Method for identifying transformer oil gas fault based on data discretization and normalization algorithm

Abstract

This invention relates to the field of online monitoring and fault identification technology for transformers and reactors, specifically a method for identifying gas faults in transformer oil based on data discretization and normalization. The method includes the following steps: obtaining characteristic quantities of dissolved gases in transformer oil; calculating normalized energy intensity using these characteristic quantities; performing fuzzy discretization and normalization on the characteristic quantities and normalized energy intensity to obtain characteristic parameter one; standardizing the ratios of different dissolved gas characteristic quantities using the three-ratio method to obtain characteristic parameter two; and constructing a fault diagnosis model based on characteristic parameters one and two to identify gas fault types. This method improves the speed, accuracy, and precision of transformer fault detection, enabling accurate judgment of transformer faults. Applying it to transformer monitoring systems can more effectively detect early latent faults in transformers, increasing the value of the monitoring system, reducing on-site maintenance work and costs, and improving the safety of maintenance work.

Description

Technical Field

[0001] This invention relates to the field of online monitoring and fault identification technology for transformers and reactors, and in particular to a method for identifying gas faults in transformer oil based on data discretization and normalization. Background Technology

[0002] As a pivotal piece of equipment in the power system, the operating status of power transformers directly affects the safety and stability of the power system. With the rapid development of ultra-high voltage and extra-high voltage transmission and transformation technologies in my country, the power grid has a large capacity and wide coverage. Transformer faults can cause significant harm and impact on the power system and users. Because the development process of power transformer faults is directly related to the operating environment and load conditions, it is often difficult to detect these faults in a timely manner using regular maintenance methods. Moreover, regular maintenance needs to be carried out offline, resulting in long power outages and significant power losses. Therefore, timely and accurate detection of early latent faults in transformers is of paramount importance.

[0003] Most transformers in China use a combination of oil and paper insulation, which can experience various types of faults during daily operation, such as abnormal operating temperature, arc discharge, partial discharge, and reduced insulation. When an internal fault occurs, the insulating oil decomposes, producing various fault gases. Real-time monitoring of these characteristic gases can help detect latent transformer faults early. However, existing threshold-based methods are insufficient for accurately analyzing the true nature of transformer faults. Consequently, transformer damage or personal injury due to delayed fault detection are frequent occurrences. Therefore, rapid and accurate fault identification of transformers is of paramount importance.

[0004] Analyzing the composition and concentration of fault gases can help determine the approximate location and severity of the fault. Currently, power transformer fault analysis primarily relies on threshold judgments based on national standards, supplemented by methods such as the modified three-ratio method, David's triangle, and cube diagrams. However, these traditional methods have limited accuracy, and some are insensitive to oil chromatography gas data, failing to provide accurate diagnoses. To effectively improve the accuracy of fault diagnosis and identification in transformer oil, scholars have conducted research on transformer fault identification using statistical analysis and machine learning methods, but the results have been limited. Discrete-normalized identification methods, on the other hand, can quickly identify the characteristic information of samples and accurately determine faults. They can effectively analyze the current operating status of transformers and provide rapid and accurate fault type identification when gas content exceeds the standard. Summary of the Invention

[0005] This invention proposes a method for identifying gas faults in transformer oil based on a data discretization and normalization algorithm. This method can quickly identify the characteristic information of samples and accurately diagnose faults. First, the value of the characteristic quantity NEI is calculated using a formula. Then, a membership function is used for normalization, transforming continuous data into discretized data. A modified three-ratio method is then used for standardization. The standardized result is input into a fault diagnosis table to obtain the diagnostic result, improving the speed and accuracy of transformer fault detection, thereby reducing on-site maintenance work and costs, and improving the safety of maintenance operations.

[0006] The technical solution adopted by the present invention to achieve the above objectives is as follows:

[0007] A method for identifying gas faults in transformer oil based on a data discretization and normalization algorithm includes the following steps:

[0008] To obtain the characteristic quantities of dissolved gases in transformer oil;

[0009] Calculate the normalized energy intensity using characteristic quantities;

[0010] The feature quantity and the normalized energy intensity are fuzzily discretized and normalized to obtain feature parameter one;

[0011] The ratios of different dissolved gas characteristic quantities were standardized using the three-ratio method to obtain characteristic parameter two.

[0012] A fault diagnosis model is constructed based on feature parameters one and two to identify gas fault types.

[0013] Use variable x i The characteristic quantities of different dissolved gases are represented as follows: Acetylene (C₂H₂) characteristic quantity is represented by variable x1; ethylene (C₂H₄) characteristic quantity is represented by variable x2; methane (CH₄) characteristic quantity is represented by variable x3; hydrogen (H₂) characteristic quantity is represented by variable x4; ethane (C₂H₆) characteristic quantity is represented by variable x5; carbon monoxide (CO) characteristic quantity is represented by variable x6; carbon dioxide (CO₂) characteristic quantity is represented by variable x7; total hydrocarbon characteristic quantity is represented by variable x8; the ratio of acetylene (C₂H₂) characteristic quantity to ethylene (C₂H₄) characteristic quantity is represented by variable x9; and the ratio of methane (CH₄) characteristic quantity to hydrogen (H₂) characteristic quantity is represented by variable x. 10 The ratio of the characteristic amount of ethylene (C2H4) to that of ethane (C2H6) is expressed by the variable x. 11 This indicates that the normalized energy intensity (NEI) characteristic is represented by the variable x. 12 express.

[0014] The calculation of the normalized energy intensity is specifically as follows:

[0015]

[0016] in, This represents the concentration of four hydrocarbon gases after degassing at 20°C, i.e., the characteristic quantity, where T is the temperature.

[0017] Use membership functions to evaluate x1 to x8, x 12 The fuzzy discretization and normalization process is performed as follows:

[0018]

[0019] Where A(x) represents the membership function value of the characteristic parameter attribute as normal. As the first characteristic parameter, x represents the characteristic quantity of dissolved gas; a1 represents 90% of the threshold of the characteristic quantity of dissolved gas, and a2 represents 110% of the threshold of the characteristic quantity of dissolved gas.

[0020] Using the three ratio method to analyze x9~x 11 The standardization process involves: determining the code for the ratio range S based on the ratio range of the gas characteristic quantities, which serves as characteristic parameter two.

[0021] When S < 0.1, the encoded value of x9 is 0, x 10 The encoded value is 1, x 11 The encoded value is 0;

[0022] When 0.1 ≤ S < 1, the encoded value of x9 is 1, x 10 The encoded value is 0, x 11 The encoded value is 0;

[0023] When 1 ≤ S < 3, the encoded value of x9 is 1, x 10 The encoded value is 2, x 11 The encoded value is 1;

[0024] When S≥3, the encoded value of x9 is 2, x 10 The encoded value is 2, x 11 The encoded value is 2.

[0025] The fault diagnosis model is x1~x 12 The mapping relationship between the value, fault type, and confidence level is as follows:

[0026] When x1 = 0, x6 = 0, x8 = 0, x3 = 0, and x4 = 0, the operation is normal.

[0027] When x1 = 0, x6 = 0, x7 = 0, x3 = 0, and x4 = 0, the operation is normal.

[0028] When x1 = 0, x5 = 1, and x3 = 0, the fault type is medium-temperature overheating.

[0029] When x9 = 0 and x 11=1 and x2=1, at this time the fault type is medium temperature overheating;

[0030] When x9 = 0 and x4 = 0 and x 12 =1 and x 11 =0, at this time the fault type is high temperature overheating;

[0031] When x9 = 0 and x7 = 0 and x 12 =1 and x6=1, at this time the fault type is high temperature overheating;

[0032] When x9 = 2, x4 = 1, x2 = 0, and x1 = 1, the fault type is low-energy discharge.

[0033] When x9 = 2 and x 10 =1 and x5=0 and x3=0, at this time the fault type is low-energy discharge;

[0034] When x 10 =0 and x8=0 and x7=1, at this time the fault type is high-energy discharge;

[0035] When x9 = 1, x4 = 0, and x7 = 1, the fault type is high-energy discharge.

[0036] When x9 = 0 and x 11 =0 and x1=0 and x6=0 and x 12 =1, at this time the fault type is low energy discharge and overheating;

[0037] When x9 = 0, x7 = 1, and x5 = 0, the fault type is low-energy discharge combined with overheating.

[0038] A transformer oil gas fault identification system based on a data discretization and normalization algorithm includes a memory and a processor; the memory is used to store a computer program; the processor is used to implement the transformer oil gas fault identification method based on the data discretization and normalization algorithm when the computer program is executed.

[0039] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the method for identifying gas faults in transformer oil based on a data discretization and normalization algorithm.

[0040] The present invention has the following beneficial effects and advantages:

[0041] This paper presents a method for identifying gas faults in transformer oil based on a data discretization and normalization algorithm. This method improves the speed and accuracy of transformer fault detection, enabling precise judgment of transformer faults. Applying it to transformer monitoring systems can more effectively detect early-stage latent faults, increasing the value of the monitoring system, reducing on-site maintenance work and costs, and improving the safety of maintenance operations. Attached Figure Description

[0042] Figure 1 This is a flowchart of a method for identifying gas faults in transformer oil based on a data discretization and normalization algorithm. Detailed Implementation

[0043] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.

[0044] A method for identifying gas faults in transformer oil based on a data discretization and normalization algorithm is presented. The implementation process is as follows: The discretization and normalization identification method can quickly identify the characteristic information of samples and accurately diagnose faults. First, the value of the feature quantity NEI is calculated using a formula. Then, a membership function is used for normalization, transforming continuous data into discretized data. A modified three-ratio method is then used for standardization. The standardized result is input into a fault diagnosis table to obtain the diagnostic result, satisfying the requirements for both speed and accuracy in transformer fault detection.

[0045] like Figure 1 As shown, a method for identifying gas faults in transformer oil based on a data discretization and normalization algorithm is presented. The technical solution mainly includes the following steps:

[0046] Step 1: Extract feature quantities and determine the transformer fault type based on the feature quantities.

[0047] Step 2: Build a normalized energy intensity calculation model.

[0048] Step 3: Use membership functions to evaluate x1 to x8, x 12 Fuzzy discretization and normalization processing.

[0049] Step 4: Based on the gas ratio range, the improved three-ratio method is used to determine the ratio range code and standardize it to obtain specific coding rules.

[0050] Step 5: Build the model, extract data, and finally obtain the fault diagnosis model. Based on the values ​​of different feature parameters, determine the final fault type and obtain the corresponding confidence level.

[0051] Step 1: Feature Extraction

[0052] The dissolved gases in transformer oil mainly include acetylene (C2H2), ethylene (C2H4), methane (CH4), hydrogen (H2), ethane (C2H6), carbon monoxide (CO), and carbon dioxide (CO2). The characteristic amounts of acetylene (C2H2) are represented by variable x1, ethylene (C2H4) by variable x2, methane (CH4) by variable x3, hydrogen (H2) by variable x4, ethane (C2H6) by variable x5, carbon monoxide (CO) by variable x6, carbon dioxide (CO2) by variable x7, total hydrocarbons by variable x8, the ratio of acetylene (C2H2) characteristic amounts to ethylene (C2H4) characteristic amounts by variable x9, and the ratio of methane (CH4) characteristic amounts to hydrogen (H2) characteristic amounts by variable x1. 10 The ratio of the characteristic properties of ethylene (C2H4) to those of ethane (C2H6) is expressed by the variable x. 11 In other words, the NEI feature quantity is represented by the variable x. 12 The details are shown in Table 1.

[0053] Table 1 Transformer Fault Characteristic Quantities

[0054]

[0055] The typical transformer faults currently include low-temperature overheating, medium-temperature overheating, high-temperature overheating, low-energy discharge, high-energy discharge, and low-energy discharge combined with overheating. By building a mathematical model, let variable y1 represent low-temperature overheating, variable y2 represent medium-temperature overheating, variable y3 represent high-temperature overheating, variable y4 represent low-energy discharge, variable y5 represent high-energy discharge, and variable y6 represent high-energy discharge, as shown in Table 2.

[0056] Table 2 Typical Transformer Faults

[0057]

[0058] Step 2: Building the NEI (Normalized Energy Intensity) Calculation Model

[0059] After analyzing the transformer parameters, the normalized energy intensity model of this design includes four hydrocarbon gases: methane, ethane, ethylene, and acetylene. The model establishment formula is shown in formula (1):

[0060]

[0061] In formula (1): φ represents the concentration (μL / L) of the four hydrocarbon gases after degassing at 20℃, measured by chromatographic analysis; the denominator 22400 is the algebraic simplification adopted to unify the units of each gas and its corresponding standard enthalpy of formation. If the gas concentration is not measured at 273K (0℃), it must be multiplied by a factor of 273 / (273+T) for temperature correction before calculating NEI.

[0062] Step 3: x1~x8, x 12 Fuzzy discretization and normalization processing

[0063] Normalization is performed using a membership function. Data within the normal threshold range will be mapped to "0", while data exceeding the threshold will be mapped to "1".

[0064] Fuzzy set theory is introduced to transform continuous data into discrete data through membership functions. Considering that in actual transformer oil chromatographic analysis, only values ​​within a small interval centered on the normal gas value are controversial and require fuzzy processing, while values ​​outside this interval can be binarized, the trapezoidal function is usually chosen as the membership function, as shown in Formula 2.

[0065]

[0066] In Formula 2, A(x) is the membership function value of the characteristic parameter attribute "normal"; x is the concentration of eight characteristic gases (x1 to x8) in μL / L; a1 is 90% μL / L of the normal threshold of the corresponding characteristic gas; and a2 is 110% μL / L of the normal threshold of the corresponding characteristic gas.

[0067] Table 4 Characteristic Gas Thresholds for Transformers (μL / L)

[0068]

[0069] Step 4: Standardize the modified three-ratio method.

[0070] Based on the gas ratio range, the improved three-ratio method is used to determine the ratio range code for standardization. The specific coding rules are shown in Table 5.

[0071] Table 5. Encoding Rules for the Improved Three-Ratio Method

[0072]

[0073]

[0074] Step 5: Establish a fault diagnosis model

[0075] Through the model construction and data extraction described above, a fault diagnosis model is finally obtained. The fault diagnosis model includes feature parameters, fault types, and confidence levels. Based on the values ​​of different feature parameters, the final fault type is determined, and the corresponding confidence level is obtained. See Table 6 for details.

[0076] Table 6. Fuzzy Association Rules for Some Transformer Fault Types

[0077]

Claims

1. A method for identifying gas faults in transformer oil based on a data discretization and normalization algorithm, characterized in that, Includes the following steps: To obtain the characteristic quantities of dissolved gases in transformer oil; Calculate the normalized energy intensity using characteristic quantities; The feature quantity and the normalized energy intensity are fuzzily discretized and normalized to obtain feature parameter one; The ratios of different dissolved gas characteristic quantities were standardized using the three-ratio method to obtain characteristic parameter two. A fault diagnosis model is constructed based on feature parameters one and two to identify gas fault types. Use variables x i Characteristic quantities representing different dissolved gases, specifically: characteristic quantities of acetylene (C₂H₂) are represented by variables. x 1 indicates that the characteristic quantities of ethylene C2H4 are represented by variables. x 2 indicates that the characteristic quantities of methane (CH4) are represented by variables. x 3 indicates that the characteristic quantity of hydrogen (H2) is represented by a variable. x 4 indicates that the characteristic quantities of ethane (C2H6) are represented by variables. x 5 indicates that the characteristic quantity of carbon monoxide (CO) is represented by a variable. x 6 indicates that the characteristic quantity of carbon dioxide (CO2) is represented by a variable. x 7 indicates that the total hydrocarbon characteristic quantity is represented by a variable. x 8 indicates that the ratio of the characteristic amount of acetylene (C2H2) to that of ethylene (C2H4) is represented by a variable. x 9 indicates that the ratio of the characteristic amount of methane (CH4) to the characteristic amount of hydrogen (H2) is represented by a variable. x 10 The ratio of the characteristic amount of ethylene (C2H4) to that of ethane (C2H6) is expressed as a variable. x 11 This indicates that the normalized energy intensity (NEI) characteristic is represented by variables. x 12 express; Use membership function to x 1~ x 8. x 12 The fuzzy discretization and normalization process is performed as follows: ； in, The characteristic parameter attribute is represented by a normal membership function value. As characteristic parameter one, x represents the characteristic quantity of dissolved gas; a1 represents 90% of the threshold of dissolved gas characteristic quantity, and a2 represents 110% of the threshold of dissolved gas characteristic quantity. The membership function is used for normalization. Data within the normal threshold range will be mapped to "0", and data exceeding the threshold will be mapped to "1". Using the three ratio method x 9~ x 11 The standardization process involves: determining the code for the ratio range S based on the ratio range of the gas characteristic quantities, which serves as characteristic parameter two. When S < 0.1, x The code value for 9 is 0. x 10 The encoded value is 1. x 11 The encoded value is 0; When 0.1 ≤ S < 1, x The code value for 9 is 1. x 10 The encoded value is 0. x 11 The encoded value is 0; When 1≤S<3 x The code value for 9 is 1. x 10 The encoded value is 2. x 11 The encoded value is 1; When S≥3 x The code value for 9 is 2. x 10 The encoded value is 2. x 11 The encoded value is 2; The fault diagnosis model is as follows: x 1~ x 12 The mapping relationship between the value, fault type, and confidence level is as follows: when x 1=0 and x 6=0 and x 8=0 and x 3=0 and x 4=0, indicating normal operation. when x 1=0 and x 6=0 and x 7=0 and x 3=0 and x 4=0, indicating normal operation. when x 1=0 and x 5=1 and x 3=0, at this time the fault type is medium temperature overheating; when x 9=0 and x 11 =1 and x 2=1, at this time the fault type is medium temperature overheating; when x 9=0 and x 4=0 and x 12 =1 and x 11 =0, at this time the fault type is high temperature overheating; when x 9=0 and x 7=0 and x 12 =1 and x 6=1, at this time the fault type is high temperature overheating; when x 9=2 and x 4=1 and x 2=0 and x 1=1, at this time the fault type is low energy discharge; when x 9=2 and x 10 =1 and x 5=0 and x 3=0, at this time the fault type is low energy discharge; when x 10 =0 and x 8=0 and x 7=1, at this time the fault type is high-energy discharge; when x 9 = 1 and x 4=0 and x 7=1, at this time the fault type is high-energy discharge; when x 9 = 0 and x 11 =0 and x 1=0 and x 6=0 and x 12 =1, at this time the fault type is low energy discharge and overheating; when x 9 = 0 and x 7 = 1 and x 5=0, at this time the fault type is low energy discharge and overheating.

2. The method for identifying gas faults in transformer oil based on a data discretization and normalization algorithm according to claim 1, characterized in that, The calculation of the normalized energy intensity is specifically as follows: ； in, This represents the concentration of four hydrocarbon gases after degassing at 20°C, i.e., the characteristic quantity, where T is the temperature.

3. A transformer oil gas fault identification system based on a data discretization and normalization algorithm, characterized in that, It includes a memory and a processor; the memory is used to store a computer program; the processor is used to implement, when executing the computer program, the method for identifying gas faults in transformer oil based on the data discretization and normalization algorithm as described in any one of claims 1-2.

4. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the method for identifying gas faults in transformer oil based on a data discretization and normalization algorithm as described in any one of claims 1-2.

Images