning sensitivity of the baseline model; The improved hybrid time series prediction algorithm is adopted, and its modeling method is as follows: The prediction results of the ARIMA model are weighted and fused with the seasonal term and trend term components of the Prophet algorithm. The value range of the weight coefficient α is 0.6 to 0.8, and the environmental variable correction term is superimposed. The coupling coefficient β of the correction term is dynamically calculated by the ridge regression algorithm. Establish a double residual feedback mechanism: The first-order residual between the real-time monitoring value and the model prediction value is calculated, the second-order residual is decomposed by wavelet transform, and the second-order residual is injected back into the model input layer for parameter fine-tuning.

3. The SF6 gas state intelligent diagnosis and early warning system based on the Internet of Things according to claim 1 is characterized in that: The anti-interference communication module comprises: The LoRa channel encoder based on convolutional code has the following encoding process: the information bit polynomial is modulo-operated with the preset generator polynomial to generate the encoded data with redundant check bits; Adaptive frequency hopping mechanism, the frequency selection method is: allocate available frequency band resources according to the real-time signal strength RSSI ratio of each frequency band, and select the communication frequency band with the least interference; Encrypted transmission protocol uses AES-256 algorithm to perform end-to-end encryption on sensor data and control instructions.

4. The SF6 gas state intelligent diagnosis and early warning system based on the Internet of Things according to claim 1 is characterized in that: The deep learning model includes: Leakage source localization sub-model: It uses an improved multi-scale attention convolutional neural network to dynamically calculate the attention weight matrix of each convolution kernel through a gated recurrent unit, and performs weighted fusion of convolution feature maps of different scales; Risk prediction sub-model: It adopts a hybrid architecture of LSTM and temporal convolutional network. The temporal convolutional network captures multi-scale time series features through an exponentially growing expansion factor. Federated learning optimization mechanism: The local model parameters of each site are weighted and aggregated, and the weight value is inversely proportional to the privacy protection strength of the site data distribution.

5. The SF6 gas state intelligent diagnosis and early warning system based on the Internet of Things according to claim 1 is characterized in that: The early warning and linkage control module includes: Adaptive warning strategy: dynamically adjust the warning level according to the leakage rate, equipment importance level and personnel presence; Air supply control logic: PID-fuzzy composite controller is used, and its parameter setting method is as follows: The proportional coefficient Kp is dynamically adjusted according to the differential value of the pressure deviation, and the integral coefficient Ki is adaptively increased according to the absolute value of the pressure deviation; Emergency ventilation mechanism: Based on the dynamic weight path planning algorithm, its path cost is composed of the actual movement cost, the estimated remaining cost and the gas concentration penalty term, among which the weight of the concentration penalty term decays exponentially with time.

6. The SF6 gas state intelligent diagnosis and early warning system based on the Internet of Things according to claim 1 is characterized by: The energy management module comprises: Photovoltaic power supply unit: adopts maximum power point tracking algorithm to minimize equivalent series resistance loss by optimizing the output voltage of photovoltaic array; Intelligent power distribution circuit: The total power is distributed proportionally according to the device priority weight, and the devices with higher priority get power supply resources first.

7. The SF6 gas state intelligent diagnosis and early warning system based on the Internet of Things according to claim 1 is characterized in that: The self-checking and fault-tolerance module comprises: Multi-index fusion diagnosis algorithm: The difference between the sensor reading and the calibration coefficient and failure threshold is activated through a modified linear unit, and then the fault probability distribution is output after dynamic weighting; Fault recovery mechanism: When communication is interrupted, it automatically switches to local cache mode, and the amount of stored data does not exceed the maximum capacity of the edge device.

Images