Modular power electronic device operation control method and system for full operating conditions

Abstract

A method and system for operating and controlling modular power electronic equipment under all operating conditions includes: sampling the physical parameters of each power module; predicting the junction temperature at the current moment using a junction temperature closed-loop prediction algorithm; constructing a feature vector using the physical parameters and the predicted junction temperature at the current moment; determining whether there are potential sub-health issues or sensor failures based on the deviation between the measured efficiency and the predicted mean efficiency obtained from the feature vector through a Gaussian process model; if so, cleaning the collected physical parameters and adjusting the loss calibration coefficient in the total loss calculation function; predicting cumulative damage; when the predicted cumulative damage exceeds a set failure threshold, the remaining lifetime is calculated by subtracting the current moment from the corresponding moment; making decisions on the number of in-service modules and power allocation; and executing a smooth control algorithm to dynamically adjust the duty cycle of the pulse width modulation signal when power modules are put into or taken out. This invention effectively solves the contradiction between energy efficiency instability and reliability heterogeneity in modular systems.

Description

Technical Field

[0001] This invention belongs to the field of power conversion technology, and more specifically, relates to a method and system for operation control of modular power electronic equipment for all operating conditions. Background Technology

[0002] With the construction of the global energy internet and the acceleration of transportation electrification, high-power modular power electronic systems (such as megawatt-level photovoltaic grid-connected inverter clusters, flexible DC transmission converter valves, and core power modules for supercharging stations) are facing unprecedented and severe operational challenges. In current industrial applications, although modular architecture provides theoretical redundancy and scalability, traditional control technologies are facing insurmountable bottlenecks in actual "electric-thermal-mechanical" multi-physics coupled operating environments. First, there is the nonlinear instability of the energy efficiency characteristics across all operating conditions. Traditional equipment is typically designed for efficiency based on the rated power point, but in actual operation, the load rate often fluctuates over a very wide range. Traditional "equal power distribution" or simple "fixed threshold switching" logic leads to a large number of switching devices operating in an inefficient no-load state under light load conditions, resulting in significant switching losses and magnetic component losses. Simultaneously, due to the lack of accurate capture of the instantaneous efficiency envelope, the system often cannot self-calibrate in real time on the efficiency envelope curve, resulting in an overall energy efficiency conversion rate far lower than the design expectation.

[0003] Secondly, there is a contradiction between macroscopic load sharing and microscopic damage heterogeneity. Existing control strategies mostly employ current sharing, blindly pursuing uniform output power across modules. However, due to differences in substrate materials during manufacturing processes, packaging thermal resistance drift, and cooling field heterogeneity caused by mounting location (such as airflow short-circuiting and thermal stacking effects), the junction temperature fluctuations induced by identical current commands vary significantly across different modules. This "microscopic thermal difference" leads to accelerated aging of some modules, resulting in irreversible physical damage such as bond-wire detachment or solder fatigue. Traditional technologies, lacking real-time quantitative prediction of this nonlinear aging trajectory, are highly susceptible to the "bottleneck effect" of the system, causing premature failure of the entire system.

[0004] Third, the Pareto optimality dilemma under multi-objective constraints. Under harsh conditions such as extreme temperatures, overloads, or input voltage dips, maximizing efficiency and maintaining thermal safety boundaries are often mutually exclusive. Traditional strategies typically employ a black-and-white approach like "power-off protection," lacking guarantees for system operational continuity. Furthermore, during step load transients, due to system inertial response lag and the energy gap during the wake-up process of inactive modules, the DC bus is highly susceptible to voltage dips or resonant overshoots. This trade-off between voltage stability and wake-up efficiency has lacked an effective mathematical analytical solution within the existing control framework. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a modular power electronic equipment operation control method and system for all operating conditions.

[0006] The present invention adopts the following technical solution.

[0007] The first invention proposes a modular power electronic equipment operation control method for all operating conditions, including: The physical parameters of each power module are sampled, including the input voltage, output voltage, inductor current, and switching frequency of the power module, as well as the case temperature of the power devices set in the power module. Establish a total loss calculation function that combines junction temperature and physical parameters. Combine the junction temperature prediction value of the previous moment with the total loss calculation function, and predict the junction temperature at the current moment through a junction temperature closed-loop prediction algorithm. The physical parameters and the predicted junction temperature at the current moment are used to form an eigenvector. A Gaussian process model between the efficiency and the eigenvector is established. Based on the deviation between the measured efficiency and the predicted mean of the efficiency obtained by the Gaussian process model from the eigenvector, it is determined whether there are sub-health risks or sensor failures. If so, the collected physical parameters are cleaned and the loss calibration coefficient in the total loss calculation function is adjusted. The single-step damage at the current moment is calculated based on the junction temperature sequence within the set period. The cumulative damage at different times is calculated based on single-step damage, and a probabilistic degradation model of cumulative damage is established. The time when the cumulative damage predicted by the probabilistic degradation model exceeds the set failure threshold is obtained, and the corresponding time is subtracted from the current time to obtain the remaining lifetime. The current operating mode range is determined based on the load rate, junction temperature, ambient temperature, and load current change rate. The number of in-service modules in the power module is controlled based on the operating mode range. The adaptive weight of each in-service module is calculated based on the remaining lifetime and junction temperature. Power commands are issued to each in-service module according to the adaptive weight as the corresponding power ratio. When a power module is put into operation or switched out, a smooth control algorithm is executed to dynamically adjust the duty cycle of the pulse width modulation signal.

[0008] Preferably, the establishment of the total loss calculation function combining junction temperature and physical parameters specifically includes: Obtain the relationship curve between on-resistance and junction temperature. The on-loss is: the square of the effective value of the device current of the power device set in the power module multiplied by the on-resistance corresponding to the junction temperature, plus the on-voltage drop multiplied by the average device current. Switching losses are calculated as follows: switching frequency multiplied by the sum of turn-on energy consumption and turn-off energy consumption under reference voltage and current, multiplied by the ratio of the DC voltage of the power device to the reference voltage, and multiplied by the ratio of the average value of the collected inductor current to the reference current. The total loss is obtained by multiplying the sum of conduction loss and switching loss by the set loss calibration factor.

[0009] Preferably, the junction temperature at the current moment is predicted using a junction temperature closed-loop prediction algorithm, specifically as follows: The prior state estimate for the current time is obtained by multiplying the posterior state estimate of the previous time step with the state transition matrix, and adding the product of the total loss and power loss matrix for the current time step. The posterior state estimate of the previous time step is a matrix composed of the predicted junction temperatures of the power devices corresponding to all power modules in the previous time step. The observation residual is obtained by subtracting the product of the observation matrix and the prior state estimate from the matrix composed of all the case temperature observations at the current time. The observation residual is then multiplied by the Kalman gain calculated iteratively based on the covariance matrix. The result of the multiplication is added to the prior state estimate at the current time to obtain the posterior state estimate at the current time. The posterior state estimate at the current time is a matrix composed of the predicted junction temperatures of the power devices corresponding to all power modules at the current time.

[0010] Preferably, the method of determining whether there is a sub-health risk or sensor failure based on the deviation between the measured efficiency and the predicted mean of the efficiency obtained by the feature vector through a Gaussian process model specifically includes: When the absolute value of the deviation between the measured efficiency and the predicted mean of the efficiency obtained by the eigenvector through the Gaussian process model is greater than 3 times the variance obtained by the eigenvector through the Gaussian process model, it is considered that there is a potential sub-health risk or sensor failure.

[0011] Preferably, the single-step damage at the current moment is calculated based on the junction temperature sequence within a set period, specifically as follows: Based on the junction temperature sequence within a set period, a complete closed thermal cycle before the current moment is identified, and the junction temperature fluctuation amplitude, absolute average junction temperature, and heating duration of the cycle are obtained. Using the junction temperature fluctuation amplitude as the base and the negative first exponent coefficient as the exponent, we perform exponential calculation to obtain the first exponent; using the natural logarithm e as the base, we divide the activation energy by the Boltzmann constant and then by the absolute average junction temperature to obtain the second exponent; using the heating time as the base and the negative second exponent coefficient as the exponent, we perform exponential calculation to obtain the third exponent. The single-step damage is obtained by multiplying the set constant, the first exponent, the second exponent, and the third exponent.

[0012] Preferably, the cumulative damage at different times is calculated based on single-step damage, and a probabilistic degradation model of the cumulative damage is established, specifically as follows: The p-th power of all single-step damages within a given time period from the initial time to a set time is summed, where p is the load interaction influence factor. The summation result is subtracted from 1, and the summation result is calculated. The power is calculated by subtracting the sum of the powers from 1. The power of this yields the cumulative damage at the set time. The probabilistic degradation model is fitted with the cumulative damage at each time step from the initial time step to the current time step.

[0013] Preferably, the current operating mode range is determined based on the load rate, junction temperature, ambient temperature, and load current change rate, and the number of in-service modules in the power module is controlled based on the operating mode range, specifically as follows: When the load rate is within the set high-efficiency load range, it is in the optimal efficiency control zone. The number of in-service modules is calculated by taking the sum of the real-time total losses of all in-service modules as the objective function and minimizing the objective function. When the load rate is consistently lower than the set load threshold within a set period, it enters the light load sleep rotation zone, and the status of each power module in operation is periodically rotated based on the cumulative sleep time of each power module. When the maximum value of the junction temperature of the power devices set by all power modules or the ambient temperature exceeds the set warning threshold, the system is in the high temperature stress avoidance zone, all physically available power modules are put into operation, and the current switching frequency is corrected to the switching frequency multiplied by the derating operator. When the load current change rate is greater than the set jump threshold, it is in the load sudden change response zone. A pre-activation command is sent to all non-inactive power modules to perform pre-charging and synchronization. During pre-charging and synchronization, the inertia constant of the control loop is adjusted so that all currently inactive modules can use their overload capacity to instantly support the dynamic power deficit. After the non-inactive power modules have completed pre-charging and synchronization, the non-inactive power modules are put into operation.

[0014] Preferably, the status of each power module in operation is periodically rotated based on the cumulative sleep time of each power module, specifically as follows: The priority of each power module is calculated based on the cumulative sleep time of each power module, and power modules with higher priority than the set priority threshold are periodically adjusted to be in service. The priority is calculated by multiplying the cumulative sleep time of the corresponding power module by the set sleep weight, and then subtracting the cumulative damage of the corresponding power module by the set damage coefficient.

[0015] Preferably, the inertia constant of the control loop is adjusted, specifically as follows: The required power after the load current change rate exceeds the set switching threshold is equal to the actual output power plus the derivative of the system DC bus voltage with respect to time, plus the real-time bus voltage deviation multiplied by the set adaptive droop coefficient.

[0016] Preferably, an adaptive weight is calculated for each in-service module based on its remaining lifetime and junction temperature. Power commands are then issued to each in-service module according to the adaptive weight as the corresponding power percentage. Specifically: For each power module, the natural logarithm e is used as the base, the junction temperature of the corresponding power device is multiplied by a negative set thermal sensitivity weighting coefficient as the exponent, and an exponential operation is performed. The result of the exponential operation is multiplied by the corresponding remaining lifetime to obtain the weight component of the corresponding power module. Divide the weight component of each power module by the sum of the weight components of all power modules to obtain the adaptive weight of the corresponding power module.

[0017] Preferably, if the load is in the sudden change response zone, within a set short period after the load current change rate exceeds the set jump threshold, the adaptive weight is temporarily disabled and switched to the instantaneous current sharing mode based on the droop coefficient.

[0018] Preferably, a smooth control algorithm is executed during the power module's activation or deactivation process, specifically as follows: When power modules are put into operation, before operation, the voltage vector of the module to be put into operation is adjusted by a controller based on a second-order transfer function, so that the difference in magnitude between the voltage vector of the module to be put into operation and the bus voltage vector is less than a set difference threshold. After the power module is put into or cut out, the natural logarithm e is used as the base, and the ratio of the negative current time to the time constant is used as the exponent. The exponent is calculated by subtracting the result of the exponent calculation from 1, and the result of the subtraction is multiplied by the set target output current to obtain the output current of the corresponding power module at the current time. Furthermore, the feedforward voltage compensation term is calculated, which is: the total demand load current sampled in real time minus the integral of the actual output current of all in-service modules over time, and then divided by the equivalent capacitance of the DC bus; the feedforward voltage compensation term is directly superimposed on the reference output command of the original voltage outer loop PI control algorithm to output the duty cycle of the pulse width modulation signal.

[0019] The second aspect of this invention proposes a modular power electronic equipment operation control system for all operating conditions based on the method described in the first aspect of this invention, including a data acquisition module, a junction temperature prediction module, a remaining lifetime prediction module, and a two-stage operation control module, specifically: Acquisition module: used to sample the physical parameters of each power module, including the input voltage, output voltage, inductor current and switching frequency of the power module, as well as the case temperature of the power devices set in the power module; Junction temperature prediction module: Used to establish a total loss calculation function that combines junction temperature and physical parameters. Combining the junction temperature prediction value of the previous moment with the total loss calculation function, the junction temperature prediction algorithm predicts the junction temperature at the current moment. The remaining lifetime prediction module is used to construct a feature vector using physical parameters and the predicted junction temperature at the current moment. It then establishes a Gaussian process model between the efficiency and the feature vector. Based on the deviation between the measured efficiency and the predicted mean efficiency obtained from the Gaussian process model using the feature vector, it determines whether there are potential sub-health issues or sensor failures. If so, it cleans the collected physical parameters and adjusts the loss calibration coefficient in the total loss calculation function. It also calculates the single-step damage at the current moment based on the junction temperature sequence within a set period. Based on the single-step damage, it calculates the cumulative damage at different moments and establishes a probabilistic degradation model for the cumulative damage. It obtains the moment when the cumulative damage predicted by the probabilistic degradation model exceeds a set failure threshold, and subtracts the current moment from the corresponding moment to obtain the remaining lifetime. The dual-stage intelligent scheduling module is used to determine the current operating mode range based on load rate, junction temperature, ambient temperature, and load current change rate, and to control the number of in-service modules in the power module based on the operating mode range; it calculates the adaptive weight of each in-service module based on the remaining lifetime and junction temperature, and issues power commands to each in-service module according to the adaptive weight as the corresponding power ratio; when a power module is put into operation or switched out, it executes a smooth control algorithm to dynamically adjust the duty cycle of the pulse width modulation signal.

[0020] A third aspect of the invention provides an apparatus comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, the processor performing steps using the operation control method for modular power electronic devices oriented towards all operating conditions described in the first aspect of the invention.

[0021] A fourth aspect of the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, uses the steps of the modular power electronic equipment operation control method for all operating conditions described in the first aspect of the present invention.

[0022] The beneficial effects of this invention are as follows: Compared with the prior art, it establishes a high-precision loss model and maps it to the device junction temperature in real time by sensing the electrothermal parameters of each module's power devices; it performs module-level efficiency assessment and fatigue damage quantification based on loss data to achieve remaining lifetime prediction; it makes dynamic decisions based on load characteristics to determine the number of in-service modules to optimize the system's energy efficiency under all operating conditions; it calculates adaptive weights based on the predicted lifetime and junction temperature, and performs power allocation according to the adaptive weights to balance the reliability of each module; and it performs pre-synchronization and charge compensation control for switching transients. This invention effectively solves the contradiction between energy efficiency instability and reliability heterogeneity in modular systems, achieving maximum energy efficiency and balanced health throughout the entire life cycle while ensuring stable bus voltage. Attached Figure Description

[0023] Figure 1 A flowchart illustrating the operation control method for modular power electronic equipment under all operating conditions; Figure 2 This is a flowchart of the in-service module quantity decision and power allocation decision of the present invention; Figure 3 This is the control diagram of the smooth switching feedforward compensation system of the present invention. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of this invention. The embodiments described in this application are merely some embodiments of this invention, and not all embodiments. Based on the spirit of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the protection scope of this invention.

[0025] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings, in order to guide those skilled in the art to implement the present invention.

[0026] like Figure 1 As shown, Embodiment 1 of the present invention provides a modular power electronic equipment operation control method for all operating conditions, including: S1. Sample the physical parameters of each power module, including the input voltage, output voltage, inductor current and switching frequency of the power module, as well as the case temperature of the power devices set in the power module. Specifically, voltage sampling circuits and current transformers are deployed at the input and output terminals of each parallel module, while temperature sensors are installed on the substrates of key power devices (such as IGBTs or SiC MOSFETs). The input voltage U of each power module is sampled. in Output voltage Uout Inductor current i L and switching frequency f sw And obtain the case temperature T of the power device. c .

[0027] S2. Establish a total loss calculation function that combines junction temperature and physical parameters. Combine the junction temperature prediction value of the previous moment with the total loss calculation function, and predict the junction temperature at the current moment through a junction temperature closed-loop prediction algorithm. S2.1 Establish a function for calculating the total loss by combining junction temperature and physical parameters; The on-resistance characteristics and switching energy curves of the device under different voltages and currents are obtained in advance through experiments. Conduction loss P con The calculation formula is:

[0028] Where R ds(on) (T j ) as the junction temperature T j Dynamically changing on-resistance, I rms V is the effective value of the device current. ce Here, I represents the zero-current on-state voltage drop of the device. avg It is the average current of the device; where I rms and I avg From the sampled inductor current i L V is calculated by combining the real-time duty cycle. ce The on-state voltage drop is an inherent physical constant obtained through prior experiments, or a lookup parameter extracted directly from the device's datasheet.

[0029] Switching loss calculation P sw The formula is:

[0030] Where E on and E off These are the reference voltages U and U, respectively. ref and reference current I ref Energy consumption during activation and deactivation; For the collected inductor current i L The average value; This refers to the actual DC voltage across the device when a switching action occurs.

[0031] The calculated values ​​are corrected using a loss calibration coefficient to compensate for the impact of ambient temperature rise on model accuracy, and a total loss compensation function is established:

[0032] in, Let σ be the total loss at time k; σ is the loss based on T. c Corrected loss calibration factor; S2.2 Combine the junction temperature prediction value of the previous moment with the total loss calculation function, and predict the junction temperature at the current moment through the junction temperature closed-loop prediction algorithm; The junction temperature is predicted at the current moment using a closed-loop junction temperature prediction algorithm: The controller derives the predicted value at the current moment from the junction temperature prediction value at the previous moment and the current loss input, and establishes a state-space model for junction temperature prediction.

[0033] in, This is a priori estimate of the junction temperature state vector; This is the junction temperature state vector from the previous moment; This is the state transition matrix; The control matrix is ​​defined as follows: Using the measured shell temperature as feedback, the predicted residual is compensated by a dynamic gain adjustment term, and a correction equation based on the residual is introduced:

[0034] Where z k K is a matrix composed of shell temperature observations. k For covariance moments P k =(IK k H)P k|k-1 The dynamic gain, P, calculated iteratively. k It is the covariance moment, K k It is a dynamic gain matrix used to approximate the junction temperature inside the chip in real time. This is the junction temperature state vector at the current moment, which is a matrix composed of the predicted junction temperatures of all power devices corresponding to all power modules at the current moment; Substituting into the corrected equation, where Let be the total loss at time k. According to S2.1, it is a function related to the junction temperature at the current time. The junction temperature at the current time can be obtained by solving for it.

[0035] This invention also provides a module-level efficiency calculation and remaining lifetime prediction, the steps of which include the following: S3. Use physical parameters and the predicted junction temperature at the current moment to form a feature vector, establish a Gaussian process model between efficiency and feature vector, and judge whether there is a sub-health risk or sensor failure based on the deviation between the measured efficiency and the predicted mean of the measured efficiency in the Gaussian process model. If so, clean the collected physical parameters and adjust the loss calibration coefficient in the total loss calculation function, and calculate the single-step damage from the corresponding moment based on the junction temperature fluctuation. S3.1 Extract features from physical parameters and the predicted junction temperature at the current moment to obtain feature vectors, establish a Gaussian process model between efficiency and feature vectors, and determine whether there are potential sub-health risks or sensor failures based on the deviation between the measured efficiency and the predicted mean of efficiency obtained by the Gaussian process model from the feature vectors. This invention utilizes a non-parametric Bayesian learning method to generate a sample set. for:

[0036] in Let be the feature vector of the i-th sample. The efficiency of the i-th sample is given by n, where n is the total number of generated samples. The feature vector specifically includes the input voltage U of each power module. in Output voltage U out Inductor current i L Switching frequency f sw Shell temperature T c And the predicted junction temperature.

[0037] Establish a Gaussian process model between efficiency and eigenvectors:

[0038] in, Represents a random process. It is a Gaussian process function; This is the mean of the efficiency distribution under the condition of the corresponding eigenvector (i.e., the efficiency predicted by the eigenvector through the Gaussian process model). For kernel functions; Define the covariance function (kernel function) κ(x,x'):

[0039] Where x and x' are two arbitrary eigenvectors, used to calculate the Gaussian kernel function to measure their spatial correlation. The length scaling parameter of the kernel function reflects the smoothness of the effect of eigenvector changes on efficiency. The standard deviation of the kernel function; The feature vector of the new working condition for which efficiency prediction is required at the current moment The posterior distribution mean of the prediction efficiency (i.e., the prediction mean). and variance satisfy:

[0040]

[0041] in, For the feature vector in the new working condition, It is a matrix composed of kernel functions of pairwise feature vectors from the previous training samples; The standard deviation of the noise; It is the identity matrix; A vector composed of the efficiency of the training samples. It is the set of kernel functions of the feature vectors of the new working condition and the feature vectors of the previous training samples.

[0042] The controller calculates the measured efficiency η. nom The deviation from the predicted mean, when |η nom - |>3 When the system determines that the module has potential sub-health issues or sensor failure, it cleans the collected physical parameters and adjusts the loss calibration coefficient in the total loss calculation function. S3.2 Based on the fluctuation of junction temperature, calculate the single-step damage at the corresponding time, calculate the cumulative damage at different times based on the single-step damage, and establish a probabilistic degradation model of the cumulative damage. The controller stores the junction temperature trajectory in real time, extracts features from the real-time junction temperature sequence, identifies a complete closed thermal cycle prior to the current moment, and obtains the junction temperature fluctuation amplitude ΔT of the cycle. j Absolute mean junction temperature T m and heating time t on The specific identification and extraction process is as follows: filtering out the minimal high-frequency junction temperature ripple caused by the switching frequency, and extracting the thermodynamic peak and valley points caused by low-frequency load fluctuations; pairing adjacent valleys and peaks in a closed loop to extract the true macroscopic thermal cycle amplitude ΔT. j To avoid the problem of severely losing the macroscopic thermal stress period due to directly subtracting the junction temperature from the previous sampling time from the current sampling time; the single-step damage rate at the k-th time is calculated. :

[0043] Where A, β1, and β2 are physical constants, and E a To activate energy, k B T is the Boltzmann constant. m For absolute average junction temperature, t on The heating duration is the duration of temperature rise in a single macroscopic thermal cycle (i.e., the time it takes for the junction temperature to rise from a trough to a peak). Establish a nonlinear cumulative model to calculate the cumulative damage at time k. :

[0044] Where p is the load interaction influence factor.

[0045] S4. Calculate the cumulative damage at different times based on single-step damage and establish a probabilistic degradation model for the cumulative damage. Obtain the time when the cumulative damage predicted by the probabilistic degradation model exceeds the set failure threshold, and subtract the current time from the corresponding time to obtain the remaining lifetime. The cumulative damage calculated at each historical time from the initial time to the current time is used as the historical observation sequence, and a probabilistic degradation model p(D) for the cumulative damage is fitted to establish it. total,k D total,0:k-1 ), where D total,0:k-1 Let be the sequence of cumulative damage from the initial time to the (k-1)th time. The model uses this probabilistic degradation model to extrapolate and predict the cumulative damage to future times. Since the probabilistic degradation model is a probabilistic model, the output is the mean of the cumulative damage in the model distribution (the mean of the cumulative damage in the output distribution is the cumulative damage predicted by the model). When the predicted mean of the cumulative damage at the k-th future time... First approaching and exceeding the set failure threshold L th At that time, the predicted failure time is obtained as the device's failure time. Subtract the current time Output the final remaining lifetime (RUL).

[0046] S5, such as Figure 2 As shown, the current operating mode range is determined based on load rate, junction temperature, ambient temperature, and load current change rate. The number N of in-service modules in the power module is controlled based on this operating mode range. An adaptive weight for each in-service module is calculated based on remaining lifetime and junction temperature. Power commands are then issued to each in-service module based on these adaptive weights. Figure 3 As shown, when the power module is put into operation or switched out, a smooth control algorithm is executed to dynamically adjust the duty cycle of the pulse width modulation signal. S5.1 Determine the current operating mode range based on load rate, junction temperature, ambient temperature and load current change rate, and control the number N of in-service modules in the power module based on the operating mode range; The controller uses real-time collected load rate β and ambient temperature T amb The load current change rate di / dt is used to determine the number of in-service modules N within the four operating mode ranges. To ensure normal system power supply, the optimization solution in all modes must satisfy the physical constraint of total power balance. (where P) out,i For single-module output power, P load (Total system load power requirement) and module number constraints. (N) total (Total number of modules physically configured for the system).

[0047] The specific decision-making logic for the four working mode intervals is as follows: 1. When the working mode range is in the optimal efficiency control zone (i.e., the load rate is in the set high efficiency range), determine the optimal number of in-service modules N with the goal of minimizing the total system loss; Solve for the objective function: ; Among them, J eff The objective function is the one that optimizes efficiency; P loss_total, i (P out / N , T j,i ) represents the real-time total loss value of the i-th in-service module, which is explicitly modeled as the single-module allocated power P. out / N Its real-time junction temperature T j,i The function, P out The objective function represents the total output power required by the system. Its physical meaning is to find the optimal module combination that, while meeting the total power requirement, allows the system to operate at the Pareto optimum (minimum overall heat loss) by iterating through and changing the number of in-service modules N.

[0048] 2. When the working mode range is in the light load sleep rotation zone (when the total load rate is continuously lower than the preset load threshold, β) min，即 β<β min The score is based on a combination of the cumulative sleep time and the cumulative damage calculation priority of each module. i Perform periodic rotation;

[0049] The system is based on the score i Regularly perform rotations to ensure that the aging rates between modules are decoupled and balanced over long-term timescales; score i Let ∑t be the priority score for the i-th module. A higher score indicates a module has been dormant for a longer period and is healthier, thus the system will prioritize its execution; conversely, a lower score indicates a module will be prioritized for dormancy. sleep,i ω: The cumulative sleep time of this module from the initial moment to the current moment. sleep The dormancy weight (a set constant coefficient). ω represents the cumulative damage level of the i-th module at the current time. damage This is the damage coefficient (a set constant coefficient).

[0050] 3. When the operating mode range is in the high-temperature stress avoidance zone (the maximum value of the junction temperature of the power devices set by all power modules or the ambient temperature exceeds the set warning threshold), the system forcibly switches to the objective function minJ=max{T j,iThe optimal solution to this objective function is to deploy all physically available modules and simultaneously change the switching frequency to the switching frequency multiplied by the set derating operator f. sw ·δ red Reduce heat production rate, where δ red The derialization operator is set; 4. When the operating mode range is in the load change response zone (the preset jump threshold K is triggered when the load current change rate is detected). jump This immediately triggers a full-module pre-activation command because the non-in-service modules are pre-activated, creating an energy gap. During this brief energy gap, the actual number of in-service modules, N, remains unchanged. However, the overload of the original in-service modules cannot be sustained; otherwise, the junction temperature will spike and the modules will burn out. Once the pre-activated modules have completed pre-charging and synchronization, they will be officially connected to the bus to share the power. At this point, the number of in-service modules, N, will increase, and the previously overloaded modules will return to the safe operating range. The virtual inertial balance equations are invoked during the overload period of the in-service module:

[0051] Among them, P req (t) represents the demand power after the abrupt change, and t represents the time after the abrupt change. out U represents the actual output power. dc The system DC bus voltage, ΔU dc H represents the real-time bus voltage deviation. vir K is the inertial constant. d This is the set adaptive droop coefficient. It should be noted that during this sudden transient, due to the inherent millisecond-level delay in the physical wake-up process of inactive modules, the actual number of active modules remains unchanged during this energy gap period. This is achieved by adjusting the inertia constant H of the control loop. vir The in-service module utilizes its overload capacity to instantaneously support dynamic power deficits, ensuring voltage stability. In this balance equation, besides the traditional inertial support term H... vir dU dc In addition to using / dt to suppress the rate of voltage change, this invention introduces a damping adjustment term K. d ·ΔU dc The function of this term is to enhance the static support capability of the system in the dynamic process through proportional feedback, effectively reduce the steady-state drop amplitude, and work in conjunction with the inertial term to achieve high-order damping suppression of bus voltage disturbances.

[0052] Voltage fluctuations are suppressed by utilizing the support and damping terms in the virtual inertial balance equation.

[0053] If not in one of the four working ranges, the number of modules in service remains unchanged.

[0054] S5.2 Calculate the adaptive weight of each in-service module based on the remaining lifetime and junction temperature, and issue power commands to each in-service module based on the adaptive weight; Under steady-state conditions, a weight vector W is established based on the predicted remaining lifetime of each module and the real-time junction temperature, and its components satisfy:

[0055] Adaptive weights W of in-service module i i This is proportional to the predicted remaining lifetime of the module. Physically, the system prioritizes power allocation to healthy modules. An exponential temperature penalty term is introduced, where γ is the heat sensitivity weighting coefficient. When the real-time junction temperature T of module i... j,i When the value increases, the denominator decreases exponentially, forcibly reducing the weight of the module.

[0056] The controller issues non-uniform power commands to each in-service module according to the corresponding adaptive weight as the corresponding proportion. This command can make the instantaneous damage increment of the in-service modules converge in a statistical sense. In this preferred embodiment, if the load is in the sudden load response zone, within a set short period after the load current change rate exceeds the set jump threshold (i.e., in the initial transient phase of the load jump), the algorithm temporarily disables the adaptive weights and switches to a weighting based on the droop coefficient K. drop In the instantaneous current sharing mode, during the initial stage of dynamic adjustment, the system prioritizes current sharing control to ensure bus stability. Once the voltage fluctuation returns to the steady-state dead zone, it resumes life-cycle equalization scheduling.

[0057] S5.3. When a power module is put into operation or switched out, a smooth control algorithm is executed to dynamically adjust the duty cycle of the pulse width modulation signal; S5.3.1. Voltage pre-synchronization and phase angle latching before operation: Establish synchronization logic based on second-order transfer function to force the voltage vector V of the module to be put into operation. mod With the generatrix vector V bus The difference in magnitude |ΔV| < 0.01V N V N The set difference threshold; The second-order transfer function is specifically: S5.3.2 Dynamic ramp function control adjusts the output current based on a preset rate of change relationship, so that the command current gradually approaches the target value according to the time constant τ, where τ is dynamically adjusted with the system impedance characteristics;

[0058] in, For the current moment, For the set target output current, This represents the output current of the corresponding power module at the current moment.

[0059] S5.3.3, Feedforward Suppression and Disturbance Countermeasure Control: Calculate the feedforward voltage compensation term Δu comp :

[0060] Among them, C dc I is the equivalent capacitance of the DC bus of the system. load The total demand load current sampled in real time by the system; I mod This represents the actual output current of a single in-service module.

[0061] The controller will include the aforementioned feedforward voltage compensation term Δu comp The feedforward instruction is directly superimposed onto the reference output command of the system's original voltage outer-loop PI control algorithm. Then, considering the error term of this feedforward instruction and the current bus voltage prediction term, the final PWM modulation duty cycle is corrected in real time and issued to reduce the transient deviation ΔU of the bus voltage. dc Minimization. Embodiment 2 of this invention proposes a modular power electronic equipment operation control system for all operating conditions based on the method described in Embodiment 1 of this invention, including a data acquisition module, a junction temperature prediction module, a remaining lifetime prediction module, and a two-stage operation control module, specifically: Acquisition module: used to sample the physical parameters of each power module, including the input voltage, output voltage, inductor current and switching frequency of the power module, as well as the case temperature of the power devices set in the power module; Junction temperature prediction module: Used to establish a total loss calculation function that combines junction temperature and physical parameters. Combining the junction temperature prediction value of the previous moment with the total loss calculation function, the junction temperature prediction algorithm predicts the junction temperature at the current moment. The remaining lifetime prediction module is used to construct a feature vector using physical parameters and the predicted junction temperature at the current moment. It then establishes a Gaussian process model between the efficiency and the feature vector. Based on the deviation between the measured efficiency and the predicted mean efficiency obtained from the Gaussian process model using the feature vector, it determines whether there are potential sub-health issues or sensor failures. If so, it cleans the collected physical parameters and adjusts the loss calibration coefficient in the total loss calculation function. It also calculates the single-step damage at the current moment based on the junction temperature sequence within a set period. Based on the single-step damage, it calculates the cumulative damage at different moments and establishes a probabilistic degradation model for the cumulative damage. It obtains the moment when the cumulative damage predicted by the probabilistic degradation model exceeds a set failure threshold, and subtracts the current moment from the corresponding moment to obtain the remaining lifetime. A two-stage intelligent scheduling module is used to determine the current operating mode range based on load rate, junction temperature, ambient temperature, and load current change rate; control the number of in-service modules in the power module based on the operating mode range; calculate the adaptive weight of each in-service module based on remaining lifetime and junction temperature; and issue power commands to each in-service module according to the corresponding power ratio based on the adaptive weight; execute a smooth control algorithm when a power module is put into operation or switched out, dynamically adjusting the duty cycle of the pulse width modulation signal. This disclosure can be a system, method, and / or computer program product. A computer program product may include a computer-readable storage medium loaded with computer-readable program instructions for causing a processor to implement various aspects of this disclosure.

[0062] Computer-readable storage media can be tangible devices capable of holding and storing instructions for use by an instruction execution device. Computer-readable storage media can be, for example—but not limited to—electrical storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of computer-readable storage media include: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable compact disc read-only memory (CD-ROM), digital multifunction disc (DVD), memory sticks, floppy disks, mechanical encoding devices, such as punch cards or recessed protrusions storing instructions thereon, and any suitable combination of the foregoing. The computer-readable storage media used herein are not to be construed as transient signals themselves, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (e.g., light pulses through fiber optic cables), or electrical signals transmitted through wires.

[0063] The computer-readable program instructions described herein can be downloaded from computer-readable storage media to various computing / processing devices, or downloaded via a network, such as the Internet, local area network, wide area network, and / or wireless network, to an external computer or external storage device. The network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards them to the computer-readable storage media in the respective computing / processing device.

[0064] Computer program instructions used to perform the operations of this disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, status setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk, C++, etc., and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The computer-readable program instructions may execute entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or may be connected to an external computer (e.g., via the Internet using an Internet service provider). In some embodiments, electronic circuitry, such as programmable logic circuitry, field-programmable gate arrays (FPGAs), or programmable logic arrays (PLAs), is personalized by utilizing the status information of the computer-readable program instructions to implement various aspects of this disclosure.

[0065] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.

Claims

1. A modular power electronic equipment operation control method for all operating conditions, characterized in that, include: The physical parameters of each power module are sampled, including the input voltage, output voltage, inductor current, and switching frequency of the power module, as well as the case temperature of the power devices set in the power module. Establish a total loss calculation function that combines junction temperature and physical parameters. Combine the junction temperature prediction value of the previous moment with the total loss calculation function, and predict the junction temperature at the current moment through a junction temperature closed-loop prediction algorithm. The physical parameters and the predicted junction temperature at the current moment are used to form an eigenvector. A Gaussian process model between the efficiency and the eigenvector is established. Based on the deviation between the measured efficiency and the predicted mean of the efficiency obtained by the Gaussian process model from the eigenvector, it is determined whether there are sub-health risks or sensor failures. If so, the collected physical parameters are cleaned and the loss calibration coefficient in the total loss calculation function is adjusted. The single-step damage at the current moment is calculated based on the junction temperature sequence within the set period. The cumulative damage at different times is calculated based on single-step damage, and a probabilistic degradation model of cumulative damage is established. The time when the cumulative damage predicted by the probabilistic degradation model exceeds the set failure threshold is obtained, and the corresponding time is subtracted from the current time to obtain the remaining lifetime. The current operating mode range is determined based on the load rate, junction temperature, ambient temperature, and load current change rate. The number of in-service modules in the power module is controlled based on the operating mode range. The adaptive weight of each in-service module is calculated based on the remaining lifetime and junction temperature. Power commands are issued to each in-service module according to the adaptive weight as the corresponding power ratio. When a power module is put into operation or switched out, a smooth control algorithm is executed to dynamically adjust the duty cycle of the pulse width modulation signal.

2. The modular power electronic equipment operation control method for all operating conditions as described in claim 1, characterized in that: The function for calculating the total loss by combining junction temperature and physical parameters is as follows: Obtain the relationship curve between on-resistance and junction temperature. The on-loss is: the square of the effective value of the device current of the power device set in the power module multiplied by the on-resistance corresponding to the junction temperature, plus the on-voltage drop multiplied by the average device current. Switching losses are calculated as follows: switching frequency multiplied by the sum of turn-on energy consumption and turn-off energy consumption under reference voltage and current, multiplied by the ratio of the DC voltage of the power device to the reference voltage, and multiplied by the ratio of the average value of the collected inductor current to the reference current. The total loss is obtained by multiplying the sum of conduction loss and switching loss by the set loss calibration factor.

3. The modular power electronic equipment operation control method for all operating conditions as described in claim 1, characterized in that: The junction temperature at the current moment is predicted using a junction temperature closed-loop prediction algorithm, specifically as follows: The prior state estimate for the current time is obtained by multiplying the posterior state estimate of the previous time step with the state transition matrix, and adding the product of the total loss and power loss matrix for the current time step. The posterior state estimate of the previous time step is a matrix composed of the predicted junction temperatures of the power devices corresponding to all power modules in the previous time step. The observation residual is obtained by subtracting the product of the observation matrix and the prior state estimate from the matrix composed of all the case temperature observations at the current time. The observation residual is then multiplied by the Kalman gain calculated iteratively based on the covariance matrix. The result of the multiplication is added to the prior state estimate at the current time to obtain the posterior state estimate at the current time. The posterior state estimate at the current time is a matrix composed of the predicted junction temperatures of the power devices corresponding to all power modules at the current time.

4. The modular power electronic equipment operation control method for all operating conditions as described in claim 1, characterized in that: The method for determining whether there are potential sub-health risks or sensor failures based on the deviation between the measured efficiency and the predicted mean of the efficiency obtained through a Gaussian process model using eigenvectors is as follows: When the absolute value of the deviation between the measured efficiency and the predicted mean of the efficiency obtained by the eigenvector through the Gaussian process model is greater than 3 times the variance obtained by the eigenvector through the Gaussian process model, it is considered that there is a potential sub-health risk or sensor failure.

5. The modular power electronic equipment operation control method for all operating conditions as described in claim 1, characterized in that: The single-step damage at the current moment is calculated based on the junction temperature sequence within a set period, specifically as follows: Based on the junction temperature sequence within a set period, a complete closed thermal cycle before the current moment is identified, and the junction temperature fluctuation amplitude, absolute average junction temperature, and heating duration of the cycle are obtained. Using the junction temperature fluctuation amplitude as the base and the negative first exponent coefficient as the exponent, we perform exponential calculation to obtain the first exponent; using the natural logarithm e as the base, we divide the activation energy by the Boltzmann constant and then by the absolute average junction temperature to obtain the second exponent; using the heating time as the base and the negative second exponent coefficient as the exponent, we perform exponential calculation to obtain the third exponent. The single-step damage is obtained by multiplying the set constant, the first exponent, the second exponent, and the third exponent.

6. The modular power electronic equipment operation control method for all operating conditions as described in claim 5, characterized in that: Based on single-step damage calculation, cumulative damage at different times is obtained, and a probabilistic degradation model of cumulative damage is established, specifically as follows: The p-th power of all single-step damages within a given time period from the initial time to a set time is summed, where p is the load interaction influence factor. The summation result is subtracted from 1, and the summation result is calculated. The power is calculated by subtracting the sum of the powers from 1. The power of this yields the cumulative damage at the set time. The probabilistic degradation model is fitted with the cumulative damage at each time step from the initial time step to the current time step.

7. The modular power electronic equipment operation control method for all operating conditions as described in claim 1, characterized in that: The current operating mode range is determined based on load rate, junction temperature, ambient temperature, and load current change rate. The number of in-service modules in the power module is then controlled based on this operating mode range. Specifically: When the load rate is within the set high-efficiency load range, it is in the optimal efficiency control zone. The number of in-service modules is calculated by taking the sum of the real-time total losses of all in-service modules as the objective function and minimizing the objective function. When the load rate is consistently lower than the set load threshold within a set period, it enters the light load sleep rotation zone, and the status of each power module in operation is periodically rotated based on the cumulative sleep time of each power module. When the maximum value of the junction temperature of the power devices set by all power modules or the ambient temperature exceeds the set warning threshold, the system is in the high temperature stress avoidance zone, all physically available power modules are put into operation, and the current switching frequency is corrected to the switching frequency multiplied by the derating operator. When the load current change rate is greater than the set jump threshold, it is in the load sudden change response zone. A pre-activation command is sent to all non-inactive power modules to perform pre-charging and synchronization. During pre-charging and synchronization, the inertia constant of the control loop is adjusted so that all currently inactive modules can use their overload capacity to instantly support the dynamic power deficit. After the non-inactive power modules have completed pre-charging and synchronization, the non-inactive power modules are put into operation.

8. The modular power electronic equipment operation control method for all operating conditions as described in claim 7, characterized in that: The periodic rotation of the in-service status of each power module is based on the cumulative sleep time of each power module, specifically as follows: The priority of each power module is calculated based on the cumulative sleep time of each power module, and power modules with higher priority than the set priority threshold are periodically adjusted to be in service. The priority is calculated by multiplying the cumulative sleep time of the corresponding power module by the set sleep weight, and then subtracting the cumulative damage of the corresponding power module by the set damage coefficient.

9. The modular power electronic equipment operation control method for all operating conditions as described in claim 7, characterized in that: Adjusting the inertia constant of the control loop, specifically: The required power after the load current change rate exceeds the set switching threshold is equal to the actual output power plus the derivative of the system DC bus voltage with respect to time, plus the real-time bus voltage deviation multiplied by the set adaptive droop coefficient.

10. The modular power electronic equipment operation control method for all operating conditions as described in claim 7, characterized in that: Based on the remaining lifetime and junction temperature, an adaptive weight is calculated for each in-service module. Power commands are then issued to each in-service module according to the corresponding power percentage based on the adaptive weight. Specifically: For each power module, the natural logarithm e is used as the base, the junction temperature of the corresponding power device is multiplied by a negative set thermal sensitivity weighting coefficient as the exponent, and an exponential operation is performed. The result of the exponential operation is multiplied by the corresponding remaining lifetime to obtain the weight component of the corresponding power module. Divide the weight component of each power module by the sum of the weight components of all power modules to obtain the adaptive weight of the corresponding power module.

11. The modular power electronic equipment operation control method for all operating conditions as described in claim 10, characterized in that: If the load is in the sudden change response zone, within a set short period after the load current change rate exceeds the set jump threshold, the adaptive weight is temporarily disabled and switched to the instantaneous current sharing mode based on the droop coefficient.

12. The modular power electronic equipment operation control method for all operating conditions as described in claim 1, characterized in that: A smooth control algorithm is executed during the power module's activation or deactivation process, specifically: When power modules are put into operation, before operation, the voltage vector of the module to be put into operation is adjusted by a controller based on a second-order transfer function, so that the difference in magnitude between the voltage vector of the module to be put into operation and the bus voltage vector is less than a set difference threshold. After the power module is put into or cut out, the natural logarithm e is used as the base, and the ratio of the negative current time to the time constant is used as the exponent. The exponent is calculated by subtracting the result of the exponent calculation from 1, and the result of the subtraction is multiplied by the set target output current to obtain the output current of the corresponding power module at the current time. Furthermore, the feedforward voltage compensation term is calculated, which is: the total demand load current sampled in real time minus the integral of the actual output current of all in-service modules over time, and then divided by the equivalent capacitance of the DC bus; the feedforward voltage compensation term is directly superimposed on the reference output command of the original voltage outer loop PI control algorithm to output the duty cycle of the pulse width modulation signal.

13. A modular power electronic equipment operation control system for all operating conditions based on the method of any one of claims 1-12, comprising a data acquisition module, a junction temperature prediction module, a remaining lifetime prediction module, and a two-stage operation control module, characterized in that: Acquisition module: used to sample the physical parameters of each power module, including the input voltage, output voltage, inductor current and switching frequency of the power module, as well as the case temperature of the power devices set in the power module; Junction temperature prediction module: Used to establish a total loss calculation function that combines junction temperature and physical parameters. Combining the junction temperature prediction value of the previous moment with the total loss calculation function, the junction temperature prediction algorithm predicts the junction temperature at the current moment. The remaining lifetime prediction module is used to construct a feature vector using physical parameters and the predicted junction temperature at the current moment. It then establishes a Gaussian process model between the efficiency and the feature vector. Based on the deviation between the measured efficiency and the predicted mean efficiency obtained from the Gaussian process model using the feature vector, it determines whether there are potential sub-health issues or sensor failures. If so, it cleans the collected physical parameters and adjusts the loss calibration coefficient in the total loss calculation function. It also calculates the single-step damage at the current moment based on the junction temperature sequence within a set period. Based on the single-step damage, it calculates the cumulative damage at different moments and establishes a probabilistic degradation model for the cumulative damage. It obtains the moment when the cumulative damage predicted by the probabilistic degradation model exceeds a set failure threshold, and subtracts the current moment from the corresponding moment to obtain the remaining lifetime. The dual-stage intelligent scheduling module is used to determine the current operating mode range based on load rate, junction temperature, ambient temperature, and load current change rate, and to control the number of in-service modules in the power module based on the operating mode range; it calculates the adaptive weight of each in-service module based on the remaining lifetime and junction temperature, and issues power commands to each in-service module according to the adaptive weight as the corresponding power ratio; when a power module is put into operation or switched out, it executes a smooth control algorithm to dynamically adjust the duty cycle of the pulse width modulation signal.

14. An apparatus comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: The processor performs the steps of using the modular power electronic device operation control method for all operating conditions as described in any one of claims 1-12.

15. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program that, when executed by a processor, uses the steps of the modular power electronic device operation control method for all operating conditions as described in any one of claims 1-12.

Images