A Model Predictive Current Control Method and System Based on Online Parameter Identification

Abstract

This invention discloses a model predictive current control method and system based on online parameter identification. The method independently identifies the resistor using steady-state average information within a switching cycle, and then identifies the inductor using transient current ripple information combined with a logarithmic linearization method. Next, the identified real-time resistance and inductance values ​​are input into a deadbeat-free model predictive current control algorithm to update the coefficients of the predictive model. Based on the latest model parameters and reference current, the controller calculates the ideal voltage vector required to eliminate current error at the next moment, and after amplitude limiting, drives the switching transistor through a PWM module, thereby achieving high-performance control of the coil current. The system includes an online parameter identification module and a model predictive control module. By using this invention, high accuracy and high stability of the magnetic bearing power amplifier are ensured under all operating conditions. This invention can be widely applied in the field of electronic power technology.

Description

Technical Field

[0001] This invention relates to the field of power electronics technology, and in particular to a model predictive current control method and system based on online parameter identification. Background Technology

[0002] Magnetic bearings, as a non-contact support component, have been widely used in high-end industrial fields such as aerospace, flywheel energy storage, and high-speed motors due to their advantages of no mechanical friction, no lubrication required, low loss, controllable dynamic characteristics, and long lifespan. A magnetic bearing system is a complex mechatronic closed-loop control system, in which the power amplifier plays a crucial role. As a bridge connecting the digital controller and the electromagnetic coil actuator, the power amplifier is responsible for converting the reference current signal into the actual current driving the electromagnet coil. Its dynamic response speed, tracking accuracy, and stability directly determine the output quality of the electromagnetic force, thus affecting the levitation performance of the entire magnetic bearing system.

[0003] Current magnetic bearing power amplifiers primarily employ switching power amplifiers, utilizing the high-frequency switching of power transistors to regulate coil current. For current loop control, since magnetic bearings are typical nonlinear systems, Model Predictive Control (MPCC) has been introduced into magnetic bearing power amplifier control in recent years due to its excellent dynamic performance and flexibility in handling nonlinear problems. The core idea of ​​MPCC is to predict the current response at the next moment based on a discrete mathematical model of the controlled object, thereby selecting the optimal control voltage vector. However, existing MPCC techniques have significant technical drawbacks in magnetic bearing applications: MPCC control performance is highly dependent on the accuracy of the system model. In actual magnetic bearing operation, the resistance and inductance parameters of the controlled object—the electromagnetic coil—are not fixed but exhibit significant time-varying and nonlinear characteristics. On the one hand, the small displacement of the rotor relative to the electromagnet stator during actual operation directly changes the air gap, causing changes in the magnetic reluctance in the magnetic circuit, leading to a significant change in inductance. On the other hand, changes in the excitation current affect the magnetic saturation of the core, resulting in nonlinear changes in inductance. Furthermore, coil resistance heating also causes nonlinear changes in resistance, leading to MPCC model parameter mismatch. Therefore, performance degradation of fixed-parameter MPCC magnetic bearing power amplifiers is inevitable. Summary of the Invention

[0004] In view of this, in order to solve the problem that in existing magnetic bearing power amplifiers, the resistance and inductance parameters of the electromagnetic coil change with operating conditions (such as air gap changes, temperature rise, magnetic saturation, etc.), leading to model mismatch in traditional fixed-parameter MPCC, which in turn causes increased current tracking steady-state error, decreased system stability, or even instability, the present invention proposes a model predictive current control method based on online parameter identification. This method includes the following steps: An adaptive closed-loop system comprising an online parameter identification module and a model predictive control module is constructed. First, a two-step decoupled identification strategy, "steady-state-transient," is employed to address the "underrank" problem when simultaneously identifying resistance and inductance parameters: the resistor is independently identified using the steady-state average value information within the switching cycle, and the inductor is identified using transient current ripple information combined with a logarithmic linearization method. Next, the identified real-time resistance and inductance values ​​are input into a deadbeat-free model predictive current control (DB-MPCC) algorithm to update the coefficients of the predictive model. Based on the latest model parameters and reference current, the controller calculates the ideal voltage vector required to eliminate current error at the next moment, and after amplitude limiting, drives the switching transistor through a PWM module, thereby achieving high-performance control of the coil current.

[0005] The online parameter identification employs a decoupling strategy: Resistance identification: Based on the physical characteristic that the average voltage of an inductor is zero during a steady-state PWM cycle, the coil resistance is calculated in real time using Ohm's law based on the average voltage and average current over one switching cycle, avoiding coupling interference from the inductance term. Inductance identification: Based on the resistance identification results, the transient exponential growth model of the coil current under switching action (current ripple) is used. This model is transformed into a linear equation through logarithmic linearization, and the least squares method (LSM) is used to perform regression analysis on the sampled data to solve for the slope, thereby accurately calculating the coil inductance.

[0006] Applying the above method, the present invention also proposes a model predictive current control system based on online parameter identification, which includes an online parameter identification module and a model predictive control module.

[0007] Based on the above scheme, this invention provides a model predictive current control method and system based on online parameter identification. By updating the resistance parameters in real time through online identification, it solves the steady-state current tracking error problem caused by the drift of the resistance thermal effect in traditional fixed-parameter MPCC, and realizes deadbeat current tracking under all operating conditions. This invention can track the nonlinear changes of inductance parameters caused by rotor displacement (air gap change) and magnetic saturation in real time, avoiding system oscillation and instability caused by the closed-loop poles shifting out of the unit circle due to model inductance parameter mismatch. The decoupled identification strategy proposed in this invention, which combines periodic steady-state analysis and transient logarithmic linearization, effectively overcomes the problem of identification results diverging or getting trapped in local optima caused by resistance-inductance coupling and mutual interference in traditional methods, ensuring the convergence and robustness of the identification algorithm. In addition, this invention retains the inherent fast dynamic response characteristics of model predictive current control (MPCC), while endowing the system with the ability to adapt to changes in the physical parameters of the controlled object, so that it can always maintain optimal control performance in the complex and ever-changing operating environment of magnetic bearings. Attached Figure Description

[0008] Figure 1 This is a typical structural diagram of a magnetic bearing power amplifier system; Figure 2 This is a flowchart illustrating the steps of a model predictive current control method based on online parameter identification according to the present invention. Figure 3 This is a schematic diagram of the main circuit of the power amplifier in an embodiment of the present invention; Figure 4 This is a flowchart illustrating the online parameter identification algorithm in an embodiment of the present invention. Figure 5 This is a schematic diagram of the steady-state current change of a two-level power amplifier during one switching cycle in an embodiment of the present invention; Figure 6 This is a schematic diagram of current sampling for coil inductance identification in an embodiment of the present invention; Figure 7 This is a framework diagram of the adaptive DB-MPCC control algorithm in an embodiment of the present invention; Figure 8 This is a diagram showing the system pole distribution under parameter mismatch conditions in an embodiment of the present invention. Detailed Implementation

[0009] See Figure 1 This is a typical structural diagram of a magnetic bearing power amplifier system. It mainly consists of a switching power amplifier main circuit, a digital control system, and an electromagnet coil as the load. Precise control of the electromagnet coil current is achieved through closed-loop control. The specific working principle is as follows: the system input signal is the target current signal given by the upstream displacement controller. The system's feedback signal is the current sensor signal transmitted by the system via an ADC. The sampled current signal is obtained by real-time sampling of the measured actual current. By comparing the two, the current tracking error can be obtained and used as the input to the controller.

[0010] The controller processes the error signal according to a preset control algorithm, calculates and generates a target voltage command to compensate for the error. To convert the digitized voltage signal into commands. This is converted into a physical signal that can drive power devices, and then processed by a PWM module. Comparing with a high-frequency triangular carrier signal, four channels with varying duty cycles are generated. The dynamically updated PWM signal is converted into a driving signal for the switching transistor by the gate drive circuit. - The gate drive signal is turned on. - The system switches the electromagnetic coil on and off according to the logic state of the drive signal, changing the actual voltage applied across the coil to drive it. Towards Convergence allows for precise adjustment of the electromagnetic force of the coil.

[0011] While existing technologies have optimized MPCC performance by increasing sampling frequency or improving modulation methods, none of them fundamentally solve the problems of model mismatch and controller performance degradation caused by time-varying parameters. Therefore, how to accurately identify the time-varying parameters of the electromagnetic coil online and update the control model in real time to eliminate the effects of model mismatch is a technical challenge that urgently needs to be solved to improve the performance of magnetic bearing power amplifiers.

[0012] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0013] It should be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings. Unless otherwise specified, the embodiments and features described in this application can be combined with each other.

[0014] It should be understood that the terms "system," "apparatus," "unit," and / or "module" used in this application are a method of distinguishing different components, elements, parts, sections, or assemblies at different levels. However, if other terms can achieve the same purpose, they may be replaced by other expressions.

[0015] As indicated in this application and claims, unless the context clearly indicates otherwise, the words "a," "an," "a," and / or "the" are not specifically singular and may include the plural. Generally, the terms "comprising" and "including" only indicate the inclusion of expressly identified steps and elements, which do not constitute an exclusive list, and the method or apparatus may also include other steps or elements. An element defined by the phrase "comprising an..." does not exclude the presence of other identical elements in the process, method, product, or apparatus that includes the element.

[0016] In the description of the embodiments of this application, "a plurality of" refers to two or more. The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature.

[0017] Furthermore, flowcharts are used in this application to illustrate the operations performed by the system according to embodiments of this application. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, the steps can be processed in reverse order or simultaneously. Additionally, other operations can be added to these processes, or one or more steps can be removed from them.

[0018] Reference Figure 2 This is a flowchart illustrating an optional example of the model prediction current control method based on online parameter identification proposed in this invention. This method can be applied to computer devices, and the control method proposed in this embodiment may include, but is not limited to, the following steps: Step S1: Identify the resistor using the steady-state average value information within the switching cycle; Step S2: Based on the resistance identification results, identify the inductance using transient current ripple information combined with the logarithmic linearization method; Step S3: Based on the resistance identification results and inductance identification results, use the discretized mathematical model to predict the current at the next moment, and calculate the target voltage command that makes the predicted current at the next moment equal to the reference current.

[0019] Figure 3 This is a schematic diagram of the main circuit of a power amplifier. Its load is an electromagnetic coil, which can be considered as an equivalent series model of inductor L and resistor R. Taking a two-level SPA as an example, within one switching cycle, the power switching transistor... and Conduction, and When the circuit is turned off, the DC power supply charges the coil, and the coil current increases rapidly; outside the duty cycle, and Conduction, and When switched off, the coil passes through. and As energy is fed back to the power source, the absolute value of the coil current decreases rapidly. According to Kirchhoff's voltage law, the voltage balance equation across the coil during one switching cycle is: (1) In the formula, U is the voltage applied across the coil. The instantaneous current flowing through the coil, and These are the forward voltage drops of the switching transistor and its parallel diode, respectively; their values ​​are generally small and can be ignored. T is the switching period. The charging time for the coil.

[0020] Solving equation (1), we can obtain the electromagnetic bearing coil current when a constant voltage U is applied. The general solution is: (2) (3) Where C represents an arbitrary constant, This exponential model accurately describes the transient behavior of the coil current and serves as the theoretical basis for subsequent parameter identification.

[0021] In some feasible embodiments, the parameter identification in steps S1 and S2 specifically includes: When performing online parameter identification of an electromagnetic coil, the ideal goal is to base it on the circuit's voltage balance equation. The method identifies two parameters, resistance R and inductance L, in real time. However, when attempting to treat both R and L as unknown parameters and directly perform multiple linear regression using transient process data, the two regressors—instantaneous current—become problematic. Its rate of change There is an inherent linear correlation between them. This multicollinearity in the data causes the coefficient matrix of the identification equation system to tend to be singular, that is, it produces the "underrank" problem.

[0022] The underrank problem leads to completely unreliable identification results, causing the identification to get stuck in a local minimum: First, extremely small noise in the input data (such as ADC quantization error or ADC measurement noise) will be drastically amplified, causing significant changes in the final calculated R and L values. Second, the model cannot distinguish the individual contributions of R and L. The identification results often show a high negative correlation between the errors in R and L. During the identification process, R is frequently overestimated, and L is correspondingly underestimated to compensate. Although the combined... It may still be able to equal U perfectly, but the values ​​of R and L are completely wrong. This fundamental flaw requires abandoning this coupled identification approach and instead seeking an identification strategy that can decouple the identification processes of R and L.

[0023] See Figure 4 This is a framework diagram of the online parameter identification algorithm of the present invention. To fundamentally solve the "underrank" problem, the present invention proposes a resistance-inductance decoupling identification strategy based on periodic steady-state analysis. Its core idea is that since the resistance change caused by resistor heating mainly occurs after the current reaches a steady state, the transient change process of the current is rapid and the power consumption is low. Therefore, see [link to relevant documentation]. Figure 5 This refers to the steady-state current change of a two-level power amplifier over one cycle. It utilizes the fact that when an RL circuit, driven by a stable PWM signal, enters periodic steady-state operation, the average voltage across the inductor is zero within a complete switching cycle, as detailed below: First, let's look at the differential equation of the coil: By averaging the left and right terms over one switching cycle T, we can obtain: (4) In the formula: The average voltage across the coil during one switching cycle. Let n be the average current across the coil, n = 0, 1, 2, ... By the principles of calculus: (5) according to Figure 3 In a steady-state process, since the inductor is an energy storage element, the energy absorbed in one cycle is equal to the energy released, therefore: (6) Therefore, we can further simplify to obtain: (7) Equation (7) is the theoretical basis for the resistance-inductance decoupling identification adopted in this invention. This equation shows that as long as the voltage applied to both ends of the coil and the current flowing through the coil are obtained, R can be calculated. This calculation process is completely independent of L.

[0024] In a PWM-driven SPA system, the average voltage It is a known quantity determined by the controller. It is equal to the DC bus voltage. Multiply by the coil charging duty cycle D, that is: (8) According to equation (7), the key to identification lies in how to determine the average current. Due to the switching action of PWM, current ripple exists in the coil circuit. This article provides three strategies for measurement and calculation. Each of the three methods has its advantages and can be used for convenient and efficient calculations. .

[0025] Method 1: High-Frequency Oversampling. This method is based on the mathematical definition of the average value. Within one PWM switching cycle, it utilizes a high-speed ADC and timer of a digital controller to perform sampling at fixed time intervals much higher than the switching frequency. This yields a large number of evenly spaced current sampling points within one cycle. Then, these N sampled values ​​are directly added together and averaged to accurately approximate the true average current, i.e.: (9) The advantages of this method are its extremely high accuracy, complete independence from the specific shape of the current waveform, wide applicability, and increased accuracy with more sampling points. The disadvantages are its extremely high performance requirements for the digital controller, consuming significant amounts of ADC and CPU resources.

[0026] Method 2: Segmented Weighted Average Method. The core idea of ​​this method is to instead of making a rough approximation of the current waveform for the entire cycle, divide a complete PWM switching cycle T into two independent parts: the coil charging stage and the coil charging stage. Coil discharge stage: The average current in each of the two phases is calculated, and then a weighted average is performed based on the proportion of time each phase occupies, so as to obtain the true average current of the entire cycle.

[0027] Right now: (10) To obtain the average current across the two stages, ADC sampling is performed at one-third and two-thirds of each stage, and the average is taken, providing a good approximation for each stage. The advantage of this approach is that the second stage can compensate for errors introduced in the first stage, making the final calculated result very close to the true average current. The advantages of this method are high accuracy; it does not require the current rise and fall slopes to be the same or the waveform to be perfectly symmetrical. Regardless of the waveform, this method yields relatively accurate results and has a very wide range of applications. The disadvantage is that it places high demands on the algorithm; accurately arranging the sampling time without interfering with other periodic control algorithms is the main challenge.

[0028] Method 3: Midpoint Sampling Method: The principle is that if the current ripple is a symmetrical or approximately symmetrical triangle, then the average value of the waveform is equal to the midpoint value. This method is more suitable for applications with relatively low resistance.

[0029] The average current during the switching cycle can be estimated using the three methods described above. Substituting into (7), we can find R in reverse.

[0030] Based on the above decoupling strategy, after successfully identifying R, it is used as a known parameter for identifying inductor L. This paper proposes an identification strategy based on a combination of logarithmic linearization and least squares method. Due to the transient current of the RL circuit under step voltage excitation... The model follows an exponential growth pattern. The inherent nonlinearity of this model is a major challenge for direct parameter fitting, rendering traditional linear methods unsuitable. To address this issue, this paper linearizes the nonlinear model into a linear one and combines it with the least squares method for efficient and accurate parameter estimation.

[0031] Dependent variable The independent variable, time t, and slope can be obtained through measurement and calculation. .

[0032] At this point, the problem of nonlinear identification has moved beyond the initial nonlinear data. Perform nonlinear fitting to transform the transformed linear data. To perform efficient linear regression analysis, the least squares method is used to identify the slope B, and then the inductance L is solved.

[0033] The inductor parameter estimation process based on LSM design in this invention is as follows: Data preparation: In the system, data is prepared by ADC at fixed time intervals. N instantaneous current samples are collected to obtain a set of discrete current measurement data. Where k = 0, 1, 2, ..., N-1. See also Figure 6 A schematic diagram for identifying current sampling for coil inductance.

[0034] According to formula (12), these original data points are converted into data points in a linear coordinate system. ,in: (13) (14) Error function definition: for any straight line The residual between it and the k-th data point is: (15) The goal of the least squares method is to find a set of parameters (A, B) that minimizes the total sum of squared errors S. Solving for the slope parameter: By taking the partial derivatives of the error function S with respect to A and B respectively, and setting them equal to 0, we can obtain a system of equations about A and B. Solving this system of equations yields the unique analytical solution for the slope B: (17) According to Equation 12, the inverse inductance L is: (18) In some feasible embodiments, step S3 specifically includes: See Figure 7 This is the adaptive DBMPCC algorithm framework of the present invention.

[0035] By forward difference analysis of the continuous differential equation of the electromagnetic coil RL series circuit, discretization and simplification yield: (19) In the formula, For the coil current of the next cycle, The coil current for the current cycle. The voltage applied across the coil during one cycle. To control the cycle, rearranging equation (19), we get: (20) From Equation 20, we can see that the current in the next cycle... It depends on the current cycle current. Applied voltage And the system parameters R and L identified online.

[0036] The goal of predictive control is to select the control voltage output by the controller. This will make it possible in the next cycle, equal to target current Substituting into equation (21), We can obtain: (twenty two) According to Equation 22, the voltage that needs to be applied in the current cycle can be calculated. By changing That can make Converges to the maximum extent Additionally, it should be noted that... It is a relatively ideal control voltage signal, but it cannot exceed the bus voltage, so it is saturated and limited.

[0037] (twenty three) in, The next step will be to further discuss the impact of errors in coil inductance and resistance parameters on the stability and accuracy of the controller.

[0038] First, based on equations (21) and (22), for ease of analysis and calculation, let: , ,have: (twenty four) To distinguish between the actual coil parameters and the parameters in the controller, the subscript "p" represents the parameters in the controller, and the subscript "0" represents the actual coil parameters. Therefore, the control voltage signal output by the controller is: (25) The actual current model of the coil is: (26) By combining equations (25) and (26), we can obtain: (27) Simplifying, we get: (28) The characteristic equation of a system is determined by its secondary part; therefore, let ,have to: (29) in: Let be the state transition matrix of this system. Find the system poles: Performing a Z-transform on equation (29), we obtain: (30) For equation (30) to have a non-zero solution, the following must be satisfied: That is, the system's poles are Right now: (31) Assuming the parameters in the controller prediction model are exactly the same as the actual coil parameters, we have: (32) at this time: (33) According to equation (33), under ideal conditions (without considering delay, voltage saturation, etc.), the poles of the closed-loop transfer function of the current loop of the coil DB-MPCC will coincide with zero, and the current will track the current reference value after one control cycle, thus achieving deadbeat control. Conversely, when the controller parameters do not match the actual coil parameters, according to equation (31), see [reference needed]. Figure 8 This is a pole distribution diagram of the controller of the present invention under parameter mismatch.

[0039] Depend on Figure 8It can be seen that when the controller inductance parameter is equal to the actual coil inductance, the system is stable. As the degree of inductance mismatch increases, the closed-loop poles of the system gradually approach the edge of the stability region. When the model inductance value in the controller reaches twice the actual inductance, the pole is exactly located on the negative stability boundary of the unit circle, and the system is in a critical stable state, which will produce violent oscillations. Further increases will lead to instability. Therefore, the accuracy of the inductance parameter plays a more decisive role in the stability of the system.

[0040] When the controller's resistance parameter is exactly equal to the actual coil resistance, the system is stable. As the resistance mismatch increases, the closed-loop poles of the system gradually approach the edge of the stability region. In terms of stability, even if the controller resistance parameter obtained by identification has some difference from the actual value, as long as it is not an extreme case, the pole position will be much smaller than ±1, and the impact on the stability of the entire system is small.

[0041] As shown in the diagram above, underestimating the resistance and overestimating the inductance will respectively lead to system oscillations. The dominant role of inductance in system stability can also be explained through the following derivation: Assuming resistor matching ( The pole formula can be simplified to: (34) According to equation (31): when the identified model inductance in the controller reaches twice the actual inductance, the pole z=-1 is located on the stable boundary of the unit circle, the system is in a critical stable state and will produce violent oscillations.

[0042] Assuming inductor matching ( The pole formula can be simplified to: (35) Due to control cycle It is usually a very small value, even with resistance error. The absolute value of the calculated pole z is relatively large, and unless it is an extreme case, it is usually much less than 1, and the pole is still near the origin. This also shows that the MPCC controller proposed in this invention has strong robustness in terms of stability, even in the case of resistor mismatch.

[0043] The core advantage of the DB-MPCC controller designed in this invention is zero steady-state error, which is due to its control objective. This characteristic is determined by parameters, but parameter mismatch will disrupt it and introduce steady-state error. This section will analyze in depth the impact of errors in coil inductance and resistance parameters on the steady-state error of the system. For the convenience of calculating steady-state error, the system input is defined as a step current signal.

[0044] When the system reaches steady state, the steady-state current is defined as follows: (36) Since the system input signal is a step signal, we have: (37) Substituting equations (36) and (37) into equation (28), we get: (38) Simplifying, we get: (39) System steady-state error: Substituting the values, we get: (40) According to (40), ideally, when the parameters in the controller prediction model are exactly the same as the actual parameters of the coil, then... .

[0045] When parameter mismatch occurs: Substitute equation (40) , We can obtain: (41) Equation (41) shows that resistor mismatch is the direct cause of static error. Resistor mismatch occurs if and only if resistors are matched: hour, The numerator is zero, and the static error is zero. If Therefore, the system will inevitably have a steady-state error that cannot be eliminated. Furthermore, inductor mismatch does not directly produce a steady-state error, but it will be reflected in the model's identification. As a result, the final magnitude of the error caused by resistor mismatch can be adjusted. Simultaneously, by increasing the system's sampling frequency, the static error can also be significantly reduced.

[0046] A model predictive current control system based on online parameter identification includes: An online parameter identification module is used to execute steps S1 and S2; The model prediction control module is used to execute step S3; The content of the above method embodiments is applicable to this system embodiment. The specific functions implemented in this system embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0047] A model-predictive current control device based on online parameter identification: At least one processor; At least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements a model predictive current control method based on online parameter identification as described above.

[0048] The content of the above method embodiments is applicable to the device embodiments. The specific functions implemented by the device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0049] A storage medium storing processor-executable instructions, which, when executed by a processor, are used to implement a model predictive current control method based on online parameter identification as described above.

[0050] The content of the above method embodiments is applicable to this storage medium embodiment. The specific functions implemented in this storage medium embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0051] The above is a detailed description of the preferred embodiments of the present invention. However, the present invention is not limited to the embodiments described. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention. All such equivalent modifications or substitutions are included within the scope defined by the claims of this application.

Claims

1. A model predictive current control method based on online parameter identification, characterized in that, Includes the following steps: The resistance is identified by using the steady-state average value information during the switching cycle, and the resistance identification result is obtained. Based on the resistor identification results, the inductor is identified by combining transient current ripple information with a logarithmic linearization method, and the inductor identification results are obtained. Based on the resistance identification results and the inductance identification results, the current at the next moment is predicted using a discretized mathematical model, and the target voltage command that makes the predicted current at the next moment equal to the reference current is calculated.

2. The model predictive current control method based on online parameter identification according to claim 1, characterized in that, The formula for resistor identification is expressed as follows: in, R represents the average voltage, and R represents the identification resistance. D represents the average current, and D represents the coil charging duty cycle. This indicates the DC bus voltage.

3. The model predictive current control method based on online parameter identification according to claim 2, characterized in that, The step of identifying the inductor based on the resistance identification result, using transient current ripple information combined with a logarithmic linearization method, specifically includes: The nonlinear model of the electromagnetic bearing coil current solution is linearized into a linear model; Acquire instantaneous current and generate current measurement data; Based on the linear model, the current measurement data is converted into data points in a linear coordinate system; Based on the data points, the slope parameter is solved using the least squares method; Combining the slope parameter and the resistance identification result, the inductance identification result is generated.

4. The model predictive current control method based on online parameter identification according to claim 3, characterized in that, The expression for the linear model is as follows: Where R represents the resistor being identified, L represents the inductance being identified, t represents the independent variable time, and C represents a preset constant. This indicates the inductor current bias.

5. The model predictive current control method based on online parameter identification according to claim 4, characterized in that, The step of predicting the current at the next moment using a discretized mathematical model based on the resistance identification results and the inductance identification results, and calculating the target voltage command that makes the predicted current at the next moment equal to the reference current, specifically includes: By forward difference analysis of the continuous differential equation of the electromagnetic coil RL series circuit, discretization and simplification are performed to obtain the expression for the current in the next cycle. Based on the expression for the next cycle current, the resistance identification result, and the inductance identification result, the control voltage output by the controller is selected so that the predicted current equals the target current in the next cycle.

6. The model predictive current control method based on online parameter identification according to claim 5, characterized in that, The expression for the current in the next cycle is as follows: in, R represents the current for the next cycle, R represents the identified resistance, and L represents the identified inductance. Indicates the control period. This represents the current in the current cycle. This represents the voltage applied across the coil during one cycle.

7. The model predictive current control method based on online parameter identification according to claim 6, characterized in that, The control voltage output by the controller is also subject to the following constraints: in, This indicates the bus voltage.

8. The model predictive current control method based on online parameter identification according to claim 2, characterized in that, The average current is calculated using methods including high-frequency oversampling, segmented weighted averaging, and midpoint sampling.

9. A model predictive current control system based on online parameter identification, characterized in that, include: The online parameter identification module is used to identify the resistance using the steady-state average value information during the switching cycle, and obtain the resistance identification result; Based on the resistor identification results, the inductor is identified by combining transient current ripple information with a logarithmic linearization method, and the inductor identification results are obtained. The model prediction control module, based on the resistance identification results and the inductance identification results, uses a discretized mathematical model to predict the current at the next moment and calculates the target voltage command that makes the predicted current at the next moment equal to the reference current.

10. A model predictive current control device based on online parameter identification, characterized in that, include: At least one processor; At least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the model predictive current control method based on online parameter identification as described in any one of claims 1-8.

Images