A transformer intelligent online monitoring method and system

Abstract

A transformer intelligent online monitoring method and system, in the method, according to historical working condition data, the historical data is divided into multiple working condition intervals, each working condition interval is a kind of working condition parameter corresponding numerical range;Based on all historical running data located in working condition interval, the dynamic reference range corresponding to the working condition interval is determined, and the dynamic reference range is a numerical interval with upper limit value and lower limit value;Real-time acquisition target transformer current operating parameter value and corresponding current working condition parameter value;According to the current working condition parameter value, the target working condition interval is determined from the multiple working condition intervals;The target dynamic reference range corresponding to the target working condition interval is obtained;In the case where it is determined that the current operating parameter value is not located in the target dynamic reference range, the transformer is determined to be abnormal, and a warning signal is generated.The application is used to improve the warning accuracy and reliability of transformer monitoring.

Description

Technical Field

[0001] This application belongs to the field of power equipment condition monitoring technology, and in particular relates to a method and system for intelligent online monitoring of transformers. Background Technology

[0002] As a core piece of equipment in the power system, the stability and reliability of power transformers directly affect the safety of the entire power grid. To ensure the stable operation of transformers, it is necessary to monitor their various operating parameters in real time in order to promptly detect potential faults and provide early warnings.

[0003] A transformer online monitoring system has been proposed in related technologies. This system collects real-time operating data from the transformer by installing various sensors on it, including those for temperature, current, voltage, and noise. The collected data is then transmitted to a monitoring center via a communication network. The monitoring center processes the received data, and when a monitored parameter, such as the transformer's main body temperature, exceeds a pre-set alarm threshold, the system triggers an alarm to alert maintenance personnel.

[0004] However, the normal operating parameters of power transformers are not constant values, but rather change dynamically depending on factors such as the load they bear and the ambient temperature. The aforementioned technologies are ill-suited to adapting to the dynamically changing normal state range of transformers under different operating conditions. This makes it difficult for the system to accurately distinguish between a normal parameter value under heavy load or high temperature conditions and an abnormal parameter value caused by an early fault, resulting in a high false alarm and false negative rate in diagnostic results, and reducing the accuracy and reliability of the monitoring system's early warning capabilities. Summary of the Invention

[0005] This application provides a method and system for intelligent online monitoring of transformers, which can improve the accuracy and reliability of early warning for transformer monitoring.

[0006] Firstly, this application provides an intelligent online monitoring method for transformers, which involves acquiring historical data of a target transformer within a preset historical time period, including historical operating data and corresponding historical operating condition data; dividing the historical operating condition data into multiple operating condition intervals, with each interval containing a numerical range corresponding to a specific operating condition parameter; determining a dynamic reference range corresponding to the operating condition interval based on all historical operating data within the interval, wherein the dynamic reference range is a numerical range with upper and lower limits; acquiring the current operating parameter values ​​of the target transformer and the corresponding current operating condition parameter values ​​in real time; determining the target operating condition interval from the multiple intervals based on the current operating condition parameter values; acquiring the target dynamic reference range corresponding to the target operating condition interval; and determining a transformer anomaly and generating an early warning signal if the current operating parameter value is not within the target dynamic reference range.

[0007] By adopting the above technical solution, historical data is acquired and divided into intervals according to operating conditions. A dynamic benchmark range is established for each operating condition interval, enabling the monitoring benchmark to be dynamically adjusted as operating conditions change. After acquiring current operating parameters and operating condition parameters, the method automatically matches and compares the corresponding dynamic benchmark range for the operating condition interval, avoiding potential misjudgments caused by using fixed thresholds. Since the benchmark range is determined based on the operating performance of a large amount of historical data under the same operating conditions, it has strong statistical representativeness and correlation with actual operation, and can better reflect the normal operating state range of the transformer under that condition. This adaptive monitoring method improves the accuracy of abnormal state identification, reduces false alarms caused by changes in operating conditions, and ensures the timely detection of true abnormal states. The timely generation of early warning signals enables maintenance personnel to quickly learn about transformer anomalies and take corresponding measures, reducing the risk of equipment failure.

[0008] In conjunction with some implementations of the first aspect, in some implementations, after determining a transformer anomaly and generating an early warning signal, the method further includes: in response to the generation of the early warning signal, collecting operating data of at least two pre-selected operating parameters within a preset time window before the generation time of the early warning signal, forming corresponding time-series data respectively; calculating the similarity between each type of time-series data and the corresponding parameter template sequence in multiple preset fault feature templates to obtain a parameter similarity score corresponding to each fault feature template; applying one or more preset association rules in each fault feature template to comprehensively calculate the corresponding parameter similarity score to obtain a comprehensive similarity score corresponding to each fault feature template; determining the target comprehensive similarity score with the largest value among the comprehensive similarity scores, and determining the fault feature template corresponding to the target comprehensive similarity score as the best matching template; determining the specific anomaly type identified by the best matching template, and appending the specific anomaly type to the early warning signal for output.

[0009] By adopting the above technical solution, after an anomaly is detected, time-series data of multiple pre-selected operating parameters are collected and compared with preset fault feature templates for similarity calculation and comprehensive scoring, thus achieving intelligent identification of anomaly types. Joint analysis of multiple parameters improves the comprehensiveness and reliability of anomaly type judgment. The method uses similarity calculation to capture the characteristic manifestations of different types of faults, achieving accurate matching even when fault characteristics differ to some extent. By setting multiple association rules for comprehensive calculation, the accuracy of anomaly type identification is improved. The final output warning signal contains specific anomaly type information, providing maintenance personnel with more targeted fault information and facilitating rapid fault handling.

[0010] In conjunction with some implementation methods of the first aspect, in some implementation methods, one or more preset association rules in each fault feature template are applied to comprehensively calculate the corresponding parameter similarity scores to obtain the comprehensive similarity score corresponding to each fault feature template. Specifically, this includes: for each fault feature template, determining the corresponding preset association rule as the parameter weight, with each parameter weight corresponding to a pre-selected operating parameter; multiplying the parameter similarity score corresponding to the pre-selected operating parameter by the parameter weight corresponding to the pre-selected operating parameter to obtain a weighted score; and summing all the values ​​in the weighted score to obtain the comprehensive similarity score corresponding to the fault feature template.

[0011] By employing the above technical solution, the importance of different operating parameters is quantified, enabling the highlighting of the impact of key parameters when matching fault feature templates. By multiplying and summing the parameter similarity scores by their corresponding weights, the final comprehensive similarity score accurately reflects the contribution of each parameter to the fault features. This weighted calculation method avoids the problem of weakening important features that may result from simple averaging, thus improving the accuracy of fault feature identification. Different parameter weight combinations can be set for each fault type, allowing the method to better distinguish different types of fault features and improving the discriminative power of fault type judgment.

[0012] In conjunction with some implementation methods of the first aspect, in some implementation methods, when it is determined that the current operating parameter value is not within the target dynamic reference range, a transformer anomaly is determined and an early warning signal is generated. Specifically, this includes: when it is determined that the current operating parameter value is not within the target dynamic reference range, defining the time corresponding to the current operating parameter value as an initial over-limit event; in response to the occurrence of the initial over-limit event, starting a preset confirmation duration timer; activating an over-limit count counter for accumulating the number of over-limit events, and setting the initial value of the over-limit count counter to 1 to count the initial over-limit event; within the preset confirmation duration, continuously acquiring subsequent current operating parameter values ​​and their corresponding current operating condition parameter values ​​at a preset sampling period; based on any subsequent current operating parameter value, determining the corresponding target dynamic reference range according to the corresponding current operating condition parameter value; when it is determined that any subsequent current operating parameter value is outside the corresponding target dynamic reference range, incrementing the over-limit count counter by one; when the preset confirmation duration timer ends and the final count value of the over-limit count counter exceeds a preset count threshold, a transformer anomaly is determined and an early warning signal is generated.

[0013] By adopting the above technical solution, and continuously sampling and counting the number of limit violations within a preset confirmation period, short-term fluctuations and persistent anomalies can be distinguished. The preset sampling period improves the continuity and timeliness of monitoring, while real-time updates to the target dynamic benchmark range ensure accurate judgment even when operating conditions change. Establishing final anomaly confirmation based on the number of limit violations exceeding a preset threshold improves the reliability of anomaly judgment and reduces the false alarm rate caused by occasional fluctuations.

[0014] In conjunction with some implementations of the first aspect, in some implementations, after determining a transformer anomaly and generating an early warning signal, the method further includes: acquiring the output side temperature data, transformer body temperature data, and ambient temperature data of the target transformer; calculating the output side temperature rise and the body temperature rise based on the output side temperature data, transformer body temperature data, and ambient temperature data; calculating a first rate of change of the output side temperature rise and a second rate of change of the body temperature rise; determining the current composite thermal operating state of the target transformer based on the output side temperature rise, body temperature rise, first rate of change, and second rate of change; and categorizing historical data into low-temperature steady state, high-temperature steady state, and transition state. The data are categorized to obtain corresponding historical status data. For each type of historical status data, a corresponding status operating range is determined. Based on the current composite thermal operating status of the target transformer, the corresponding status operating range is selected as multiple operating ranges. When a change in the current composite thermal operating status is detected, the status operating range before and after the switch is obtained, and the status switching reference range is calculated. During the thermal operating status switch, the status switching reference range is used as the dynamic reference range. When the fluctuation amplitude of the operating parameter value after the switch is less than the preset stability threshold, the dynamic reference range of the anomaly monitoring is updated to the status operating range corresponding to the thermal operating status after the switch.

[0015] By adopting the above technical solution, and acquiring data on the transformer's output temperature, body temperature, and ambient temperature, the temperature rise on the output side and the body temperature, along with their rate of change, can be calculated to determine the transformer's current composite thermal operating state. Historical data is categorized into low-temperature steady-state, high-temperature steady-state, and transitional states, and corresponding operating condition intervals are determined. When a change in thermal operating state is detected, a state switching reference range is calculated and used as a dynamic reference range during state switching. When the fluctuation amplitude of the operating parameter value after the switch is less than a preset stability threshold, the range corresponding to the new thermal operating state is updated. This method of classifying and dynamically adjusting the reference range based on composite thermal operating states can more accurately reflect the transformer's operating characteristics under different thermal states, improving the accuracy of the dynamic reference range. Simultaneously, considering the parameter fluctuation characteristics during state switching, the introduction of the state switching reference range as a basis for judging the transition period reduces the probability of false alarms caused by drastic parameter fluctuations during state switching, improves the adaptability of transformer anomaly monitoring, and makes the monitoring results more consistent with the actual operating patterns of the transformer under different thermal states.

[0016] In conjunction with some implementations of the first aspect, in some implementations, the current composite thermal operating state of the target transformer is determined based on the output side temperature rise, the body temperature rise, the first rate of change, and the second rate of change. Specifically, this includes: when the absolute value of the first rate of change is less than a first preset rate threshold and the absolute value of the second rate of change is less than a second preset rate threshold, the transformer is determined to be in a thermally stable stage; in the thermally stable stage, if the output side temperature rise is less than a preset output temperature rise threshold and the body temperature rise is less than a preset body temperature rise threshold, the current composite thermal operating state is determined to be a low-temperature steady state; in the thermally stable stage, if the output side temperature rise is greater than or equal to a preset output temperature rise threshold, or the body temperature rise is greater than or equal to a preset body temperature rise threshold, the current composite thermal operating state is determined to be a high-temperature steady state; when the absolute value of the first rate of change is greater than or equal to the first preset rate threshold, or the absolute value of the second rate of change is greater than or equal to the second preset rate threshold, the current composite thermal operating state is determined to be a transitional state.

[0017] By adopting the above technical solution, the thermal stability stage is identified by determining whether the absolute values ​​of the temperature rise rates on the output side and the transformer body exceed preset rate thresholds. Within the thermal stability stage, low-temperature steady state and high-temperature steady state are distinguished based on whether the temperature rise rates on the output side and the transformer body exceed corresponding temperature rise thresholds. When either temperature rise rate exceeds the preset rate threshold, it is determined to be a transition state. This approach more comprehensively characterizes the thermal state of the transformer and improves the accuracy of thermal operation state determination. By introducing a rate-of-change criterion to identify the transition state, the dynamic process of rapid temperature rise changes can be effectively captured, improving the sensitivity of identifying the transformer's thermal state switching process. Using multiple threshold constraints for state division improves the reliability of thermal operation state determination and makes the state determination results closer to the actual thermodynamic characteristics of the transformer.

[0018] In conjunction with some implementation methods of the first aspect, in some implementation methods, for historical data of each operating mode, the corresponding operating condition range is determined, specifically including: grouping the historical data into low temperature steady state, high temperature steady state and transition state to obtain several groups of historical data; calculating the average value and standard deviation of the operating parameters for each group of historical data; constructing confidence intervals for different thermal operating states based on the average value and standard deviation; and using the upper and lower boundaries of the confidence intervals as the upper and lower limits of the corresponding state operating condition range.

[0019] By employing the above technical solution, historical data is grouped into low-temperature steady-state, high-temperature steady-state, and transitional states. The average and standard deviation of the operating parameters for each group are calculated, and confidence intervals for different thermal operating states are constructed based on these statistical characteristics. The upper and lower boundaries of the confidence intervals are used as the upper and lower limits of the corresponding operating condition intervals. This interval construction method based on statistical principles can objectively reflect the distribution pattern of transformer parameters under different thermal operating states, improving the rationality of the operating condition interval boundaries. Establishing confidence intervals through grouped statistics reduces the interference of data cross-contamination under different thermal operating states on interval division, improving the distinguishability of the operating condition intervals. Using the average and standard deviation as the basis for constructing confidence intervals improves the statistical reliability of the operating condition interval boundaries, making the interval range more accurately reflect the fluctuation characteristics of transformer parameters under various thermal operating states.

[0020] Secondly, embodiments of this application provide a transformer intelligent online monitoring system, which includes: one or more processors and a memory; the memory is coupled to one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and one or more processors call the computer instructions to cause the system to perform the method described in the first aspect and any possible implementation thereof.

[0021] Thirdly, embodiments of this application provide a computer-readable storage medium including instructions that, when executed on a system, cause the system to perform the method described in the first aspect and any possible implementation thereof.

[0022] Fourthly, embodiments of this application provide a computer program product that, when run on a system, causes the system to execute the method described in any possible implementation of the first aspect.

[0023] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages: 1. This application provides an intelligent online monitoring method for transformers. By acquiring historical data and dividing the data into intervals according to operating conditions, a dynamic benchmark range is established for each operating condition interval, enabling the monitoring benchmark to be dynamically adjusted as operating conditions change. After acquiring current operating parameters and operating condition parameters, the method automatically matches and compares the dynamic benchmark range of the corresponding operating condition interval, avoiding potential misjudgments caused by using fixed thresholds. Since the benchmark range is determined based on the operating performance of a large amount of historical data under the same operating conditions, it has strong statistical representativeness and actual operational relevance, and can better reflect the normal operating state range of the transformer under that condition. This adaptive monitoring method improves the accuracy of abnormal state identification, reduces false alarms caused by changes in operating conditions, and ensures the timely detection of real abnormal states. Through the timely generation of early warning signals, maintenance personnel can quickly learn about transformer abnormalities and take corresponding measures, reducing the risk of equipment failure.

[0024] 2. This application provides an intelligent online monitoring method for transformers. After detecting an anomaly, it collects time-series data of multiple pre-selected operating parameters and performs similarity calculation and comprehensive scoring with preset fault feature templates to achieve intelligent identification of the anomaly type. The joint analysis of multiple parameters improves the comprehensiveness and reliability of anomaly type judgment. The method uses similarity calculation to capture the characteristic manifestations of different types of faults, achieving accurate matching even when fault characteristics differ. By setting multiple association rules for comprehensive calculation, the accuracy of anomaly type identification is improved. The final output warning signal contains specific anomaly type information, providing maintenance personnel with more targeted fault information and facilitating rapid fault handling.

[0025] 3. This application provides an intelligent online monitoring method for transformers. Historical data is categorized according to different operating modes, establishing operating condition ranges corresponding to main grid operation, ring network operation, and interconnection operation. This allows the benchmark range for anomaly monitoring to be specifically matched to the current actual operating mode. When an operating mode changes, the system calculates the benchmark range for the switching process and uses it as a dynamic benchmark range. Simultaneously, it monitors the fluctuations in operating parameter values ​​after the switch. When the fluctuation amplitude is less than a preset stability threshold, it updates the operating condition range corresponding to the new operating mode. This allows the system to adapt to the characteristic differences of transformers under different operating modes, avoiding false alarms caused by operating mode switching and improving the accuracy of anomaly monitoring. Furthermore, through special processing of the switching process, the continuity and reliability of anomaly monitoring during operating mode transitions are improved, ensuring that genuine anomalies are not missed and that normal fluctuations caused by operating mode switching are not misjudged. Attached Figure Description

[0026] Figure 1This is a flowchart illustrating a method for intelligent online monitoring of transformers in an embodiment of this application.

[0027] Figure 2 This is another flowchart illustrating a method for intelligent online monitoring of transformers in an embodiment of this application.

[0028] Figure 3 This is a schematic diagram of the physical device structure of a transformer intelligent online monitoring system provided in an embodiment of this application. Detailed Implementation

[0029] The terminology used in the following embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification and appended claims of this application, the singular expressions “a,” “an,” “the,” “the,” “the,” and “this” are intended to include the plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this application refers to any or all possible combinations including one or more of the listed items.

[0030] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature, and in the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.

[0031] With the transformation of the global energy structure and the accelerated construction of smart grids, the stability and reliability of power systems have become crucial. As a core hub in the power grid, the safe and stable operation of transformers is the cornerstone of ensuring reliable power transmission and distribution. Modern transformers are generally equipped with numerous online monitoring sensors, capable of collecting massive amounts of operational data in real time, including voltage, current, temperature, and vibration. This provides an unprecedented data foundation for using big data and artificial intelligence technologies to achieve accurate assessment of transformer status and early warning of faults.

[0032] Currently, the most common early warning method in transformer online monitoring systems is the alarm mechanism based on fixed thresholds. Maintenance personnel or system experts set one or more fixed alarm thresholds (e.g., alarm value 85℃, trip value 95℃) for key operating parameters (such as top oil temperature) based on equipment manufacturer specifications, industry standards, and limited operational experience. When the real-time monitored parameter value exceeds this fixed threshold, the system triggers an alarm.

[0033] The technical solution of this application is applicable to health status monitoring and fault early warning of large power transformers with complex and variable operating conditions. A specific scenario is, for example, a power transformer in a main substation supplying power to a large industrial park and surrounding residential areas. During weekdays, this transformer needs to bear heavy loads to meet peak industrial electricity demand; while at night and on holidays, the load is mainly for residential electricity consumption, dropping to a lower level. Simultaneously, its operating environment also varies significantly with the seasons, with high ambient temperatures in summer and low temperatures in winter. Therefore, the two key operating parameters of this transformer—load rate and ambient temperature—exhibit wide-range, periodic, and drastic fluctuations within a day, a week, and a year.

[0034] In the scenarios described above, using a fixed threshold monitoring method presents significant technical problems. For example, setting a fixed alarm threshold of 85°C for the top-layer oil temperature presents challenges. During full-load operation in summer, even if the transformer itself is healthy, its oil temperature could easily reach 80°C or even higher due to the combined effects of high load and high ambient temperature, very close to the preset alarm threshold. This makes it highly susceptible to false alarms due to even minor fluctuations, leading to frequent handling of invalid alarms by maintenance personnel and alarm fatigue. Conversely, during light-load operation in winter, a potential, nascent internal fault (such as a minor blockage in the cooling system) might only cause the oil temperature to rise slightly from the normal 40°C to 50°C. This 50°C temperature is far below the fixed alarm threshold of 85°C, thus completely drowning out this early fault signal and preventing timely detection, resulting in missed alarms. Over time, this hidden danger could develop into a serious fault, causing unplanned outages and significant economic losses. Therefore, the existing one-size-fits-all monitoring method based on fixed thresholds cannot adapt to the dynamic changes in transformer operating conditions. In scenarios with variable operating conditions, the accuracy and timeliness of its early warnings are difficult to guarantee. This is the core technical problem that the technical solution of this application aims to solve.

[0035] The following example is used in conjunction with Figure 1 The present application describes a method for intelligent online monitoring of transformers in its embodiments: Please refer to Figure 1 This is a flowchart illustrating a method for intelligent online monitoring of transformers in an embodiment of this application.

[0036] S101. Obtain historical data of the target transformer within a preset historical time period. The historical data includes historical operating data and corresponding historical operating condition data. The target transformer refers to one or more specific transformer devices that require intelligent online monitoring. The preset historical time period is a time span predefined by the user or system, such as the past year, three years, or the entire period since the equipment was put into operation. This time period is chosen to ensure coverage of various typical operating modes and seasonal changes experienced by the transformer. Historical data refers to the collection and storage of all relevant data records within this time period. Historical operating data specifically refers to the measured values ​​of parameters reflecting the transformer's operating status. Historical operating condition data refers to parameter values ​​recorded at the same time as the operating data, reflecting the transformer's operating environment and load conditions, including, but not limited to, the transformer's load current or load rate, ambient temperature, ambient humidity, and cooler operating status (such as the start / stop status of fans or oil pumps). One-to-one correspondence means that for each historical operating data record at a given time point, there exists a corresponding historical operating condition data record at the same moment; the two are strictly aligned in timestamps, forming a data pair. Essentially, this step involves the system accessing the data storage center to prepare a complete, time-synchronized, comprehensive historical dataset containing the transformer's status and its operating conditions for subsequent analysis and modeling. The data acquisition device included in this application has the following specifications: It supports full power parameter measurement (U, I, P, Q, S, PF, F); supports bidirectional active and reactive energy metering; supports active and reactive pulse output; supports multi-rate metering (historical 12 months), with multi-rate settings for 4 time zones, 10 time periods, and four rates (peak and off-peak); supports power quality analysis functions: voltage and current imbalance measurement, voltage and current phase angle measurement, 2nd-63rd order harmonics, total (odd and even) harmonics measurement, voltage crest factor, telephone crest factor, and current K-coefficient measurement; supports event logging (128 records). Supports TF card expansion), alarm logs (including 66 types of alarms, with the most recent 16 records for each type, and supports TF card expansion), maximum demand (historical data for 12 months, forward and reverse), and extreme value statistics (this month and last month); supports 2-channel switch output (expandable to 8 channels with optional MD82 module) and 2-channel switch input (expandable to 26 channels with optional MD82 module). Switch outputs can be configured as alarm outputs or remote control outputs. When DO is used as an alarm output, alarm content can be freely associated; supports TF card storage, which can be used for timed storage of data such as electrical parameters, power, and harmonics, as well as waveform storage.

[0037] Supports up to 12 analog outputs (optional MA84 module), with arbitrarily configurable output electrical parameters; RS485 communication, supporting Modbus RTU protocol / DL / T645-07 protocol (with freeze function) adaptive; Technical specifications: Voltage and current accuracy: 0.2 level; Active energy accuracy: 0.5s level, reactive energy accuracy: 0.5s level 2; Harmonic accuracy: ±1% (2nd to 45th order), ±2% (46th to 63rd order); The first method is direct query based on the database interface. In a modern power monitoring system, all sensor data are usually aggregated into a central time-series database (TSDB) or relational database. The system can act as a client, connecting to this database through standard database connection protocols (such as ODBC, JDBC). Subsequently, the system executes a Structured Query Language (SQL) or time-series database-specific query command. This command explicitly specifies the unique identifier of the target transformer, the start and end timestamps of the preset historical time period, and a parameter list (or measurement point labels) of all historical operating data and historical condition data to be extracted. Upon receiving the request, the database server retrieves all data records that meet the criteria and returns the result set to the system. After receiving the data, the system loads it into memory or temporary storage, forming a structured data table or data frame for subsequent processing. The second method is file-based batch processing. In some systems, historical data may be periodically archived and stored as files on a file server or distributed file system (such as HDFS). These files are typically organized and named according to device name and date (e.g., T1_2023-10-26.csv). The system first locates all relevant historical data files based on the target transformer and the preset historical time period. Then, the system reads these files one by one, parses their contents, filters the data within the specified time period based on timestamps, and extracts all required operating parameters and condition parameters. Finally, the system will merge, sort, and integrate the data extracted from all files into a unified historical dataset.

[0038] S102. Based on historical operating data, the historical data is divided into multiple operating ranges, and the numerical range of a certain operating parameter is defined for each operating range. A working condition interval refers to a specific set of operating conditions defined by the numerical range of one or more working condition parameters. For example, a working condition interval can be defined as a load rate between 50% and 75% and an ambient temperature between 15°C and 25°C. The purpose of this division is to discretize the entire complex and continuously changing operating history into several representative and relatively stable working condition patterns. Each working condition interval corresponds to a unique range of working condition parameter combinations. The core idea of ​​this step is to no longer treat all historical data as a whole, but to classify it according to the external conditions (working conditions) at the time of its occurrence, laying the foundation for establishing a dynamic monitoring benchmark adapted to the working conditions. The system only uses historical working condition data (such as load rate and ambient temperature) as the division criteria to allocate the entire historical dataset (including historical operating data and historical working condition data) into different working condition intervals.

[0039] The first approach is a grid partitioning method based on expert experience. This method first requires domain experts or maintenance engineers to determine the key operating parameters affecting the transformer's condition based on their understanding of the transformer's operating characteristics (e.g., selecting load rate and ambient temperature as two dimensions). Then, several discrete numerical ranges are defined for each selected operating parameter. For example, the load rate is divided into three ranges: light load (0-30%), medium load (30-70%), and heavy load (70-100%); the ambient temperature is divided into three ranges: low temperature (<0℃), normal temperature (0-25℃), and high temperature (>25℃). By combining these different dimensional ranges, a two-dimensional grid is formed, with each cell representing an operating condition interval (e.g., medium load to normal temperature interval). The system then iterates through all historical data points and assigns each data point to the corresponding grid (operating condition interval) based on its historical operating parameter value. The second approach is based on a data-driven unsupervised clustering algorithm. This method does not require pre-setting of ranges but allows the algorithm to automatically discover natural operating condition clustering patterns from the data. The system combines multiple historical operating condition parameters (such as load rate, ambient temperature, humidity, etc.) corresponding to each historical data point into a multi-dimensional vector. Then, using the operating condition vectors of all historical data points as input, a clustering algorithm, such as K-Means clustering or the density-based DBSCAN algorithm, is applied. The algorithm automatically aggregates data points with similar operating conditions together, forming several clusters. Each cluster represents an operating condition interval. For example, the K-Means algorithm will find K cluster centers, each representing a typical operating condition, and all data points are assigned to the operating condition interval represented by the nearest cluster center.

[0040] S103. Based on all historical operating data within the operating condition range, determine the dynamic reference range corresponding to the operating condition range. The dynamic reference range is a numerical range with an upper limit value and a lower limit value. The dynamic reference range is a normal or healthy numerical fluctuation range derived through statistical analysis within a specific operating condition interval for a particular operating parameter (such as top oil temperature). This range is defined by an upper limit and a lower limit, forming a closed or semi-closed numerical interval. It is dynamic because the system calculates different reference ranges for different operating condition intervals. For example, in the summer-heavy load operating condition interval, the dynamic reference range for top oil temperature might be [75℃, 85℃], while in the winter-light load operating condition interval, its range might be [35℃, 45℃]. The purpose of this step is to establish a dedicated set of quantitative standards that reflect the healthy operating status of the transformer under each operating condition interval defined in S102. For each operating condition interval, the system extracts all historical operating data belonging to that interval and performs statistical modeling on this data to determine its upper and lower limits.

[0041] The first approach uses the percentile method from statistics. Given an operating condition range and an operating parameter (e.g., winding temperature), the system first collects all historical winding temperature data within that range. Then, these data are sorted, and specific percentiles of their probability distribution are calculated. For example, the system can set the lower limit as the 5th percentile (P5) and the upper limit as the 95th percentile (P95). This means that historically, under this operating condition, 90% of healthy operating data falls within the range [P5, P95]. This method is independent of the specific shape of the data distribution and has strong robustness, especially suitable for non-normally distributed data. The second approach is a statistical parameter method based on the Gaussian distribution (normal distribution). This method assumes that under stable operating conditions, healthy operating parameter values ​​roughly follow a Gaussian distribution. The system first calculates the mean (μ) and standard deviation (σ) of all historical data for a given operating parameter within a given operating condition range. Then, a baseline range is defined based on the 3-sigma criterion or a more lenient / stricter criterion. For example, the lower limit can be set to μ-3σ and the upper limit to μ+3σ. Based on the properties of the Gaussian distribution, this interval can theoretically cover approximately 99.7% of the health data points. This method is very effective and easy to compute when the data truly conforms to or approximately conforms to a Gaussian distribution.

[0042] S104. Real-time acquisition of the current operating parameter values ​​and corresponding current operating condition parameter values ​​of the target transformer; Real-time acquisition means the system reads the latest data from field monitoring equipment at a very high time frequency (e.g., every second, every ten seconds, or every minute). Current operating parameter values ​​refer to the instantaneous measurements of various status parameters of the target transformer at the current moment, such as the current top oil temperature of 78.2℃. Current operating condition parameter values ​​refer to the instantaneous values ​​of parameters reflecting the transformer's operating environment and load conditions at the same time, such as the current load rate of 88% and the ambient temperature of 31℃. The purpose of this step is to acquire input data for real-time evaluation, ensuring that subsequent judgments are based on the transformer's latest actual condition. The system needs to establish a continuous communication connection with the transformer's online monitoring system (such as a SCADA system, data acquisition unit, or IoT gateway) to ensure the timeliness and synchronization of data.

[0043] The first approach uses a publish-subscribe communication protocol, such as MQTT or OPC-UA. In this model, the system, acting as a subscriber (Client), registers its list of interested parameters (i.e., all operating and condition parameters to be monitored) with the data source (e.g., an MQTT Broker or OPC-UA Server). Each time the data source receives a new measurement, it proactively pushes that value to all clients that have subscribed to that parameter. This method is efficient, has low latency, and enables near real-time information delivery, making it ideal for monitoring applications requiring rapid response. The second approach uses a polling communication protocol, such as Modbus TCP / IP or HTTP-based API calls. The system proactively sends data request commands to the data acquisition device or data interface server at a preset fixed time interval (polling cycle). Upon receiving the request, the device or server queries the latest parameter values ​​and packages them in a response message, returning them to the system. This method is relatively simple to implement and has good compatibility, but the real-time performance of the data is limited by the length of the polling cycle and incurs continuous network communication overhead.

[0044] S105. Determine the target operating condition interval from the plurality of operating condition intervals based on the current operating condition parameter values; The system compares the real-time operating condition parameter values ​​(such as the current load rate and ambient temperature) obtained in S104 with the multiple operating condition intervals defined in S102. The purpose is to accurately identify which mode the transformer is currently in among many predefined operating condition modes. The interval that is successfully matched is determined as the target operating condition interval. This precise matching is a key step in realizing dynamic early warning. It ensures that subsequent health status assessments will call upon the dynamic benchmark range (determined in S103) that corresponds exactly to the current operating condition, thereby avoiding false alarms or missed alarms caused by changes in operating conditions due to traditional fixed threshold methods.

[0045] The first method is the interval range traversal matching method, which mainly corresponds to the expert experience-based grid partitioning in S102. The system traverses all predefined operating condition intervals. For each interval, the system checks whether all current operating condition parameter values ​​fall within the numerical range defined for each operating condition parameter in that interval. For example, if the current load rate is 88% and the ambient temperature is 31℃, the system will check all intervals such as medium load-normal temperature and heavy load-high temperature one by one. When the heavy load-high temperature interval is checked (assuming it is defined as load rate > 70% and temperature > 25℃), the system finds that 88% and 31℃ both meet the conditions, and immediately determines this interval as the target operating condition interval. Since the division of operating condition intervals is usually mutually exclusive, the matching result is unique. The second method is the nearest centroid matching method, which mainly corresponds to the data-driven clustering algorithm in S102. In this method, each operating condition interval has a centroid vector representing the center of the interval generated during partitioning. The system first combines the current multiple operating condition parameter values ​​into a real-time operating condition vector with the same dimension as the centroid vector. Then, the system calculates the geometric distance (e.g., Euclidean distance) between this real-time operating condition vector and the centroids of all pre-stored operating condition intervals. The operating condition interval represented by the centroid with the smallest distance is determined as the current target operating condition interval.

[0046] S106. Obtain the target dynamic reference range corresponding to the target operating condition range; The target operating condition range refers to the pre-defined operating condition range in S102 into which the current operating condition parameter values ​​(e.g., current load rate 88%, current ambient temperature 31℃) acquired in real-time in S104 fall. For example, if an operating condition range is defined as heavy load (>70%) - high temperature (>25℃), then the current operating condition belongs to this target operating condition range. The target dynamic reference range is the specific upper and lower limit values ​​calculated and stored in S103 for this target operating condition range, targeting specific operating parameters (e.g., top oil temperature). This step serves as a bridge, connecting the real-time operating condition with the historical experience model library to find the correct benchmark that matches the current operating condition for the upcoming real-time status assessment.

[0047] The first approach is fast lookup based on hash tables or dictionaries. After S103 is completed, the system can store all operating condition interval definitions and their corresponding dynamic baseline ranges (upper and lower limits) for each operating parameter in one or more hash table (or dictionary) data structures. The definition of the operating condition interval can be encoded into a unique key, for example, a string Load_High_Temp_High composed of various operating condition parameter categories. During real-time monitoring, the system first generates the corresponding key based on the current operating condition parameter value, and then uses this key to search in the hash table. Since the average lookup time complexity of a hash table is O(1), this method can retrieve the required target dynamic baseline range extremely quickly. The second approach is nearest neighbor search based on a spatial index structure. If the operating condition intervals are obtained through a clustering algorithm (such as K-Means), then each interval can be represented by its cluster center (a multidimensional operating condition vector). The system can store all cluster centers in an efficient multidimensional spatial index structure, such as a KD-tree (K-Dimensional Tree). During real-time monitoring, the system performs a nearest neighbor search operation in the KD-tree using the vector composed of the current operating condition parameter values. The nearest cluster center returned by the search result represents the target operating condition interval to which the current operating condition belongs, and the system then obtains the dynamic benchmark range associated with that cluster center. This method is particularly suitable for situations where the operating condition parameters have high dimensionality and the interval boundaries are irregular.

[0048] S107. If it is determined that the current operating parameter value is not within the target dynamic reference range, an abnormality is identified in the transformer and an early warning signal is generated.

[0049] If the current operating parameter value is determined to be outside the target dynamic reference range, a transformer anomaly is identified, and an early warning signal is generated. Specifically, this includes: defining the moment corresponding to the current operating parameter value as an initial over-limit event; in response to the occurrence of the initial over-limit event, starting a preset confirmation period; activating an over-limit count counter to accumulate the number of over-limit events, and setting the initial value of the over-limit count counter to 1 to count the initial over-limit event; continuously acquiring subsequent current operating parameter values ​​and their corresponding current operating condition parameter values ​​at a preset sampling period within the preset confirmation period; determining the corresponding target dynamic reference range based on any subsequent current operating parameter value and the corresponding current operating condition parameter value; incrementing the over-limit count counter by one if any subsequent current operating parameter value is outside the corresponding target dynamic reference range; and identifying a transformer anomaly and generating an early warning signal when the preset confirmation period ends and the final count value of the over-limit count counter exceeds a preset threshold.

[0050] The initial out-of-limit event refers to the moment during routine monitoring when the system first detects that the real-time value of a certain current operating parameter falls outside the dynamic reference range for its corresponding operating condition. This event marks the starting point of the entire verification process. In response to this event, the system immediately initiates a timing process for a preset verification duration. This duration is a pre-defined observation window, such as 30 seconds or 1 minute, to give the system sufficient time to observe whether the anomaly is persistent. Simultaneously, the system activates a dedicated out-of-limit counter to accumulate the number of out-of-limit occurrences, setting its initial value to 1, indicating that the initial out-of-limit event triggering this verification process has been included. Within this preset verification duration window, the system does not wait passively but continuously and frequently collects subsequent current operating parameter values ​​and their corresponding operating condition parameters at a preset sampling period much shorter than the verification duration (e.g., once every 5 seconds). For each newly collected subsequent operating parameter value, the system repeats the matching process in S106 to find its corresponding dynamic reference range under the new operating condition and determine whether it exceeds the limit. If the limit is exceeded again, the over-limit count counter is incremented. Finally, when the preset confirmation time expires, the system checks the final count value of the over-limit count counter. Only when this count value exceeds a preset threshold (e.g., more than 8 over-limits within 1 minute) will the system ultimately determine that the transformer is indeed abnormal and officially generate a warning signal. This threshold, together with the confirmation time and sampling period, defines the strictness of the abnormality judgment.

[0051] The first approach is based on a state machine model. The system can maintain an independent finite state machine for each monitored operating parameter. This state machine contains at least two states: normal monitoring and confirmation. In the normal monitoring state, the system continuously performs basic comparisons from S104 to S107. Once an initial limit violation event is detected, the state machine of the corresponding parameter switches from normal monitoring to confirmation. Upon entering the confirmation state, the system records the current timestamp as the timing start point and initializes a counter bound to the parameter to 1. A background timer task periodically (with a period equal to the preset sampling period) checks all parameters in the confirmation state. For each parameter in this state, the timer task obtains its latest value and operating condition, determines whether it exceeds the limit, and increments its dedicated counter if it does. Simultaneously, the timer task checks whether the difference between the current time and the timing start point has exceeded the preset confirmation duration. If it has exceeded the timeout, the count value of the parameter is compared with the count threshold. If it exceeds the threshold, an alarm generation module is triggered, and the state machine is switched back to normal monitoring; if it does not exceed the threshold, no alarm is triggered, and the state machine is simply switched back to normal monitoring. The second approach is based on an asynchronous task queue. When an initial limit-crossing event occurs, the main monitoring thread is not blocked. Instead, it creates a confirmation task containing information such as parameter identifiers, confirmation duration, sampling period, and count threshold, and pushes this task to an asynchronous task executor or thread pool. Upon receiving the task, the task executor immediately begins execution. The task's internal logic is as follows: First, a local counter is initialized to 1. Then, a loop is entered, the total loop duration of which is determined by the confirmation duration. Within the loop, the task sleeps for one sampling period, then is awakened and performs data acquisition, baseline matching, and limit-crossing judgment. If a limit-crossing occurs, the local counter is incremented. When the loop ends (i.e., the confirmation duration is exhausted), the task performs a final judgment, comparing the final value of the local counter with the count threshold. If the threshold is exceeded, a global warning service interface is called to generate a warning signal. The task self-destructs after execution. This approach decouples the main monitoring process from the confirmation process, providing excellent concurrency handling capabilities and system scalability.

[0052] In the above embodiments, historical data is acquired and divided into intervals according to operating conditions. A dynamic benchmark range is established for each operating condition interval, enabling the monitoring benchmark to be dynamically adjusted as operating conditions change. After acquiring the current operating parameters and operating condition parameters, the method automatically matches and compares the dynamic benchmark range of the corresponding operating condition interval, avoiding potential misjudgments caused by using fixed thresholds. Since the benchmark range is determined based on the operating performance of a large amount of historical data under the same operating conditions, it has strong statistical representativeness and actual operational relevance, and can better reflect the normal operating state range of the transformer under that condition. This adaptive monitoring method improves the accuracy of abnormal state identification, reduces false alarms caused by changes in operating conditions, and ensures the timely detection of real abnormal states. Through the timely generation of early warning signals, maintenance personnel can quickly learn about transformer abnormalities and take corresponding measures, reducing the risk of equipment failure.

[0053] Furthermore, after determining the transformer anomaly and generating an early warning signal, the system can also, in response to the generation of the early warning signal, collect operating data of at least two pre-selected operating parameters within a preset time window before the generation time of the early warning signal, forming corresponding time-series data; calculate the similarity between each type of time-series data and the corresponding parameter template sequence in multiple preset fault feature templates to obtain a parameter similarity score for each fault feature template; apply one or more preset association rules in each fault feature template to comprehensively calculate the corresponding parameter similarity score to obtain a comprehensive similarity score for each fault feature template; and determine the comprehensive similarity score. The maximum target comprehensive similarity score in the degree score is used to determine the fault feature template corresponding to the target comprehensive similarity score as the best matching template. Specifically, this includes: for each fault feature template, determining the corresponding preset association rule as parameter weight, with each parameter weight corresponding to a pre-selected operating parameter; multiplying the parameter similarity score corresponding to the pre-selected operating parameter by the parameter weight corresponding to the pre-selected operating parameter to obtain a weighted score; summing all values ​​in the weighted score to obtain the comprehensive similarity score corresponding to the fault feature template; determining the specific anomaly type identified by the best matching template, and adding the specific anomaly type to the warning signal for output.

[0054] First, after the warning signal is triggered, the system's primary task is to immediately backtrack and collect dynamic data of key operating parameters prior to the fault. Using the warning time as a baseline, the system traces back a preset time window (e.g., 30 minutes) to capture the continuous changes of multiple pre-selected parameters during this period, forming their respective time series. Data can be obtained from a fast-responding real-time memory cache or by querying a historical database. To ensure the accuracy of subsequent analysis, the collected raw data undergoes cleaning, such as interpolation to fill missing points or filtering to smooth abnormal fluctuations.

[0055] Next, the system will quantitatively compare the collected field time-series data with various standard fault modes stored in the knowledge base. Specifically, the time series of each field parameter will be compared with the corresponding parameter template sequences in all templates in the fault template library. This process mainly uses the Dynamic Time Warping (DTW) algorithm, which can effectively measure the similarity in shape between two sequences with different rates or lengths. To eliminate the interference of differences in units and absolute values, all data will be normalized before calculation, so that the comparison focuses on the trend of change itself, and finally a series of similarity scores with different fault templates will be generated for each parameter.

[0056] Subsequently, to arrive at a comprehensive diagnostic judgment, the system needs to merge the scattered similarity scores of various parameters into a total score for each fault template. This step is achieved through weighted fusion, where each fault feature template has a pre-defined set of weights corresponding to the parameters. These weights reflect the importance of different parameters in diagnosing that specific fault. The system multiplies the similarity score of each parameter by its weight under that template, and then sums all the multiplications to obtain a comprehensive similarity score. This total score integrates all relevant evidence and highlights the influence of key indicators.

[0057] After calculating the overall similarity score of all candidate fault templates, the system will proceed to the best match determination stage. By comparing all scores, the system will find the maximum value, and the fault feature template with the highest score will be identified as the best matching template that best matches the current anomaly. To improve the reliability of the diagnosis, the system will also perform confidence checks, such as determining whether the highest score exceeds a preset threshold, or analyzing the difference between the highest and second-highest scores. If the difference is too small, it may indicate that the diagnostic result is ambiguous.

[0058] Finally, the system transforms its internal diagnostic conclusions into output information that provides clear guidance for maintenance personnel. The system extracts a text description of the specific anomaly type represented by the determined best-matching template, such as a short circuit between turns in the high-voltage side winding. This specific diagnostic conclusion is appended to the initial warning message, forming a richer and more targeted intelligent alarm, thus providing clear and accurate decision support for subsequent fault handling.

[0059] In the above embodiments, after an anomaly is detected, time-series data of multiple pre-selected operating parameters are collected and compared with preset fault feature templates for similarity calculation and comprehensive scoring, thereby achieving intelligent identification of the anomaly type. Joint analysis of multiple parameters improves the comprehensiveness and reliability of anomaly type judgment. The method employs similarity calculation, which can capture the characteristic manifestations of different types of faults, achieving accurate matching even when fault features differ to some extent. By setting multiple association rules for comprehensive calculation, the accuracy of anomaly type identification is improved. The final output warning signal contains specific anomaly type information, providing maintenance personnel with more targeted fault information and facilitating rapid fault handling.

[0060] Considering the significant differences in transformer operating characteristics under different operating modes, relying solely on the aforementioned dynamic reference range based on operating condition division may not fully adapt to the state changes during transformer operating mode switching. For example, even with the same operating parameters, the normal operating parameter range of the transformer may differ between main grid operation and ring network operation. To further improve the accuracy of anomaly monitoring, this application also proposes another intelligent online monitoring method for transformers. This method identifies the transformer's operating mode, establishes an operating condition interval corresponding to the operating mode, and employs a special reference range calculation method during operating mode switching, achieving more precise monitoring of the transformer's operating status. The following section combines... Figure 2 Another intelligent online monitoring method for transformers in this application is described below: Please refer to Figure 2 This is another flowchart illustrating a method for intelligent online monitoring of transformers in this application.

[0061] S201. Obtain the output side temperature data, transformer body temperature data, and ambient temperature data of the target transformer; Outgoing line temperature data specifically refers to the real-time temperature values ​​at the low-voltage or high-voltage side outgoing terminals, wiring terminals, or cable connections of the transformer, locations prone to electrical connection faults. Transformer body temperature data reflects the thermal state of the transformer's core heat-generating components, typically including winding temperatures (A, B, and C phases), core temperatures, or top oil temperatures. Ambient temperature data refers to the air temperature surrounding the transformer room or installation location, used as a reference for heat exchange. The system continuously reads these temperature parameters in real-time through a sensor network deployed at key locations on the transformer. These sensors can be contact-type resistance temperature detectors (RTDs) (such as PT100), thermocouples, or non-contact infrared temperature probes or fiber optic sensors. When acquiring data, the system follows a preset sampling frequency to ensure data timeliness and synchronization, providing accurate raw data for subsequent temperature rise calculations. The system not only reads instantaneous values ​​but also performs analog-to-digital conversion, filtering, noise reduction, and outlier removal on the acquired analog signals to ensure the quality of the input data.

[0062] To achieve this, the system can employ wired transmission-based industrial bus data acquisition technology. Specifically, the system connects temperature sensors distributed on the transformer's output side, inside the windings, and on the transformer room walls via RS485 or CAN bus. The sensors convert the acquired analog temperature signals into digital signals and, following communication protocols such as Modbus RTU or IEC60870-5-104, package and send the data to the field monitoring terminal or edge computing gateway. The gateway timestamps the multiple data streams to ensure that the output side, transformer body, and ambient temperatures are acquired simultaneously, thus eliminating analysis errors caused by time deviations. Another approach is to use wireless IoT sensing technology. The system deploys passive wireless temperature sensors (such as SAW surface acoustic wave sensors or RFID temperature tags) on the high-voltage side output end and the transformer body surface, using wireless radio frequency signals to transmit temperature data, avoiding the insulation risks associated with wiring in high-voltage areas. Ambient temperature is acquired using independent wireless thermo-hygrometers. All wireless nodes converge to a data concentrator via a ZigBee or LoRaWAN network, and the system obtains real-time temperature data streams from the concentrator through a publish-subscribe mechanism (such as MQTT).

[0063] S202. Calculate the temperature rise on the outgoing side and the temperature rise on the transformer body based on the outgoing line temperature data, the transformer body temperature data, and the ambient temperature data. After acquiring the raw temperature data, the system performs temperature rise calculations. Temperature rise refers to the difference between the temperature of a specific part of the equipment and the ambient temperature. It reflects the actual heat generation of the equipment more accurately than absolute temperature because it eliminates the influence of seasonal or diurnal temperature variations. The temperature rise on the outgoing line side is the difference between the outgoing line temperature data and the ambient temperature data. This indicator is mainly used to measure whether the contact resistance of electrical connections has increased and whether there are contact defects such as oxidation or overheating. The temperature rise of the transformer body is the difference between the transformer body temperature data and the ambient temperature data. This indicator mainly reflects the heat accumulation caused by copper losses in the transformer windings and iron losses in the core, as well as the efficiency of the cooling system. By calculating the temperature rise in these two dimensions in real time, the system can decouple the thermal state of the transformer into the connection thermal state and the internal load thermal state, providing orthogonalized feature inputs for subsequent composite state judgment. During the calculation process, the system handles possible negative values ​​(such as those caused by sensor errors under extreme weather conditions), usually returning them to zero or maintaining their original values ​​for subsequent algorithm processing.

[0064] S203. Calculate the first rate of change of the temperature rise on the line side and the second rate of change of the temperature rise on the body. The system then performs differential processing on the temperature rise data to calculate its rate of change over time. The first rate of change is the derivative of the line-side temperature rise with respect to time, characterizing the rapid degree of heating at the contact points on the line side; the second rate of change is the derivative of the body temperature rise with respect to time, characterizing the rate of heat absorption or release of the transformer's overall thermal capacity system. The rate of change is a key indicator for predicting the thermal state trend of the transformer; a positive rate of change indicates that the temperature is rising, while a negative rate of change indicates that the temperature is falling. The absolute value reflects the speed of fault development or the severity of load changes. By calculating the rate of change, the system can transform static temperature monitoring into dynamic trend monitoring, thereby identifying early faults (such as rapid overheating caused by a sudden short circuit) where the temperature has not yet exceeded the limit but the rate of temperature rise is abnormal. During the calculation process, the system needs to set an appropriate time window to balance sensitivity to rapid changes and noise suppression capabilities.

[0065] S204. Determine the current composite thermal operating state of the target transformer based on the temperature rise on the outgoing side, the temperature rise of the transformer body, the first rate of change, and the second rate of change. The system determines the current composite thermal operating state of the target transformer based on the outgoing line temperature rise, the transformer body temperature rise, a first rate of change, and a second rate of change. Specifically, this includes: when the absolute value of the first rate of change is less than a first preset rate threshold and the absolute value of the second rate of change is less than a second preset rate threshold, the transformer is determined to be in a thermally stable stage; in the thermally stable stage, if the outgoing line temperature rise is less than a preset outgoing line temperature rise threshold and the transformer body temperature rise is less than a preset transformer body temperature rise threshold, the current composite thermal operating state is determined to be a low-temperature steady state; in the thermally stable stage, if the outgoing line temperature rise is greater than or equal to a preset outgoing line temperature rise threshold, or the transformer body temperature rise is greater than or equal to a preset transformer body temperature rise threshold, the current composite thermal operating state is determined to be a high-temperature steady state; when the absolute value of the first rate of change is greater than or equal to the first preset rate threshold, or the absolute value of the second rate of change is greater than or equal to the second preset rate threshold, the current composite thermal operating state is determined to be a transitional state.

[0066] Based on the four core parameters calculated above, the system performs multi-dimensional logical judgments to determine the current composite thermal operating state of the transformer. This state is a comprehensive description, encompassing the transformer's overall load level, connection health, and thermal dynamic trends. The system defines three main states: low-temperature steady state, high-temperature steady state, and transitional state. The first and second preset rate thresholds are key boundaries for determining whether the transformer is in thermal equilibrium (thermal stability). Values ​​below these thresholds indicate slow temperature changes, considered a stable phase; values ​​above them indicate an unstable phase. Preset outgoing line temperature rise thresholds and preset body temperature rise thresholds distinguish between low and high temperatures. Notably, the system uses an OR logic when determining the high-temperature steady state; if either the outgoing line side or the body temperature is too high, it is considered a high-temperature steady state, reflecting equal emphasis on connection faults and overload faults. This composite state classification logic breaks the limitations of traditional methods that rely solely on load rate to determine operating conditions, incorporating the physical health of the equipment itself (such as high temperatures caused by poor contact) into the definition of operating conditions.

[0067] In its implementation, the system first checks the thermodynamic characteristics: it compares the absolute value of the first rate of change with a first preset rate threshold (e.g., 0.5℃ / min) and the absolute value of the second rate of change with a second preset rate threshold (e.g., 0.2℃ / min). If both are less than their respective thresholds, the system determines that the transformer has entered the thermal stability stage. Subsequently, within the thermal stability stage, the system further checks the thermal amplitude characteristics: it reads the current output temperature rise and the transformer body temperature rise. If the output temperature rise is less than a preset output temperature rise threshold (e.g., 30K) and the transformer body temperature rise is less than a preset transformer body temperature rise threshold (e.g., 50K), the system outputs the current state as a low-temperature steady state, which typically corresponds to a healthy operating period with light load and good connections. Conversely, if, within the thermal stability stage, the output temperature rise is greater than or equal to 30K, or the transformer body temperature rise is greater than or equal to 50K, the system outputs the current state as a high-temperature steady state, which corresponds to a period of heavy load operation or a period with potential overheating issues. If any rate of change exceeds the threshold in the initial rate of change judgment, the system will directly determine the current state as a transitional state, which corresponds to the process of large load fluctuations, transformer start-up and shutdown, or rapid development of faults.

[0068] S205. Classify the historical data according to low temperature steady state, high temperature steady state and transition state to obtain the corresponding state historical data; The system performs in-depth mining and organization of historical operating data accumulated over a long period, and performs data classification operations. This historical data includes all temperature records, temperature rise records, and corresponding rate of change records over a past period. Based on the state determination logic determined in step S204, the system labels each historical record and categorizes it into one of three types: low-temperature steady-state, high-temperature steady-state, or transitional state. This process is equivalent to constructing a supervised learning training dataset, where the state is the label and the operating parameters are the features. Through this classification, the system can separate the transformer's behavior patterns under different physical scenarios. For example, the high-temperature steady-state dataset reflects the transformer's extreme performance distribution under heavy load or high-heat environments, while the low-temperature steady-state dataset reflects its baseline noise level under no-load or light-load conditions. This classification is a prerequisite for establishing a high-precision dynamic benchmark because it avoids mixing data from different operating conditions and prevents abnormal data at low loads from being masked by normal fluctuations at high loads.

[0069] One technical approach to achieving this step is to utilize the query and aggregation functions of a time-series database. When storing raw data, the system calculates and stores state labels in real time. During classification, the system only needs to execute database queries to quickly extract all historical data fragments belonging to the low-temperature steady state. The system can export these data fragments as independent structured files or in-memory data structures for subsequent statistical analysis. Another approach is offline analysis based on batch processing scripts. The system periodically starts a background task that iterates through all raw data files over a past period. The script reads the data line by line, reuses the logic in S204 to recalculate the state at each moment, and distributes the data to three different storage buckets. This method allows the system to reclassify historical data after algorithm logic updates, ensuring consistency between historical benchmarks and the current algorithm.

[0070] S206. For each type of historical data, determine the corresponding operating condition range; The system determines the corresponding operating condition range for each operating mode based on historical data. Specifically, this includes: grouping historical data into low-temperature steady-state, high-temperature steady-state, and transitional state to obtain several sets of historical data; calculating the average and standard deviation of operating parameters for each set of historical data; constructing confidence intervals for different thermal operating states based on the average and standard deviation; and using the upper and lower boundaries of the confidence intervals as the upper and lower limits of the corresponding operating condition range.

[0071] After classifying historical data, the system uses statistical methods to construct a dedicated operating condition range for each thermal operating state. This range defines the normal fluctuation range of transformer operating parameters under a specific state. The system first calculates the statistical characteristics of each set of historical data, with the core components being the mean and standard deviation. The mean reflects the central tendency under that state, while the standard deviation reflects the dispersion of the data. Based on these two parameters, the system constructs confidence intervals, typically in the form of the mean plus or minus a certain multiple of the standard deviation. The upper boundary of the confidence interval is the upper limit of the operating condition range for that state, and the lower boundary is the lower limit. The core of this step is tailoring the approach to specific conditions, acknowledging that the normal standards for transformers differ under different states. For example, the temperature fluctuation range under high-temperature steady state is naturally larger than that under low-temperature steady state, therefore its operating condition range bandwidth should be correspondingly wider.

[0072] One specific approach to this step is the parameter estimation method based on the normal distribution assumption. The system assumes that the temperature rise data under each condition follows a Gaussian distribution and calculates the sample mean and sample standard deviation. With a high confidence level, the upper limit of the operating condition interval is the mean plus three times the standard deviation, and the lower limit is the mean minus three times the standard deviation. This method is simple to calculate, has low computational cost, and is suitable for situations where the data distribution is relatively symmetrical. Another approach is the percentile method based on nonparametric statistics. Considering that actual operating data may have a skewed or long-tailed distribution, the system does not assume a data distribution model but directly sorts each set of historical data. The system selects a smaller percentile as the lower limit and a larger percentile as the upper limit, thereby constructing an operating condition interval that covers the vast majority of historical data.

[0073] S207. Based on the current composite thermal operating state of the target transformer, select the corresponding state condition intervals as multiple condition intervals. During real-time monitoring, the system dynamically retrieves matching operating condition intervals from the interval library constructed in S206 based on the current composite thermal operating state determined in S204. This is a lookup or mapping process. For example, if the transformer is currently determined to be in a low-temperature steady state, the system immediately loads the upper and lower limits corresponding to the low-temperature steady state as the current monitoring standard; if the state changes to a high-temperature steady state, it seamlessly switches to the interval parameters for the high-temperature steady state. The multiple operating condition intervals here refer to the corresponding set of intervals selected for different monitoring indicators (such as outgoing line temperature rise, body temperature rise, and rate of change). Through this dynamic selection mechanism, the system achieves real-time adaptation of the monitoring benchmark, ensuring that each key opens a specific lock, avoiding missed reports (missing minor overheating at low loads) or false alarms (misjudging normal fluctuations as faults at high loads) caused by using a single broad standard.

[0074] One specific technique for implementing this step is to utilize an in-memory hash map or lookup table. During initialization, the system loads the calculated state interval parameters into a cache, with a structure like {State_ID:[Limit_Min, Limit_Max]}. In each monitoring cycle, the algorithm directly indexes and reads the corresponding limit based on the current state ID, achieving a time complexity of O(1), greatly ensuring real-time performance. Another implementation method is configuration-based distribution using a rule engine. The system configures the interval parameters for different states into different policy groups in the rule base. When the state determination module outputs a new state event, it triggers the rule engine to switch the active policy group, pushing the new alarm threshold parameters to the front-end monitoring logic. This approach facilitates the management of complex associated rules, such as adjusting the alarm delay time or priority while switching intervals.

[0075] S208. When a change in the current composite thermal operating state is detected, the state condition interval before and after the switch is obtained, and the state switch reference range is calculated. The system monitors changes in the composite thermal operating state in real time. Once a state transition is detected (e.g., from a low-temperature steady state to a transitional state, or from a transitional state to a high-temperature steady state), the system identifies this as a special switching event. At this point, the system does not simply replace the old interval with the new one, but initiates an intermediate calculation process. The system simultaneously acquires the operating interval corresponding to the state before the switch (the old interval) and the operating interval corresponding to the state after the switch (the new interval). Based on these two intervals, the system calculates a special state switching reference range. This reference range is typically a dynamically changing envelope designed to bridge the gap between two static intervals. Because in the physical world, the temperature and parameters of a transformer do not change instantaneously, but rather have an inertial response process. Directly hard-switching the threshold might cause monitoring indicators to jump out of range (if the new interval is narrower) or create a monitoring vacuum at the moment of switching. The purpose of designing the state switching reference range is to smooth this process and provide a reasonable monitoring standard for the transition period.

[0076] S209. During the switching of hot operating states, the state switching reference range shall be used as the dynamic reference range. After confirming the transition to hot operation mode, the system formally applies the state transition reference range calculated in S208 as the current anomaly monitoring standard, i.e., the dynamic reference range. During this period, the system no longer uses a single steady-state range, but instead uses this transition range that dynamically changes over time or logic to verify real-time operating parameters. If the real-time collected temperature rise on the outgoing line side or the body temperature rise exceeds the upper or lower boundaries of this dynamic reference range, the system will determine it as an anomaly and issue a warning. The key to this step is the word "dynamic," which acknowledges the special nature of the transition process. For example, at the moment a transformer is put into heavy load, the temperature will rise rapidly. Using the low-temperature steady-state threshold at this time would obviously result in a false alarm, while immediately using the high-temperature steady-state threshold might be too wide (failing to detect the problem of excessively rapid temperature rise). The state transition reference range precisely fills this gap; it allows parameter changes while restricting the trajectory of the changes to conform to the expected physical laws.

[0077] S210. When the fluctuation amplitude of the operating parameter value after switching is less than the preset stable threshold, the dynamic benchmark range of the abnormal monitoring will be updated to the state condition range corresponding to the hot operation state after switching.

[0078] While monitoring using the dynamic reference range, the system continuously assesses whether the transformer has completed its state transition and entered a new stable state. The assessment is based on the fluctuation amplitude of operating parameter values. The system calculates in real-time the fluctuation of temperature rise data (e.g., standard deviation, range, or peak-to-peak value) over a recent period (e.g., the past 5 minutes). If this fluctuation amplitude is less than a preset stability threshold, it indicates that the transformer's thermal process has converged, and internal heat exchange has reached a new equilibrium point. At this point, the system determines that the transition process has ended and immediately performs a reference update operation: it stops using the transitional state transition reference range calculated in S208 and officially locks the dynamic reference range for anomaly monitoring to the state condition interval corresponding to the post-transition thermal operating state selected in S207. This step marks the end of a complete state transition closed loop, and the system returns to steady-state monitoring mode, awaiting the next state change.

[0079] In the above embodiments, by acquiring data on the transformer's output temperature, body temperature, and ambient temperature, and calculating the output temperature rise and body temperature rise and their rate of change, the current composite thermal operating state of the transformer can be determined. Historical data is categorized into low-temperature steady-state, high-temperature steady-state, and transitional state, and corresponding operating condition intervals are determined. When a change in thermal operating state is detected, a state switching reference range is calculated and used as a dynamic reference range during state switching. When the fluctuation amplitude of the operating parameter value after switching is less than a preset stability threshold, the range corresponding to the thermal operating state after switching is updated. This method of classifying and dynamically adjusting the reference range based on composite thermal operating states can more accurately reflect the transformer's operating characteristics under different thermal states and improve the accuracy of the dynamic reference range. Simultaneously, considering the parameter fluctuation characteristics during state switching, by introducing the state switching reference range as a basis for judging the transition period, the probability of false alarms caused by drastic parameter fluctuations during state switching is reduced, improving the adaptability of transformer anomaly monitoring and making the monitoring results more consistent with the actual operating patterns of the transformer under different thermal states.

[0080] The system in the embodiments of this invention is described below from the perspective of hardware processing. Please refer to [link / reference needed]. Figure 3 This is a schematic diagram of the physical device structure of a transformer intelligent online monitoring system provided in an embodiment of this application.

[0081] It should be noted that, Figure 3 The structure of the system shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of the present invention.

[0082] like Figure 3As shown, the system includes a Central Processing Unit (CPU) 301, which can perform various appropriate actions and processes based on a program stored in Read-Only Memory (ROM) 302 or a program loaded from storage portion 308 into Random Access Memory (RAM) 303, such as executing the methods described in the above embodiments. The RAM 303 also stores various programs and data required for system operation. The CPU 301, ROM 302, and RAM 303 are interconnected via a bus 304. An Input / Output (I / O) interface 305 is also connected to the bus 304.

[0083] The following components are connected to I / O interface 305: input section 306 including a camera, infrared sensor, etc.; output section 307 including a liquid crystal display (LCD) and speakers, etc.; storage section 308 including a hard disk, etc.; and communication section 309 including a network interface card such as a LAN (Local Area Network) card and a modem, etc. Communication section 309 performs communication processing via a network such as the Internet. Drive 310 is also connected to I / O interface 305 as needed. Removable media 311, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 310 as needed so that computer programs read from them can be installed into storage section 308 as needed.

[0084] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing computer programs for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 309, and / or installed from removable medium 311. When the computer program is executed by central processing unit (CPU) 301, it performs the various functions defined in the present invention.

[0085] It should be noted that the computer-readable medium shown in the embodiments of the present invention can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In the present invention, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, wherein a computer-readable computer program is carried. The transmitted data signal can take many forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof.

[0086] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. Each block in a flowchart or block diagram may represent a module, segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0087] In another aspect, the present invention also provides a computer-readable storage medium, which may be included in the system described in the above embodiments; or it may exist independently and not assembled into the system. The storage medium carries one or more computer programs that, when executed by a processor of a system, cause the system to implement the methods provided in the above embodiments.

[0088] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit it. Although this application 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 scope of the technical solutions of the embodiments of this application.

[0089] As used in the above embodiments, depending on the context, the term "when..." can be interpreted as "if...", "after...", "in response to determining...", or "in response to detecting...". Similarly, depending on the context, the phrase "when determining..." or "if (the stated condition or event) is interpreted as "if determining...", "in response to determining...", "when (the stated condition or event) is detected", or "in response to detecting (the stated condition or event)".

[0090] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state drive), etc.

[0091] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM or random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A method for intelligent online monitoring of transformers, characterized in that, include: Acquire historical data of the target transformer within a preset historical time period, including historical operating data and corresponding historical working condition data. Based on the historical operating data, the historical data is divided into multiple operating intervals, and each operating interval has a numerical range corresponding to a certain operating parameter. Based on all historical operating data within the operating condition range, a dynamic reference range corresponding to the operating condition range is determined. The dynamic reference range is a numerical range with an upper limit value and a lower limit value. The current operating parameter values ​​and corresponding current operating condition parameter values ​​of the target transformer are obtained in real time. Based on the current operating condition parameter values, determine the target operating condition interval from the plurality of operating condition intervals; Obtain the target dynamic reference range corresponding to the target operating condition range; If the current operating parameter value is determined to be outside the target dynamic reference range, the transformer is identified as abnormal, and an early warning signal is generated. Acquire the output side temperature data, transformer body temperature data, and ambient temperature data of the target transformer; Based on the outgoing line temperature data, the transformer body temperature data, and the ambient temperature data, the outgoing line temperature rise and the transformer body temperature rise are calculated respectively. Calculate the first rate of change of the temperature rise on the outlet side and the second rate of change of the temperature rise on the body. The current composite thermal operating state of the target transformer is determined based on the temperature rise on the outgoing side, the temperature rise of the transformer body, the first rate of change, and the second rate of change. The historical data is categorized into low-temperature steady state, high-temperature steady state, and transition state to obtain the corresponding state historical data; For each type of historical data of the aforementioned state, the corresponding state condition range is determined; Based on the current composite thermal operating state of the target transformer, the corresponding state condition interval is selected as the plurality of operating condition intervals; When a change in the current composite thermal operating state is detected, the state condition interval before and after the switch is obtained, and the state switch reference range is calculated. During the switching of hot operating states, the state switching reference range is used as the dynamic reference range; When the fluctuation amplitude of the operating parameter value after switching is less than the preset stable threshold, the dynamic benchmark range of the anomaly monitoring will be updated to the state condition range corresponding to the hot operation state after switching.

2. The method according to claim 1, characterized in that, After determining that the transformer is abnormal and generating an early warning signal, the method further includes: In response to the generation of the warning signal, at least two pre-selected operating parameters are collected within a preset time window before the time of the warning signal generation, and corresponding time-series data are generated respectively. The similarity of each type of time series data with the corresponding parameter template sequence in each of the multiple preset fault feature templates is calculated to obtain the parameter similarity score corresponding to each fault feature template. One or more preset association rules in each fault feature template are applied to calculate the corresponding parameter similarity score to obtain the comprehensive similarity score for each fault feature template. The target with the highest overall similarity score among the overall similarity scores is determined, and the fault feature template corresponding to the target overall similarity score is determined as the best matching template; The anomaly type identified by the best matching template is determined, and the anomaly type is appended to the warning signal for output.

3. The method according to claim 2, characterized in that, The step of applying one or more preset association rules from each fault feature template to comprehensively calculate the corresponding parameter similarity scores, thereby obtaining a comprehensive similarity score for each fault feature template, specifically includes: For each fault feature template, a corresponding preset association rule is determined as a parameter weight, and the parameter weight corresponds one-to-one with the pre-selected operating parameters. The similarity score of the parameter corresponding to the pre-selected operating parameter is multiplied by the parameter weight corresponding to the pre-selected operating parameter to obtain the weighted score; The comprehensive similarity score corresponding to the fault feature template is obtained by summing all the values ​​in the weighted score.

4. The method according to claim 1, characterized in that, The step of determining that the transformer is abnormal and generating an early warning signal when the current operating parameter value is not within the target dynamic reference range specifically includes: When it is determined that the current operating parameter value is not within the target dynamic reference range, the moment corresponding to the current operating parameter value is defined as an initial limit violation event; In response to the occurrence of the initial out-of-limit event, a timer for a preset confirmation duration is started; Enable the over-limit count counter for accumulating the number of over-limit events, and set the initial value of the over-limit count counter to 1 to count the initial over-limit event; Within the preset confirmation time, the current operating parameter values ​​and their corresponding current working condition parameter values ​​are continuously acquired at a preset sampling period. Based on any of the subsequent current operating parameter values, the corresponding target dynamic reference range is determined according to the corresponding current operating condition parameter values. If it is determined whether any of the subsequent current operating parameter values ​​is outside the corresponding target dynamic reference range, the over-limit count counter is incremented by one; If the preset confirmation time expires and the final count value of the over-limit counter exceeds the preset threshold, the transformer is determined to be abnormal, and an early warning signal is generated.

5. The method according to claim 1, characterized in that, The step of determining the current composite thermal operating state of the target transformer based on the outgoing line temperature rise, the body temperature rise, the first rate of change, and the second rate of change specifically includes: When the absolute value of the first rate of change is less than the first preset rate threshold and the absolute value of the second rate of change is less than the second preset rate threshold, the transformer is determined to be in the thermal stability stage. During the thermal stability phase, if the temperature rise on the outgoing line side is less than a preset outgoing line temperature rise threshold and the temperature rise on the body is less than a preset body temperature rise threshold, the current composite thermal operating state is determined to be the low-temperature steady state. During the thermal stability phase, if the temperature rise on the outlet side is greater than or equal to the preset outlet temperature rise threshold, or the temperature rise on the body is greater than or equal to the preset body temperature rise threshold, the current composite thermal operating state is determined to be the high-temperature steady state. When the absolute value of the first rate of change is greater than or equal to the first preset rate threshold, or when the absolute value of the second rate of change is greater than or equal to the second preset rate threshold, the current composite thermal operating state is determined to be the transition state.

6. The method according to claim 1, characterized in that, For each operating mode's historical data, the corresponding operating condition range is determined, specifically including: The historical data is grouped according to the low-temperature steady state, the high-temperature steady state, and the transition state to obtain several groups of historical data; For each set of historical data, calculate the mean and standard deviation of the operating parameters; Based on the mean and the standard deviation, confidence intervals are constructed for different thermal operating states; The upper and lower boundaries of the confidence interval are used as the upper and lower limits of the corresponding state condition interval.

7. A transformer intelligent online monitoring system, characterized in that, The system includes: One or more processors and a memory; the memory is coupled to the one or more processors, the memory being used to store computer program code, the computer program code including computer instructions, the one or more processors invoking the computer instructions to cause the system to perform the method as described in any one of claims 1-6.

8. A computer-readable storage medium comprising instructions, characterized in that, When the instructions are executed on the system, the system performs the method as described in any one of claims 1-6.

9. A computer program product, characterized in that, When the computer program product is run on the system, the system performs the method as described in any one of claims 1-6.

Images