Monitoring method for oil-immersed power transformer

Abstract

The invention discloses an oil-immersed power transformer monitoring method, and relates to the field of power systems, and the method comprises the steps: collecting data, collecting historical operation data, and carrying out the preprocessing; performing feature extraction on the preprocessed sensor data and historical operation data to form a sensor data feature vector and a historical operation data feature vector; fusing the sensor data feature vector and the historical operation data feature vector to form a comprehensive feature data set; establishing a fault threshold based on the historical operation data; establishing a transformer fault prediction model based on the comprehensive feature data set, and outputting a transformer fault occurrence probability; automatically adjusting a fault threshold according to a transformer fault prediction result; generating an environment state report based on the current environment state; formulating and executing a dynamic compensation strategy according to the environment state report; according to the invention, early warning of potential faults of the transformer is realized. Preventive measures can be taken before the fault occurs, and the influence of the fault on the power grid is reduced.

Description

Technical Field

[0001] The invention relates to the field of power systems, in particular to a monitoring method for an oil-immersed power transformer. Background Art

[0002] As a key device in the power system, the stability and reliability of the operating status of oil-immersed power transformers have an important impact on the safety and efficiency of the entire power grid. With the continuous development of sensor technology, signal processing technology and machine learning algorithms, transformer monitoring methods have been significantly improved. Traditional transformer monitoring methods mainly rely on regular manual inspections and simple online monitoring systems. Although these methods can detect abnormal conditions of transformers to a certain extent, they have problems such as long response time and untimely fault detection.

[0003] Although existing transformer monitoring technology has made some progress, it still has some shortcomings. Traditional methods usually only focus on a single type of sensor data, resulting in low accuracy of fault diagnosis. Many existing fault prediction models rely on simple statistical methods or a single type of sensor data, which leads to limited prediction accuracy and difficulty in accurately identifying and predicting complex fault modes. Summary of the invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides an oil-immersed power transformer monitoring method to solve the problems of insufficient comprehensive analysis of multi-source data and inaccurate fault prediction in existing transformer monitoring methods.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, an embodiment of the present invention provides a method for monitoring an oil-immersed power transformer, which includes collecting data while collecting historical operation data and performing preprocessing;

[0008] The preprocessed sensor data and historical operation data are respectively subjected to feature extraction to form a sensor data feature vector and a historical operation data feature vector;

[0009] The sensor data feature vector and the historical operation data feature vector are fused to form a comprehensive feature data set;

[0010] Establish fault thresholds based on historical operating data;

[0011] A transformer fault prediction model is established based on the comprehensive feature data set to output the probability of transformer fault occurrence;

[0012] The fault threshold will be automatically adjusted according to the transformer fault prediction results;

[0013] Generate an environmental status report based on the current environmental status;

[0014] Develop and execute dynamic compensation strategies based on environmental status reports.

[0015] As a preferred solution of the oil-immersed power transformer monitoring method of the present invention, wherein: installing sensors to collect data and collect historical operation data at the same time, and pre-processing, including the following steps,

[0016] The collected data includes oil temperature data, winding temperature data, vibration data, abnormal noise data, dissolved gas composition data, transformer ambient temperature data, transformer ambient humidity data and transformer ambient air quality data;

[0017] The historical operation data refers to the historical operation data of the transformer;

[0018] Remove noise and smooth sensor data and historical operation data.

[0019] As a preferred solution of the oil-immersed power transformer monitoring method of the present invention, wherein: feature extraction is performed on the pre-processed sensor data and historical operation data to form a sensor data feature vector and a historical operation data feature vector, including the following steps:

[0020] Use Fourier transform to extract temperature, vibration, sound, and gas composition characteristic values ​​from sensor data, and extract voltage, current, vibration, noise, temperature, and humidity characteristic values ​​from historical operation data;

[0021] The various sensor feature values ​​extracted are formed into a sensor feature vector, and the feature values ​​extracted from the historical operation data are formed into a historical operation data feature vector.

[0022] As a preferred solution of the oil-immersed power transformer monitoring method of the present invention, the sensor data feature vector and the historical operation data feature vector are fused to form a comprehensive data set, including the following steps:

[0023] The sensor data feature vector and the historical operation data feature vector are fused to form a comprehensive feature data set.

[0024] As a preferred solution of the oil-immersed power transformer monitoring method of the present invention, wherein: a fault threshold is established based on historical operation data, a transformer fault prediction model is established based on a comprehensive feature data set, and the transformer fault occurrence probability is output, including the following steps:

[0025] Establishing a transformer fault threshold based on the historical operating data;

[0026] Initialize a weight matrix containing the number of historical operation data features and the total number of fault types;

[0027] Use cross entropy to measure the difference between the probability distribution predicted by the model and the actual faults;

[0028] The weight matrix is ​​iteratively optimized and the probability of transformer failure is generated based on the comprehensive feature dataset.

[0029] As a preferred solution of the oil-immersed power transformer monitoring method of the present invention, the fault threshold is automatically adjusted according to the transformer fault prediction result, including the following steps:

[0030] Automatically adjust the corresponding fault threshold according to the probability of transformer failure;

[0031] The monitoring system will continuously monitor the adjusted fault thresholds and record the time, reason, and new threshold settings of the threshold adjustments.

[0032] As a preferred solution of the oil-immersed power transformer monitoring method of the present invention, wherein: generating an environmental status report based on the current environmental status includes the following steps:

[0033] Assess the current state of the environment based on current environmental data;

[0034] The environmental status report includes current environmental parameters, potential risks and recommended measures.

[0035] As a preferred solution of the oil-immersed power transformer monitoring method of the present invention, wherein: formulating and executing a dynamic compensation strategy according to an environmental status report comprises the following steps:

[0036] Formulate dynamic compensation based on environmental status report and output dynamic compensation strategy;

[0037] According to the output dynamic compensation strategy, corresponding compensation measures are executed.

[0038] In a second aspect, an embodiment of the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the oil-immersed power transformer monitoring method as described in the first aspect of the present invention is implemented.

[0039] In a third aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the oil-immersed power transformer monitoring method as described in the first aspect of the present invention is implemented.

[0040] The beneficial effects of the present invention are as follows: by fusing sensor data and historical operation data, a comprehensive and integrated feature data set is formed. This not only retains the complementarity of real-time data and historical data, but also enhances the representativeness of the data. The comprehensive feature data set can more comprehensively reflect the operating status of the transformer, thereby improving the accuracy and generalization ability of the fault prediction model. A transformer fault prediction model is established based on the comprehensive feature data set, and the probability of transformer failure is output, which realizes early warning of potential transformer failures and early warning of potential failures. This helps to take preventive measures before a failure occurs and reduce the impact of the failure on the power grid. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0042] Figure 1 This is a flow chart of the oil-immersed power transformer monitoring method in Example 1.

[0043] Figure 2 This is the decision diagram of the transformer fault prediction model in Example 1. DETAILED DESCRIPTION

[0044] Example 1, reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, provides a method for monitoring an oil-immersed power transformer, comprising the following steps:

[0045] S1. Temperature sensors, vibration sensors, sound sensors, gas analysis sensors, temperature sensors, humidity sensors and air quality sensors are installed. Temperature sensors are installed at key locations on the transformer casing and inside to monitor oil temperature and winding temperature; vibration sensors are installed on the transformer body to monitor vibration conditions; sound sensors are installed near the transformer to monitor abnormal noise; gas analysis sensors are installed on the top of the oil tank to monitor dissolved gas composition; temperature sensors, humidity sensors and air quality sensors are installed around the transformer to collect environmental data around the transformer. These sensors should be distributed in multiple locations to ensure coverage of the entire working area.

[0046] S1.1. The historical operating data refers to the electrical parameters, mechanical parameters and environmental parameters of the transformer.

[0047] It is further explained that by collecting various types of data, the status of the transformer and its surrounding environment can be fully reflected, and the accuracy of fault diagnosis can be improved.

[0048] S1.2. Remove noise and smooth the sensor data and historical operation data.

[0049] The noise removal refers to removing high frequency noise in the sensor data by using a digital filter.

[0050] The smoothing process refers to smoothing the data using a smoothing algorithm to reduce the impact of short-term fluctuations and make the data more stable. Smoothing can better reflect the true trend of the data.

[0051] It is further shown that noise removal and smoothing can significantly improve the quality of data, reduce false positives and false negatives, and improve the reliability of fault detection.

[0052] S2. Calculate the mean and standard deviation of all sensor data and historical operation data, weight them, normalize the weighted summation result, and obtain the sensor data feature vector and the historical operation data feature vector, which are expressed as:

[0053]

[0054] Among them, v s represents the sensor data feature vector, v h represents the feature vector of historical operation data, n represents the number of sensor data, x t represents the original value of the tth sensor data, z st represents the characteristic value of the t-th sensor data, μ s represents the mean of the sensor data, σ s represents the standard deviation of the sensor data, z hl represents the characteristic value of the lth historical running data, m represents the number of historical running data, y l represents the original value of the lth historical running data, μ h represents the mean of historical running data, σ h Indicates the standard deviation of historical running data.

[0055] It is further explained that the unified format and efficient processing flow simplify the data processing work, and the high-quality feature vectors enhance the training effect and generalization ability of the model.

[0056] S3. Add the corresponding elements of the sensor data feature vector and the historical operation data feature vector to obtain a new vector v s +v h ;

[0057] Calculate the new vector v s +v h The Euclidean norm of ;

[0058] The new vector v s +v h Divide by its norm to get the normalized comprehensive feature data, the expression is:

[0059]

[0060] Where D represents the comprehensive feature dataset, v s represents the sensor data feature vector, v h represents the feature vector of historical operation data, p represents the dimension of the feature vector, v sk represents the kth element of the sensor data feature vector, v hk Represents the kth element of the historical operation data feature vector.

[0061] It is further explained that by fusing sensor data and historical operation data, a comprehensive feature dataset containing multi-source information is formed. This helps to more comprehensively reflect the current status and historical operation of the transformer.

[0062] S4. By analyzing historical operation data, determine the normal and abnormal ranges of various fault types, and establish fault thresholds. These thresholds are used to determine whether the current operation status exceeds the normal range.

[0063] For example, suppose the focus is on transformer oil temperature overtemperature faults.

[0064] The oil temperature data of the transformer was collected over the past year, recorded once a day, for a total of 365 data points.

[0065] Perform statistical analysis on historical data and calculate the mean and standard deviation of oil temperature data.

[0066] Assume that the mean of the oil temperature data obtained is 70°C and the standard deviation is 5°C.

[0067] Set the initial fault threshold. When the oil temperature exceeds the mean plus two standard deviations, there may be a fault risk.

[0068] Therefore, the initial failure threshold can be set to 70°C+2×5°C=80°C.

[0069] It is further explained that through historical data analysis, the fault threshold can be determined more accurately and the false positives and negatives can be reduced.

[0070] S4.1. Determine the number of historical operation data features and the total number of fault types, and initialize a weight matrix whose dimension is the total number of fault types multiplied by the number of historical operation data features.

[0071] Each row vector corresponds to the weight of a fault type.

[0072] Initialize the bias term, whose dimension is the total number of fault types multiplied by 1.

[0073] Assume that there are 4 types of faults (overheating, abnormal vibration, gas leakage, electrical fault), and the number of features corresponding to each fault type is 5. Then the weight matrix is ​​initialized to a matrix with a dimension of 4×5; the same is true for the bias term.

[0074] It is further shown that reasonable initialization can speed up the convergence of the model.

[0075] S4.1.1. The cross entropy loss function is used to measure the difference between the probability distribution predicted by the model and the actual fault label, thereby guiding the optimization of the model parameters. Its expression is:

[0076]

[0077] Among them, y r represents the true fault label, and P(y=r|D) is the fault probability predicted by the model.

[0078] Assume that there are four fault types (overheating, abnormal vibration, gas leakage, electrical fault), and the true fault type of the current sample is "overheating".

[0079] True fault label y=1000;

[0080] Assume that the probability distribution predicted by the model is P(y=1|D)=0.8, P(y=2|D)=1, P(y=3|D)=0.05, P(y=4|D)=0.05.

[0081] Substitute and calculate Cross-Entropy Loss = -log(0.8)≈0.223.

[0082] It is further shown that the cross entropy loss function can effectively measure the difference between the probability distribution predicted by the model and the true fault label, and guide the optimization of the model parameters.

[0083] S4.2. Through optimization algorithms such as gradient descent, the weight matrix and bias terms are continuously adjusted to make the probability distribution predicted by the model as close as possible to the actual fault label, and the probability of transformer failure is generated based on the comprehensive feature data set, which is expressed as:

[0084]

[0085] Where P(y=c|D) represents the probability of transformer fault type c when given a comprehensive feature data set D, y represents the category of fault type, c represents the category of specific fault type, W c represents the weight vector corresponding to the fault type c, b crepresents the bias term corresponding to fault type c, W i represents the weight vector corresponding to the i-th fault type, b i represents the bias term corresponding to the i-th fault type, and a represents the total number of fault types.

[0086] It is further shown that by iteratively optimizing the weight matrix and bias term, the model can gradually improve the accuracy of prediction and enhance the reliability of the system. These steps work together to not only improve the overall performance of transformer monitoring, but also provide important support for the long-term stable operation of the power system.

[0087] S5. Calculate the probability of each failure type according to the model;

[0088] S5.1. Set corresponding threshold adjustment rules according to the magnitude of the fault probability. For example, if the probability of a certain fault type exceeds a preset threshold, it is considered that the risk of the fault type is high and the corresponding fault threshold needs to be adjusted.

[0089] S5.1.1. Automatically adjust the corresponding fault threshold according to the adjustment rules. For example, if the probability of "overheating" fault is high, the temperature threshold can be lowered to trigger the alarm earlier.

[0090] For example, suppose there are 4 types of faults (overheating, abnormal vibration, gas leakage, electrical fault), and the probability of each type of fault has been calculated by the model. For example, P(y=1|D)=0.85, P(y=2|D)=0.05, P(y=3|D)=0.07, P(y=4|D)=0.03;

[0091] S5.1.2. Set adjustment rules. If the probability of a certain fault type exceeds 0.8, adjust the corresponding fault threshold. For example, for an "overheating" fault, the initial temperature threshold is 80°C.

[0092] S5.1.3. Since the probability of an “overheating” fault is 0.85, which exceeds 0.8, the temperature threshold is adjusted from 80°C to 75°C to detect overheating earlier.

[0093] S5.2. The monitoring system will continuously monitor the operating status of the transformer and use the adjusted fault threshold to make judgments.

[0094] S5.2.1. Each time a threshold is adjusted, the system will record the specific timestamp of the threshold adjustment, the reason for the adjustment, such as the failure probability exceeding the preset threshold, and the new threshold after adjustment.

[0095] For example, the monitoring system continuously monitors the oil temperature data of the transformer.

[0096] Adjusted time: 2024-09-29 10:00:00;

[0097] Adjustment reason: "overheating" failure probability exceeds 0.8;

[0098] New threshold, the temperature threshold is adjusted from 80℃ to 75℃.

[0099] Further explanation: Dynamically adjusting the fault threshold according to the fault probability enables the system to better adapt to different operating environments and working conditions, improving the flexibility and adaptability of the system. Dynamically adjusting the threshold can detect potential fault risks earlier, reduce false positives and false negatives, and improve the accuracy of fault detection. Continuously monitor the adjusted fault threshold to ensure that the system can respond to the new threshold setting in real time, and record the time, reason and new threshold setting of the threshold adjustment for subsequent analysis and auditing.

[0100] S6. Analyze temperature, humidity, and air quality data to assess the current environmental status of the transformer;

[0101] S6.1 Generate an environmental status report based on the current environmental status, such as current environmental parameters, potential risks, and recommended measures.

[0102] S6.1.1. The current environmental parameters refer to the current working environment of the transformer, such as the current working environment temperature of the transformer is 35.2°C, the current working environment humidity of the transformer is 79.5%, and the current working environment PM2.5 is 48μg / m 3 and CO2 concentration of 595ppm.

[0103] S6.1.2. The potential risk is that if the temperature of the current transformer working environment is too high, it may cause poor heat dissipation of the transformer and increase the risk of overheating; if the humidity of the current transformer working environment is too high, it may cause condensation on the surface of electrical equipment and affect the insulation performance; if the air quality of the current transformer working environment is very poor, it will have an adverse effect on human health and equipment operation.

[0104] It is further explained that through environmental status assessment, potential environmental risks can be discovered in advance, supporting preventive maintenance strategies and reducing the impact of sudden failures on the system.

[0105] S6.1.3, The recommended measures are to start the cooling system to reduce the ambient temperature around the transformer. Turn on the dehumidification equipment to reduce the relative humidity. Strengthen ventilation to improve air quality. Regularly inspect and maintain the equipment to ensure its normal operation in harsh environments.

[0106] Further explanation: Generate a detailed environmental status report to record current environmental parameters, potential risks and recommended measures to facilitate subsequent analysis and audit. The environmental status report provides an important basis for decision-making for operation and maintenance personnel, and supports scientific management and optimization of operation and maintenance strategies.

[0107] S7. Setting an adjustment parameter γ for controlling the intensity of the exponential operation and an adjustment parameter δ for controlling the intensity of the logarithmic operation.

[0108] S7.1 calculates the impact of the environmental status report on the compensation strategy, then formulates dynamic compensation based on the environmental status report and outputs the dynamic compensation strategy, which is expressed as:

[0109]

[0110] Where C represents the dynamic compensation strategy, γ represents the adjustment parameter used to control the intensity of the exponential operation, E represents the environmental status report, m represents the number of compensation strategies, δ represents the adjustment parameter used to control the intensity of the logarithmic operation, and ∈ represents a constant;

[0111] S7.1.1. The dynamic compensation strategy includes the type, intensity, priority, and duration of compensation measures.

[0112] S7.1.2 The types of compensation measures include cooling system adjustment, such as increasing or decreasing the operating intensity of the cooling system. Dehumidification equipment startup, such as starting or increasing the operation of the dehumidification equipment. Ventilation system optimization, such as increasing or decreasing the air volume of the ventilation system. Other related measures, such as adjusting the workload of the transformer and changing the operating mode.

[0113] S7.1.3 The intensity of the compensatory measures shall refer to the specific operating power and frequency set for the cooling system, dehumidification equipment, ventilation system and other related measures.

[0114] S7.1.4, the priority of the compensation measures refers to the classification of the measures into high priority measures, medium priority measures and low priority measures;

[0115] The high-priority measures are critical measures that need to be implemented immediately, such as emergency cooling.

[0116] The medium priority measures are important measures that need to be implemented as soon as possible, such as increasing ventilation.

[0117] The low priority measures are general measures that can be performed later, such as scheduled maintenance.

[0118] S7.1.5 The duration of the compensatory measures refers to short-term measures and long-term measures.

[0119] The short-term measures are temporary measures that are only implemented within a specific period of time, such as high-intensity cooling within a short period of time.

[0120] The long-term measures refer to long-term measures that need to be implemented continuously for a period of time, such as long-term low-intensity ventilation.

[0121] It is further explained that dynamically adjusting the compensation strategy according to the current environmental status report enables the system to better adapt to different environmental conditions, improving the flexibility and adaptability of the system. Through the dynamic compensation strategy, the operating status of the transformer can be optimized and the risk of failure caused by environmental factors can be reduced.

[0122] This embodiment also provides a computer device, which is applicable to the case of an oil-immersed power transformer monitoring method, including: a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the oil-immersed power transformer monitoring method proposed in the above embodiment.

[0123] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0124] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the oil-immersed power transformer monitoring method proposed in the above embodiment is implemented; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, referred to as EPROM), programmable read-only memory (Programmable Red-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0125] In summary, the present invention forms a comprehensive and integrated feature data set by fusing sensor data and historical operation data. This not only retains the complementarity of real-time data and historical data, but also enhances the representativeness of the data. The comprehensive feature data set can more comprehensively reflect the operating status of the transformer, thereby improving the accuracy and generalization ability of the fault prediction model. A transformer fault prediction model is established based on the comprehensive feature data set, and the probability of transformer failure is output, thereby achieving early warning of potential transformer failures. This helps to take preventive measures before a failure occurs and reduce the impact of the failure on the power grid.

[0126] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for monitoring an oil-immersed power transformer, characterized in that: include, While collecting data, historical operation data is collected and pre-processed; The preprocessed sensor data and historical operation data are respectively subjected to feature extraction to form a sensor data feature vector and a historical operation data feature vector; The sensor data feature vector and the historical operation data feature vector are fused to form a comprehensive feature data set; Establish fault thresholds based on historical operating data; A transformer fault prediction model is established based on the comprehensive feature data set to output the probability of transformer fault occurrence; The fault threshold will be automatically adjusted according to the transformer fault prediction results; Generate an environmental status report based on the current environmental status; Develop and execute dynamic compensation strategies based on environmental status reports.

2. The oil-immersed power transformer monitoring method according to claim 1, characterized in that: Install sensors to collect data and collect historical operation data, and perform preprocessing, including the following steps: The collected data includes oil temperature data, winding temperature data, vibration data, abnormal noise data, dissolved gas composition data, transformer ambient temperature data, transformer ambient humidity data and transformer ambient air quality data; The historical operation data refers to the historical operation data of the transformer; Remove noise and smooth sensor data and historical operation data.

3. The oil-immersed power transformer monitoring method according to claim 2, characterized in that: The pre-processed sensor data and historical operation data are subjected to feature extraction to form sensor data feature vectors and historical operation data feature vectors. The following steps are included: Use Fourier transform to extract temperature, vibration, sound, and gas composition characteristic values ​​from sensor data, and extract voltage, current, vibration, noise, temperature, and humidity characteristic values ​​from historical operation data; The various sensor feature values ​​extracted are formed into a sensor feature vector, and the feature values ​​extracted from the historical operation data are formed into a historical operation data feature vector.

4. The oil-immersed power transformer monitoring method according to claim 3, characterized in that: The sensor data feature vector and the historical operation data feature vector are fused to form a comprehensive data set, including the following steps: The sensor data feature vector and the historical operation data feature vector are fused to form a comprehensive feature data set.

5. The oil-immersed power transformer monitoring method according to claim 4, characterized in that: Establish fault thresholds based on historical operation data, establish transformer fault prediction models based on comprehensive feature data sets, and output the probability of transformer fault occurrence. The following steps are included: Establishing a transformer fault threshold based on the historical operating data; Initialize a weight matrix containing the number of historical operation data features and the total number of fault types; Use cross entropy to measure the difference between the probability distribution predicted by the model and the actual faults; The weight matrix is ​​iteratively optimized and the probability of transformer failure is generated based on the comprehensive feature dataset.

6. The oil-immersed power transformer monitoring method according to claim 5, characterized in that: The fault threshold will be automatically adjusted according to the transformer fault prediction results, including the following steps: Automatically adjust the corresponding fault threshold according to the probability of transformer failure; The monitoring system will continuously monitor the adjusted fault thresholds and record the time, reason, and new threshold settings of the threshold adjustments.

7. The oil-immersed power transformer monitoring method according to claim 6, characterized in that: Generate an environmental status report based on the current environmental status, including the following steps: Assess the current state of the environment based on current environmental data; The environmental status report includes current environmental parameters, potential risks and recommended measures.

8. The oil-immersed power transformer monitoring method according to claim 7, characterized in that: Develop and implement dynamic compensation strategies based on environmental status reports, including the following steps: Formulate dynamic compensation based on environmental status report and output dynamic compensation strategy; According to the output dynamic compensation strategy, corresponding compensation measures are executed.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the oil-immersed power transformer monitoring method according to any one of claims 1 to 8 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the oil-immersed power transformer monitoring method according to any one of claims 1 to 8 are implemented.

Images