A method for predicting slow-blow high-voltage fuses in generator excitation voltage transformers

Abstract

The present invention relates to a method for predicting slow-blow of a high-voltage fuse of a generator excitation voltage transformer, comprising: periodically collecting three voltage data of the unit's two groups of excitation PT secondary voltages and thyristor anode voltages through a data acquisition device, calculating the voltage effective value and negative-sequence voltage value and storing them in a data storage device; sequentially performing feature extraction and dynamic feature analysis on the data stored in the data storage device through a big data streaming algorithm model to obtain the change characteristics of the dual PT voltage difference, the PT and thyristor anode voltage difference, and the negative-sequence voltage value; and an early warning output device performing threshold judgment on the change characteristics output by the algorithm server, and outputting early warning information after performing logical combination on the judgment results. The present invention introduces a real-time big data streaming algorithm model, extracts the sensitive characteristic values of six parameters using big data classification and aggregation methods, and can monitor small changes in the characteristic values of the selected parameters when slow-blow occurs in the early stages, thereby achieving early warning.

Description

Technical Field

[0001] The present invention relates to the technical field of electrical equipment, and in particular to a method for predicting slow-blow high-voltage fuses of a generator excitation voltage transformer. Background Art

[0002] A generator excitation PT is a specialized voltage transformer installed at the generator outlet for use by the excitation regulator. It measures the generator voltage and compares it with the setpoint. The excitation regulator then adjusts and stabilizes the generator voltage through its proportional (P), integral (I), and differential (D) functions. If the excitation PT fails during operation, causing the generator voltage to drop or disappear, the excitation system will immediately re-excite until the generator overvoltage relay or other protective measures trip. This can cause unplanned unit shutdown and, in severe cases, even damage the generator set. Therefore, the excitation regulator must include excitation PT disconnect protection. If a fault occurs within the PT, the high-voltage fuse on the primary side of the excitation high-voltage PT will quickly open and disconnect the power circuit, protecting the equipment and preventing further accidents. High-voltage fuses operate by melting the fuse element, which has a distinct ampere-second characteristic, also known as an inverse time characteristic: a long opening time for small overload currents and a short opening time for large overload currents. According to current technology, a fuse can typically blow within 0.1 seconds after arcing. If it operates in less than 0.1 seconds, it's called a fast blow; if it operates in more than 0.1 seconds, it's called a slow blow. If it's a fast blow, a voltage difference greater than the PT disconnect setting will occur, prompting the disconnect to trigger a quick trip, ensuring continued normal operation of the excitation system. If it's a slow blow, the generator terminal voltage and reactive power will fluctuate dramatically over a prolonged period, potentially leading to false excitation or even a tripping shutdown.

[0003] The traditional methods for judging whether the excitation high-voltage PT fuse is slow to blow are: 1. Negative sequence voltage detection method. This method uses a sampling system to calculate the positive sequence component and negative sequence component of the three-phase voltage of the generator stator, and the positive sequence component and negative sequence component of the three-phase current of the stator. When the PT high-voltage fuse is slow to blow, a negative sequence component of voltage will appear, but there will be no negative sequence component of current, so as to judge that the high-voltage fuse is slow to blow. This method is mainly based on the negative sequence voltage of the PT line break for judgment. However, since the DSP chip is a fixed-point processor, the calculation method adopts dq coordinate Z transformation to calculate it. Under normal operation, the negative sequence voltage is calculated to be about 5%, resulting in a large calculation error. It is often difficult to reach the low threshold value of the negative sequence voltage during slow blowing to achieve high-voltage protection. 1. Double PT comparison method: This method uses two groups of PT measurement values through AC, and calculates through DSP chip. One cycle can use 12 or 16 points at equal intervals, and calculates the real and imaginary parts of the AC voltage through Fourier algorithm. Then the voltage effective values U1, U2 and U=|U1-U2| are calculated. When U is greater than a certain value, it can be judged that a group of PTs min(U1, U2) is broken. The double PT voltage comparison method calculates the voltage effective value through DSP chip sampling, and judges the PT line break by comparing the voltage difference between the two groups of PTs. However, this method has the problem that the judgment threshold is difficult to set. If it is set too small, it is easy to trigger the alarm by mistake. If it is set too large, it is difficult to immediately detect the occurrence of slow melting. Summary of the Invention

[0004] The purpose of the present invention is to overcome the shortcomings of the prior art and provide a method for predicting the slow-blow of a high-voltage fuse of a generator excitation voltage transformer, thereby solving the shortcomings of the traditional method for judging the slow-blow.

[0005] The object of the present invention is achieved by the following technical solution: a method for predicting the slow-blow of a high-voltage fuse of a generator excitation voltage transformer, the method comprising:

[0006] The data acquisition device periodically collects three voltage data of the two sets of excitation PT secondary voltages and thyristor anode voltages of the unit, and calculates the voltage effective value and negative sequence voltage value and stores them in the data storage device;

[0007] The algorithm server uses a big data streaming algorithm model to perform feature extraction and dynamic feature analysis on the data stored in the data storage device to obtain the change characteristics of the dual PT voltage difference, the PT and thyristor anode voltage difference, and the negative sequence voltage value;

[0008] The early warning output device performs threshold judgment on the change characteristics output by the algorithm server, and outputs early warning information after logically combining the judgment results.

[0009] The calculation of the voltage effective value and the negative sequence voltage value includes: calculating the real part and the imaginary part of the AC voltage by the Fourier algorithm, and then calculating the voltage effective values U1, U2, and U3, and then calculating ΔU1=|U1-U2|, ΔU2=|U1-U3|, ΔU3=|U2-U3|, and the negative sequence voltage values V12, V22, and V32.

[0010] The feature extraction and dynamic feature analysis obtain the following change characteristics of the dual PT voltage difference, the PT and thyristor anode voltage difference, and the negative sequence voltage value:

[0011] The algorithm model extracts the differential voltage ΔU and negative sequence voltage V from the collected data and extracts the characteristic values. These two data are sensitive characteristic parameters that reflect the occurrence of slow melting.

[0012] The fuzzy adaptive neural network algorithm is used to fit the differential voltage ΔU and negative sequence voltage V curves of the samples of the normal operation state data of the high-voltage fuse to obtain the baseline model of the differential voltage ΔU and baseline model of negative sequence voltage V curve Then the model is used to estimate the output values of the differential voltage ΔU and the negative sequence voltage V under abnormal operation, where ΔU r Indicates the actual operating differential voltage; V r Indicates the actual operating negative sequence voltage, f c Indicates the voltage measurement factor of the difference in the degree of high-voltage fuse melting, f d The negative sequence voltage measurement factor that indicates the degree of high voltage fuse blowing.

[0013] The warning output device performs threshold judgment on the change characteristics output by the algorithm server, and outputs warning information after logically combining the judgment results, including:

[0014] Determine whether the voltage difference ΔU1 between PT1 and PT2, the voltage difference ΔU2 between PT1 and the thyristor anode, and the negative voltage V12 of PT1 under abnormal operation of the model estimation output are all outside the threshold range. If so, determine that the PT1 fuse slow blow occurs and output a warning message. If any one of them is within the threshold range, no warning message is output.

[0015] Determine whether the model estimates that the voltage difference ΔU1 between PT1 and PT2, the voltage difference ΔU3 between PT2 and the thyristor anode, and the PT2 negative sequence voltage V22 under abnormal output operation are all outside the threshold range. If so, determine that a slow blow of the PT2 fuse occurs and output an early warning message. If one of them is within the threshold range, no early warning message is output.

[0016] The big data streaming algorithm model establishment includes: establishing a heat model Q=0.24I according to the relationship between heat and resistance during the fuse blowing process 2RT, where I represents the current flowing through the fuse, R represents the resistance, and T represents the time. According to the relationship between the differential voltage ΔU and the negative-sequence voltage V and the resistance, the corresponding models ΔU∝α1α2R and V∝β1β2R are established, where α1 and α2 represent the differential voltage influence coefficients, and β1 and β2 represent the negative-sequence voltage influence coefficients.

[0017] The present invention has the following advantages: a method for predicting slow-blow of high-voltage fuses of generator excitation voltage transformers, which selects the thyristor anode voltage as a control parameter for the occurrence of slow-blow of high-voltage fuses, and extracts six judgment parameters, namely ΔU1, ΔU2, ΔU3 and negative-sequence voltage values V12, V22, and V32. Compared with traditional reference point selection, the diagnosis of slow-blow fuses is more accurate and reliable; a real-time big data streaming algorithm model is introduced, and the sensitive characteristic values of the six parameters are extracted by big data classification and aggregation methods. This method can monitor slight changes in the characteristic values of the selected parameters when slow-blow occurs in the early stage, thereby achieving early warning. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a schematic diagram of the process of the present invention;

[0019] Figure 2 ΔU and V characteristic curve fitting diagram;

[0020] Figure 3 This is a logic diagram of early warning diagnosis. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the detailed description of the embodiments of the present application provided below in conjunction with the drawings is not intended to limit the scope of protection of the present application for which protection is claimed, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of the present application. The present invention is further described below in conjunction with the drawings.

[0022] like Figure 1As shown, a method for predicting slow-blow high-voltage fuses in generator excitation voltage transformers is proposed. This method extracts characteristic values after computation using a big data streaming algorithm model. The big data streaming algorithm model consists of physical modeling, feature extraction, dynamic characteristic analysis, model building, and diagnostic analysis. The mechanism of slow-blow fuses is analyzed and used as the basis for big data physical modeling and feature extraction. The fuse melting process can be viewed as a process of heat accumulation and conversion. The degree and impact of fuse melting are related to ambient temperature, current, and material. During the slow-blow process, the heating of the fuse induces a change in resistance. Practical operation has shown that the slow-blow process is accompanied by a slow decrease in the secondary output voltage, which is caused by the resistance change during the slow-blow process. Therefore, the voltage difference and negative-sequence voltage are used as characteristic parameters to extract the occurrence of slow-blow fuses. The algorithm model uses a fuzzy adaptive neural network to fit the voltage difference and negative-sequence voltage characteristic curves to normal samples of normal data to obtain a baseline model of the characteristic curves at two voltage values. This model is then used to output the difference voltage and negative-sequence voltage characteristics in real-time big data. Specifically, the method includes the following:

[0023] The data acquisition device periodically collects three voltage data of the two sets of excitation PT secondary voltages and thyristor anode voltages of the unit, and calculates the voltage effective value and negative sequence voltage value and stores them in the data storage device. 16 points are sampled in one cycle.

[0024] The algorithm server uses a big data streaming algorithm model to perform feature extraction and dynamic feature analysis on the data stored in the data storage device to obtain the change characteristics of the dual PT voltage difference, the PT and thyristor anode voltage difference, and the negative sequence voltage value;

[0025] The early warning output device performs threshold judgment on the change characteristics output by the algorithm server, and outputs early warning information after logically combining the judgment results.

[0026] Furthermore, calculating the effective value of the voltage and the negative-sequence voltage value includes: calculating the real part and the imaginary part of the AC voltage by the Fourier algorithm, and then calculating the effective values of the voltage U1, U2, and U3, and then calculating ΔU1=|U1-U2|, ΔU2=|U1-U3|, ΔU3=|U2-U3|, and the negative-sequence voltage values V12, V22, and V32.

[0027] Furthermore, feature extraction and dynamic feature analysis reveal that the change characteristics of the dual PT voltage difference, the PT and thyristor anode voltage difference, and the negative sequence voltage value include:

[0028] The algorithm model extracts the differential voltage ΔU and negative sequence voltage V from the collected data and extracts the characteristic values. These two data are sensitive characteristic parameters that reflect the occurrence of slow melting.

[0029] like Figure 2As shown in the figure, the fuzzy adaptive neural network algorithm is used to fit the difference voltage ΔU and negative sequence voltage V curves of the samples of the normal operation state data of the high-voltage fuse to obtain the baseline model of the difference voltage ΔU and baseline model of negative sequence voltage V curve Then the model is used to estimate the output values of the differential voltage ΔU and the negative sequence voltage V under abnormal operation, where ΔU r Indicates the actual operating differential voltage; V r Indicates the actual operating negative sequence voltage, f c Indicates the voltage measurement factor of the difference in the degree of high-voltage fuse melting, f d The negative sequence voltage measurement factor that indicates the degree of high voltage fuse blowing.

[0030] Furthermore, if Figure 3 As shown, the warning output device performs threshold judgment on the change characteristics output by the algorithm server, and logically combines the judgment results to output warning information including:

[0031] Determine whether the voltage difference ΔU1 between PT1 and PT2, the voltage difference ΔU2 between PT1 and the thyristor anode, and the negative voltage V12 of PT1 under abnormal operation of the model estimation output are all outside the threshold range. If so, determine that the PT1 fuse slow blow occurs and output a warning message. If any one of them is within the threshold range, no warning message is output.

[0032] Determine whether the model estimates that the voltage difference ΔU1 between PT1 and PT2, the voltage difference ΔU3 between PT2 and the thyristor anode, and the PT2 negative sequence voltage V22 under abnormal output operation are all outside the threshold range. If so, determine that a slow blow of the PT2 fuse occurs and output an early warning message. If one of them is within the threshold range, no early warning message is output.

[0033] Among them, V32 represents the negative sequence voltage of the thyristor anode voltage, which has no physical connection with the PT slow blow and is therefore not used as a basis for judgment.

[0034] The big data streaming algorithm model establishment includes: establishing a heat model Q=0.24I according to the relationship between heat and resistance during the fuse blowing process 2 RT, where I represents the current flowing through the fuse, R represents the resistance, and T represents the time. According to the relationship between the differential voltage ΔU and the negative-sequence voltage V and the resistance, the corresponding models ΔU∝α1α2R and V∝β1β2R are established, where α1 and α2 represent the differential voltage influence coefficients, and β1 and β2 represent the negative-sequence voltage influence coefficients.

[0035] The present invention uses the thyristor anode voltage as a reference parameter for the occurrence of a slow-blow high-voltage fuse. Traditional slow-blow judgments generally only select the voltages of two groups of PTs as a reference. The thyristor anode voltage is selected as a reference parameter for the occurrence of a slow-blow high-voltage fuse, and six judgment parameters, ΔU1, ΔU2, ΔU3, and negative sequence voltage values V12, V22, and V32, are extracted. Through the logical combination of the sensitive characteristic values of multiple parameters after exceeding the limit, an accurate early warning of the occurrence of a slow-blow is achieved. A real-time big data streaming algorithm model is introduced to extract the characteristic values of the six parameters. This method can monitor slight changes in the characteristic values of the selected parameters when a slow-blow occurs in the early stages, thereby achieving early warning.

[0036] The foregoing description is merely a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the form disclosed herein and should not be construed as excluding other embodiments. Rather, the present invention can be used in various other combinations, modifications, and environments and can be modified within the scope of the concept described herein through the above teachings or techniques or knowledge in the relevant field. Modifications and variations made by those skilled in the art that do not depart from the spirit and scope of the present invention are intended to be protected by the appended claims.

Claims

1. A method for predicting slow-blow high-voltage fuses of generator excitation voltage transformers, characterized by: The slow melting prediction method comprises: The data acquisition device periodically collects three voltage data of the two sets of excitation PT secondary voltages and thyristor anode voltages of the unit, and calculates the voltage effective value and negative sequence voltage value and stores them in the data storage device; The algorithm server uses a big data streaming algorithm model to perform feature extraction and dynamic feature analysis on the data stored in the data storage device to obtain the change characteristics of the dual PT voltage difference, the PT and thyristor anode voltage difference, and the negative sequence voltage value; The warning output device performs threshold judgment on the change characteristics output by the algorithm server, and outputs warning information after logically combining the judgment results; Calculating the voltage effective value and the negative sequence voltage value includes: calculating the real part and the imaginary part of the AC voltage by Fourier algorithm, and then calculating the voltage effective values U1, U2, and U3, and then calculating ∆U1=|U1-U2|, ∆U2=|U1-U3|, ∆U3=|U2-U3|, and the negative sequence voltage values V12, V22, and V32; The feature extraction and dynamic feature analysis obtain the following change characteristics of the dual PT voltage difference, the PT and thyristor anode voltage difference, and the negative sequence voltage value: The algorithm model extracts the differential voltage ∆U and negative sequence voltage V from the collected data and extracts the characteristic values. These two data are sensitive characteristic parameters that reflect the occurrence of slow melting. The fuzzy adaptive neural network algorithm is used to fit the differential voltage ∆U and negative sequence voltage V curves to the samples of the normal operating state data of the high-voltage fuse, and the baseline model of the differential voltage ∆U is obtained. and baseline model of negative sequence voltage V curve , and then use the model to estimate the output values of the differential voltage ∆U and negative sequence voltage V under abnormal operation, where △ U r Indicates the actual operating differential voltage; V r Indicates the actual operating negative sequence voltage, f c Indicates the voltage measurement factor of the difference in the degree of high-voltage fuse melting, f d The negative sequence voltage measurement factor that indicates the degree of high voltage fuse blowing.

2. The method for predicting slow-blow high-voltage fuses of a generator excitation voltage transformer according to claim 1, characterized in that: The warning output device performs threshold judgment on the change characteristics output by the algorithm server, and outputs warning information after logically combining the judgment results, including: Determine whether the PT1 and PT2 voltage difference ∆U1, the PT1 and thyristor anode voltage difference ∆U2, and the PT1 negative voltage V12 are all outside the threshold range when the model estimates the output abnormal operation. If so, determine that the PT1 fuse is slow-blowing and output a warning message. If any one of them is within the threshold range, no warning message is output. Determine whether the PT1 and PT2 voltage difference ∆U1, the PT2 and thyristor anode voltage difference ∆U3, and the PT2 negative sequence voltage V22 under abnormal model estimated output operation are all outside the threshold range. If so, determine that the PT2 fuse slow blow has occurred and output a warning message. If any one of them is within the threshold range, no warning message is output.

3. A method for predicting slow-blow high-voltage fuses of a generator excitation voltage transformer according to claim 1 or 2, characterized in that: The big data streaming algorithm model establishment includes: establishing a heat model based on the relationship between heat and resistance during the fuse blowing process. , where I represents the current flowing through the fuse, R represents the resistance, and T represents the time. The corresponding models are established based on the relationship between the difference voltage ∆U and the negative sequence voltage V and the resistance. and , where α1 and α2 represent the differential voltage influence coefficients, and β1 and β2 represent the negative sequence voltage influence coefficients.

Images