A modeling system for start-up of a closed-loop dc-dc converter and method of use

Abstract

The application belongs to the technical field of DC-DC converter modeling, and discloses a modeling system and use method for starting of a closed-loop DC-DC converter, comprising a multi-parameter sensing module, an enhanced state space modeling module, a multi-algorithm fusion solving module, a multi-topology adaptation module and a model iteration optimization module. A three-order nonlinear model is constructed by taking the control input (duty ratio) as a new state variable, the multi-algorithm fusion solving module is used to improve the transient prediction accuracy and anti-interference ability, the enhanced state space modeling module is used to correct the model deviation, and the model iteration optimization module is used to support fast switching of multiple topologies.

Description

Technical Field

[0001] This invention belongs to the field of DC-DC converter modeling technology, specifically relating to a modeling system and method for starting a closed-loop DC-DC converter. Background Technology

[0002] As a core component of power electronic systems, the DC-DC converter is responsible for realizing the buck-boost conversion and stable output of DC voltage. The transient characteristics of its closed-loop startup process (such as voltage dip, current peak, and settling time) directly determine the stability of the load power supply. Especially for sensitive loads such as medical equipment and aerospace electronics, excessive voltage fluctuations or current spikes during startup may lead to equipment failure, data loss, or even system crashes. With the development of the energy internet and new energy technologies, the application scenarios of DC-DC converters are becoming increasingly complex, placing higher demands on modeling techniques. However, existing modeling methods still have the following key technical defects: insufficient accuracy of large-signal transient modeling, poor adaptability to non-ideal parameters and environmental factors, weak real-time performance and anti-interference ability of the solution algorithm, poor multi-topology compatibility, and insufficient long-term accuracy and stability of the model.

[0003] Chinese patent publication number CN116757141A, entitled "A Small-Signal Modeling Method for a Single-Inductor Multi-Output DC-DC Converter," describes a method that includes analyzing a six-switch DC-DC converter. The working principle of a DC converter involves classifying it into Boost and Buck types based on the inductor current after three-way power transmission. The piecewise average value of the inductor current is calculated, and a nonlinear state equation is established modally using the piecewise averaging method. Then, the state-space equation within one switching cycle is obtained using the state averaging method. Small-signal perturbations are injected into each input, output, state, and control quantity. The steady-state operating point and higher-order perturbations are then eliminated from the perturbation equation, resulting in a linearized system state equation. A Laplace transform is performed on the system state equation to obtain the frequency domain state-space equation. Solving this equation yields the transfer function of the influence of input and control perturbations on the output perturbation. However, this patent application fails to address the problems of low modeling accuracy, poor environmental adaptability, insufficient real-time performance, and weak multi-topology compatibility. Summary of the Invention

[0004] In order to overcome the problems existing in the prior art, the purpose of this invention is to provide a modeling system and method for starting a closed-loop DC-DC converter. By constructing a complete process of multi-parameter perception, enhanced modeling, fusion solution, dynamic compensation, multi-topology adaptation and iterative optimization, it can solve the problems of low modeling accuracy, poor environmental adaptability, insufficient real-time performance and weak multi-topology compatibility in the existing systems.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: In a first aspect, the present invention provides a modeling system for the startup of a closed-loop DC-DC converter, comprising: The multi-parameter sensing module is used to collect circuit parameters, initial operating condition parameters, real-time operating parameters and environmental parameters of the DC-DC converter from all dimensions to build a basic dataset for modeling. The enhanced state-space modeling module is used to construct a third-order nonlinear enhanced model that integrates the dynamic characteristics of control inputs based on the modeling base dataset. The multi-algorithm fusion solution module is used to integrate numerical solution and filtering optimization algorithms to balance the modeling accuracy, real-time performance and anti-interference capability of the third-order nonlinear enhancement model; and feeds back the solution step size to the multi-parameter sensing module and the filtering correction results to the enhancement state space modeling module. The multi-topology adaptation module is used to support the modeling and switching of DC-DC converter topologies, and dynamically switches the structure of the state equations of the third-order nonlinear enhancement model and the solution logic of the controller according to the topology type. The model iterative optimization module is used to calculate the error based on the output of the multi-algorithm fusion solution module and the measured values ​​of the multi-parameter sensing module through real-time error monitoring and incremental learning. The optimized model parameters and controller gain are then written back to the enhanced state space modeling module.

[0006] Optionally, the circuit parameters include power supply voltage, load resistance, inductor parameters, capacitor parameters, and switching parameters; the initial operating condition parameters include initial output voltage, initial inductor current, and initial duty cycle; the real-time operating parameters include real-time inductor current, real-time capacitor voltage, and real-time duty cycle; and the environmental parameters include ambient temperature and relative humidity, used for component parameter temperature drift compensation.

[0007] Optionally, the enhanced state-space modeling module includes a basic model building unit, a control input stateification unit, and a nonlinear coupling unit. The basic model building unit is used to establish a second-order circuit dynamic equation based on Kirchhoff's voltage law and current law, using inductor current and capacitor voltage as state variables. The control input stateification unit is used to add duty cycle as a new state variable, treating the duty cycle as a limitation of external fixed input, and constructing a third-order nonlinear enhanced state-space model containing inductor current, capacitor voltage, and duty cycle. The nonlinear coupling unit is used to introduce the dynamic mapping relationship between ambient temperature and component parameters, quantifying and embedding the drift effect of temperature on inductor and capacitor parameters into the model.

[0008] Optionally, for the boost converter, the model expression constructed by the control input state unit is:

[0009]

[0010] Where 'a' is the load voltage division factor, and the formula is: R is the resistance value of the upper voltage divider resistor, r_C is the equivalent series resistance value of the output capacitor, R_P is the resistance value of the lower voltage divider resistor; k_I is the gain of the integral controller, V o , e f is the output reference voltage, and L(T) is the temperature-corrected inductance value. The formula is: C(T) is the temperature-corrected capacitance value, and the formula is C(T)=C0(1+α_C(T-T0)), where α_T is the temperature coefficient of inductance, α_C is the temperature coefficient of capacitance, T is the current operating temperature, and T0 is the reference temperature.

[0011] Optionally, the multi-algorithm fusion solution module includes: an RK4 numerical solution unit, an adaptive Kalman filter optimization unit, a solution step size dynamic adjustment unit, and a real-time optimization unit; the RK4 numerical solution unit uses the fourth-order Runge-Kutta algorithm to solve the enhanced state-space model, and realizes the time-domain evolution prediction of state variables through iterative calculation; the AKF optimization unit uses the real-time operating parameters collected by the multi-parameter sensing module as observation values, and constructs observation equations to filter and correct the solution results of the numerical solution unit of the RK4 numerical solution unit; the solution step size dynamic adjustment unit is used to adaptively adjust the step size of RK4 according to the absolute value of the inductor current; the real-time optimization unit is used to allocate different algorithm tasks to different processor cores, and realize task scheduling through a real-time operating system.

[0012] Optionally, the observation equation is: ,in For the observation vector, For the observation matrix, To observe noise.

[0013] Optionally, the solution step size dynamic adjustment unit is used to reduce the step size to 3 μs when |(iL)|>0.5A / ms; keep the step size constant when 0.1A / ms<|(iL)|≤0.5A / ms; and increase the step size when |(iL)|≤0.1A / ms.

[0014] Optionally, the multi-topology adaptation module supports modeling and switching between three mainstream DC-DC converter topologies: boost, buck, and buck-boost. The multi-topology adaptation module includes a topology parameter matrix unit, a control strategy switching unit, a model structure adjustment unit, and a topology dynamic switching unit. The topology parameter matrix unit is used to pre-store circuit parameter templates, state equation coefficient matrices, and default values ​​of controller parameters for the three topologies. The control strategy switching unit automatically switches the controller type according to the topology type: boost topology uses integral control, buck topology uses proportional-integral control, and buck-boost topology uses proportional-integral-derivative control. The model structure adjustment unit is used to optimize the state equation form according to the topology characteristics. The topology dynamic switching unit is used to support real-time topology switching during operation.

[0015] Optionally, the model iterative optimization module includes an error monitoring unit, an incremental learning unit, a historical data feedback unit, and a model health assessment unit. The error monitoring unit calculates the error index between the model's predicted value and the measured value of the multi-parameter sensing module in real time, and sets a threshold to trigger the model optimization process. The incremental learning unit fine-tunes the key parameters of the model based on newly collected operating data, and uses the gradient descent method to update the parameters. The historical data feedback unit optimizes the basic parameters of the model using historical operating data, and adjusts the default controller gain in the topology parameter template based on the modeling error under different load rates. The model health assessment unit is used to periodically conduct a comprehensive evaluation of the model's operating status, prompting users to check the sensor status or recalibrate the component parameters.

[0016] Secondly, the present invention provides a method for using the modeling system for starting up the closed-loop DC-DC converter, comprising the following steps: The multi-parameter sensing module collects circuit parameters, initial operating parameters, real-time operating parameters, and environmental parameters of the DC-DC converter from all dimensions to construct a basic dataset for modeling. The enhanced state-space modeling module constructs a third-order nonlinear enhanced model that integrates the dynamic characteristics of the control input based on the modeling dataset. The multi-algorithm fusion solution module integrates numerical solution and filtering optimization algorithms to balance the modeling accuracy, real-time performance and anti-interference capability of the third-order nonlinear enhancement model; and feeds back the solution step size to the multi-parameter sensing module and the filtering correction result to the enhancement state space modeling module. The multi-topology adaptation module is used to model and switch the DC-DC converter topology, and dynamically switches the structure of the state equation of the third-order nonlinear enhancement model and the solution logic of the controller according to the topology type. The model iteration optimization module uses real-time error monitoring and incremental learning; based on the output of the multi-algorithm fusion solution module and the measured values ​​of the multi-parameter perception module, the optimized model parameters and controller gain are written back to the enhanced state space modeling module.

[0017] Compared with the prior art, the present invention has the following beneficial effects: The present invention discloses a modeling system and method for starting a closed-loop DC-DC converter. It systematically integrates six major modules: a multi-parameter sensing module, an enhanced state-space modeling module, a multi-algorithm fusion solution module, a multi-topology adaptation module, and a model iteration optimization module. It constructs a third-order nonlinear model by using the control input (duty cycle) as an additional state variable. The multi-algorithm fusion solution module improves transient prediction accuracy and anti-interference capability. The enhanced state-space modeling module can correct model deviations. The model iteration optimization module supports rapid switching between multiple topologies.

[0018] Furthermore, this invention combines adaptive Kalman filtering (AKF) with the fourth-order Runge-Kutta (RK4) algorithm to improve transient prediction accuracy and anti-interference capability, introduces environmental temperature compensation and online parasitic parameter identification mechanisms to correct model bias, and supports fast switching between boost, buck, and buck-boost topologies. Attached Figure Description

[0019] The accompanying drawings described herein are for illustrative purposes only and are not intended to limit the scope of the invention in any way. Furthermore, the shapes and proportions of the components in the drawings are merely schematic to aid in understanding the invention and do not specifically limit the shapes and proportions of the components. In the drawings: Figure 1 This is a flowchart of the system usage method steps according to an embodiment of the present invention; Figure 2 This is a schematic diagram comparing the output voltage during the startup process of the Boost converter according to an embodiment of the present invention; Figure 3 This is a schematic diagram comparing the inductor current during the startup process of the Boost converter according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the output voltage prediction error of the Boost converter according to an embodiment of the present invention. Detailed Implementation

[0020] To enable those skilled in the art to better understand the technical solutions of this invention, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this invention.

[0021] Therefore, the following detailed description of embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention.

[0022] In the description of the embodiments of the present invention, it should be noted that if terms such as "upper", "lower", "horizontal", "inner" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship in which the product of the invention is usually placed during use, they are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention.

[0023] When an element is referred to as being "set on" another element, it can be directly on the other element or there may be an intervening element. When an element is referred to as being "connected to" another element, it can be directly connected to the other element or there may be an intervening element. The terms "vertical," "horizontal," "left," "right," and similar expressions used herein are for illustrative purposes only and do not represent the only embodiments. The use of the term "horizontal" does not imply that the component is required to be absolutely horizontal, but rather that it may be slightly tilted. "Horizontal" simply means that its direction is more horizontal than "vertical," and does not mean that the structure must be completely horizontal, but rather that it may be slightly tilted.

[0024] It should be noted that similar reference numerals and letters in the following figures denote similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. In the description of this invention, it should be understood that the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or collections thereof.

[0025] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the specification of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used in the specification and appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0026] The present invention will now be described in detail with reference to the accompanying drawings.

[0027] Traditional modeling methods are based on small-signal linearization theory, assuming that the system state variables only have small disturbances around the steady-state operating point. They simplify the model complexity by ignoring higher-order nonlinear terms through Taylor expansion. However, during closed-loop startup, the inductor current needs to rise rapidly from 0 to its steady-state value, and the capacitor voltage may fluctuate significantly, representing a typical large-signal dynamic process. In this case, the approximation error of the linearized model increases sharply, making it unable to accurately guide overshoot suppression and steady-state control design.

[0028] Existing models often assume that inductors, capacitors, and switching transistors are ideal components, neglecting non-ideal characteristics in actual applications: inductors have equivalent series resistance (r_L) and parasitic inductance (L_p), capacitors have equivalent series resistance (r_C) and parasitic resistance (R_p), and switching transistors have on-state voltage drop (V_DS,on) and switching losses; at the same time, changes in ambient temperature will cause inductance and capacitance values ​​to drift, further aggravating modeling bias and failing to meet the requirements of wide-temperature industrial applications.

[0029] Existing modeling methods mostly employ a single numerical solution algorithm, which can guarantee accuracy but has high computational complexity and is difficult to meet the real-time modeling requirements on embedded processors. At the same time, electromagnetic interference in industrial environments can cause noise in sensor data, which a single solution algorithm cannot effectively filter, further reducing prediction accuracy.

[0030] Existing models are mostly designed for a single topology (such as supporting only boost or buck). When the converter needs to switch topologies, the model structure needs to be reconstructed manually and the controller parameters need to be adjusted. The operation is complicated and the switching is time-consuming, which cannot adapt to multi-topology mixed power supply scenarios.

[0031] The existing model parameters are mostly fixed values, and the impact of component aging and changes in operating conditions on modeling accuracy is not considered. After long-term operation, errors gradually accumulate, and the model needs to be manually recalibrated regularly, resulting in high maintenance costs.

[0032] The present invention provides a modeling system for the startup of a closed-loop DC-DC converter, comprising: The multi-parameter sensing module is used to collect circuit parameters, initial operating condition parameters, real-time operating parameters and environmental parameters of the DC-DC converter from all dimensions to build a basic dataset for modeling. The enhanced state-space modeling module is used to construct a third-order nonlinear enhanced model that integrates the dynamic characteristics of control inputs based on the modeling base dataset. The multi-algorithm fusion solution module is used to integrate numerical solution and filtering optimization algorithms to balance the modeling accuracy, real-time performance and anti-interference capability of the third-order nonlinear enhancement model; and feeds back the solution step size to the multi-parameter sensing module and the filtering correction results to the enhancement state space modeling module. The multi-topology adaptation module is used to support the modeling and switching of DC-DC converter topologies, and dynamically switches the structure of the state equations of the third-order nonlinear enhancement model and the solution logic of the controller according to the topology type. The model iterative optimization module is used to calculate the error based on the output of the multi-algorithm fusion solution module and the measured values ​​of the multi-parameter sensing module through real-time error monitoring and incremental learning. The optimized model parameters and controller gain are then written back to the enhanced state space modeling module.

[0033] The output of the multi-parameter sensing module is connected to the input of the enhanced state space modeling module.

[0034] The output of the enhanced state-space modeling module is connected to the input of the multi-algorithm fusion solution module.

[0035] The multi-algorithm fusion solution module achieves dynamic matching of sampling and modeling during the solution process.

[0036] The multi-topology adaptation module is bidirectionally connected to the enhanced state space modeling module and the multi-algorithm fusion solution module, and is used to dynamically switch the structure of the state equation and the solution logic of the controller according to the topology type.

[0037] The model iterative optimization module forms a closed loop with the multi-parameter sensing module, the multi-algorithm fusion solution module, and the enhanced state space modeling module. Based on the error between the output of the multi-algorithm fusion solution module and the measured value of the multi-parameter sensing module, the optimized model parameters and controller gain are written back to the enhanced state space modeling module, forming a closed loop of modeling, solving, optimizing, and iterating throughout the entire process.

[0038] Example 1 A modeling system for starting a closed-loop DC-DC converter includes: a multi-parameter sensing module, an enhanced state-space modeling module, a multi-algorithm fusion solution module, a multi-topology adaptation module, and a model iterative optimization module.

[0039] The multi-parameter sensing module is used to collect circuit parameters, initial operating parameters, real-time operating parameters and environmental parameters of DC-DC converters from all dimensions to build a basic dataset for modeling.

[0040] The circuit parameters include power supply voltage V9, load resistance R, inductance parameters (nominal inductance L0, equivalent series resistance r_L, parasitic inductance L_p), capacitance parameters (nominal capacitance C0, equivalent series resistance r_C, parasitic resistance R_p), and switching parameters (on-state voltage drop V_DS,on, off-state time t_off, switching frequency f). s ).

[0041] The initial operating parameters include the initial output voltage V. o0 Initial inductor current I_L0, initial duty cycle D0, where D0 is set to 0.5±0.02.

[0042] The real-time operating parameters include real-time inductor current i_L, real-time capacitor voltage v_C, and real-time duty cycle d.

[0043] The environmental parameters include ambient temperature T and relative humidity H, which are used for component parameter temperature drift compensation.

[0044] The multi-parameter sensing module also includes a parameter calibration unit, which is used to periodically and automatically perform sensor zero-point calibration and parameter reference calibration to ensure that the parameter acquisition accuracy meets the modeling requirements.

[0045] The enhanced state-space modeling module is used to construct a third-order nonlinear enhanced model that integrates the dynamic characteristics of control inputs, achieving accurate transient modeling of large signals and breaking through the limitations of traditional second-order models.

[0046] The enhanced state-space modeling module includes: a basic model building unit, a control input state unit, and a nonlinear coupling unit. The basic model building unit uses inductor current i_L and capacitor voltage v_C as core state variables, and establishes second-order circuit dynamic equations based on Kirchhoff's voltage law KVL and current law KCL to describe the basic coupling relationship between inductor charging and discharging and capacitor charging and discharging.

[0047] The control input state unit adds the duty cycle d as a new state variable, breaking through the limitation of the traditional model that treats d as an external fixed input. It constructs a third-order nonlinear enhanced state-space model that includes inductor current i_L, capacitor voltage v_C, and duty cycle d. For the boost converter, the model expression is:

[0048]

[0049] Where 'a' is the load voltage division factor, and the formula is: R is the resistance value of the upper voltage divider resistor, r_C is the equivalent series resistance value of the output capacitor, R_P is the resistance value of the lower voltage divider resistor; k_I is the gain of the integral controller (range 50-100, adaptively adjusted according to the topology type); V o , e f is the output reference voltage, and L(T) is the temperature-corrected inductance value. The formula is: C(T) is the temperature-corrected capacitance value, and the formula is C(T)=C0(1+α_C(T-T0)), where α_T is the temperature coefficient of inductance, α_C is the temperature coefficient of capacitance, T is the current operating temperature, and T0 is the reference temperature (default 25℃).

[0050] The nonlinear coupling unit is used to introduce a dynamic mapping relationship between ambient temperature and component parameters, quantify the drift effect of temperature on inductor and capacitor parameters and embed it into the model, thus solving the problem of decreased modeling accuracy under wide temperature conditions.

[0051] The multi-algorithm fusion solution module is used to integrate numerical solution and filtering optimization algorithms to balance modeling accuracy, real-time performance and anti-interference capability.

[0052] The multi-algorithm fusion solution module includes an RK4 numerical solution unit, an adaptive Kalman filter (AKF) optimization unit, a solution step size dynamic adjustment unit, and a real-time optimization unit.

[0053] The RK4 numerical solution unit uses the fourth-order Runge-Kutta algorithm to solve the enhanced state-space model, and realizes the time-domain evolution prediction of state variables (inductor current i_L, capacitor voltage v_C and duty cycle d) through iterative calculation.

[0054] The AKF optimization unit is used to construct observation equations by using the real-time operating parameters collected by the multi-parameter sensing module as observation values. .

[0055] in For the observation vector, For the observation matrix, (for observation noise), used to filter and correct the solution results of the RK4 numerical solution unit, reducing prediction deviations caused by electromagnetic interference and sensor noise.

[0056] The solution step size dynamic adjustment unit adjusts the step size based on the absolute value of the inductor current. Adaptive adjustment of RK4 step size: when >0.5A / ms (during rapid current changes, such as the initial startup phase), the step size is reduced to 3μs to capture key transient details such as current peaks and voltage dips; when 0.1A / ms < ≤0.5A / ms (current change phase), step size remains constant; when ≤0.1A / ms (close to steady state), increase the step size to reduce computational resource consumption.

[0057] The real-time optimization unit adopts a multi-core parallel computing architecture, which allocates different algorithm tasks to different processor cores and implements task scheduling through a real-time operating system to meet the requirements of closed-loop startup real-time modeling.

[0058] The multi-topology adaptation module supports modeling and switching of three mainstream DC-DC converter topologies: boost, buck, and buck-boost, improving system versatility. The multi-topology adaptation module includes a topology parameter matrix unit, a control strategy switching unit, a model structure adjustment unit, and a topology dynamic switching unit.

[0059] The topology parameter matrix unit pre-stores circuit parameter templates, state equation coefficient matrices, and default values ​​of controller parameters for three topologies. Users can call the corresponding template through the topology selection interface (hardware DIP switch or software command) without having to manually reconstruct the model.

[0060] The control strategy switching unit automatically switches the controller type according to the topology type: the boost topology uses integral control (because the output voltage needs to be increased, integral control can achieve zero steady-state error tracking), the buck topology uses proportional-integral (PI) control (because the output voltage needs to be reduced, PI control takes into account both response speed and steady-state accuracy), and the buck-boost topology uses proportional-integral-derivative (PID) control (because the voltage gain range is wide, PID control improves dynamic stability).

[0061] The model structure adjustment unit optimizes the state equation form based on topological characteristics. For example, the output voltage of the buck converter is directly equal to the capacitor voltage (without a voltage divider), and its enhanced state-space model expression is:

[0062]

[0063] Where k_I and k_p are the integral controller gains, V o , e f is the output reference voltage, v_C is the real-time capacitor voltage, y is the system output variable, vo is the actual output voltage of the DC-DC converter, Vg is the input power supply voltage, VDS,on is the saturation voltage drop when the switch is turned on, fs is the switching frequency, iL is the inductor current, R is the load resistance, d is the duty cycle, and L(T) is the temperature-corrected inductance value. The formula is: C(T) is the temperature-corrected capacitance value, expressed by the formula C(T) = C0(1 + α_C(T - T0)), where α_T is the inductor temperature coefficient, α_C is the capacitor temperature coefficient, T is the current operating temperature, and T0 is the reference temperature. This adjustment simplifies the output layer calculation and reduces model complexity.

[0064] The dynamic topology switching unit supports real-time topology switching during operation, quickly completes parameter template calling, control strategy updates and model structure adjustments, without restarting the modeling process, and is suitable for multi-topology mixed power supply scenarios.

[0065] The model iteration and optimization module ensures the long-term accuracy and environmental adaptability of the model through real-time error monitoring and incremental learning. The model iteration and optimization module includes an error monitoring unit, an incremental learning unit, a historical data feedback unit, and a model health assessment unit.

[0066] The error monitoring unit calculates the error indices between the model's predicted values ​​and the measured values ​​from the multi-parameter sensing module in real time: root mean square error (RMSE), peak error (PE), and steady-state error (SE), and sets thresholds to trigger the model optimization process.

[0067] The incremental learning unit fine-tunes key model parameters based on newly acquired operational data: the process noise covariance matrix Q and the observation noise covariance matrix R of AKF filtering, the step size threshold of RK4, and the controller gain (k_P, k_I). The gradient descent method is used to update the parameters without requiring a full retraining of the model.

[0068] The historical data feedback unit optimizes the basic parameters of the model using historical operating data: it refits the temperature characteristic curves of L(T) and C(T) using the least squares method, and updates the component parameter temperature coefficients (α_T, α_C); based on the modeling error under different load rates, it adjusts the default controller gain in the topology parameter template to make the model adaptable to a wider range of operating conditions.

[0069] The model health assessment unit periodically performs a comprehensive evaluation of the model's operating status, including the statistical error index compliance rate, parameter update frequency, and topology switching success rate, and generates a health report (divided into three levels: "Excellent", "Good", and "Requires Maintenance"). When the health status is "Requires Maintenance", the user is prompted to check the sensor status or recalibrate the component parameters.

[0070] The AKF optimization unit of the multi-algorithm fusion solution module adopts an adaptive noise covariance adjustment mechanism: when the predicted residual... When the residual increases, the weight of the observation noise covariance matrix R is automatically increased (adjustment formula: R(k)=R(k-1)×(1+0.1×|ε(k)| / σ_ε), where σ_ε is the residual standard deviation), enhancing the correction effect of the observation on the filtering result; when the residual decreases, the weight of the process noise covariance matrix Q is increased (adjustment formula: Q(k)=Q(k-1)×(1-0.05×|ε(k)| / σ_ε)), improving the model's ability to track dynamic changes and ensuring stable filtering effect.

[0071] The multi-topology adaptation module also includes a topology condition adaptive adjustment unit: for different operating conditions of the same topology (such as light load, heavy load, and input voltage fluctuation scenarios of a boost converter), it pre-stores multiple sets of optimized controller parameters and solution step size configurations, and monitors the load rate (R_load=V) in real time. o / I o The system automatically calls up the matching parameter configuration based on the input voltage fluctuation (ΔV9=|V9-V9_nom|, where V9_nom is the nominal input voltage).

[0072] The advantages of the multi-parameter sensing module of the present invention include: Extended parameter acquisition range: Breaking through the limitations of traditional methods that only acquire circuit parameters, new parasitic parameters (L_p, R_p), switching parameters (V_DS,on), and environmental parameters (T, H) are added to construct a three-dimensional parameter system of "circuit-component-environment", providing comprehensive input for high-precision modeling.

[0073] Dynamic calibration mechanism: Adopting a dual calibration strategy of "zero-point calibration + reference calibration", the calibration process is automatically executed periodically to eliminate the effects of temperature drift and attenuation of acquisition accuracy, and to ensure the accuracy and stability of parameter acquisition.

[0074] The enhanced state-space modeling module of this invention is constructed using a third-order nonlinear model, and its advantages include: Control input stateification: The duty cycle (d) is upgraded from an external fixed input in the traditional model to a new state variable, and a third-order nonlinear model including inductor current, capacitor voltage and duty cycle is constructed to fully capture the dynamic coupling relationship between control input and circuit state.

[0075] Nonlinear coupling of ambient temperature: By establishing a dynamic mapping relationship between inductance and capacitance parameters and temperature (L(T)=L0(1+α_T(T-T0)), C(T)=C0(1+α_C(T-T0))), the drift effect of temperature on inductance and capacitance parameters is quantified and embedded into the model, thereby improving the modeling adaptability under wide temperature environments.

[0076] Non-ideal parameters are fully embedded: non-ideal parameters such as parasitic inductance of inductors, parasitic resistance of capacitors, and voltage drop during switch conduction are fully incorporated into the state equation to correct the dynamic characteristics of the circuit and closely approximate the actual working state of components.

[0077] The multi-algorithm fusion solution module of this invention achieves a balance between high accuracy, real-time performance, and anti-interference capabilities, with the following advantages: The integration of RK4 and AKF algorithms: The fourth-order Runge-Kutta (RK4) numerical solution algorithm and the adaptive Kalman filter (AKF) optimization algorithm are combined. RK4 ensures the numerical solution accuracy of the nonlinear model, while AKF achieves filtering optimization by introducing measured observations, reducing the impact of electromagnetic interference and sensor noise, thus balancing accuracy and anti-interference capability.

[0078] Adaptive step size adjustment: An adaptive step size adjustment mechanism based on the inductor current change rate is designed to dynamically adjust the solution step size according to the transient characteristics of the circuit, thereby reducing the consumption of computing resources while accurately capturing key details.

[0079] Multi-core parallel computing optimization: A multi-core parallel computing architecture is adopted to allocate different algorithm tasks to different processor cores, and combined with a real-time operating system to achieve efficient task scheduling and improve the real-time performance of modeling.

[0080] The non-ideal parameter dynamic compensation module of the present invention can achieve online identification and real-time correction, and its advantages include: Hybrid parameter identification algorithm: The hybrid parameter identification algorithm is constructed by combining the least squares method and the recursive least squares method (RLS) to achieve high-precision, real-time identification of non-ideal parameters and dynamically track the drift of component parameters.

[0081] Outlier handling and rationality verification: An outlier handling and rationality verification mechanism is added. Outliers are removed through data preprocessing, and the identification parameters are verified within the specification range to avoid model inaccuracies caused by abnormal data.

[0082] Real-time full-parameter correction: The non-ideal parameters after verification are fed into the enhancement model in real time, and the relevant terms of the state equation are corrected in real time to achieve dynamic matching between the model and the actual circuit characteristics.

[0083] The multi-topology adaptation module of this invention can achieve multi-topology compatibility and dynamic switching, and its advantages include: Topology parameter templates: Pre-store parameter templates for three mainstream topologies: boost, buck, and buck-boost, including core configurations such as circuit parameters, controller parameters, and state equation coefficients. Users can call the corresponding templates through simple operations without manually reconstructing the model.

[0084] Automatic control strategy switching: The controller type is automatically switched according to the topology type, matching the appropriate control strategy for different topologies, meeting the dynamic characteristics requirements of different topologies, and achieving a smooth transition during the switching process.

[0085] Dynamic topology switching and operating condition adaptation: Supports real-time topology switching during operation, quickly completes parameter template calling, control strategy updates and model structure adjustments, and adapts to multi-topology mixed power supply scenarios; at the same time, for different operating conditions of the same topology, multiple sets of optimized parameters are pre-stored to achieve adaptive adjustment of operating conditions.

[0086] The model iterative optimization module of this invention can achieve long-term accuracy stability and health management, with the following advantages: Real-time error monitoring and incremental learning: Real-time monitoring of the error index between the model's predicted values ​​and the measured values, setting an error threshold, triggering the model optimization process when the threshold is exceeded, performing incremental learning based on newly collected running data, fine-tuning the key parameters of the model, without the need to retrain the model completely.

[0087] Historical data feedback optimization: Utilize historical operating data to periodically optimize the model's basic parameters, refit the temperature characteristic curve, adjust the default values ​​of the topology parameter template, and improve the model's long-term operating accuracy and adaptability to operating conditions.

[0088] Model health assessment and maintenance tips: Regularly generate model health reports, statistically analyze indicators such as error compliance rate, parameter update frequency, and topology switching success rate, evaluate the model's operating status in a tiered manner and provide maintenance suggestions to reduce maintenance costs and ensure long-term stable operation of the model.

[0089] Example 2 The method of using a closed-loop DC-DC converter startup modeling system according to this embodiment includes the following steps: S1: Multi-parameter acquisition and initialization: S11: Collect circuit parameters, initial operating condition parameters, and environmental parameters of the DC-DC converter through a multi-parameter sensing module.

[0090] S12: Perform parameter calibration: Ensure data acquisition accuracy through zero-point calibration and reference calibration procedures.

[0091] S13: Initialize the model: Calculate the load voltage division coefficient a=R / (R+r_C+R_p); determine the temperature-corrected inductance value L(T)=L0(1+α_T(T-T0)) and capacitance value C(T)=C0(1+α_C(T-T0)); initialize AKF parameters Q, R and observation matrix.

[0092] S2: Enhanced State-Space Model Construction: S21: Basic Model Construction: Establishing the second-order dynamic equations of I_L and v_C based on KVL and KCL:

[0093]

[0094] in, The rate of change of inductor current. This represents the rate of change of voltage on the capacitor side.

[0095] S22: Control input stateification: The duty cycle d is taken as a new state variable, and combined with the integral control characteristics, it is integrated with the second-order equation to form a third-order enhanced state-space model. Substituting the non-ideal parameters and temperature-corrected L(T) and C(T), the corresponding topological complete model is obtained.

[0096] S23: Model Construction and Verification: Verify the correctness of the model structure and ensure that the equation derivation and parameter embedding are error-free.

[0097] S3: Multi-algorithm fusion solution and real-time modeling: S31: Numerical solution of RK4 (Fourth-Order Runge-Kutta Method): Initialize the state vector with a base step size of 5μs. Calculate the state variables at each time step using the RK4 iterative formula:

[0098] Where h is the solution step size, f(·) is the state transition function of the augmented state-space model, and a set of predicted values ​​of i_L, v_C, and d are output for each iteration.

[0099] S32: AKF filter optimization: Using the real-time i_L,meas and v_C,meas acquired by the multi-parameter sensing module as observations Perform the AKF filtering steps: predict: ( (This is the state transition matrix).

[0100] renew: , , Optimized predicted value after output filtering .

[0101] S33: Dynamic step size adjustment: Real-time calculation of inductor current change rate The step size is dynamically adjusted according to the set threshold range.

[0102] S34: Parallel task scheduling: The RK4 solving task and AKF filtering task are allocated to different cores of the STM32H743 through FreeRTOS, while the parameter acquisition task is allocated to idle time periods to ensure the real-time performance of the modeling process.

[0103] S4: Dynamic compensation for non-ideal parameters: S41: Data preprocessing: Perform outlier removal and smoothing on the i_L,meas and v_C,meas data collected by the multi-parameter sensing module.

[0104] S42: Non-ideal parameter identification: The RLS algorithm is used to identify parameters in the preprocessed data, and L_p, R_p, V_DS,on, and α_T are updated every 10ms.

[0105] Constructing the regression vector: Substitute into the RLS iterative formula to calculate This yields the non-ideal parameter values ​​at the current moment.

[0106] S43: Parameter rationality verification: Compare the identified parameters with the specification range. If they are within the range, proceed to S44; if they are outside the range, replace them with the historical best parameters.

[0107] S44: Real-time model correction: Substitute the verified non-ideal parameters into the enhanced state-space model, update the resistance term (r_L+L_p×2πf_s), capacitance term (r_C+R_p), and voltage term (V9-V_DS,on) in the state equation, and correct the temperature coefficient α_T of L(T) to ensure that the model is consistent with the actual circuit characteristics.

[0108] S5: Multi-topology adaptation and model optimization: S51: Topology Identification and Parameter Loading: Based on the converter topology type, call the corresponding parameter template and switch the control strategy and model structure.

[0109] S52: Error Monitoring and Optimization Trigger: Calculate the Predicted Value After Filtering , Compared with the measured values ​​after preprocessing , RMSE (Root Mean Square Error):

[0110] If the RMSE exceeds the threshold within a consecutive set period, incremental learning is triggered.

[0111] S53: Incremental parameter update: Based on the newly acquired data, the gradient descent method is used to fine-tune the Q and R matrices of AKF and the RK4 step size threshold.

[0112] Define loss function Calculate the partial derivatives of J with respect to Q, R, and the step size threshold, according to... (η is the learning rate) Update the parameters until the RMSE meets the requirements.

[0113] S54: Historical data feedback optimization: Refit the temperature characteristic curves of L(T) and C(T) using historical data: Perform polynomial fitting on the measured values ​​of L and C at different temperature points (e.g., L(T) = a0 + a1T + a2T²); update the α_T and α_C parameters in the model to ensure that the accuracy of L(T) and C(T) after temperature correction meets the requirements.

[0114] S6: Modeling Results Output and Application S61: Time-domain waveform generation: Generate time-domain waveforms from the predicted i_L, v_C, and d values ​​output by S3 according to the time series, and mark key indicators: minimum voltage dip and its occurrence time, peak inductor current and its occurrence time, and settling time (1% error band).

[0115] S62: Key Performance Indicator Statistics: Outputs statistical results of core performance indicators.

[0116] S63: Data Interaction and Control Application: Outputs modeling results to external control systems via standardized interfaces (CAN bus, Ethernet, or JSON protocol): When the predicted voltage dip exceeds the set threshold, it outputs instructions to adjust controller parameters; when the predicted current peak exceeds the set threshold, it outputs instructions to adjust the switching frequency; it supports external systems calling model prediction results to formulate load scheduling strategies.

[0117] S64: Model Health Report Generation: Regularly calculate the model error compliance rate, parameter update frequency, and topology switching success rate, generate a health report, and provide corresponding maintenance suggestions.

[0118] Optionally, in step S32, the observation matrix of the AKF optimization unit... It can adaptively adjust according to the topology type: for buck topologies, because v o =v_C, observation vector Observation matrix For the buck-boost topology, because v o =-d / (1-d) (V9-i_L r_L), observation matrix Introducing a dynamic factor of d, and adjusting it to This ensures that the observation equation matches the topological characteristics, thereby improving the filtering accuracy.

[0119] Optionally, in step S42, the non-ideal parameter identification unit also supports temperature compensation for switching parameters: by establishing a relationship model between the switch on-state voltage drop V_DS,on and temperature (V_DS,on(T)=V_DS,on0×(1+α_V(T-T0)), where α_V is the voltage temperature coefficient), the influence of temperature is incorporated into the parameter identification model, thereby improving the modeling adaptability under wide temperature environments.

[0120] Optionally, in step S5, the multi-topology adaptation module also supports user-defined topology extension: it provides a model editing interface, allowing users to input the circuit structure of the custom topology (such as the connection method of inductors, capacitors, and switches), the derivation results of KVL / KCL equations, and the system automatically generates the corresponding enhanced state space model structure, imports the default controller parameters and solution configuration, realizes personalized topology modeling, and expands the system application scenarios (such as isolated DC-DC converters).

[0121] Example 3 This embodiment provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the steps of the method described in Embodiment 2. The processor includes an embedded processor, and the storage medium includes a non-volatile storage medium to ensure long-term stable storage of the program and data.

[0122] Example 4 This embodiment of a DC-DC converter control system includes a hardware layer, a software layer, an application layer, and a modeling system for starting up a closed-loop DC-DC converter as described in Embodiment 1.

[0123] Hardware layer: Composed of DC-DC converter main circuit (inductor, capacitor, switch), multi-parameter sensing module sensor array (current, voltage, temperature and humidity sensors), embedded processor, and communication interface, it realizes parameter acquisition, modeling calculation and data interaction.

[0124] The software layer includes a real-time operating system, drivers, modeling algorithm libraries (RK4, AKF, RLS implementation code), and a parameter database (storing circuit parameters and historical running data), providing automated execution of the modeling process and parameter management.

[0125] Application layer: Provides a human-machine interface (displaying modeling waveforms, key indicators, and health reports), an external system interface (communicating with UPS and BMS), and a fault diagnosis module (identifying sensor faults or component parameter drift based on modeling errors).

[0126] This control system dynamically adjusts the controller parameters (k_P, k_I, switching frequency) of the DC-DC converter based on the modeling results, supports a wide operating temperature range, and ensures modeling accuracy under different environments through temperature compensation and non-ideal parameter compensation mechanisms. It can be directly applied to UPS, on-board power supply for new energy vehicles, aerospace electronic equipment and other fields.

[0127] Example 5 like Figure 2 The diagram shows a comparison of the output voltage during the startup process of the Boost converter. The black dashed line (with noise) simulates the actual sensor measurement, the blue solid line represents the predicted value from the proposed third-order enhanced state-space model, and the red solid line represents the predicted value from the traditional second-order small-signal model. The true value is composed of the output of the proposed third-order model superimposed with random noise, simulating the voltage signal in actual measurement. The proposed third-order model (blue solid line) accurately tracks the dynamic changes of the true value, including the rapid voltage rise during startup, the overshoot amplitude (approximately 0-2ms), and the steady-state convergence process, almost perfectly matching the true value. The traditional second-order model (red solid line) deviates significantly from the true value during the initial startup phase (0-2ms), with an excessively slow rise rate and an inability to accurately reflect voltage overshoot details. While it approaches the true value after steady-state, the transient error is significant.

[0128] Figure 3 Comparison of inductor currents: The current waveform of the third-order model (blue solid line) of this invention closely matches the actual value with noise (black dashed line), especially during the rapid current rise phase (0~1ms) and near the peak value, where the error is extremely small. The current response of the traditional second-order model (red solid line) is significantly lagging, the peak amplitude is low, and there is a large deviation between the rise process and the actual value, indicating that its modeling ability under large-signal dynamics is insufficient.

[0129] Figure 4 Output voltage prediction error: The error of the model in this invention (blue curve) fluctuates within a small range (approximately ±0.2V) throughout the startup process and quickly converges to near zero, indicating that the model has good transient and steady-state accuracy. The error of the traditional model (red curve) is as high as -2V or more in the early stage of startup (0~2ms), with violent fluctuations, and only gradually decreases when approaching steady state, proving that the traditional small-signal model cannot accurately describe the large-signal nonlinear behavior during startup.

[0130] Unless otherwise specified, the equipment components involved in the above embodiments are all conventional equipment components, and the structural settings, working methods or control methods involved are all conventional settings, working methods or control methods in the art unless otherwise specified.

[0131] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Any other modifications or equivalent substitutions made by those skilled in the art to the technical solutions of the present invention, as long as they do not depart from the spirit and scope of the technical solutions of the present invention, should be covered within the scope of the claims of the present invention.

Claims

1. A modeling system for the startup of a closed-loop DC-DC converter, characterized in that, include: The multi-parameter sensing module is used to collect circuit parameters, initial operating condition parameters, real-time operating parameters and environmental parameters of the DC-DC converter from all dimensions to build a basic dataset for modeling. The enhanced state-space modeling module is used to construct a third-order nonlinear enhanced model that integrates the dynamic characteristics of control inputs based on the modeling base dataset. The multi-algorithm fusion solution module is used to integrate numerical solution and filtering optimization algorithms to balance the modeling accuracy, real-time performance and anti-interference capability of the third-order nonlinear enhancement model; and feeds back the solution step size to the multi-parameter sensing module and the filtering correction results to the enhancement state space modeling module. The multi-topology adaptation module is used to support the modeling and switching of DC-DC converter topologies, and dynamically switches the structure of the state equations of the third-order nonlinear enhancement model and the solution logic of the controller according to the topology type. The model iterative optimization module is used for real-time error monitoring and incremental learning. Based on the error between the output of the multi-algorithm fusion solution module and the measured value calculation of the multi-parameter sensing module, the optimized model parameters and controller gain are written back to the enhanced state space modeling module.

2. The modeling system for the startup of a closed-loop DC-DC converter according to claim 1, characterized in that, The circuit parameters include power supply voltage, load resistance, inductor parameters, capacitor parameters, and switching parameters; the initial operating parameters include initial output voltage, initial inductor current, and initial duty cycle. The real-time operating parameters include real-time inductor current, real-time capacitor voltage, and real-time duty cycle; the environmental parameters include ambient temperature and relative humidity, which are used for component parameter temperature drift compensation.

3. The modeling system for the startup of a closed-loop DC-DC converter according to claim 1, characterized in that, The enhanced state-space modeling module includes a basic model building unit, a control input stateification unit, and a nonlinear coupling unit. The basic model building unit is used to establish a second-order circuit dynamic equation based on Kirchhoff's voltage law and current law, using inductor current and capacitor voltage as state variables. The control input stateification unit is used to add duty cycle as a new state variable, treating the duty cycle as a limitation of external fixed input, and constructing a third-order nonlinear enhanced state-space model containing inductor current, capacitor voltage, and duty cycle. The nonlinear coupling unit is used to introduce the dynamic mapping relationship between ambient temperature and component parameters, quantifying and embedding the drift effect of temperature on inductor and capacitor parameters into the model.

4. The modeling system for the startup of a closed-loop DC-DC converter according to claim 3, characterized in that, For the boost converter, the model expression for constructing the stateful control input unit is: Where 'a' is the load voltage division factor, and the formula is: R is the resistance value of the upper voltage divider resistor, r_C is the equivalent series resistance value of the output capacitor, R_P is the resistance value of the lower voltage divider resistor; k_I is the gain of the integral controller, V o , e f is the output reference voltage, and L(T) is the temperature-corrected inductance value. The formula is: C(T) is the temperature-corrected capacitance value, and the formula is C(T)=C0(1+α_C(T-T0)), where α_T is the temperature coefficient of inductance, α_C is the temperature coefficient of capacitance, T is the current operating temperature, and T0 is the reference temperature.

5. The modeling system for the startup of a closed-loop DC-DC converter according to claim 1, characterized in that, The multi-algorithm fusion solution module includes: an RK4 numerical solution unit, an adaptive Kalman filter optimization unit, a solution step size dynamic adjustment unit, and a real-time optimization unit. The RK4 numerical solution unit uses the fourth-order Runge-Kutta algorithm to solve the enhanced state-space model, and realizes the time-domain evolution prediction of state variables through iterative calculation. The AKF optimization unit uses the real-time operating parameters collected by the multi-parameter sensing module as observation values, and constructs observation equations to filter and correct the solution results of the RK4 numerical solution unit. The solution step size dynamic adjustment unit is used to adaptively adjust the step size of RK4 according to the absolute value of the inductor current. The real-time optimization unit is used to allocate different algorithm tasks to different processor cores and realize task scheduling through a real-time operating system.

6. The modeling system for the startup of a closed-loop DC-DC converter according to claim 5, characterized in that, The observation equation is: ,in For the observation vector, For the observation matrix, To observe noise.

7. The modeling system for the startup of a closed-loop DC-DC converter according to claim 5, characterized in that, The solution step size dynamic adjustment unit is used to reduce the step size to 3μs when |(iL)|>0.5A / ms; keep the step size constant when 0.1A / ms<|(iL)|≤0.5A / ms; and increase the step size when |(iL)|≤0.1A / ms.

8. The modeling system for the startup of a closed-loop DC-DC converter according to claim 1, characterized in that, The multi-topology adaptation module supports modeling and switching of three mainstream DC-DC converter topologies: boost, buck, and buck-boost. The multi-topology adaptation module includes a topology parameter matrix unit, a control strategy switching unit, a model structure adjustment unit, and a topology dynamic switching unit. The topology parameter matrix unit is used to pre-store circuit parameter templates, state equation coefficient matrices, and default values ​​of controller parameters for three topologies; the control strategy switching unit automatically switches the controller type according to the topology type: boost topology uses integral control, buck topology uses proportional-integral control, and buck-boost topology uses proportional-integral-derivative control; the model structure adjustment unit is used to optimize the form of the state equation according to the topology characteristics; the topology dynamic switching unit is used to support real-time topology switching during operation.

9. The modeling system for the startup of a closed-loop DC-DC converter according to claim 1, characterized in that, The model iterative optimization module includes an error monitoring unit, an incremental learning unit, a historical data feedback unit, and a model health assessment unit; the error monitoring unit calculates the error index between the model's predicted value and the measured value of the multi-parameter perception module in real time, and sets a threshold to trigger the model optimization process; The incremental learning unit fine-tunes key model parameters based on newly acquired operational data and uses gradient descent to update parameters; the historical data feedback unit optimizes basic model parameters using historical operational data and adjusts the default controller gain in the topology parameter template based on modeling errors under different load rates; the model health assessment unit is used to periodically conduct a comprehensive assessment of the model's operational status and prompts for checking sensor status or recalibrating component parameters.

10. A method of using the modeling system for starting a closed-loop DC-DC converter according to any one of claims 1 to 9, characterized in that, Includes the following steps: The multi-parameter sensing module collects circuit parameters, initial operating parameters, real-time operating parameters, and environmental parameters of the DC-DC converter from all dimensions to construct a basic dataset for modeling. The enhanced state-space modeling module constructs a third-order nonlinear enhanced model that integrates the dynamic characteristics of the control input based on the modeling dataset. The multi-algorithm fusion solution module integrates numerical solution and filtering optimization algorithms to balance the modeling accuracy, real-time performance and anti-interference capability of the third-order nonlinear enhancement model; and feeds back the solution step size to the multi-parameter sensing module and the filtering correction result to the enhancement state space modeling module. The multi-topology adaptation module is used to model and switch the DC-DC converter topology, and dynamically switches the structure of the state equation of the third-order nonlinear enhancement model and the solution logic of the controller according to the topology type. The model iterative optimization module utilizes real-time error monitoring and incremental learning. Based on the error between the output of the multi-algorithm fusion solution module and the measured value calculation of the multi-parameter sensing module, the optimized model parameters and controller gain are written back to the enhanced state space modeling module.

Images