A method for predicting the life of a water-cooled damping resistor for a converter valve

Abstract

The application relates to a water-cooled damping resistor life prediction method for a converter valve, which comprises the following steps: S01: preparing samples and collecting aging data of the water-cooled damping resistor at different preset time points under each temperature; S02: obtaining a plurality of nonlinear fitting relationship formulas by using the collected aging data; S03: determining the life end point of the water-cooled damping resistor, and inversely calculating the life end value at multiple temperature points by using the nonlinear fitting formula; S04: obtaining a life prediction model by nonlinearly fitting the life end values at the multiple temperature points based on an Arrhenius equation; S05: optimizing parameters of the life prediction model to obtain optimal fitting coefficients; and S06: using the life prediction model with the optimal fitting coefficients as an optimal life prediction model. The application can accurately predict the life of the water-cooled damping resistor, and greatly improves the safe application of the water-cooled damping resistor in actual life.

Description

Technical Field

[0001] This invention relates to the field of resistor lifetime prediction, and in particular to a method for predicting the lifetime of water-cooled damping resistors in converter valves. Background Technology

[0002] Water-cooled damping resistors are key components of high-voltage DC converter valves, playing a crucial role in their operation. A water-cooled damping resistor consists of an external PVDF shell and an internal resistance wire circuit. The resistance wire is spirally wound internally and leads out to connect to external metal connectors at both ends. When the thyristor is turned on, the water-cooled damping resistor effectively reduces the extremely fast-rising-edge (large di / dt) current generated by the discharge of the converter valve's damping capacitor, while also effectively suppressing high-frequency oscillations caused by capacitance and inductance in the valve circuit. The stable operation of the water-cooled damping resistor is critical to the safe turn-on and turn-off of the thyristor.

[0003] During thyristor operation, the switching occurs with a period of 20ms, and the damping resistor is subjected to pulse current surges in each cycle. To ensure good damping performance, the resistance value of the water-cooled damping resistor must be within the normal range. The resistance value of the water-cooled damping resistor is typically 36Ω or 45Ω. To ensure safe operation, its resistance threshold is ±3%, meaning the normal range for a 36Ω damping resistor is [34.92Ω, 37.08Ω], and the normal range for a 45Ω damping resistor is [43.65Ω, 46.35Ω].

[0004] Water-cooled damping resistors dissipate heat from the resistance wire using deionized water, preventing overheating and damage to the PVDF casing. However, due to the long-term exposure to high temperature and humidity, the resistance wire begins to age, gradually increasing in resistance beyond the normal range, thus ending the "lifespan" of the water-cooled damping resistor and rendering it inoperable. Current methods cannot accurately predict the "lifespan" of water-cooled damping resistors, making it highly susceptible to overheating as the resistor's temperature rises, potentially causing the PVDF casing and internal resistance wire to burn, posing a significant danger. Summary of the Invention

[0005] In view of the above-mentioned problems in the existing technology, the technical problem to be solved by the present invention is: how to accurately predict the final life value of the water-cooled damping resistor in the converter valve.

[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: a method for predicting the lifespan of a water-cooled damping resistor in a converter valve, comprising the following steps:

[0007] S100: Select a water-cooled damping resistor. The water-cooled damping resistor has a true final life value at several temperature points. Deionized water is passed through the water-cooled damping resistor, and the inlet and outlet of the water-cooled damping resistor are sealed.

[0008] Let the initial resistance value of the water-cooled damping resistor be r. Select P temperature points and Q time points to heat the water-cooled damping resistor with deionized water flowing through it. Collect aging data of the water-cooled damping resistor at Q time points at each temperature point.

[0009] S200: Select Q aging data points at the p-th temperature point for nonlinear fitting calculation to obtain the nonlinear fitting relationship of the water-cooled damping resistance value at the p-th temperature point. The calculation expression is as follows:

[0010] R = a + b * ln(t″ + c); (1)

[0011] Where R represents the resistance value of the water-cooled damping resistor, t″ represents the aging time, and a, b, and c all represent fitting coefficients;

[0012] S210: Traverse all temperature points to obtain P nonlinear fitting equations for the resistance values ​​of water-cooled damping resistors;

[0013] S300: Take 103% of the initial water-cooled damping resistor value r as the end of the lifespan, substitute it into formula (1) for reverse calculation, and obtain the final lifespan value of the water-cooled damping resistor at P temperature points based on the reverse calculation result and the nonlinear fitting relationship of the P water-cooled damping resistor values.

[0014] S400: Based on the Arrhenius equation, the final lifetime value of the water-cooled damping resistor at P temperature points is nonlinearly fitted to obtain the lifetime prediction model, the specific expression of which is as follows:

[0015]

[0016] Where t represents lifetime; T represents thermodynamic temperature; A and B represent fitting coefficients;

[0017] S500: Batch gradient descent is used to optimize A and B. The specific optimization steps are as follows:

[0018] S510: Select a publicly available sample dataset, wherein each sample in the publicly available sample dataset includes a temperature point and the actual final value of the water-cooled damping resistor at that temperature point.

[0019] N samples are randomly selected from the public dataset as the training set, and the remaining samples are used as the test set.

[0020] S520: Initialize the lifespan prediction model and train the lifespan prediction model;

[0021] S530: Let i = 1;

[0022] S540: Select the i-th temperature point T from N.i , will T i As input to the lifetime prediction model, the output yields the predicted lifetime t at the i-th temperature point. i ';

[0023] S550: Calculate the error function between the actual final lifetime value and the predicted final lifetime value at the i-th temperature point. The expression for the error function is as follows:

[0024]

[0025] Among them, J i () represents the error function, t i This represents the true final value of the water-cooled damping resistor at the i-th temperature point, where i = 1, 2, ..., N;

[0026] S560: Calculate the optimized fitting coefficient A i and B i The specific calculation expression is as follows:

[0027]

[0028]

[0029] A i =A i-1 -α·dA i (6)

[0030] B i =B i-1 -α·dB i (7)

[0031] Where α represents the learning rate, A i-1 and B i-1 This represents the optimized fitting coefficient value for the (i-1)th iteration. When i = 1, A i-1 and B i-1 Indicates the initial value;

[0032] S570: Stop the calculation when i≥N, and output the optimized fitting coefficient A. i and B i If the condition is met, proceed to the next step; otherwise, set i = i + 1 and return to S520.

[0033] S600: Optimize the fitting coefficients A i and B i Substitute into formula (2), then select any test sample from the test set, use the temperature of the test sample as the input of the lifetime prediction model, and output the predicted final lifetime value of the water-cooled damping resistor at that temperature.

[0034] S700: Preset difference threshold S. The predicted lifetime end value is calculated by subtracting the actual lifetime end value of the test sample. When the difference is less than S, then A... i and B i For the optimal fitting coefficients, use A i and B i If the lifetime prediction model is the optimal lifetime prediction model, proceed to the next step; otherwise, return to S520.

[0035] S800: The target temperature of the water-cooled damping resistor to be predicted is used as the input of the optimal lifetime prediction model, and the output is the predicted lifetime value of the water-cooled damping resistor at the target temperature.

[0036] Preferably, the P temperature points selected in S100 are 55°C, 85°C, and 95°C.

[0037] It includes two high-temperature points to ensure the aging effect of accelerated aging tests, and one lower-temperature point, which is slightly higher than the actual operating temperature, to provide a set of data that is more in line with reality for the prediction model.

[0038] Preferably, the Q times selected in S100 are 1d, 2d, 3d, 7d, 17d and 30d, where the time unit d represents a day.

[0039] The fitting relationship approximates the relationship between R and lnt. This time setting ensures uniform data distribution while reducing unnecessary workload.

[0040] In S120, the selected time for 55°C is specifically 1d, 2d, 3d, 7d, 17d, 30d, and 42d.

[0041] The fitting relationship is approximately the relationship between R and lnt. This time setting ensures uniform data distribution and reduces unnecessary workload. At the same time, since the aging temperature is relatively low, the aging time is extended to ensure the aging effect.

[0042] Compared with the prior art, the present invention has at least the following advantages:

[0043] 1. This invention analyzes and models the aging data of actual water-cooled damping resistors. It uses the inverse operation of a nonlinear fitting algorithm to accurately calculate the lifespan endpoints of the water-cooled damping resistors at multiple temperatures. Based on the Arrhenius equation, it constructs a lifespan prediction model using the fitted lifespan endpoints at multiple temperatures. This method can simply and accurately predict the lifespan of a resistor using the relationship between temperature and resistor lifespan. The fitting coefficients A and B of the lifespan model are optimized using the actual lifespan of the water-cooled damping resistor. This optimization method employs a batch gradient descent algorithm to obtain the optimal fitting coefficients, thus yielding the optimal lifespan prediction model. The optimization model based on the batch gradient descent algorithm optimizes the fitting coefficients using real historical failure data, making the prediction model more accurate, reliable, and closely aligned with actual operating conditions.

[0044] 2. This invention is the first to study the lifespan problem of water-cooled damping resistors in converter valves, and the actual prediction accuracy is good.

[0045] 3. This invention employs a media aging test, with the aging medium being deionized water used in actual operation. The aging environment is similar to the actual operating environment, ensuring the reliability of the aging data obtained from the test.

[0046] 4. An optimized model based on the batch gradient descent algorithm is used to correct and optimize the resistance life model parameters of the water-cooled damping resistor by utilizing historical failure data of the water-cooled damping resistor, making the prediction model more accurate, reliable, and closer to actual operating conditions.

[0047] 5. The life prediction model used in this technical solution mainly uses temperature as the primary variable for prediction. Obtaining the operating environment temperature allows for rapid and accurate prediction of the life of the water-cooled damping resistor. It eliminates the need for complex parameters and complicated processing, making it easier to understand and operate. Attached Figure Description

[0048] Figure 1 This is a schematic diagram of the process of the present invention.

[0049] Figure 2 This is the fitting result of the water-cooled damping resistor value and aging time at 55℃ in the embodiment of the present invention.

[0050] Figure 3 This is the fitting result of the water-cooled damping resistor value and aging time at 85℃ in an embodiment of the present invention.

[0051] Figure 4 This is the fitting result of the water-cooled damping resistor value and aging time at 95℃ in an embodiment of the present invention.

[0052] Figure 5 This is the fitting result of aging time lnt and the reciprocal of temperature 1 / T in the embodiments of the present invention. Detailed Implementation

[0053] The present invention will now be described in further detail.

[0054] This invention proposes a method for predicting the lifetime of water-cooled damping resistors in converter valves based on the Arrhenius equation and a BP neural network optimized by a genetic algorithm. The water-cooled damping resistor is aged through accelerated damp heat aging tests at multiple temperatures, recording the resistance increase rate and corresponding aging time. Based on the aging data, lifetime equations are obtained for each temperature, calculating the time required to reach the resistance threshold at each temperature. The aging lifetimes at each test temperature are then fitted into the Arrhenius equation to obtain a preliminary lifetime equation for the water-cooled damping resistor. Simultaneously, the resistance lifetime model is corrected and optimized by combining historical real lifetime final value data of the water-cooled damping resistor with an optimization model based on batch gradient descent algorithm, ultimately yielding the optimal lifetime prediction model for the water-cooled damping resistor.

[0055] See Figure 1 A method for predicting the lifespan of a converter valve using a water-cooled damping resistor includes the following steps:

[0056] S100: Select a water-cooled damping resistor. The water-cooled damping resistor has a true final life value at several temperature points. Deionized water is passed through the water-cooled damping resistor, and the inlet and outlet of the water-cooled damping resistor are sealed. The sealed water-cooled damping resistor is filled with aging medium, and no medium will be lost during the aging process.

[0057] Let the initial resistance value of the water-cooled damping resistor be r. Select P temperature points and Q time points to heat the water-cooled damping resistor with deionized water flowing through it. Collect aging data of the water-cooled damping resistor at Q time points at each temperature point.

[0058] The P temperature points selected in S100 are specifically 55℃, 85℃ and 95℃.

[0059] The Q times selected in S100 are specifically 1d, 2d, 3d, 7d, 17d, and 30d, where the time unit d represents a day;

[0060] In S100, the specific times selected for 55°C are 1d, 2d, 3d, 7d, 17d, 30d, and 42d.

[0061] S200: Select Q aging data points at the p-th temperature point for nonlinear fitting calculation to obtain the nonlinear fitting relationship of the water-cooled damping resistance value at the p-th temperature point. The calculation expression is as follows:

[0062] R = a + b * ln(t″ + c); (1)

[0063] Where R represents the resistance value of the water-cooled damping resistor, in Ω; t″ represents the aging time, in d; a, b, and c are all fitting coefficients;

[0064] S210: Traverse all temperature points to obtain P nonlinear fitting equations for the resistance values ​​of water-cooled damping resistors;

[0065] S300: Take 103% of the initial water-cooled damping resistor value r as the end of the lifespan, substitute it into formula (1) for reverse calculation, and obtain the final lifespan value of the water-cooled damping resistor at P temperature points based on the reverse calculation result and the nonlinear fitting relationship of the P water-cooled damping resistor values.

[0066] S400: Based on the Arrhenius equation, a current technology, the life prediction model is obtained by nonlinearly fitting the final life values ​​of the water-cooled damping resistor at P temperature points. The specific expression is as follows:

[0067]

[0068] Where t represents lifetime in days; T represents thermodynamic temperature in K; and A and B represent fitting coefficients.

[0069] S500: Batch gradient descent is used to optimize A and B. Batch gradient descent is an existing technology. The specific optimization steps are as follows:

[0070] S510: Select a publicly available sample dataset, wherein each sample in the publicly available sample dataset includes a temperature point and the actual final value of the water-cooled damping resistor at that temperature point.

[0071] N samples are randomly selected from the public dataset as the training set, and the remaining samples are used as the test set.

[0072] S520: Initialize the lifespan prediction model and train the lifespan prediction model;

[0073] S530: Let i = 1;

[0074] S540: Select the i-th temperature point T from N. i , will T i As input to the lifetime prediction model, the output yields the predicted lifetime t at the i-th temperature point. i ';

[0075] S550: Calculate the error function between the actual final lifetime value and the predicted final lifetime value at the i-th temperature point. The expression for the error function is as follows:

[0076]

[0077] Among them, J i () represents the error function, t i This represents the true final value of the water-cooled damping resistor at the i-th temperature point, where i = 1, 2, ..., N;

[0078] S560: Calculate the optimized fitting coefficient A i and B i The specific calculation expression is as follows:

[0079]

[0080]

[0081] A i =A i-1 -α·dA i (6)

[0082] B i =B i-1 -α·dB i (7)

[0083] Where α represents the learning rate, A i-1 and B i-1 This represents the optimized fitting coefficient value for the (i-1)th iteration. When i = 1, A i-1 and B i-1 Indicates the initial value;

[0084] S570: Stop the calculation when i≥N, and output the optimized fitting coefficient A. i and B i If the condition is met, proceed to the next step; otherwise, set i = i + 1 and return to S520.

[0085] S600: Optimize the fitting coefficients A i and B i Substitute into formula (2), then select any test sample from the test set, use the temperature of the test sample as the input of the lifetime prediction model, and output the predicted final lifetime value of the water-cooled damping resistor at that temperature.

[0086] S700: Preset difference threshold S. The predicted lifetime end value is calculated by subtracting the actual lifetime end value of the test sample. When the difference is less than S, then A... i and B i For the optimal fitting coefficients, use A i and B i If the lifetime prediction model is the optimal lifetime prediction model, proceed to the next step; otherwise, return to S520.

[0087] S800: The target temperature of the water-cooled damping resistor to be predicted is used as the input of the optimal lifetime prediction model, and the output is the predicted lifetime value of the water-cooled damping resistor at the target temperature.

[0088] Experimental data

[0089] Taking a 45Ω water-cooled damping resistor as an example, the failure value is 46.35Ω. The relationship between the resistance value of the water-cooled damping resistor and time at different temperatures was obtained through experimental data fitting, such as... Figure 2 , Figure 3 , Figure 4 As shown in Table 1, the data for the reciprocal of the thermodynamic temperature (1 / T), the end time t, and lnt at three aging temperatures are presented below.

[0090] Table 1. Lifespan of water-cooled damping resistors at test temperatures

[0091] Test temperature (°C) 1 / T(K) Lifetime t(d) lnt(d) 55 0.003047 30846.79 10.337 85 0.002792 19340.64 9.256 95 0.27163 8037.87 8.992

[0092] The lifetime prediction model for the water-cooled damping resistor in this embodiment is as follows:

[0093]

[0094] Figure 2 The relationship between the resistance R and the aging time t is obtained by fitting the aging data at an aging temperature of 55℃. Using this relationship, it can be calculated that the time to reach a resistance of 46.35Ω (initial resistance 45Ω) at 55℃ is 30846 days.

[0095] Figure 3 The relationship between the resistance R and the aging time t is obtained by fitting the aging data at an aging temperature of 85℃. Using this relationship, it can be calculated that the time to reach a resistance of 46.35Ω (initial resistance 45Ω) at 85℃ is 19340 days.

[0096] Figure 4 The relationship between the resistance R and the aging time t is obtained by fitting the aging data at an aging temperature of 95℃. Using this relationship, it can be calculated that the time to reach a resistance of 46.35Ω (initial resistance 45Ω) at 55℃ is 8037 days.

[0097] Figure 5 The data fitting results under this lifespan model are shown. Based on formula (2), with 1 / T as the variable and lnt as the dependent variable, the parameters A and B are obtained by linear fitting, resulting in formula (4), which in turn yields the lifespan prediction equation.

[0098] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for predicting the lifespan of a converter valve using a water-cooled damping resistor, characterized in that: Includes the following steps: S100: Select a water-cooled damping resistor. The water-cooled damping resistor has a true final life value at several temperature points. Deionized water is passed through the water-cooled damping resistor, and the inlet and outlet of the water-cooled damping resistor are sealed. Let the initial resistance value of the water-cooled damping resistor be r. Select P temperature points and Q time points to heat the water-cooled damping resistor with deionized water flowing through it. Collect aging data of the water-cooled damping resistor at Q time points at each temperature point. S200: Select Q aging data points at the p-th temperature point for nonlinear fitting calculation to obtain the nonlinear fitting relationship of the water-cooled damping resistance value at the p-th temperature point. The calculation expression is as follows: ；（1） Where R represents the resistance value of the water-cooled damping resistor. denoted as aging time, and a, b, and c are all fitting coefficients; S210: Traverse all temperature points to obtain P nonlinear fitting equations for the resistance values ​​of water-cooled damping resistors; S300: Take 103% of the initial water-cooled damping resistor value r as the end of the lifespan, substitute it into formula (1) for reverse calculation, and obtain the final lifespan value of the water-cooled damping resistor at P temperature points based on the reverse calculation result and the nonlinear fitting relationship of the P water-cooled damping resistor values. S400: Based on the Arrhenius equation, the final lifetime value of the water-cooled damping resistor at P temperature points is nonlinearly fitted to obtain the lifetime prediction model, the specific expression of which is as follows: ；（2） Where t represents lifetime; T represents thermodynamic temperature; A and B represent fitting coefficients; S500: Batch gradient descent is used to optimize A and B. The specific optimization steps are as follows: S510: Select a publicly available sample dataset, wherein each sample in the publicly available sample dataset includes a temperature point and the actual final value of the water-cooled damping resistor at that temperature point. N samples are randomly selected from the public dataset as the training set, and the remaining samples are used as the test set. S520: Initialize the lifespan prediction model and train the lifespan prediction model; S530: Let i=1; S540: Select the i-th temperature point T from N. i , will T i As input to the lifetime prediction model, the output yields the predicted lifetime at the i-th temperature point. ; S550: Calculate the error function between the actual final lifetime value and the predicted final lifetime value at the i-th temperature point. The expression for the error function is as follows: ；（3） Among them, J i () represents the error function, t i This represents the true final value of the water-cooled damping resistor at the i-th temperature point, where i = 1, 2, ... N; S560: Calculate the optimized fitting coefficient A i and B i The specific calculation expression is as follows: ；（4） ；（5） ；（6） ；（7） Where α represents the learning rate, A i-1 and B i-1 This represents the optimized fitting coefficient value for the (i-1)th iteration. When i=1, A i-1 and B i-1 Indicates the initial value; S570: Stop the calculation when i≥N, and output the optimized fitting coefficient A. i and B i If the condition is met, proceed to the next step; otherwise, set i = i + 1 and return to S540. S600: Optimize the fitting coefficients A i and B i Substitute into formula (2), then select any test sample from the test set, use the temperature of the test sample as the input of the lifetime prediction model, and output the predicted final lifetime value of the water-cooled damping resistor at that temperature. S700: Preset difference threshold S. The predicted lifetime end value is calculated by subtracting the actual lifetime end value of the test sample. When the difference is less than S, then A... i and B i For the optimal fitting coefficients, use A i and B i If the lifetime prediction model is the optimal lifetime prediction model, proceed to the next step; otherwise, return to S530. S800: The target temperature of the water-cooled damping resistor to be predicted is used as the input of the optimal lifetime prediction model, and the output is the predicted lifetime value of the water-cooled damping resistor at the target temperature.

2. The method for predicting the lifespan of a converter valve using a water-cooled damping resistor as described in claim 1, characterized in that: The P temperature points selected in S100 are specifically 55℃, 85℃, and 95℃.

3. The method for predicting the lifespan of a converter valve using a water-cooled damping resistor as described in claim 2, characterized in that: The Q time points selected in S100 are specifically 1d, 2d, 3d, 7d, 17d and 30d, where the time unit d represents a day.

4. The method for predicting the lifespan of a converter valve using a water-cooled damping resistor as described in claim 3, characterized in that: In S100, the selected time for 55°C is specifically 1d, 2d, 3d, 7d, 17d, 30d, and 42d.

Images