A partial overheating positioning method for a converter transformer

Abstract

The application discloses a partial overheating positioning method for a converter transformer, which comprises the following steps: firstly, temperature monitoring is performed on an oil tank temperature probe, a sleeve temperature probe, an upper clamp temperature probe, a lower clamp temperature probe, a pull plate temperature probe, a core temperature probe and a winding temperature probe, and real-time gas production monitoring is performed on a gas production regularity monitoring device; secondly, whether partial overheating occurs is determined based on a threshold value determined by a BP neural network algorithm; according to the determination result, whether the temperature of a clamp, a pull plate, a core, a winding, an oil tank and a sleeve structure is locally overheated is determined in sequence; finally, a result of a local overheating position is outputted and a prompt is given. The method can display the temperature of each structure in real time and determine whether a local overheating fault occurs. For the problem of local overheating of the converter transformer, a heating point can be quickly found, so that the converter transformer can be protected to operate at a safe working temperature.

Description

Technical Field

[0001] This invention relates to the field of electrical engineering technology, specifically to a new method for locating and analyzing the heating characteristics of various structures of a converter transformer and for diagnosing local overheating faults. Background Technology

[0002] During operation, converter transformers generate heat in various structures. Different converter transformer designs and materials affect the heat generation in each part. Structurally, the main body of a converter transformer consists of the core, which comprises the core column, tie plate, clamps, pads, and ferrules; the coil, which comprises the high-voltage coil, low-voltage coil, regulating coil, upper pressure plate, support bars, and insulation structures of each coil; and the tank, which comprises the tank cover, tank walls, tank bottom, outer casing, and magnetic shielding.

[0003] During operation, converter transformers radiate heat outwards, causing the overall operating temperature to rise. Various research institutions have conducted studies on transformer operating temperatures and established relevant regulations. The International Electrotechnical Commission (IEC) considers that within the temperature range of 80-140℃, for every 6℃ increase in temperature, the effective service life of transformer insulation doubles—this is the 6℃ rule for transformer operation. IEC 354, "Guidelines for Transformer Operation under Load," states that when the transformer hotspot temperature reaches 140℃, bubbles will form in the transformer oil. These bubbles can reduce insulation or trigger flashover, causing transformer damage. The Chinese national standard GB 1094 stipulates that the average temperature rise limit for oil-immersed transformer windings is 65℃, the top oil temperature rise is 55℃, and the core and tank temperature rise is 80℃. The IEC also stipulates that the coil hotspot temperature should never exceed 140℃, with 130℃ generally used as the design value.

[0004] Currently, the acquisition of thermal field data for various structural components of converter transformers mainly utilizes gas analysis, infrared thermography, and probe methods. Infrared thermography is widely used in high-voltage applications such as substations for real-time monitoring of transformer core temperature. However, due to the presence of the converter transformer tank and transformer oil, this method often only monitors the temperature of the tank's outer surface, making it difficult to directly and accurately detect the temperature and operating status of the core and structural components. Furthermore, it cannot accurately determine whether a localized overheating fault has occurred when assessing the presence of a localized thermal fault. Summary of the Invention

[0005] The purpose of this invention is to provide a method for locating local overheating in converter transformers. This method addresses the limitations of existing real-time monitoring methods, which can only monitor the temperature of the outer surface of the tank and struggle to directly and accurately detect the temperature and operating status of the core and structural components. Furthermore, it cannot accurately determine whether a local overheating fault has occurred. This method collects and analyzes temperature and gas data from the converter transformer, summarizes relevant patterns, and thus monitors for local overheating faults in the converter transformer.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A method for locating local overheating in a converter transformer, comprising the following steps:

[0008] Step 1: Temperature monitoring is performed on the oil tank, bushings, upper clamp, lower clamp, tie plate, core, and windings of the converter transformer using oil tank temperature probes, bushing temperature probes, upper clamp temperature probes, lower clamp temperature probes, tie plate temperature probes, core temperature probes, and winding temperature probes, respectively. Real-time monitoring of gas production at the gas production outlet is also performed using a gas production pattern monitoring device. The converter transformer is equipped with an oil tank temperature probe, a gas production pattern monitoring device at the gas production outlet, a bushing temperature probe on the bushing, an upper clamp temperature probe on the upper clamp, a lower clamp temperature probe on the lower clamp, a tie plate temperature probe on the tie plate, a core temperature probe on the core, and a winding temperature probe on the windings. All the oil tank temperature probes, gas production pattern monitoring device, bushing temperature probes, upper clamp temperature probes, lower clamp temperature probes, tie plate temperature probes, core temperature probes, and winding temperature probes are electrically connected to a data analysis terminal.

[0009] Step 2: Determine whether local overheating has occurred based on the threshold determined by the BP neural network algorithm for the collected real-time temperature data and gas production data;

[0010] Step 3: Based on the judgment results, determine whether the oil tank, bushing, upper clamp, lower clamp, pull plate, core and winding are locally overheated, and output the results and prompts for the locally overheated parts respectively.

[0011] Furthermore, the gas production monitoring device monitors the hydrogen, methane, acetylene, carbon monoxide, and carbon dioxide in the converter transformer oil in real time.

[0012] Furthermore, the BP neural network algorithm uses the hydrogen, methane, acetylene, carbon monoxide, and carbon dioxide gas contents measured by the transformer oil gas production monitoring device as input values, and the operating temperatures measured by the tank temperature probe, bushing temperature probe, upper clamping component temperature probe, lower clamping component temperature probe, pull plate temperature probe, core temperature probe, and winding temperature probe as output values, and establishes a weighted relationship between the input and output values. The calculation process between the input and output values ​​is called the hidden layer, and the weights of the hidden layer are updated after each iteration. The least squares method of the difference between the ideal output and the actual output is propagated from back to front, and the iteration ends after a given number of iterations or the maximum value of the difference is a threshold, outputting an ideal and true result.

[0013] Furthermore, the BP neural network algorithm has 5 input neurons, 2 output neurons, 10 hidden layer neurons, 1000 iterations, and an error target of 10. -5 .

[0014] Compared with the prior art, the beneficial effects of the present invention are:

[0015] This invention monitors the local overheating problem of converter transformers. It combines temperature probes and gas analysis methods to measure, collect, and monitor the temperature of various structures and locations of the converter transformer. It also uses a neural network algorithm to analyze the temperature and gas data of the converter transformer and summarizes the laws of the operating temperature of various structures of the converter transformer and the different gas contents of the transformer oil over time. Through logical comparison, it diagnoses the local overheating of the converter transformer.

[0016] The converter transformer local overheating location method of the present invention effectively monitors the temperature of each structure of the converter transformer, can display the temperature of each structure in real time, and determine whether a local overheating fault has occurred in each structure. For the problem of local overheating in converter transformers, it can quickly locate the heat source, thereby protecting the converter transformer to operate at a safe operating temperature and improving the insulation performance and service life of the converter transformer. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the structure of the present invention;

[0018] Figure 2 This is a schematic diagram of the local exothermic diagnostic process of the present invention;

[0019] Figure 3 This is a schematic diagram of the neural network algorithm of the present invention;

[0020] 1. Gas production monitoring device; 2. Upper clamp temperature probe; 3. Lower clamp temperature probe; 4. Pull plate temperature probe; 5. Iron core temperature probe; 6. Winding temperature probe; 7. Oil tank temperature probe; 8. Bushing temperature probe; 9. Data analysis terminal; 10. Gas production outlet; 11. Bushing; 12. Upper clamp; 13. Lower clamp; 14. Pull plate; 15. Iron core; 16. Winding; 17. Oil tank. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] like Figure 1 As shown, a converter transformer structure is provided. The converter transformer is provided with an oil tank 17. A gas outlet 10 and a bushing 11 are installed outside the oil tank 17. The oil tank is provided with an upper clamp 12, a lower clamp 13, a tie plate 14, an iron core 15 and windings 16. The tie plate 14 is fixed on the iron core 15. The upper clamp 12 and the lower clamp 13 are respectively fixed to the upper end of the iron core 15 and the lower end of the tie plate 14. Multiple windings 16 are wound on the iron core 15. The oil tank 17 is equipped with an oil tank temperature probe 7, the gas production outlet 10 is equipped with a gas production pattern monitoring device 1, the bushing 11 is equipped with a bushing temperature probe 8, the upper clamp 12 is equipped with an upper clamp temperature probe 2, the lower clamp 13 is equipped with a lower clamp temperature probe 3, the pull plate 14 is equipped with a pull plate temperature probe 4, the iron core 15 is equipped with an iron core temperature probe 5, and the winding 16 is equipped with a winding temperature probe 6. The oil tank temperature probe 7, the gas production pattern monitoring device 1, the bushing temperature probe 8, the upper clamp temperature probe 2, the lower clamp temperature probe 3, the pull plate temperature probe 4, the iron core temperature probe 5, and the winding temperature probe 6 are all electrically connected to the data analysis terminal 9.

[0023] like Figure 2 As shown, a method for locating local overheating in a converter transformer includes the following steps:

[0024] First, the oil tank 17, bushing 11, upper clamp 12, lower clamp 13, pull plate 14, core 15 and winding 16 of the converter transformer are monitored by oil tank temperature probe 7, bushing temperature probe 8, upper clamp temperature probe 2, lower clamp temperature probe 3, pull plate temperature probe 4, core temperature probe 5 and winding temperature probe 6 respectively. Then, the gas production at the gas production outlet 10 is monitored in real time by gas production pattern monitoring device 1.

[0025] Secondly, the collected real-time data is used to determine whether local overheating has occurred based on the threshold determined by the BP neural network algorithm;

[0026] The gas production monitoring device 1 monitors the hydrogen, methane, acetylene, carbon monoxide, and carbon dioxide in the converter transformer oil in real time. The model of the gas production monitoring device 1 is TROM-600.

[0027] like Figure 3 As shown, the BP neural network algorithm takes the content of hydrogen, methane, acetylene, carbon monoxide, and carbon dioxide gas measured by the transformer oil gas production monitoring device 1 as input values, and takes the operating temperatures measured by the oil tank temperature probe 7, bushing temperature probe 8, upper clamp temperature probe 2, lower clamp temperature probe 3, pull plate temperature probe 4, core temperature probe 5, and winding temperature probe 6 as output values, and obtains a weight relationship between the input and output values. The calculation process between the input and output values ​​is called the hidden layer, and the weights of the hidden layer are updated after each iteration to reduce the error.

[0028] Backpropagation (BP) neural networks, proposed in 1986 by scientists led by Rumelhart and McClelland, are multilayer feedforward neural networks trained using an error backpropagation algorithm. They are one of the most widely used neural network models. The principle behind BP neural networks is that they are multilayer perceptrons containing an input layer, hidden layers, and an output layer. They contain one or more hidden layers, capable of handling complex nonlinear mapping relationships. Neurons between layers are fully connected, but neurons within the same layer are independent and unconnected. Initially, the system randomly assigns weights. Input information is entered into the input layer and combined with the corresponding weights. After being processed by the transfer function in the hidden layer, the output signal is combined with the weights in the output layer and enters the output layer. The output layer's transfer function performs calculations to obtain a result. The network calculates the error between the current output and the expected output value. When the error value drops to an acceptable range, the training process stops, and learning ends. When the error value exceeds the expected range, the network will backpropagate the error, correcting it layer by layer from back to front, adjusting the connection weights between each layer to reduce the error. After weight adjustment, the network will start from the input layer and repeat the above process to obtain a new output value, recalculate the error range, and repeat this cycle until the error value drops to an acceptable range, at which point training will end. In addition, training will also terminate when the maximum number of iterations is reached. This allows the BP neural network to effectively learn the converter transformer temperature data collected by each probe under normal conditions and the data during local overheating faults, thereby identifying local overheating faults in the converter transformer.

[0029] This invention uses existing BP neural network algorithms to learn and accurately identify local overheating faults in converter transformers. It then detects local overheating faults at each stage and ultimately pinpoints the local overheating hotspots in the converter transformer.

[0030] The principle of the BP neural network used in this invention is to use the real-time measured content of various gases (H2, CH4, C2H4, C2H6, C2H2) in transformer oil and the fault data used for neural network training as input values ​​to propagate to the hidden and output layers. However, the output results have errors with the actual output and cannot achieve the ideal effect. Therefore, the least squares method of the difference between the ideal output and the actual output will propagate from the back to the front. After a given number of iterations or the maximum value of the difference is the threshold, the iteration ends and the ideal and true result is output.

[0031] The BP neural network algorithm in this invention has 5 input neurons and 2 output neurons (including a bias value). Through preliminary exploratory diagnostic experiments, the number of hidden layer neurons can be set to 10, the number of iterations to 1000, and the error target to 10. -5 .

[0032] Finally, based on the judgment results, the temperature of the oil tank 17, bushing 11, upper clamp 12, lower clamp 13, pull plate 14, core 15, and winding 16 is judged to determine whether there is local overheating. The results are output and prompts are given for each locally overheated part. The process is carried out in the following manner: prompt 1 for upper clamp 12 and lower clamp 13, prompt 2 for pull plate 14, prompt 3 for core 15, prompt 4 for winding 16, prompt 5 for oil tank 17, and prompt 6 for bushing 11. The test ends when no component is overheated.

[0033] This method for locating localized overheating in converter transformers utilizes data obtained from temperature monitoring devices at various points on the transformer and from monitoring the gas production patterns of the transformer oil. This data is then analyzed using a neural network algorithm. When a localized overheating fault occurs in the converter transformer, the method can promptly determine the location of the overheating based on data uploaded from temperature probes on various structural components. This method can effectively and promptly identify localized overheating faults in converter transformers under operating conditions, ensuring proper working conditions, extending the insulation life and service life of the converter transformer, and providing economic benefits. It has significant practical implications for the entire power system.

Claims

1. A partial overheating location method for a converter transformer, characterized by: The steps are: Step 1: The oil tank (17), the bushing (11), the upper clamp (12), the lower clamp (13), the pull plate (14), the core (15) and the winding (16) of the converter transformer are monitored in temperature by the oil tank temperature probe (7), the bushing temperature probe (8), the upper clamp temperature probe (2), the lower clamp temperature probe (3), the pull plate temperature probe (4), the core temperature probe (5) and the winding temperature probe (6), and the gas production outlet (10) is monitored in real time by the gas production regularity monitoring device (1); the oil tank temperature probe (7) is arranged on the oil tank (17) of the converter transformer, the gas production regularity monitoring device (1) is installed on the gas production outlet (10), the bushing temperature probe (8) is arranged on the bushing (11), the upper clamp temperature probe (2) is arranged on the upper clamp (12), the lower clamp temperature probe (3) is arranged on the lower clamp (13), the pull plate temperature probe (4) is arranged on the pull plate (14), the core temperature probe (5) is arranged on the core (15), the winding temperature probe (6) is arranged on the winding (16), and the oil tank temperature probe (7), the gas production regularity monitoring device (1), the bushing temperature probe (8), the upper clamp temperature probe (2), the lower clamp temperature probe (3), the pull plate temperature probe (4), the core temperature probe (5) and the winding temperature probe (6) are all electrically connected to the data analysis terminal (9); Step 2: The collected real-time temperature data and gas production data are compared with the threshold value determined based on the BP neural network algorithm to determine whether local overheating occurs; Step 3: According to the determination result, whether the oil tank (17), the bushing (11), the upper clamp (12), the lower clamp (13), the pull plate (14), the core (15) and the winding (16) are locally overheated is determined, and the local overheating position is output and prompted; At the beginning, the system randomly assigns a weight value, the input information is input by the input layer and combined with the corresponding weight value, after being processed by the transfer function in the hidden layer, a signal is output and combined with each weight value of the output layer, and then enters the output layer, and a result is obtained through the output layer transfer function; the network calculates the error between the current output result and the expected output value, when the error value is reduced to an acceptable range, the training process stops, and the learning ends; when the error value exceeds the expected range, the network will correct each layer of connection weight value from back to front layer by layer in the way of error back propagation, in order to reduce the error and adjust the connection weight value between each layer. The BP neural network algorithm takes the hydrogen, methane, acetylene, carbon monoxide and carbon dioxide gas content measured by the gas production regularity monitoring device (1) in the transformer oil as input values, takes the working temperature measured by the oil tank temperature probe (7), the sleeve temperature probe (8), the upper clamp temperature probe (2), the lower clamp temperature probe (3), the pull plate temperature probe (4), the core temperature probe (5) and the winding temperature probe (6) as output values, and obtains the weight relationship between the input values and the output values; the calculation process between the input values and the output values is a hidden layer, and the hidden layer updates the weight after each iteration; The least square method of the difference between the ideal output and the actual output is propagated from back to front, and the iteration is ended when the given iteration number or the maximum difference value is the threshold value, and the ideal and real result is output; The input nerve node number of the BP neural network algorithm is 5, the output nerve node number is 2, the number of hidden layer nerve nodes is set to 10, the iteration number is set to 1000, and the error target is 10 -5 ; The gas production regularity monitoring device (1) respectively monitors the hydrogen, methane, acetylene, carbon monoxide and carbon dioxide in the converter transformer oil in real time.

Images