Active clamping three-level converter power module layout method based on working condition driving

Abstract

The invention relates to the technical field of power electronics, in particular to an active clamping three-level converter power module layout method based on working condition driving, and the method comprises the steps: obtaining working condition parameters of an active clamping three-level converter in a preset time period, and building a historical working condition data set; calculating the maximum power loss value of each power module in the historical working condition data set according to the historical working condition data set; obtaining the maximum temperature of the power module corresponding to the multiple groups of coordinate parameters through finite element simulation calculation, and forming a temperature coordinate data set; predicting the maximum temperature of the power module under given input coordinate parameters by using the trained maximum temperature prediction model of the power module; and constructing a target function, searching an optimal solution of the target function, and obtaining an optimal power module layout coordinate parameter. By means of the layout method, the layout efficiency can be remarkably improved, the highest temperature and the mean square error of the power module are reduced, and therefore the optimal power module layout parameters are obtained.

Description

Technical Field

[0001] The invention relates to the technical field of power electronics, and in particular to a layout method of a power module of an active embedded three-level converter based on working condition driving. Background Art

[0002] Active clamped three-level converters have more flexible switch switching options due to the addition of controllable clamping switching devices. By reasonably selecting the redundant switch state, the loss distribution between the switching devices in the topology can be changed. Compared with traditional diode clamped three-level converters, they can achieve higher power density. Therefore, they are increasingly widely used in high-power power electronic power conversion scenarios. In active clamped three-level converters, the layout design of power modules is a key link to ensure the thermal performance and reliability of the system. With the continuous improvement of power density and power module integration, the heat dissipation problem of power modules has become an important bottleneck limiting its performance improvement. The ability of power modules to handle heat limits their high-power applications.

[0003] In the design stage, it is a hot topic and difficulty to optimize the layout of power modules to improve heat dissipation efficiency and reduce overall temperature rise. Existing power module layout design usually relies on empirical rules or numerical simulation tools such as finite element analysis to simulate the thermal effects and temperature field distribution of power modules. Due to the large amount of calculation in the simulation process, this type of method is difficult to iterate quickly in actual design, especially in scenarios where multiple tests of different boundary conditions and power module configurations are required. It is inefficient and difficult to quickly respond to diverse working conditions. Summary of the invention

[0004] To solve the above technical problems, the present invention proposes an active embedded three-level converter power module layout method based on working condition drive, which can significantly improve the layout efficiency, reduce the maximum temperature of the power module after layout and its mean square error, and obtain the optimal power module layout parameters.

[0005] The present invention is achieved by adopting the following technical solutions:

[0006] The method for layout of power modules of an active clamped three-level converter based on working condition drive comprises the following steps:

[0007] Step S 1 . Obtain the operating parameters of the active clamped three-level converter within a predetermined time period and establish a historical operating condition data set;

[0008] Step S 2 . According to the historical operating condition data set, the maximum power loss of each power module in the historical operating condition data set is calculated;

[0009] Step S 3. Obtain the maximum temperature of the power module corresponding to multiple sets of coordinate parameters through finite element simulation calculation to form a temperature coordinate data set;

[0010] Step S 4 .Use the trained power module maximum temperature prediction model to predict the maximum temperature of the power module under given input coordinate parameters;

[0011] Step S 5 .According to the predicted maximum temperature of the power module under the given input coordinate parameters, the objective function is constructed, the optimal solution of the objective function is found, and the optimal power module layout coordinate parameters are obtained.

[0012] The operating parameters include the ambient temperature T j , Active power P j , reactive power Q j , AC voltage measurement AC side current And DC side voltage Wherein, j is the count value of the number of collections in a statistical period.

[0013] Step S 1 It also includes preprocessing the acquired operating parameters.

[0014] Step S 1 The establishment of historical operating condition data set specifically includes:

[0015] Based on the preprocessed operating condition parameters, periodic statistical values ​​of the operating condition parameters are calculated;

[0016] The periodic statistical values ​​are screened to obtain operating parameters that meet the data quality judgment standard; based on the operating parameters that meet the data quality judgment standard, the power module power loss value corresponding to the operating parameters is calculated to establish a historical operating condition data set.

[0017] Calculating the power loss value of the power module corresponding to the operating condition parameter specifically refers to: calculating the power loss values ​​of the IGBT transistor and the anti-parallel diode respectively, and then calculating the power loss value of each power module.

[0018] Step S 2 The calculation method of the maximum power loss value of each power module in the historical operating condition data set is: the maximum value of the power loss value of each power module in the periodic statistical value is determined as the maximum power loss value of the power module.

[0019] Step S 3Specifically, the method includes: determining the maximum power loss value of each power module as a power parameter; obtaining the maximum temperature of the power module corresponding to multiple sets of coordinate parameters through finite element simulation calculation; wherein the vertical and horizontal spacing between power module edges and power module edges, and the spacing between power module edges and active embedded three-level edges are used as coordinate parameters.

[0020] Step S 4 The training method of the medium power module maximum temperature prediction model is as follows: construct a neural network model and establish a nonlinear mapping through the neural network; based on the temperature coordinate data set obtained by finite element simulation calculation and the historical power module coordinate parameters as the input layer, the training model minimizes the error of the historical maximum temperature in the predicted value and the periodic statistical value.

[0021] Step S 5 Specifically include:

[0022] Based on the maximum temperature of the power module predicted by the power module maximum temperature prediction model, the maximum temperature average value and the maximum temperature variance are calculated;

[0023] Determine the weight coefficients corresponding to the maximum temperature average and the maximum temperature variance according to actual needs;

[0024] Construct the objective function;

[0025] The objective function is searched and optimized using an optimization algorithm to obtain the coordinate parameters that minimize the objective function value.

[0026] The objective function is:

[0027]

[0028] In the formula, is the average maximum temperature of the power module under the jth set of coordinate parameters, is the maximum temperature variance of the power module under the jth set of coordinate parameters; δ 1 and δ 2 is the weight coefficient, and δ 1 +δ 2 =1.

[0029] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0030] 1. This layout method obtains historical operating data, establishes a historical operating data set, and combines finite element simulation with the power module maximum temperature prediction model. Through numerical analysis methods and machine learning, it can quickly optimize the power module layout coordinate parameters. This method can significantly improve layout efficiency, reduce the power module maximum temperature and its mean square error, and thus obtain the optimal power module layout parameters.

[0031] 2. In the present invention, by selecting the maximum temperature average value and the maximum temperature variance, and dynamically selecting the corresponding weight coefficient based on the actual situation, it is possible to decide whether the optimization process is more biased towards reducing the average temperature or reducing the temperature difference according to the actual situation, so that selective optimization can be achieved for different purposes.

[0032] 3. In the present invention, a neural network model of the maximum power loss of the power module is established for the active embedded three-level converter, by establishing a nonlinear mapping; and based on the loss minimization prediction value of the active embedded three-level converter calculated by finite element simulation; and then further using the geometric parameters of the active embedded three-level converter and the power module to quickly determine the power module position. Therefore, this combined algorithm can provide a faster and more accurate power module layout. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments, wherein:

[0034] Figure 1 It is a schematic diagram of the process of the present invention;

[0035] Figure 2 It is a schematic diagram of some processes in the present invention;

[0036] Figure 3 It is a schematic diagram of the structure of one-phase bridge arm of the active embedded three-level converter in the present invention. DETAILED DESCRIPTION

[0037] Example 1

[0038] As a basic implementation mode of the present invention, the present invention includes a method for layout of power modules of an active clamped three-level converter based on working condition drive, comprising the following steps:

[0039] Step S 1 . Obtain the operating parameters of the active clamp three-level converter within a predetermined time period and establish a historical operating condition data set.

[0040] Step S 2 .Based on the historical operating condition data set, the maximum power loss of each power module in the historical operating condition data set is calculated.

[0041] Step S 3 .The maximum temperature of the power module corresponding to multiple sets of coordinate parameters is obtained through finite element simulation calculation to form a temperature coordinate data set.

[0042] Step S 4 .Use the trained power module maximum temperature prediction model to predict the maximum temperature of the power module under given input coordinate parameters.

[0043] Step S5 .According to the predicted maximum temperature of the power module under the given input coordinate parameters, the objective function is constructed, the optimal solution of the objective function is found, and the optimal power module layout coordinate parameters are obtained.

[0044] Example 2

[0045] As a preferred embodiment of the present invention, the present invention includes a method for layout of power modules of an active clamped three-level converter based on working condition drive, comprising the following steps:

[0046] Step S 1 . Obtain the operating parameters of the active clamp three-level converter within a predetermined time period and establish a historical operating condition data set.

[0047] Specifically, the operating condition data of the active clamped three-level converter power module is collected within the statistical period, and the collected operating condition data is preprocessed. The operating condition parameters include the ambient temperature t j , Active power P j , reactive power Q j , AC voltage measurement AC side current And DC side voltage Wherein, j is the count value of the number of collections in a statistical period.

[0048] Based on the preprocessed operating condition data, the periodic statistical value of the operating condition data is calculated.

[0049] Based on the preprocessed operating condition data and the periodic statistical values ​​of the calculated operating condition data, the power loss values ​​of each power module corresponding to the operating condition parameters are calculated to establish a historical operating condition data set. Each element in the historical operating condition data set includes at least the calculated operating condition parameters and the corresponding power loss values ​​of each power module. Specifically, based on the ambient temperature T, active power P, reactive power Q, AC voltage U abc , AC side current I abc And the DC side voltage U dc , recorded as a set of operating parameters. According to the AC side current data collection points, the current expression i is fitted a (α), i b (α), i c (α). The bridge arm of one phase of the active clamped three-level converter includes IGBT transistors T1 to T6 and anti-parallel diodes D1 to D6. Taking phase A as an example, the power loss values ​​of T1 to T6 and D1 to D6 are calculated, and finally the power loss value of each power module is calculated and recorded as P n_kloss , n is the power module number, n = 1, 2, 3. k is the number of power modules connected in parallel, k = 1, 2, ….

[0050] Step S2 .Based on the historical operating condition data set, the maximum power loss of each power module in the historical operating condition data set is calculated.

[0051] Step S 3 .The maximum temperature of the power module corresponding to multiple sets of coordinate parameters is obtained through finite element simulation calculation to form a temperature coordinate data set.

[0052] Step S 4 .Use the trained power module maximum temperature prediction model to predict the maximum temperature of the power module under given input coordinate parameters.

[0053] Step S 5 .According to the predicted maximum temperature of the power module under the given input coordinate parameters, the objective function is constructed, the optimal solution of the objective function is found, and the optimal power module layout coordinate parameters are obtained.

[0054] Example 3

[0055] As another preferred embodiment of the present invention, the present invention includes a method for layout of power modules of an active clamped three-level converter based on working condition drive, comprising the following steps:

[0056] Step S 1 . Obtain operating parameters of the active clamped three-level converter within a predetermined time period, and establish a historical operating condition data set, wherein the historical operating condition data set includes the operating condition parameters and the power module power loss value corresponding to the operating condition parameters.

[0057] Among them, it can be based on the IGBT initial saturation conduction voltage drop V CE0 , IGBT on-state equivalent resistance r CE , the initial saturation conduction voltage drop of the anti-parallel diode V D0 , the on-state equivalent resistance of the anti-parallel diode r CE , IGBT turn-on and turn-off energy E under specific test conditions on_ref 、E off_ref , test current and voltage I ref , U ref , diode reverse recovery energy under certain test conditions, E Qrr_ref The power loss value of each power module corresponding to the operating condition parameters is calculated and recorded as P n_kloss .

[0058] Step S 2 . According to the historical operating condition data set, the maximum power loss of each power module in the historical operating condition data set is calculated. Specifically, the maximum value of the power loss value of each power module in the periodic statistical value is determined as the maximum power loss of the power module. The maximum loss value of each power module in the historical data set is recorded as Pklossmax .

[0059] Step S 3 .The maximum temperature of the power module corresponding to multiple sets of coordinate parameters is obtained through finite element simulation calculation to form a temperature coordinate data set.

[0060] The maximum power loss of each power module is determined as the power parameter. The corresponding maximum temperature of the power module under multiple sets of coordinate parameters of the power module is obtained through finite element simulation calculation. A temperature coordinate data set of coordinate parameters and corresponding maximum temperature of the power module is established, wherein each element in the temperature coordinate data set includes at least the calculated maximum temperature of the power module and the power module coordinate parameters.

[0061] The method for determining the power module coordinate parameters includes: determining the length parameter based on the horizontal and vertical distances between the power module edge and the nearest power module edge or the distance between the power module edge and the nearest active clamping three-level edge. Width parameter The length parameter includes the horizontal distance between each power module and the nearest power module on the left. If there is no left power module, it is the distance from the power module to the left edge of the plane; the horizontal distance from the right power module If there is no right power module, it is the distance from the power module to the right edge of the plane. The width parameter includes the vertical distance between each power module and the power module above it. If there is no upper power module, it is the distance from the power module to the upper boundary of the plane; the vertical distance to the lower power module If there is no power module below, it is the distance from the power module to the lower boundary of the plane.

[0062] Step S 4 .Use the trained power module maximum temperature prediction model to predict the maximum temperature of the power module under given input coordinate parameters.

[0063] Step S 5 .According to the predicted maximum temperature of the power module under the given input coordinate parameters, the objective function is constructed, the optimal solution of the objective function is found, and the optimal power module layout coordinate parameters are obtained.

[0064] Example 4

[0065] As another preferred embodiment of the present invention, the present invention includes a method for arranging power modules of an active clamped three-level converter based on working condition driving, comprising the following steps:

[0066] Step S 1. Obtain the operating parameters of the active clamped three-level converter within a predetermined time period, pre-process the operating parameters, calculate the corresponding power loss values ​​of each power module, and establish a historical operating data set. Each element in the historical data set at least includes the power loss value of each power module corresponding to the calculated operating parameter.

[0067] Step S 2 .Based on the historical operating condition data set, the maximum power loss of each power module in the historical operating condition data set is calculated.

[0068] Step S 3 .The maximum temperature of the power module corresponding to multiple sets of coordinate parameters is obtained through finite element simulation calculation to form a temperature coordinate data set.

[0069] Step S 4 .Use the trained power module maximum temperature prediction model to predict the maximum temperature of the power module under given input coordinate parameters.

[0070] Specifically, a neural network model can be constructed based on the temperature coordinate data set obtained by finite element simulation, the historical power module coordinate parameters as the input layer, and the input includes the length parameter and width parameters Combine the inputs into a single vector:

[0071]

[0072] Where X is a 4k-dimensional input vector. Target output power module temperature: Where T is a k-dimensional output vector.

[0073] A nonlinear mapping is established through a neural network: T = f(X; Θ). Among them, Θ is a trainable parameter of the neural network. The neural network model includes an input layer that receives a 4k-dimensional coordinate parameter vector; several hidden layers using the activation function ReLU; and a k-dimensional vector of the predicted maximum temperature: The output layer of .

[0074] Furthermore, the training model minimizes the error between the predicted value and the historical maximum temperature in the periodic statistical value, so that the neural network model can efficiently predict the maximum temperature of the power module under given input coordinate parameters. Multiple sets of power module coordinate parameters are used as the input layer of the neural network to obtain the maximum temperature predicted by the neural network model.

[0075] When training a neural network, the goal is to minimize the error between the predicted value and the true value. The mean squared error can be used as the loss function is the predicted value of the neural network. The neural network model is trained using the training set data, and the network weights and biases are optimized by minimizing the loss function. After each training cycle, the prediction performance of the model is evaluated using the validation set data, and the model structure or training parameters are adjusted based on the validation results. Finally, after reaching the expected error threshold on the validation set, the model training is determined to be complete, and the trained neural network model is saved.

[0076] Step S 5 .According to the predicted maximum temperature of the power module under the given input coordinate parameters, the objective function is constructed, the optimal solution of the objective function is found, and the optimal power module layout coordinate parameters are obtained. Specifically, based on the judgment of the overall heat dissipation performance of the active embedded three-level converter power module, target 1 is determined to be the average value of the maximum temperature predicted by the neural network model. Based on the judgment of the overall heat dissipation balance of the active embedded three-level converter power module, target 2 is determined to be the variance of the maximum temperature predicted by the neural network model.

[0077] Therefore, the average value and variance of the maximum temperature predicted by the neural network model can be calculated, and the weighted sum of the average value and variance of the maximum temperature can be calculated. The weight coefficient of the average value and the variance degree can be selected according to actual needs. Specifically, the power module maximum temperature prediction model predicts the maximum temperature of the power module under given input coordinate parameters, and the average value of the maximum temperature predicted by the power module maximum temperature prediction model under different coordinate parameters is calculated. and variance The weighted sum of the average and variance of the maximum temperature of the power module corresponding to the coordinate parameters obtained is δ 1 With δ 2 The two weight coefficients determine whether the optimization process is more inclined to reduce the average temperature or reduce the temperature difference. 1 +δ 2 =1 ensures that the sum of the weight coefficients is 1, making the weight distribution of the objective function reasonable.

[0078] The designed objective function is searched and optimized using optimization algorithms, such as genetic algorithms, particle swarm optimization or gradient descent, to find the optimal solution of the weighted sum of the average and variance of the highest temperature predicted by the neural network model, and finally determine that the coordinate parameters corresponding to the global minimum of the objective function are the optimal power module layout coordinate parameters.

[0079] Specifically, the condition for whether the layout design of the power module of the active clamped three-level converter is optimal is that the weighted sum of the average value and variance of the maximum temperature of the power module is used as the objective function, and the optimization algorithm is used to search and optimize the objective function of the design. It can be expressed as: The specific judgment on whether the power module layout of the active embedded three-level converter is optimal is as follows: if the weighted sum of the average value and the variance of the maximum temperature of the power module currently predicted by the neural network model is the minimum value within the statistical period, that is, when the objective function reaches the global minimum value, the corresponding input coordinate parameters and the corresponding power module layout coordinate parameters are optimal, and the power module layout of the active embedded three-level converter is optimal at this time.

[0080] Example 5

[0081] As the most preferred embodiment of the present invention, the present invention includes a method for arranging power modules of an active clamped three-level converter based on working condition driving, referring to the attached specification. Figure 1 , including the following steps:

[0082] Step S 1 .Acquire the operating parameters of the active clamped three-level converter within a predetermined time period, pre-process the acquired operating parameters, and establish a historical operating data set.

[0083] Among them, for the active embedded three-level converter power module, its thermal effect is affected by the operating parameters of the ambient temperature T j , Active power P j , reactive power Q j , AC voltage measurement and DC side voltage Therefore, the operating parameters of the power module of the active embedded three-level converter can be collected multiple times in a statistical cycle: the ambient temperature T of the active embedded three-level converter is collected j , Active power P j , reactive power Q j , AC voltage measurement and DC side voltage The sampling times count value j=1, 2, ..., N; N is the sampling times threshold value in a statistical cycle.

[0084] Preprocess the data to determine the periodic statistical value of the data. Filter the periodic statistical value to obtain the operating condition data that meets the data quality judgment standard. Calculate the power module power loss value corresponding to the operating condition parameter based on the operating condition data that meets the data quality judgment standard to establish a historical operating condition data set.

[0085] The power module power loss value corresponding to the calculated operating condition parameters specifically refers to:

[0086] Refer to the instruction manual Figure 3 , the bridge arm of one phase of the active clamped three-level converter includes IGBT transistors T1~T6 and anti-parallel diodes D1~D6. Taking phase A as an example, the power loss values ​​of T1~T6 and D1~D6 are obtained by calculation. Finally, the power loss value of each power module is calculated and recorded as Pn_kloss , n is the power module number, n = 1, 2, 3. k is the number of power modules in parallel, k = 1, 2, .... The loss calculation expression is:

[0087]

[0088] Among them, V CE0 is the initial saturation conduction voltage drop of IGBT, r CE is the on-state equivalent resistance of the IGBT, V D0 is the initial saturation conduction voltage drop of the anti-parallel diode, r CE E is the on-state equivalent resistance of the anti-parallel diode, on_ref 、E off_ref I is the IGBT turn-on and turn-off energy under specific test conditions, ref , U ref To test current and voltage, E Qrr_ref is the reverse recovery energy of the diode under given test conditions. Step S 2 . According to the historical operating condition data set, the maximum power loss value of each power module in the historical operating condition data set is calculated. Specifically, the maximum value of the power loss value of each power module in the periodic statistical value is determined as the maximum power loss value P of the power module. klossmax and store it in the historical operating condition data set.

[0089] Step S 3 . Collect the power module coordinate parameters, based on the maximum power loss of each power module in the historical operating condition data set, and obtain the corresponding maximum temperature of the power module under multiple sets of coordinate parameters through finite element simulation to form a temperature coordinate data set.

[0090] The present invention uses the vertical and horizontal power module edge-to-power module edge spacing, and the power module edge-to-active embedded three-level edge spacing as coordinate parameters. Specifically, the power module coordinate parameters can be obtained: Among them, the length parameter Includes the horizontal distance between each power module and the nearest power module on the left. If there is no power module on the left, it is the distance from the power module to the left boundary of the plane; the horizontal distance with the power module on the right. If there is no power module on the right, it is the distance from the power module to the right boundary of the plane. Width parameter This includes the vertical distance between each power module and the power module above it. If there is no power module above it, it is the distance from the power module to the upper boundary of the plane. The vertical distance between each power module and the power module below it is the distance from the power module to the lower boundary of the plane if there is no power module below it. All coordinate parameters should be greater than 0 so that there is no overlap between power modules and between power modules and boundaries.

[0091] Step S4 .Use the trained power module maximum temperature prediction model to predict the maximum temperature of the power module under given input coordinate parameters.

[0092] Specifically, a neural network model is constructed, based on the temperature coordinate data set obtained by finite element simulation, the historical power module coordinate parameters as the input layer, and the input contains the length parameter Width parameter Combine the inputs into a single vector:

[0093]

[0094] Where X is a 4k-dimensional input vector. Target output power module temperature: T is a k-dimensional output vector. A nonlinear mapping is established through a neural network: T = f(X; Θ), where Θ is a trainable parameter of the neural network. The training model minimizes the error between the predicted value and the historical maximum temperature in the periodic statistical value, so that the neural network model can efficiently predict the maximum temperature of the power module under given input coordinate parameters. Multiple sets of coordinate parameters of the power module are used as the input layer of the neural network to obtain the maximum temperature predicted by the neural network model.

[0095] When training a neural network, the goal is to minimize the error between the predicted values ​​and the true values. As an example, you can use mean squared error as the loss function:

[0096]

[0097] in, is the predicted value of the neural network.

[0098] Construct a neural network model with an input layer that receives a 4k-dimensional coordinate parameter vector; several hidden layers using the activation function ReLU:

[0099] First layer: h 1 =ReLU(W 1 X+b 1 )

[0100] Second layer: h 2 =ReLU(W 2 h 1 +b 2 )

[0101]

[0102] Output the k-dimensional vector of predicted maximum temperature:

[0103] Step S 5. Calculate different coordinate parameters of the power module to obtain the optimal power module layout coordinate parameters. Specifically, an objective function can be constructed to find the optimal solution of the objective function to obtain the optimal power module layout coordinate parameters.

[0104] Refer to the instruction manual Figure 2 , specifically including the following steps:

[0105] Based on the maximum temperature predicted by the power module maximum temperature prediction model, the target 1 of the power module maximum temperature prediction model is calculated as the average value of the power module maximum temperature. Target 2 is the maximum temperature variance of the power module

[0106]

[0107] Construct the objective function:

[0108] Where: 1 +δ 2 =1 (15)

[0109] in, is the average value of the maximum temperature of the power module under the jth set of coordinate parameters, which aims to reflect the overall maximum temperature; is the variance of the maximum temperature of the power module under the jth set of coordinate parameters, which aims to reflect the overall heat dissipation balance. 1 With δ 2 are two weight coefficients that determine whether the optimization process is more inclined to reduce the average temperature or reduce the temperature difference. Constraint δ 1 +δ 2 =1 ensures that the sum of the weight coefficients is 1, making the weight distribution of the objective function reasonable. The weighted sum of the average value and variance of the maximum temperature of the power module corresponding to the coordinate parameters is obtained.

[0110] As an example, if the main goal of the cooling system is to extend the life of the equipment, then determining δ 1 =0.4,δ 2 = 0.6. If the main goal of the cooling system is to improve performance or efficiency, then δ 1 =0.7,δ 2 =0.3. If the goal is unclear or needs to be balanced: δ 1 =0.5,δ 2 =0.5. The weight coefficient can be dynamically adjusted through the optimization algorithm to balance performance and uniformity in actual engineering. The condition for whether the layout design of the power module of the active clamped three-level converter is optimal is that the weighted sum of the average value and variance of the maximum temperature of the power module is the minimum value.

[0111] The optimal solution of the objective function can be found through optimization algorithms such as genetic algorithm, particle swarm optimization or gradient descent, that is, the input coordinate parameters that minimize the objective function value can be obtained. The condition for whether the layout design of the power module of the active clamped three-level converter is optimal is that the weighted sum of the average value and variance of the maximum temperature of the power module is used as the objective function, and the optimization algorithm is used to search and optimize the designed objective function. It can be expressed as:

[0112]

[0113] That is: the objective function Optimize the target to find X * , so that Minimum. Finally determine the input coordinate parameters when the objective function reaches the global minimum.

[0114] The specific judgment on whether the power module layout of the active embedded three-level converter is optimal is as follows: if the weighted sum of the average value and the variance of the power module maximum temperature predicted by the power module maximum temperature prediction model is the minimum value within the statistical period, then the corresponding power module layout coordinate parameter is optimal, and the power module layout of the active embedded three-level converter is optimal at this time.

[0115] As an example, the genetic algorithm in the optimization algorithm is used to search and optimize the designed objective function, and finally determine the input coordinate parameters when the objective function reaches the global minimum. The genetic algorithm is used to randomly generate an initial population, and each individual represents an input coordinate parameter combination. The representation of each individual is a vector. The weighted sum of the mean and variance of the results of the power module maximum temperature prediction model is used as the objective function to calculate the fitness of each individual.

[0116] Define the fitness function Since the goal is to minimize The fitness function takes a negative value, and the optimization problem is transformed into maximizing practicality. The higher the fitness, the better the individual, that is, the better the input function coordinate parameters. Use roulette selection or tournament selection to select individuals with high fitness from the population as the parents of the next generation until the fitness of the best individual in the population no longer increases significantly. When the training obtains the target with the highest fitness, that is, when the weighted sum of the mean and variance of the target maximum temperature is minimized, the optimal coordinate parameters are obtained, and the layout at this time is optimal.

[0117] In summary, after reading the present invention document, ordinary technicians in this field can make various other corresponding transformation schemes based on the technical scheme and technical concept of the present invention without creative mental labor, which all fall within the scope of protection of the present invention.

Claims

1. A method for layout of power modules of an active clamped three-level converter based on working condition drive, characterized in that: The following steps are involved: Step S1. Obtaining operating parameters of the active clamped three-level converter within a predetermined time period and establishing a historical operating data set; Step S2. Calculate the maximum power loss of each power module in the historical operating condition data set according to the historical operating condition data set; Step S3. Obtain the maximum temperature of the power module corresponding to the power module under multiple sets of coordinate parameters through finite element simulation calculation to form a temperature coordinate data set; Step S4. Predicting the maximum temperature of the power module under given input coordinate parameters using the trained power module maximum temperature prediction model; Step S5. According to the predicted maximum temperature of the power module under the given input coordinate parameters, an objective function is constructed, and an optimal solution of the objective function is found to obtain the optimal power module layout coordinate parameters.

2. The method for layout of power modules of an active clamped three-level converter based on working condition drive according to claim 1, characterized in that: The operating parameters include the ambient temperature T j , Active power P j , reactive power Q j , AC voltage measurement AC side current And DC side voltage Wherein, j is the count value of the number of collections in a statistical period.

3. The method for layout of power modules of an active clamped three-level converter based on working condition drive according to claim 2, characterized in that: Step S1 also includes preprocessing the acquired operating condition parameters.

4. The method for layout of power modules of an active clamped three-level converter based on working condition drive according to claim 3 is characterized in that: The establishment of the historical operating condition data set in step S1 specifically includes: Based on the preprocessed operating condition parameters, periodic statistical values ​​of the operating condition parameters are calculated; Screen the periodic statistical values ​​to obtain the operating parameters that meet the data quality judgment standards; Based on the operating condition parameters that meet the data quality judgment standard, the power module power loss value corresponding to the operating condition parameters is calculated to establish a historical operating condition data set.

5. The method for layout of power modules of an active clamped three-level converter based on working condition drive according to claim 4 is characterized in that: Calculating the power loss value of the power module corresponding to the operating condition parameter specifically refers to: calculating the power loss values ​​of the IGBT transistor and the anti-parallel diode respectively, and then calculating the power loss value of each power module.

6. The method for layout of power modules of an active clamped three-level converter based on working condition drive according to claim 4 is characterized in that: The method for calculating the maximum power loss of each power module in the historical operating condition data set in step S2 is: determining the maximum value of the power loss value of each power module in the periodic statistical value as the maximum power loss value of the power module.

7. The method for layout of power modules of an active clamped three-level converter based on working condition drive according to claim 4, characterized in that: Step S3 specifically includes: determining the maximum power loss value of each power module as a power parameter; obtaining the maximum power module temperature corresponding to the power module under multiple sets of coordinate parameters through finite element simulation calculation; wherein the vertical and horizontal power module edge-to-power module edge spacing, and the power module edge-to-active embedded three-level edge spacing are used as coordinate parameters.

8. The method for layout of power modules of an active clamped three-level converter based on working condition drive according to claim 1, characterized in that: The training method of the power module maximum temperature prediction model in step S4 is: constructing a neural network model and establishing a nonlinear mapping through the neural network; Based on the temperature coordinate data set obtained by finite element simulation and the historical power module coordinate parameters as the input layer, the model is trained to minimize the error of the historical maximum temperature in the predicted value and the periodic statistical value.

9. The method for layout of power modules of an active clamped three-level converter based on working condition drive according to claim 1, characterized in that: Step S5 specifically includes: Based on the maximum temperature of the power module predicted by the power module maximum temperature prediction model, the maximum temperature average value and the maximum temperature variance are calculated; Determine the weight coefficients corresponding to the maximum temperature average and the maximum temperature variance according to actual needs; Construct the objective function; The objective function is searched and optimized using an optimization algorithm to obtain the coordinate parameters that minimize the objective function value.

10. The method for layout of power modules of an active clamped three-level converter based on working condition drive according to claim 9, characterized in that: The objective function is: In the formula, is the average maximum temperature of the power module under the jth set of coordinate parameters, is the maximum temperature variance of the power module under the jth set of coordinate parameters; δ1 and δ2 are weight coefficients, and δ1+δ2=1.

Images