A sulfur hexafluoride gas data monitoring management system

Abstract

The application relates to the technical field of data processing, and discloses a sulfur hexafluoride gas data monitoring management system. The system comprises the following units: a collecting unit that collects SF6 gas density values and temperature values, generates a monitoring original data set in combination with a device identification code and a time stamp; a screening unit that calculates a density-temperature rho-T correlation coefficient and screens reliable monitoring data; a prediction unit that inputs the reliable data into an SF6-NN neural network to predict density values and temperature compensation amounts; a division unit that stores data in a graded manner according to density ranges; and an analysis unit that calculates density stability indexes and device health degree scores and generates SF6 management data. The application solves the technical problems of incomplete data collection coverage, difficult guarantee of data reliability, lack of prediction capability and low data management efficiency in a traditional SF6 gas monitoring system, and improves the data quality, prediction accuracy and management efficiency of SF6 device monitoring.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a sulfur hexafluoride gas data monitoring and management system. Background Technology

[0002] With the rapid development of power systems, sulfur hexafluoride (SF6) gas-insulated switchgear is widely used in high-voltage substations. SF6 gas has excellent insulation and arc-extinguishing properties, and its density, temperature, and other state parameters directly affect the safe operation of the equipment. Traditional SF6 gas monitoring technology mainly uses mechanical density relays and digital pressure gauges for single-point monitoring, and equipment condition assessment is achieved through periodic manual inspections and simple threshold alarms. Existing monitoring methods mostly adopt a centralized architecture for data acquisition, collecting sensor signals to a central control unit for unified processing. Data storage typically uses a single database structure, with all monitoring data stored in the same data table in chronological order.

[0003] However, existing technologies have many shortcomings in practical applications. First, traditional single-point monitoring methods cannot comprehensively reflect the overall status of SF6 equipment, and incomplete monitoring point coverage leads to insufficient data representativeness. Second, existing systems lack effective data quality assessment mechanisms, and factors such as sensor failure, drift, and environmental interference make it difficult to guarantee the reliability of monitoring data. Third, traditional monitoring methods mainly rely on historical data analysis and simple statistical indicators, lacking the ability to predict the changing trends of SF6 gas states and thus failing to achieve preventative maintenance of equipment. In addition, existing data management methods employ a unified storage strategy, mixing data of different importance, resulting in low query efficiency and high storage costs.

[0004] Based on the analysis of the aforementioned technical shortcomings, it can be found that building a complete SF6 gas data monitoring and management system requires solving a series of progressive technical problems. First, a distributed monitoring network needs to be established to address the data acquisition coverage issue; however, distributed acquisition inevitably leads to inconsistent data quality from multiple sources, thus necessitating the design of a reliable data filtering mechanism. Secondly, based on reliable data, an effective prediction model needs to be built to address the state prediction problem; however, the importance of prediction data differs from that of historical data, requiring the establishment of a hierarchical storage mechanism to optimize data management efficiency. Finally, comprehensive analysis of the stored multi-level data is needed to establish a quantitative equipment status evaluation system, addressing the issues of strong subjectivity and lack of unified standards in traditional equipment health assessment methods. Summary of the Invention

[0005] This application provides a sulfur hexafluoride gas data monitoring and management system to solve the technical problems of incomplete data acquisition coverage, difficulty in ensuring data reliability, lack of predictive ability and low data management efficiency in traditional SF6 gas monitoring systems, thereby improving the data quality, prediction accuracy and management efficiency of SF6 equipment monitoring.

[0006] This application provides a sulfur hexafluoride gas data monitoring and management system, which includes:

[0007] The acquisition unit is used to collect the density and temperature values ​​of SF6 gas, and combine them with the equipment identification code and timestamp to obtain the raw dataset of SF6 monitoring.

[0008] The filtering unit is used to calculate the ρ-T correlation coefficient between the density value and the temperature value in the original SF6 monitoring data, and to filter data whose correlation coefficient meets a preset threshold as reliable monitoring data.

[0009] The prediction unit is used to input the reliable monitoring data into the SF6-NN neural network, predict the density value at the next moment and calculate the temperature compensation amount to obtain SF6 state prediction data.

[0010] The partitioning unit is used to divide the SF6 state prediction data into normal, attention and abnormal levels according to the density value in the data and establish a hierarchical SF6 database.

[0011] The analysis unit is used to analyze the time series data in the SF6 database, calculate the density stability index and equipment health score, and generate SF6 management data.

[0012] Optionally, the acquisition unit is used for:

[0013] The density value of SF6 gas is collected by each sensor node in the distributed intelligent density sensor network. Each sensor node reads data according to the set acquisition frequency to obtain the raw sensing data containing the density value.

[0014] Based on the temperature sensor in the original sensing data, the temperature value at the corresponding time is collected synchronously, and the density value and temperature value are paired to obtain a density-temperature data pair.

[0015] Each sensor node is assigned a unique device identifier code according to the sensor network topology. The device identifier code is then associated with the corresponding density-temperature data pair to obtain monitoring data with device identification information.

[0016] Based on the system clock, a precise timestamp is added to each group of monitoring data containing device identification information to complete the data encapsulation process and obtain the original SF6 monitoring dataset.

[0017] Optionally, the filtering unit is used for:

[0018] Based on the density and temperature values ​​in the original SF6 monitoring dataset, data normalization was performed, and density and temperature deviations were calculated to obtain standardized density-temperature deviation data.

[0019] The density-temperature deviation data is subjected to covariance calculation. The covariance numerator is calculated by summing the products of density deviation and temperature deviation. The covariance denominator is obtained by taking the square root of the product of the sum of squares of density deviation and the sum of squares of temperature deviation. The ρ-T correlation values ​​of each sensor node are then obtained.

[0020] The ρ-T correlation values ​​are compared and judged based on a preset correlation coefficient threshold. Data with correlation values ​​greater than the threshold are marked as valid identifiers, and the sensor reliability evaluation results are obtained.

[0021] Based on the sensor reliability evaluation results, monitoring data marked as valid identifiers are selected, and abnormal data with correlation coefficients below the threshold are removed to obtain reliable monitoring data.

[0022] Optionally, the prediction unit is used for:

[0023] The density, temperature, pressure, and humidity values ​​in the reliable monitoring data are normalized and mapped to a value range of 0 to 1 to obtain standardized network input data.

[0024] The input layer of the SF6-NN neural network is constructed based on the standardized network input data. The input layer contains four neurons corresponding to the density, temperature, pressure and humidity parameters, respectively, to obtain the network input layer data.

[0025] The network input layer data is passed to a hidden layer containing 12 neurons for forward propagation calculation. The hidden layer output value is generated by weighted summation and activation function processing to obtain the intermediate layer calculation result.

[0026] The calculation results from the intermediate layer are passed to the output layer, which contains two neurons, for final calculation. The predicted density value and temperature compensation value are output respectively to obtain the network output data.

[0027] The network output data is denormalized to restore the predicted density and temperature compensation values ​​to the range of actual physical quantities, thus obtaining SF6 state prediction data.

[0028] Optionally, the partitioning unit is used for:

[0029] Based on the density values ​​in the SF6 state prediction data, density range judgment processing is performed. The density values ​​are compared with preset density thresholds for normal, attention, and abnormal levels to obtain density level classification identifiers.

[0030] The SF6 state prediction data is grouped according to the density level classification identifier, and the data with the same level identifier are classified into the corresponding data set to obtain SF6 data groups divided by level.

[0031] Based on the SF6 data groups classified by level, establish data table structures with different storage strategies, create independent data storage tables for normal level, attention level and abnormal level respectively, and obtain a hierarchical data table architecture;

[0032] The SF6 data groups, divided by level, are written into the corresponding hierarchical data tables according to the corresponding storage strategies to complete the data entry operation and obtain a hierarchical SF6 database.

[0033] Optionally, the analysis unit is used for:

[0034] The extraction sub-unit is used to perform time series extraction processing on historical density data in the hierarchically stored SF6 database, arrange the density values ​​in the order of timestamps to obtain a continuous time series, and obtain SF6 density time series data.

[0035] The analysis subunit is used to perform density fluctuation analysis and calculation on the SF6 density time series data, and to obtain the density stability coefficient by calculating the ratio of the density standard deviation to the density mean, thereby obtaining the density stability index.

[0036] The weighted subunit is used to perform weighted calculations based on the density stability index, density change trend, and abnormal frequency. By setting different weight coefficients, it performs a comprehensive scoring calculation on each index to obtain the equipment health score.

[0037] The integration subunit is used to integrate and process data based on the density stability index and equipment health score, and associate and combine the score results with the equipment identification code and timestamp to obtain SF6 management data.

[0038] Optionally, the analysis subunit is used for:

[0039] The arithmetic mean of each density value in the SF6 density time series data is calculated by adding all density values ​​in the time series and dividing by the total number of data to obtain the density mean.

[0040] Based on the density mean, the deviation of each density value in the SF6 density time series data is calculated, the difference between each density value and the density mean is calculated and squared, and a density deviation squared value sequence is obtained.

[0041] The density deviation square value sequence is processed by variance calculation. The average of all deviation square values ​​is then taken by square root to obtain the density standard deviation.

[0042] The density stability index is obtained by calculating the ratio between the density standard deviation and the density mean, dividing the density standard deviation by the density mean, and obtaining the coefficient of variation.

[0043] The technical solution provided in this application obtains the technical characteristics of the original SF6 monitoring dataset by collecting SF6 gas density and temperature values ​​and combining them with equipment identification codes and timestamps. This solves the problem of incomplete coverage in traditional single-point monitoring. The distributed acquisition architecture ensures the spatial integrity and temporal continuity of the monitoring data, and the introduction of equipment identification codes enables accurate attribution and tracking management of multi-source data. Based on the calculation and screening techniques of the ρ-T correlation coefficient between density and temperature values, an objective data quality assessment mechanism is established. The degree of correlation between density and temperature is quantified through covariance calculation, effectively identifying sensor faults and measurement anomalies, and significantly improving the reliability of monitoring data. The application of SF6-NN neural network prediction technology features constructs a prediction model specifically optimized for SF6 gas characteristics. The 4-12-2 network structure design fully considers the nonlinear relationships of multi-dimensional parameters such as SF6 gas density, temperature, pressure, and humidity. The output of predicted density values ​​and temperature compensation values ​​provides a scientific basis for equipment status prediction. Tiered storage technology categorizes data into normal, critical, and abnormal levels based on density range, enabling differentiated storage strategies and access optimizations. Normal-level data is compressed for cost reduction, while abnormal-level data is accessed in real-time to ensure fast response times, significantly improving data management efficiency. Time-series analysis and health scoring technologies establish a quantitative equipment status evaluation system. The calculation of density stability indicators and equipment health scores eliminates the subjectivity of traditional assessment methods, providing objective data support for equipment operation and maintenance decisions.

[0044] The ρ-T correlation coefficient algorithm fully utilizes the physical correlation between SF6 gas density and temperature. The mathematical model established through covariance calculation can accurately identify the sensor's operating status, achieving higher fault detection accuracy compared to traditional single-parameter monitoring methods. The SF6-NN neural network algorithm is specifically optimized for the properties of the SF6 gas equation of state. The selection of the number of hidden layer neurons and activation functions considers the response characteristics of SF6 molecules, resulting in better adaptability and accuracy in SF6 state prediction. The density stability index calculation in the time series analysis algorithm uses the coefficient of variation to quantify the fluctuation of SF6 gas density. This algorithm design is particularly suitable for long-term trend analysis of SF6 equipment operating status. The weighted calculation algorithm comprehensively evaluates density stability, change trend, and anomaly frequency. The weight coefficients are set based on the fault modes and operating characteristics of SF6 equipment, ensuring the effectiveness and accuracy of health scores in SF6 equipment status assessment. The synergistic effect of these algorithmic features achieves intelligent and precise SF6 gas data monitoring and management. Attached Figure Description

[0045] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0046] Figure 1 This is a schematic diagram of one embodiment of the sulfur hexafluoride gas data monitoring and management system in this application.

[0047] Figure 2 This is a schematic diagram of the SF6 gas data screening unit evaluating sensor reliability based on the ρ-T correlation coefficient in an embodiment of this application. Detailed Implementation

[0048] This application provides a sulfur hexafluoride gas data monitoring and management system. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or device that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0049] For ease of understanding, the specific process of the embodiments of this application is described below. Please refer to [link / reference]. Figure 1 One embodiment of the sulfur hexafluoride gas data monitoring and management system in this application includes:

[0050] The acquisition unit is used to collect the density and temperature values ​​of SF6 gas, and combine them with the equipment identification code and timestamp to obtain the raw dataset of SF6 monitoring.

[0051] The filtering unit is used to calculate the ρ-T correlation coefficient between density and temperature values ​​in the original SF6 monitoring dataset and filter data whose correlation coefficient meets a preset threshold as reliable monitoring data.

[0052] The prediction unit is used to input reliable monitoring data into the SF6-NN neural network, predict the density value at the next moment and calculate the temperature compensation amount to obtain SF6 state prediction data.

[0053] The partitioning unit is used to divide the SF6 status prediction data into normal, attention, and abnormal levels according to the density range, and to establish a hierarchical SF6 database.

[0054] The analysis unit is used to analyze time-series data in the SF6 database, calculate density stability indicators and equipment health scores, and generate SF6 management data.

[0055] It is understood that the executing entity of this application can be a sulfur hexafluoride gas data monitoring and management system, or it can be a terminal or a server; the specific implementation is not limited here. This application's embodiment uses a server as an example for illustration.

[0056] Specifically, the acquisition unit achieves comprehensive data acquisition of SF6 gas through a distributed intelligent density sensor network. Each sensor node integrates a piezoresistive density sensor and a temperature sensor. The piezoresistive density sensor detects density values ​​based on the thin-film strain caused by changes in SF6 gas density, while the temperature sensor synchronously measures ambient temperature using the thermistor principle. During acquisition, sensor nodes read data at a preset acquisition frequency. Each node is assigned a unique 16-bit device identifier based on the network topology, and the system clock adds a precise timestamp to each data set. The raw sensor data includes density values, and the temperature sensor synchronously acquires the corresponding temperature values. The density and temperature values ​​are paired to form density-temperature data pairs. The device identifier is associated with the density-temperature data pairs to obtain monitoring data with device identification information. Finally, the data is encapsulated to obtain the raw SF6 monitoring dataset.

[0057] The screening unit assesses the reliability of monitoring data using the ρ-T correlation coefficient algorithm, which calculates the correlation coefficient based on the covariance relationship between density and temperature values. Data normalization maps density and temperature values ​​to standard intervals, calculating density and temperature deviations separately. Density deviation equals each density value minus the mean density, and temperature deviation equals each temperature value minus the mean temperature. Covariance calculation involves summing the products of density and temperature deviations to obtain the numerator, and taking the square root of the product of the sum of squared density and temperature deviations to obtain the denominator. Dividing these two values ​​yields the ρ-T correlation coefficient. This correlation coefficient is compared to a preset threshold. Data with correlation coefficients greater than the threshold are marked as valid, forming the sensor reliability evaluation result. Finally, the monitoring data marked as valid are selected, while outlier data with correlation coefficients below the threshold are removed to obtain reliable monitoring data.

[0058] The prediction unit employs an SF6-NN neural network to predict the state of reliable monitoring data. The neural network structure consists of three parts: an input layer, hidden layers, and an output layer. Data normalization maps density, temperature, pressure, and humidity values ​​to a range of 0 to 1. The input layer contains four neurons corresponding to these four parameters. The hidden layer contains twelve neurons, using a forward propagation computation method. Each hidden layer neuron receives a weighted signal from the input layer and processes it through an activation function to generate an output value. The output layer contains two neurons, outputting the predicted density value and the temperature compensation value, respectively. During forward propagation, the input signal undergoes weight matrix calculation and activation function transformation, passing layer by layer to the output layer. Inverse normalization restores the predicted density and temperature compensation values ​​output by the network to the actual physical quantity values, forming the SF6 state prediction data.

[0059] The partitioning unit performs hierarchical management based on the density values ​​in the SF6 status prediction data. Density range determination compares each density value with preset density thresholds for normal, watchful, and abnormal levels. The normal level corresponds to the density range under normal equipment operation; the watchful level corresponds to the density range requiring monitoring but not reaching alarm standards; and the abnormal level corresponds to density values ​​exceeding safe operating limits. Data grouping categorizes data with the same level identifier into corresponding data sets, forming SF6 data groups divided by level. The hierarchical data table architecture creates independent data storage tables for normal, watchful, and abnormal levels, each with different storage strategies and index structures. Data import operations write the level-divided SF6 data groups into the corresponding hierarchical data tables, establishing a hierarchically stored SF6 database.

[0060] The analysis unit performs in-depth analysis of time-series data in the SF6 database. Time-series extraction and processing arranges density values ​​in timestamp order to form a continuous time series. Density fluctuation analysis quantifies the stability of density changes using statistical methods. The arithmetic mean is calculated by summing all density values ​​within the time series and dividing by the total number of data points. Deviation calculation calculates the difference between each density value and the density mean, squares it, and obtains a sequence of density deviation squares. Variance calculation averages all deviation squares and takes the square root to obtain the density standard deviation. Dividing the standard deviation by the density mean yields the coefficient of variation, which serves as a density stability indicator. Equipment health scoring uses a weighted calculation process, combining density stability indicators with density change trends and anomaly frequencies, setting different weight coefficients to comprehensively score various indicators. Data integration processing associates and combines the density stability indicators and equipment health scores with equipment identification codes and timestamps to generate SF6 management data.

[0061] In one specific embodiment, the acquisition unit is used for:

[0062] The density value of SF6 gas is collected by each sensor node in the distributed intelligent density sensor network. Each sensor node reads data according to the set acquisition frequency to obtain the raw sensing data containing the density value.

[0063] Based on the temperature sensor synchronously acquiring the temperature value at the corresponding moment from the original sensing data, the density value and the temperature value are paired to obtain density-temperature data pairs.

[0064] Each sensor node is assigned a unique device identifier code according to the sensor network topology. The device identifier code is then associated with the corresponding density-temperature data pair to obtain monitoring data with device identification information.

[0065] Based on the system clock, a precise timestamp is added to each set of monitoring data containing device identification information to complete the data encapsulation process and obtain the original dataset of SF6 monitoring.

[0066] Specifically, the data acquisition process of the acquisition unit is based on a distributed intelligent density sensor network. This network consists of multiple sensor nodes deployed at different gas chamber locations within the SF6 gas-insulated switchgear, each with independent data acquisition and preliminary processing capabilities. Each sensor node integrates a piezoresistive density sensor and a thermistor temperature sensor. The piezoresistive density sensor detects the strain on a sensitive thin film caused by changes in SF6 gas density and outputs a corresponding voltage signal. This voltage signal is then converted into a digital density value by an analog-to-digital converter. The set acquisition frequency refers to the data reading operation performed by the sensor nodes at pre-configured time intervals. The acquisition frequency directly affects the temporal resolution of the monitoring data and the system response speed. During data reading, the microcontroller of the sensor node periodically triggers measurement commands from the density and temperature sensors. The sensor chip outputs an analog voltage signal, which is converted into a digital value by the built-in 12-bit analog-to-digital converter. The digital value is then processed by a calibration algorithm to obtain the actual density value. These density values, along with the sensor status information, constitute the raw sensing data.

[0067] Synchronous acquisition by the temperature sensor refers to the simultaneous measurement and output of the temperature value by the temperature sensor at the same time as acquiring the density value. The synchronization mechanism ensures the temporal consistency of density and temperature measurements through a shared clock signal. Data pairing processing establishes a correspondence between density and temperature values ​​acquired from the same sensor node at the same time. The pairing process ensures the correctness of the data association through timestamp matching and node identifier verification. The density-temperature data pair is a data structure containing density value, temperature value, and measurement time information. The establishment of data pairs eliminates the time deviation between density and temperature values, ensuring the accuracy of subsequent correlation analysis. The data pairing algorithm first checks the timestamp difference between the density and temperature values. When the timestamp difference is less than the preset synchronization tolerance, the two values ​​are combined into a data pair; otherwise, they are marked as asynchronous data and a re-acquisition process is triggered.

[0068] The sensor network topology describes the physical location and communication connections of various sensor nodes. The topology includes node coordinates, communication paths, and hierarchical relationships. Device identification code allocation is based on the location and functional characteristics of nodes in the network topology. The allocation algorithm uses a 16-bit binary encoding method, with the first eight bits representing the device type and functional area, and the last eight bits representing the specific node number. The association and binding process establishes a mapping relationship between the device identification code as the primary key and the corresponding density-temperature data pair. The binding operation is implemented through database table association, ensuring that each data pair can be accurately traced back to the specific sensor node that generated the data. Monitoring data with device identification information is an extended data structure that adds a device identification code field to the density-temperature data pair. This data structure supports unified management and tracking analysis of multi-node data.

[0069] The system clock provides a unified time base, and the timestamp uses a 64-bit integer format to record the precise moment the data was generated. The timestamp includes year, month, day, hour, minute, second, and millisecond-level precision information. Data encapsulation organizes the monitoring data, which contains device identification information, according to a predefined data frame format. The data frame includes fields such as a frame header, device identification code, timestamp, density value, temperature value, checksum, and frame trailer. The encapsulation algorithm first calculates the cyclic redundancy check (CRC) code to ensure data integrity, then combines the fields into a complete data frame according to a fixed byte order, and finally writes the data frame to a buffer for transmission or storage. The SF6 monitoring raw dataset is a collection of encapsulated data frames generated by all sensor nodes within a certain time period. The dataset is stored in a time-series database format, supporting fast querying and retrieval based on time range and device identification.

[0070] In one specific embodiment, the filtering unit is used for:

[0071] Data normalization was performed on the density and temperature values ​​in the original SF6 monitoring dataset. The density and temperature deviations were calculated separately to obtain standardized density-temperature deviation data.

[0072] Covariance calculation is performed on the density-temperature deviation data. The covariance numerator is calculated by summing the products of density deviation and temperature deviation. The covariance denominator is obtained by taking the square root of the product of the sum of squares of density deviation and the sum of squares of temperature deviation. The ρ-T correlation values ​​of each sensor node are then obtained.

[0073] The correlation coefficient threshold is used to compare and judge the ρ-T correlation values. Data with correlation values ​​greater than the threshold are marked as valid, and the sensor reliability evaluation results are obtained.

[0074] Based on the sensor reliability evaluation results, monitoring data marked as valid are selected, and abnormal data with correlation coefficients below the threshold are removed to obtain reliable monitoring data.

[0075] Specifically, the data normalization process in the screening unit performs standardization operations on the density and temperature values ​​in the original SF6 monitoring dataset. Data normalization is a preprocessing method that transforms data with different dimensions and numerical ranges to a unified standard interval, eliminating the impact of dimensional differences on subsequent correlation analysis. The density value normalization process first calculates the maximum and minimum density values ​​for all sensor nodes, then subtracts the minimum value from each density value and divides by the difference between the maximum and minimum values ​​to obtain a standardized density value within the zero-to-one interval. Temperature value normalization uses the same processing logic, mapping temperature values ​​to the zero-to-one standard interval through a maximum-to-minimum scaling method. Density value deviation calculation involves subtracting the arithmetic mean of the standardized density values; the deviation reflects the degree of deviation of a single density value from the overall average level. Temperature value deviation calculation uses the same method, subtracting the arithmetic mean of the standardized temperature values ​​to obtain the deviation amount for each temperature value. The standardized density-temperature deviation data is a set of data pairs containing density deviation values ​​and temperature deviation values; these deviation data provide a standardized input data source for subsequent covariance calculation.

[0076] Covariance calculation, based on standardized density-temperature deviation data, calculates the degree of linear correlation between density and temperature values. Covariance is a statistic that measures the degree of coordinated change between two random variables. The core step in covariance calculation is summing the products of density and temperature deviations. The density deviation value of each sensor node is multiplied by its corresponding temperature deviation value, and then the products of all sensor nodes are summed to obtain the numerator of the covariance. The numerator of the covariance reflects the coordinated trend of density and temperature changes; positive values ​​indicate changes in the same direction, negative values ​​indicate changes in opposite directions, and the absolute value indicates the strength of the coordination. The density deviation sum of squares calculation involves squaring each density deviation value and summing the results. The temperature deviation sum of squares is calculated using the same method. The square root of the product of the two sums of squares yields the denominator of the covariance. The denominator of the covariance represents the overall fluctuation of density and temperature changes; a larger denominator indicates a higher degree of data dispersion. The ρ-T correlation coefficient is calculated by dividing the covariance numerator by the covariance denominator. The correlation coefficient ranges from -1 to +1. The closer the absolute value is to one, the stronger the linear correlation between density and temperature.

[0077] The correlation coefficient threshold is a pre-set critical value for judging the reliability of sensor data. The threshold is set based on the physical properties of SF6 gas and statistical analysis of historical monitoring data. The numerical comparison judgment compares the ρ-T correlation value calculated for each sensor node with the preset threshold. A floating-point comparison algorithm is used to ensure accuracy. The valid identifier is a Boolean flag used to mark data that has passed the reliability test. When the correlation value is greater than the preset threshold, the corresponding monitoring data is marked as a valid true value; otherwise, it is marked as a false value. The sensor reliability evaluation result is a set of evaluation data containing the valid identifier status of all sensor nodes. The evaluation result reflects the data quality status of each sensor node within the current monitoring period. The reliability evaluation algorithm identifies abnormal conditions such as sensor failure, drift, or environmental interference through the statistical characteristics of the correlation coefficient. Sensor nodes with correlation coefficients significantly deviating from the normal range are judged to be unreliable.

[0078] The data screening process performs quality control on the monitoring data based on the sensor reliability evaluation results. The screening algorithm traverses all monitoring data and checks the corresponding valid identification status. Monitoring data marked as valid are retained and copied to the reliable monitoring data set. These data have passed the density-temperature correlation test and possess the data quality required for subsequent analysis and prediction. The outlier removal operation removes monitoring data with correlation coefficients below a threshold from the dataset. The removal process includes two steps: data deletion and index update. The reliable monitoring data is a high-quality monitoring data set retained after ρ-T correlation coefficient screening. Each record in the dataset has a reasonable correlation between density and temperature values, meeting the theoretical expectations of the physical properties of SF6 gas. The screened dataset eliminates the impact of sensor failure, measurement errors, and environmental noise on monitoring accuracy.

[0079] Figure 2 This diagram illustrates the trend of the ρ-T correlation coefficient and the reliability evaluation process of an SF6 circuit breaker in a substation during a 24-hour monitoring period, calculated based on the covariance of density-temperature deviation data. The solid curves in the diagram represent the calculated correlation values ​​at each time point, while the horizontal dashed line indicates the preset reliability threshold of 0.95. When the correlation value is above 0.95, the corresponding monitoring data is marked as valid and retained as reliable monitoring data. When the correlation value is between 0.85 and 0.95, indicating a need for calibration, it suggests sensor drift and requires calibration.

[0080] In one specific embodiment, the prediction unit is used for:

[0081] The density, temperature, pressure, and humidity values ​​in the reliable monitoring data are normalized and mapped to a numerical range of 0 to 1 to obtain standardized network input data.

[0082] The input layer of the SF6-NN neural network is constructed based on standardized network input data. The input layer contains four neurons corresponding to the density, temperature, pressure and humidity parameters, respectively, to obtain the network input layer data.

[0083] The input layer data of the network is passed to the hidden layer containing 12 neurons for forward propagation calculation. The output value of the hidden layer is generated by weighted summation and activation function, and the calculation result of the intermediate layer is obtained.

[0084] The calculation results from the intermediate layer are passed to the output layer, which contains two neurons, for final calculation. The predicted density value and temperature compensation value are output respectively, thus obtaining the network output data.

[0085] The network output data is denormalized to restore the predicted density and temperature compensation values ​​to the range of actual physical quantities, thus obtaining SF6 state prediction data.

[0086] Specifically, the data normalization process of the prediction unit performs standardization operations on four key parameters in the reliable monitoring data. Density normalization maps SF6 gas density values ​​to a zero-to-one range using a maximum-to-minimum scaling method. The maximum and minimum density values ​​are determined based on the theoretical density boundaries of SF6 gas within its normal operating pressure range. Temperature normalization uses the same scaling algorithm to convert ambient temperature values ​​to a standard numerical range covering the normal operating temperature window of the SF6 equipment. Pressure normalization processes the absolute pressure data of SF6 gas, measured by a pressure sensor within the gas chamber. This normalization process eliminates the influence of pressure units and numerical ranges on neural network training. Humidity normalization processes ambient humidity data, reflecting the moisture content in SF6 gas. Excessive humidity can affect the insulation performance of SF6 and equipment safety. The standardized network input data is a normalized four-dimensional data vector, with each dimension's value restricted to a zero-to-one range. Data normalization eliminates dimensional and numerical scale differences between different physical quantities, solving the technical problem of difficult multi-source data fusion in traditional monitoring methods.

[0087] The input layer of the SF6-NN neural network is constructed based on standardized network input data. The input layer is the first layer of the neural network, responsible for receiving external input signals and passing them to subsequent layers. The input layer contains four neurons, each corresponding to an input parameter: the first neuron receives the density value, the second the temperature value, the third the pressure value, and the fourth the humidity value. A neuron is the basic computational unit of the neural network; each neuron stores a value and passes that value to all neurons in the next layer. The network input layer data is a vector composed of the values ​​stored by the four neurons in the input layer. The vector has four dimensions, corresponding to the standardized values ​​of density, temperature, pressure, and humidity, respectively. The order and precision of the input layer data directly affect the computational results of the neural network; floating-point format is used during data transmission to ensure computational accuracy.

[0088] Forward propagation is the core algorithm of neural network information processing, representing the computational process of passing input layer data from the input layer to the output layer. The hidden layer contains twelve neurons, each receiving weighted signals from four neurons in the input layer. The weighted sum operation multiplies the four input values ​​by their corresponding weight coefficients and then sums them to obtain the weighted sum. The weight coefficients are parameters learned during neural network training, reflecting the degree of influence of the input parameters on the output result. The magnitude and sign of the weight values ​​determine the amplification or suppression effect of the signal. The activation function performs a non-linear transformation on the weighted sum. The sigmoid function is used, mapping any real value to the zero-to-one interval. The introduction of the activation function enables the neural network to handle non-linear relationships. The hidden layer output value is the value after the activation function transformation; the twelve hidden layer neurons produce twelve output values, forming the hidden layer output vector. The intermediate layer computation results include the hidden layer output vector and intermediate computation states, providing input data for the output layer computation.

[0089] The output layer performs final signal processing based on the intermediate layer's calculation results. The output layer contains two neurons, responsible for calculating the predicted density value and the temperature compensation value, respectively. The first output neuron receives a weighted signal of the twelve output values ​​from the hidden layer, processes it through weighted summation and an activation function to generate the predicted density value, which reflects the density state of SF6 gas at the next time step. The second output neuron uses the same calculation process to generate the temperature compensation value, which corrects for the impact of temperature changes on density measurement. The network output data is a two-dimensional vector composed of the values ​​generated by the two neurons in the output layer. The first element of the vector is the predicted density value, and the second element is the temperature compensation value; both values ​​are within the normalized range of zero to one.

[0090] Inverse normalization restores the network output data from the standardized range to the numerical range of the actual physical quantities. The inverse normalization algorithm employs a mathematical transformation that is the inverse of normalization. Inverse normalization of the predicted density value involves multiplying the standardized prediction result by the difference between the maximum and minimum density values, then adding the minimum value to obtain the actual density prediction value in kilograms per cubic meter. Inverse normalization of the temperature compensation value uses the same transformation method, converting the standardized compensation value into a temperature compensation amount in degrees Celsius. The SF6 state prediction data includes the inverse normalized predicted density value and the temperature compensation value, reflecting the future state change trend of SF6 gas.

[0091] In one specific embodiment, the partitioning unit is used for:

[0092] Based on the density values ​​in the SF6 state prediction data, density range judgment is performed. The density values ​​are compared with preset density thresholds for normal, attention, and abnormal levels to obtain density level classification labels.

[0093] The SF6 state prediction data is grouped according to the density level classification label. Data with the same level label are classified into the corresponding data set to obtain SF6 data groups divided by level.

[0094] Based on the SF6 data groups divided by level, data table structures with different storage strategies are established. Independent data storage tables are created for normal level, attention level and abnormal level respectively, resulting in a hierarchical data table architecture.

[0095] The SF6 data groups, divided by level, are written into the corresponding hierarchical data tables according to the corresponding storage strategies to complete the data entry operation and obtain the hierarchical SF6 database.

[0096] Specifically, the density range judgment process for unit division performs hierarchical decision-making based on the density values ​​in the SF6 state prediction data. The density range judgment is a classification standard established based on the density characteristics of SF6 gas under different operating conditions. The normal level density threshold corresponds to the density range of SF6 equipment under standard operating conditions, reflecting the normal operating status of the equipment. The upper and lower limits of the normal level threshold are determined based on the rated density and allowable fluctuation range of SF6 gas. The attention level density threshold is set near the boundary of the normal range, indicating that the equipment status needs monitoring but has not yet reached the alarm standard. The setting of the attention level threshold takes into account the natural fluctuations and measurement errors of SF6 gas density. The abnormal level density threshold covers density values ​​that exceed the safe operating range. Abnormal level data indicates that the equipment has a fault or leakage risk, requiring immediate maintenance measures. The comparison operation uses a numerical comparison algorithm to compare the predicted density value with the threshold ranges of the three levels one by one. The comparison process uses greater than or equal to and less than or equal to logical judgments to determine the level range to which the density value belongs. The density level classification identifier is a numerical code representing the level to which the density value belongs: 1 for normal level, 2 for attention level, and 3 for abnormal level. A one-to-one mapping relationship is established between the classification identifier and the density value.

[0097] Data grouping processing performs set partitioning operations on SF6 state prediction data based on density level classification identifiers. The data grouping algorithm traverses all prediction data records and checks the level identifier of each record. Data with the same level identifier are grouped into the same data set through equality matching. The grouping process creates three independent data sets corresponding to the normal, concern, and abnormal levels, respectively. The data classification operation copies the complete prediction data record, including density value, temperature compensation value, timestamp, and device identifier code, to the corresponding level data set, ensuring data integrity after grouping. The SF6 data group divided by level is the collective name for the three data sets formed after grouping processing. Each data set contains all prediction data records of the same level. The data grouping eliminates the problem of mixed storage of data of different levels and solves the technical problem of low data management efficiency in traditional monitoring systems. The grouping algorithm also includes a data counting function to count the number of records in each level data set. The counting results are used for storage capacity planning and query optimization.

[0098] The hierarchical data table architecture is based on the characteristics and access patterns of SF6 data groups divided into levels, with different storage strategies designed for each level. These strategies include table structure design, index configuration, and data compression schemes. Normal-level data storage tables employ a compressed storage strategy. Due to the large volume and relatively low access frequency of normal-level data, a columnar storage format is used to reduce storage space usage, and the index design uses timestamps as the primary key to support time range queries. Attention-level data storage tables employ a fast access strategy, using a row-based storage format to support fast random access, and a composite index of device identifiers and timestamps is created to accelerate query operations. Abnormal-level data storage tables employ a real-time access strategy, with the table structure optimized for in-memory storage to ensure the fastest access speed, and multiple indexes created to support complex query conditions. The hierarchical data table architecture also includes the definition of inter-table relationships, ensuring data consistency between tables of different levels through foreign key constraints. The architecture design considers data lifecycle management and automatic archiving requirements.

[0099] The data ingestion operation writes SF6 data groups, categorized by tier, into corresponding hierarchical data tables according to a predefined storage strategy. The ingestion process includes three stages: data verification, format conversion, and transaction management. The data verification stage checks the integrity and validity of each record, verifying the numerical range of density values, the correctness of timestamp formats, and the uniqueness of device identifiers. The format conversion stage converts the in-memory data structure into the storage format of the database tables. This conversion process includes data type mapping, character encoding processing, and precision adjustment. Transaction management ensures the atomicity and consistency of the data ingestion operation. If data ingestion fails at a certain tier, the entire batch of ingestion operations is rolled back to its initial state. The hierarchical SF6 database is a multi-table database system formed after the data ingestion operation is completed. The database contains three data tables: normal level, attention level, and exception level, along with related indexes and constraint structures.

[0100] In one specific embodiment, the analysis unit is used for:

[0101] Extraction sub-units are used to perform time series extraction processing on historical density data in the hierarchically stored SF6 database. The density values ​​are arranged in the order of timestamps to obtain continuous time series, thus obtaining SF6 density time series data.

[0102] The analysis subunit is used to perform density fluctuation analysis on SF6 density time series data. The density stability coefficient is obtained by calculating the ratio of the density standard deviation to the density mean, thus obtaining the density stability index.

[0103] The weighted sub-unit is used to perform weighted calculations based on density stability indicators, density change trends, and abnormal frequency. By setting different weight coefficients, it performs comprehensive scoring calculations on various indicators to obtain the equipment health score.

[0104] The integration subunit is used to integrate and process data based on density stability indicators and equipment health scores. It associates and combines the score results with equipment identification codes and timestamps to obtain SF6 management data.

[0105] Specifically, the time series extraction processing of the sub-units performs data retrieval and sorting operations based on historical density data in the hierarchically stored SF6 database. Time series extraction is the process of reorganizing density data stored in different hierarchical tables according to the time dimension. Historical density data retrieval uses SQL queries to extract density values ​​and corresponding timestamps from three data tables: normal, attention, and abnormal. Query conditions include time range filtering and device identification code matching to ensure that the extracted data belongs to the same monitoring device and is within a specified time window. Timestamp order sorting uses a sorting algorithm to sort the extracted density data according to the order of the timestamps. The sorting process uses a quicksort algorithm to ensure sorting efficiency, and the sorted result forms a density value sequence arranged in ascending order of time. Continuous time series refers to a data sequence in which density values ​​are arranged in chronological order and the time intervals are relatively uniform. A continuity check algorithm identifies missing values ​​and abnormal intervals in the time series. When a time interval exceeds a preset threshold, data interpolation is triggered. SF6 density time series data is a standard time series data structure formed after extraction, sorting, and continuity processing. The data structure contains a timestamp array and a corresponding density value array. Time series data solves the technical problem of difficult historical data analysis in traditional monitoring systems.

[0106] The density fluctuation analysis of the sub-unit is based on statistical analysis of SF6 density time series data. Density fluctuation analysis is a statistical method to quantify the stability of SF6 gas density changes. The arithmetic mean is calculated by summing all density values ​​in the time series and dividing by the total number of data points; the average reflects the central trend and overall level of the density data. Variance is calculated by squaring the difference between each density value and the average and then averaging the result; variance quantifies the dispersion of density values ​​relative to the average. The density standard deviation is obtained by taking the square root of the variance; the standard deviation has the same dimensions as the original data, directly reflecting the absolute amplitude of density fluctuations. The ratio operation divides the density standard deviation by the density mean to obtain the dimensionless coefficient of variation, which eliminates the comparability problem of standard deviations at different density levels. The density stability coefficient is calculated by subtracting the coefficient of variation from one; the stability coefficient ranges from zero to one, with values ​​closer to one indicating greater density stability. The density stability index is a quantitative representation of the density stability coefficient; its value directly reflects the operational stability of SF6 equipment, solving the technical problem of strong subjectivity in equipment condition evaluation in traditional monitoring methods.

[0107] The weighted calculation of the weighted sub-unit is based on a comprehensive evaluation performed by combining the density stability index with other operating parameters. Weighted calculation is a mathematical method that combines multiple evaluation indicators according to their importance. Density change trend analysis uses a linear regression algorithm to calculate the slope of the density time series; a positive slope indicates an upward trend, and a negative slope indicates a downward trend. The absolute value of the slope reflects the rate of change. Anomaly frequency calculation is the ratio of the number of data points exceeding the normal density range in the statistical time series to the total number of data points. Anomaly frequency reflects the probability of abnormal conditions occurring during equipment operation. Weight coefficients are set based on the operating characteristics and failure mode analysis of SF6 equipment. The density stability index is weighted at 0.4, the density change trend at 0.3, and the anomaly frequency at 0.3. The sum of the weight coefficients equals 1.0 to ensure the normalization of the score. The comprehensive score calculation multiplies each of the three indicators by its corresponding weight coefficient and then sums them. The calculation process uses a weighted average algorithm to ensure that the contribution of each indicator is proportional to its importance. The equipment health score is the result of a comprehensive scoring calculation. The score ranges from zero to one. The higher the score, the better the equipment health status. The scoring system solves the technical problem of the difficulty of comprehensive evaluation of multiple indicators in traditional monitoring systems.

[0108] The data integration processing of the integration subunit performs the final data organization operation based on the density stability index and equipment health score. Data integration is the process of associating analysis results with equipment identity information and time information. The scoring results include the numerical values ​​of the density stability index and the equipment health score, which constitute the core data for equipment status evaluation. Equipment identification code association establishes a mapping relationship between the scoring results and the specific sensor equipment that generated the data. The association operation ensures the accurate attribution of the scoring results through matching the equipment identification codes. Timestamp association establishes a correspondence between the scoring results and the end time of the analysis time window. The timestamp marks the generation time and validity period of the scoring results. The association combination operation organizes the density stability index, equipment health score, equipment identification code, and timestamp into a unified data record. The data record adopts a structured format to ensure the efficiency of subsequent query and processing. SF6 management data is the final output of the data integration processing. Management data includes equipment status evaluation information and equipment identity information. The data format supports direct access and display by the equipment operation and maintenance management system.

[0109] In one specific embodiment, the analysis subunit is used for:

[0110] The arithmetic mean of the density values ​​in the SF6 density time series data is calculated by adding all the density values ​​in the time series and dividing by the total number of data points to obtain the density mean.

[0111] Based on the density mean, the deviation of each density value in the SF6 density time series data is calculated. The difference between each density value and the density mean is calculated and squared to obtain the density deviation square value sequence.

[0112] The density deviation square value sequence is processed by variance calculation, and the square root is taken after averaging all deviation square values ​​to obtain the density standard deviation.

[0113] The density stability index is obtained by calculating the ratio between the density standard deviation and the density mean, dividing the density standard deviation by the density mean, and obtaining the coefficient of variation.

[0114] Specifically, the arithmetic mean calculation process involves summing the density values ​​in the SF6 density time series data. The arithmetic mean is a fundamental measure of central tendency in statistics, used to calculate the center of a dataset. The density value accumulation process uses the first density value in the time series as the initial accumulation value, then sequentially adds the second, third, and so on, up to the last density value. Floating-point addition is used in the accumulation operation to ensure calculation accuracy. The total number of data points is counted by traversing the density time series, simultaneously performing data validity checks to remove null and outlier values ​​to ensure accuracy. The division operation divides the accumulated density values ​​by the total number of data points to obtain the arithmetic mean. High-precision floating-point arithmetic is used in the division operation to avoid precision loss. The density mean is the result of the arithmetic mean calculation. The mean reflects the average level of SF6 gas density during the monitoring period, solving the technical problem of the lack of density trend analysis in traditional monitoring systems. The physical meaning of the density mean represents the density baseline value of the SF6 equipment under normal operating conditions. This baseline value is used for subsequent deviation analysis and anomaly detection.

[0115] The deviation calculation process performs a difference operation on each density value in the time series based on the density mean. Deviation calculation is a statistical method to measure the degree of deviation of an individual data point from the average level of the dataset. The difference calculation subtracts the density mean from each density value to obtain the deviation value. The sign of the deviation value indicates the direction of deviation from the mean; a positive value indicates above the mean, and a negative value indicates below the mean. The squaring operation calculates each deviation value by raising it to a second power. The squaring operation eliminates the influence of the sign of the deviation value, ensuring that all deviation contributions are positive, while amplifying the influence weight of larger deviations. The density deviation squared value sequence is a new data sequence formed after squaring all deviation values. Each element in the sequence corresponds to the squared deviation of the corresponding density value in the original density time series. The mathematical properties of the deviation squared value ensure the statistical significance of the variance calculation, and the squaring operation makes the deviation values ​​cumulative and direction-independent. The deviation squared value sequence maintains the same data order and time correspondence as the original density time series, and the sequence structure supports subsequent statistical analysis operations.

[0116] The variance calculation process involves averaging and square rooting the density deviation squared value sequence. Variance is an important statistic for measuring the dispersion of data, reflecting the degree of dispersion of data values ​​relative to the mean. The average calculation involves summing all elements in the deviation squared value sequence and dividing by the sequence length. High-precision floating-point arithmetic is used in the summation process to ensure accuracy. The variance value is obtained through the average calculation; its magnitude directly reflects the fluctuation range of the density data. A larger variance indicates more drastic density fluctuations, while a smaller variance indicates more stable density changes. The square root operation calculates the standard deviation by taking the square root of the variance value. This operation restores the dimension of the variance from the square of density to the original dimension of density, making the standard deviation have the same physical meaning and numerical comparability with the original density data. The density standard deviation is a direct indicator of the fluctuation range of SF6 gas density. A smaller standard deviation indicates more stable equipment operation, while a larger standard deviation indicates fluctuations or anomalies. Standard deviation calculation solves the technical problem of quantifying density stability in traditional monitoring methods.

[0117] Ratio calculation is based on a division operation between the density standard deviation and the density mean. Ratio calculation is a mathematical method that converts absolute values ​​into relative values, eliminating the comparability issues of standard deviations at different density levels. The division operation uses the density standard deviation as the dividend and the density mean as the divisor, performing floating-point division to obtain a dimensionless ratio. The coefficient of variation (CV) is the statistical term for the ratio of the standard deviation to the mean. The CV eliminates the influence of data dimensions, making density fluctuations across different devices and time periods comparable. The CV typically ranges from zero to one; the closer the value is to zero, the more stable the data; the closer the value is to one, the greater the data fluctuation. The density stability index is calculated by subtracting the CV from one. This subtraction operation transforms the meaning of the CV from the degree of fluctuation to the degree of stability. The density stability index also ranges from zero to one; the closer the index value is to one, the more stable the SF6 gas density and the better the equipment's operating condition; the closer the index value is to zero, the greater the density fluctuation and the greater the risk of equipment malfunction.

[0118] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A sulfur hexafluoride gas data monitoring management system, characterized by, The system comprises: The acquisition unit is configured to acquire a density value and a temperature value of SF6 gas, combine a device identification code and a time stamp, and obtain an SF6 monitoring original data set; The screening unit is configured to calculate a density-temperature correlation coefficient of the density value and the temperature value in the SF6 monitoring original data set, screen data with a correlation coefficient meeting a preset threshold as reliable monitoring data, and include: performing data normalization processing based on the density value and the temperature value in the SF6 monitoring original data set, calculating a density value deviation and a temperature value deviation respectively, and obtaining standardized density-temperature deviation data; performing covariance operation on the density-temperature deviation data, calculating a covariance numerator by summing the product of the density deviation and the temperature deviation, calculating a covariance denominator by taking the square root of the product of the density deviation square sum and the temperature deviation square sum, and obtaining a ρ-T correlation coefficient value of each sensor node; performing numerical comparison on the ρ-T correlation coefficient value based on a preset correlation coefficient threshold, marking data with a correlation coefficient value greater than the threshold as valid identification, and obtaining a sensor reliability evaluation result; screening monitoring data marked as valid identification according to the sensor reliability evaluation result, eliminating abnormal data with a correlation coefficient lower than the threshold, and obtaining reliable monitoring data; The prediction unit is configured to input the reliable monitoring data into an SF6-NN neural network, predict a density value at a next time, calculate a temperature compensation amount, and obtain SF6 state prediction data; The division unit is configured to divide the density value in the SF6 state prediction data into a normal level, an attention level and an abnormal level according to a density range, and establish a hierarchical storage SF6 database; The analysis unit is configured to analyze time series data in the SF6 database, calculate a density stability index and a device health score, and generate SF6 management data.

2. The sulfur hexafluoride gas data monitoring management system of claim 1, wherein, The acquisition unit is configured to: acquire SF6 gas density values through each sensor node in a distributed intelligent density sensor network, each sensor node reads data at a set acquisition frequency to obtain original sensor data containing density values; synchronously acquire temperature values at corresponding times based on temperature sensors in the original sensor data, pair the density values and the temperature values for data processing to obtain density-temperature data pairs; assign a unique device identification code to each sensor node according to a sensor network topology, associate and bind the device identification code with the corresponding density-temperature data pairs, and obtain monitoring data with device identity information; add accurate time stamps to each group of monitoring data with device identity information based on a system clock, complete data encapsulation processing, and obtain an SF6 monitoring original data set.

3. The SF6 gas data monitoring management system according to claim 1, wherein, The prediction unit is configured to: perform data normalization processing on the density value, the temperature value, the pressure value and the humidity value in the reliable monitoring data, map them to a numerical interval of 0 to 1, and obtain standardized network input data; construct an input layer of an SF6-NN neural network based on the standardized network input data, the input layer contains four neurons corresponding to density, temperature, pressure and humidity parameters, and obtain network input layer data; The network input layer data is transmitted to a hidden layer containing 12 neurons for forward propagation calculation, and a hidden layer output value is generated by weighted summation and an activation function to obtain an intermediate layer calculation result; Based on the intermediate layer calculation result, the final calculation is performed on an output layer containing 2 neurons to respectively output a predicted density value and a temperature compensation value, and network output data is obtained; The network output data is subjected to inverse normalization processing to restore the predicted density value and the temperature compensation value to the actual physical quantity value range, and SF6 state prediction data is obtained.

4. The sulfur hexafluoride gas data monitoring management system of claim 1, wherein, The division unit is configured to: Based on the density value in the SF6 state prediction data, density range judgment processing is performed, and the density value is compared with preset normal level, attention level and abnormal level density thresholds to obtain a density level classification identifier; According to the density level classification identifier, data grouping processing is performed on the SF6 state prediction data, and data of the same level identifier is classified into a corresponding data set to obtain SF6 data groups divided by levels; Based on the SF6 data groups divided by levels, a data table structure of different storage strategies is established, and independent data storage tables are respectively created for the normal level, the attention level and the abnormal level to obtain a hierarchical data table architecture; The SF6 data groups divided by levels are written into the corresponding hierarchical data tables according to the corresponding storage strategies to complete data storage operation, and a hierarchical stored SF6 database is obtained.

5. The sulfur hexafluoride gas data monitoring management system of claim 1, wherein, The analysis unit is configured to: The extraction subunit is configured to perform time series extraction processing on historical density data in the hierarchical stored SF6 database, arrange the density values in chronological order to obtain a continuous time series, and obtain SF6 density time series data; The analysis subunit is configured to perform density fluctuation analysis calculation on the SF6 density time series data, obtain a density stability coefficient by calculating the ratio of the density standard deviation to the density mean, and obtain a density stability index; The weighting subunit is configured to perform weighting calculation processing on the density stability index in combination with a density change trend and an abnormal frequency, and obtain a device health score by setting different weight coefficients to perform comprehensive scoring operation on each index; The integration subunit is configured to perform data integration processing on the density stability index and the device health score, associate and combine the score results with a device identifier code and a timestamp, and obtain SF6 management data.

6. The sulfur hexafluoride gas data monitoring management system of claim 5, wherein, The analysis subunit is configured to: Based on each density value in the SF6 density time series data, arithmetic mean calculation processing is performed, and the sum of all density values in the time series is divided by the total number of data to obtain a density mean; According to the density mean, deviation calculation processing is performed on each density value in the SF6 density time series data, the difference between each density value and the density mean is calculated and squared to obtain a density deviation square value sequence; The density deviation square value sequence is subjected to variance calculation processing, and the average of all deviation square values is calculated to obtain a density standard deviation; Based on the density standard deviation and the density mean, ratio calculation processing is performed, and the density standard deviation is divided by the density mean to obtain a coefficient of variation, and a density stability index is obtained.

Images