Model-free predictive current control method for T-type three-level inverter
Through the model-free prediction current control method, the autoregressive exogenous input parameters are updated in real time, which solves the problem of parameter mismatch sensitivity in traditional inverter control, improves the system's robustness and control accuracy, and meets the needs of high-frequency real-time.
Patent Information
- Application Number
- CN202510559935.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-30
- Publication Date
- 2025-07-25
AI Technical Summary
In the prior art, traditional multi-level inverter control methods rely on accurate models, resulting in parameter mismatch sensitivity and inability to effectively deal with changes in operating conditions such as load mutations and device aging.
The model-free prediction current control method is adopted, and the inverter state data is collected in real time, and the autoregressive exogenous input parameters are updated using the recursive least squares algorithm to predict the three-phase current value of the next control cycle. The multi-objective cost function is combined to optimize current tracking and midpoint potential balance to avoid dependence on the precise model.
It realizes the robustness of the inverter in the case of load abrupt and device aging, reduces the computational complexity, improves control accuracy and power quality, and meets the real-time requirements of high-frequency control.
Smart Images
Figure CN120377689A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of motor drive control, and particularly relates to a model-free predictive current control method for a T-type three-level inverter. Background Art
[0002] In some high-voltage and high-power application scenarios, traditional two-level inverters cannot meet the requirements due to the voltage withstand limitation of switching devices. In this case, how to apply low-voltage withstand switching devices to high-voltage and high-power scenarios has become a research hotspot. Thus, multilevel inverter technology has emerged. The concept of multilevel was first proposed by Japanese experts A. Nabae et al. in 1980. By changing the topology of the main circuit and increasing the number of switching devices, the DC voltage is dispersed across the two ends of each device when the switching device is turned off, realizing the application of low-voltage withstand switching devices in high-power scenarios. Multilevel technology enables multiple power devices to share the DC bus voltage, with small switching losses, high efficiency, and an output voltage waveform that is closer to a sine wave than that of a two-level inverter when the switching frequency is the same, and does not require an isolation transformer. These advantages have made multilevel technology widely regarded. In actual applications, three-level technology is more commonly used.
[0003] There are mainly three common multilevel circuit topologies: diode-clamped inverters, flying-capacitor clamped inverters, and cascaded inverters with independent DC power supplies. The T-type three-level inverter used in the present invention can be said to be an improved topology of the neutral-point clamped inverter, and its advantages are mainly reflected in reducing the number of switching devices in the current path and reducing conduction losses. Moreover, compared with the diode-clamped three-level inverter, each leg of the T-type three-level inverter uses two fewer clamping diodes, and its control method is similar to that of the diode-clamped three-level inverter. The T-type three-level inverter combines the advantages of two-level and three-level inverters, having both the advantages of low conduction losses and few device numbers of two-level inverters, and the advantages of good output waveforms and high efficiency of three-level inverters. It is a three-level inverter topology with great development prospects.
[0004] In the prior art, finite control set model predictive control (FCS-MPC) combined with the switching state PWM method has been widely applied to inverter control. This strategy requires a detailed discrete-time model of the converter to operate as a function of the switching state. The controller finds the optimal switching state in each sampling period to meet the aforementioned objectives. Due to its characteristics such as high-speed response, inherent constraint handling, no need for a modulator, and multi-objective realization, FCS-MPC has been widely used in low, medium, and high-power applications.
[0005] However, the model predictive strategy requires a detailed dynamic model of the converter to calculate the control actions that drive the future behavior of the converter to a specific value. The performance of FCS-MPC depends to a large extent on the fidelity of the model in representing the converter. For example, parameter mismatches or uncertainties in modeling can significantly degrade the performance of FCS-MPC. Summary of the Invention
[0006] Aiming at the above existing technical deficiencies, the purpose of the present invention is to provide a model-free predictive current control method for a T-type three-level inverter, which solves the problems of relying on an accurate model and being sensitive to parameter mismatches in the prior art.
[0007] To solve the above technical problems, the present invention adopts the following technical solutions: In a first aspect, the present invention provides a model-free predictive current control method for a T-type three-level inverter, and the method includes: Real-time collect the real-time state data of the T-type three-level inverter, and update the autoregressive exogenous input parameters through the recursive least squares algorithm according to the real-time state data. The real-time state data includes the DC bus capacitor voltage, three-phase output current, and current switch state signal of the T-type three-level inverter; Predict the three-phase current values in the next control cycle according to the updated autoregressive exogenous input parameters; Traverse all switch state combinations of the T-type three-level inverter to generate a candidate voltage vector set containing 27 voltage vectors. For each candidate voltage vector, calculate the corresponding three-phase output voltage, and combine the predicted three-phase current values and the DC bus capacitor voltage difference to calculate the cost function value through a multi-objective cost function. The control objectives of the multi-objective cost function are current tracking error and midpoint potential balance; Compare the cost function values of all candidate voltage vectors, and select the candidate voltage vector with the minimum cost function value as the optimal voltage vector; Output the switch state signal corresponding to the optimal voltage vector to the power switch device of the inverter, complete the operation of the current control cycle, and update the autoregressive exogenous input parameters to enter the iterative process of the next control cycle.
[0008] Preferably, in a possible implementation manner of the first aspect, the structure definition of the autoregressive exogenous input is: The estimated values of the three-phase currents of the inverter are expressed as:
[0009] Where is the input polynomial, representing the dynamic response of the phase current to the phase voltage, is the autoregressive polynomial, describing Autoregressive characteristics of phase current is the output voltage of the phase with respect to the neutral point , represents the unit delay operator , , respectively represent phase A, phase B, and phase C of the three-phase inverter
[0010] Preferably, in a possible implementation manner of the first aspect, the recursive least squares algorithm includes Based on the measurement value at the current moment , the historical input-output data feature vector and the parameter estimation vector at the previous moment , calculate the prediction error , is the transpose of ; Update the parameter estimation vector through the gain matrix , is the covariance matrix at the previous moment is the forgetting factor Update the covariance matrix , where is the identity matrix
[0011] Preferably, in a possible implementation manner of the first aspect, the prediction of the three-phase current values in the next cycle is
[0012] where is the phase autoregressive exogenous input parameter vector is the historical input-output data feature vector at time is the transpose of
[0013] Preferably, in a possible implementation manner of the first aspect, the method for generating the candidate voltage vector set includes Generate 27 voltage vectors according to the three-phase switch state combination , and the output voltage of each phase is calculated by the following formula
[0014] where is the output voltage of the phase with respect to the neutral point , is the DC bus voltage , , is the switching state of the , , phase, taking values 1, 0, or -1, corresponding to output levels , 0, or , , , representing the A-phase, B-phase, and C-phase of the three-phase inverter respectively.
[0015] Preferably, in a possible implementation manner of the first aspect, the mapping relationship between the switching state and the output level is: The output level corresponding to each phase switching combination is determined by the conduction states of four groups of switching devices : When and are conducting, the output level is ; When and are conducting, the output level is 0; When and are conducting, the output level is .
[0016] Preferably, in a possible implementation manner of the first aspect, the multi-objective cost function is defined as:
[0017] where is the reference value of the phase current at time , is the predicted current value of the phase current at time , and are the upper and lower capacitor voltages on the DC side at time , is the weight coefficient used to adjust the priority of current tracking error and midpoint potential balance, , , representing the A-phase, B-phase, and C-phase of the three-phase inverter respectively.
[0018] Preferably, in a possible implementation manner of the first aspect, the implementation method of the midpoint potential balance is: According to the capacitor voltage dynamic equation Update the capacitor voltage, where is the capacitor voltage at time is the capacitor voltage at time is the capacitance value, is the capacitor index, is the sampling period, is the capacitor current at time, which is calculated by the following formula:
[0019] where is the DC bus current at time is the phase switch state function, is at time the phase current value.
[0020] The beneficial effects of the present invention are as follows: By introducing a dynamic parameter update mechanism of autoregressive exogenous input and recursive least squares algorithm, the dependence on the accurate mathematical model of the inverter is eliminated. This method only needs to collect real-time voltage and current signals to identify the dynamic characteristics of the system online, effectively overcoming the problem that traditional model predictive control is sensitive to parameter mismatch, and showing excellent robustness under working condition changes such as load mutation and device aging.
[0021] The adopted recursive least squares algorithm updates parameters through the gain matrix iteration, avoiding the complex matrix inversion operation in the traditional algorithm, and significantly reducing the computational complexity. Combined with the parallel prediction structure of 27 voltage vectors, while maintaining the high-frequency control of 8.51 kHz, the single-cycle calculation time is shortened by 38% compared with the traditional method, meeting the real-time requirements of the embedded system.
[0022] By synchronously optimizing current tracking and neutral point potential balance through a multi-objective cost function, on the basis of accurately tracking the current reference value, the total harmonic distortion of the output current is reduced.
[0023] In summary, the present invention comprehensively improves the control accuracy and power quality of the system. BRIEF DESCRIPTION OF THE DRAWINGS
[0024] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.
[0025] Figure 1This application provides a topological diagram of a three-phase T-type three-level inverter.
[0026] Figure 2 This application provides a voltage vector diagram of a three-phase T-type three-level inverter.
[0027] Figure 3 This application provides a model-free predictive current control diagram of a three-phase T-type three-level inverter.
[0028] Figure 4 This application provides a current waveform diagram of phase a when there are parameter mismatches.
[0029] Figure 5 This application provides a transient response diagram when the reference current suddenly changes from 10 A to 20 A. Detailed implementation manners
[0030] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.
[0031] Embodiment 1: The present invention provides a model-free predictive current control method for a T-type three-level inverter, including: (1) Establish a mathematical model of a three-level T-type inverter In the three-phase three-level T-type inverter model as Figure 1 shown, the output voltages corresponding to different switching states are different, as shown in Table 1.
[0032] Table 1 Corresponding relationship between switching states and output levels
[0033] Among them, the variable represents the switching function of the x-th phase of the inverter; P, O, and N respectively represent the states corresponding to the output levels of , 0, and . At this time, the voltage can be expressed as
[0034] , , respectively represent the A-phase, B-phase, and C-phase of the three-phase inverter.
[0035] According to Kirchhoff's voltage law, the output voltage of the three-phase inverter can be expressed as
[0036] Assume that the three-phase voltages are symmetrical, then , then there is
[0037] The three-phase voltages of the inverter represented by the switch states can be obtained
[0038] Considering the dynamic model of the DC-side capacitor, the capacitor current can be expressed as
[0039] In the formula, are the capacitance values of the upper and lower busbars, , after discretization, the DC-side capacitor voltage based on time is
[0040] In the formula, is the sampling time, and respectively represent the capacitor voltage at time and is the capacitance value, is the capacitor index, represents the capacitor current at time k.
[0041] The capacitor current is jointly determined by the DC bus current and the output current, and is expressed as
[0042] In the formula, is the DC bus current at time is the phase current value at time, the parameter is the phase switch state function, and is expressed as
[0043] (2) Adopt model-free predictive current control based on ARX Due to the limitations of traditional FCS-MPC, the present invention proposes to use a general structure based on autoregressive exogenous input (ARX) representation to predict the output current of a three-level topology. By using these general ARX models, detailed knowledge of the physical system is not required. The input and output data are used to estimate the parameters of the ARX model. In this way, the proposed model-free predictive current control scheme tracks the unmodeled dynamics or changes in the physical system parameters. The transfer function of each stage of the proposed strategy for controlling the output current is shown below
[0044]
[0045]
[0046] where , and represent the estimated values of the three-phase currents of the inverter; , and are the three-phase voltage values on the output side of the inverter.
[0047] The polynomials in the ARX model are defined as
[0048]
[0049] where the superscript of the polynomial represents the phase index of the estimated variable, ; the superscript of the polynomial represents the phase index of the estimated variable, , the superscript represents the phase voltage of the three-level inverter, ; and represent the orders of the polynomials. In the present invention, and are 8 and 7 respectively, to reduce the computational burden while achieving good performance.
[0050] To estimate the coefficients of the ARX model, the present invention proposes a recursive least squares (RLS) algorithm, which uses the bus capacitance of the system and the measured values of the output current. The equations of the RLS estimator can be expressed as
[0051]
[0052]
[0053]
[0054] Wherein, is the estimation error, is the instantaneous measured value of the quantity to be estimated, and the vector contains the measured values of the previous input and output signals of the system, is the model parameter estimation vector at the previous moment; is the updated model parameter vector, is the gain matrix; is the covariance matrix at the previous moment, is the forgetting factor; is the covariance matrix, is the identity matrix. To ensure good tracking of the estimated quantity, cannot become too small. By changing the forgetting factor the value of can be changed. A smaller has a faster tracking speed, and a larger has better noise immunity.
[0055] The above estimator is used to predict the future value of each phase current. The predicted current of the three-phase T-type three-level inverter can be expressed as
[0056]
[0057]
[0058] Where is the historical input-output data feature vector at the moment, is the transpose of
[0059] (III) Controller Design For the model-free predictive current control method used in the present invention, its optimization process is to predict the current through ARX, and then traverse all 27 possible voltage vectors. The 27 voltage vectors are as shown in Figure 2 . The voltage vector that minimizes the cost function in the next cycle is used as the optimal voltage vector. In addition, by redesigning the cost function, multiple constraint conditions can be included at the same time, so that multiple control objectives can be satisfied through one cost function.
[0060] For the current predictive control of the T-type three-level inverter, its control objectives are current tracking and DC bus midpoint potential balance control. To meet the above control objectives, the cost function is expressed as:
[0061] In the formula, is the reference value of the three-phase current at moment. For a short sampling time, it can be considered that , is the predicted current value of the -phase current at and are the upper and lower capacitor voltages on the DC side at is the weight coefficient, The larger the value of , the greater the weight of the corresponding term.
[0062] The control diagram of the model-free predictive current control of the T-type three-level inverter is as shown in Figure 3 First, the measured values and at time k are obtained, a transfer function is established, the RLS method is used to obtain the parameters and the current prediction value at time k+1, and then the optimal voltage vector is selected according to the cost function.
[0063] After determining the optimal voltage vector, the corresponding switch state signal is output to the power switch device of the T-type three-level inverter in real time. This switch state signal directly controls the on-off combination of the four groups of switch devices in each phase bridge arm to ensure that the inverter outputs a three-phase voltage vector that meets the target. After the signal output is completed, the operation of the current control cycle ends, and the inverter operates according to the predicted optimal voltage vector to achieve accurate tracking of the output current and dynamic balance of the neutral point potential.
[0064] Subsequently, based on the real-time state data collected in the current cycle, the parameters of the autoregressive exogenous input model are dynamically updated by the recursive least squares algorithm. This process uses the latest data to online correct the model parameters, making the current prediction in the next cycle more in line with the actual operating characteristics of the inverter, thereby improving the robustness and adaptability of the control system.
[0065] After completing the parameter update, the control system immediately enters the iterative process of the next control cycle. In the new cycle, the system re-collects real-time data, predicts future currents, traverses all voltage vectors and calculates the cost function, and repeatedly executes the closed-loop control strategy of "prediction-optimization-execution". This process effectively responds to dynamic changes such as load fluctuations and device aging through continuous parameter updates and rolling optimizations, ensuring the stable operation of the system under all operating conditions.
[0066] Embodiment 2: The present invention provides a simulation of a model-free predictive current control method for a T-type three-level inverter, and the simulation parameters are shown in Table 2.
[0067] Table 2 Simulation parameters
[0068] Obviously, those skilled in the art can make various modifications and variations to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and its equivalent technologies, the present invention also intends to encompass these modifications and variations.
Claims
1. A model-free predictive current control method for a T-type three-level inverter, characterized in that The method includes: Collecting the real-time status data of the T-type three-level inverter in real time, and updating the autoregressive exogenous input parameters through the recursive least squares algorithm according to the real-time status data. The real-time status data includes the DC bus capacitor voltage, the three-phase output current, and the current switch status signal of the T-type three-level inverter; Predicting the three-phase current values in the next control cycle according to the updated autoregressive exogenous input parameters; Traversing all switch status combinations of the T-type three-level inverter to generate a candidate voltage vector set containing 27 voltage vectors. For each candidate voltage vector, calculate the corresponding three-phase output voltage, and combine the predicted three-phase current values and the DC bus capacitor voltage difference to calculate the cost function value through a multi-objective cost function. The control objectives of the multi-objective cost function are current tracking error and neutral point potential balance; Comparing the cost function values of all candidate voltage vectors, and selecting the candidate voltage vector with the minimum cost function value as the optimal voltage vector; Outputting the switch status signal corresponding to the optimal voltage vector to the power switch device of the inverter, completing the operation of the current control cycle, and updating the autoregressive exogenous input parameters to enter the iterative process of the next control cycle.
2. The model-free predictive current control method for a T-type three-level inverter as described in claim 1, characterized in that, The structure definition of the autoregressive exogenous input is: The estimated values of the three-phase currents of the inverter are expressed as: Among them, is the input polynomial, representing the dynamic response of the phase current to the phase voltage, is the autoregressive polynomial, describing the autoregressive characteristics of the phase current, is the output voltage of the phase relative to the neutral point, represents the unit delay operator, , , respectively represent phase A, phase B, and phase C of the three-phase inverter.
3. The model-free predictive current control method for a T-type three-level inverter according to claim 2, wherein The recursive least squares algorithm includes: Based on the measured values at the current moment , the historical input-output data feature vectors and the parameter estimation vector at the previous moment , calculate the prediction error , where is the transpose of Update the parameter estimation vector through the gain matrix where , is the covariance matrix at the previous moment, is the forgetting factor; Update the covariance matrix , where is the identity matrix.
4. The model-free predictive current control method for a T-type three-level inverter according to claim 3, characterized in that, The prediction of the three-phase current values in the next cycle is Among them is the vector of exogenous input parameters of the autoregressive model, is the eigenvector of the historical input-output data at time is the transpose of 5. The model-free predictive current control method for a T-type three-level inverter according to claim 1, characterized in that The method for generating the candidate voltage vector set includes: According to the three-phase switch state combination Generate 27 voltage vectors, and the output voltage of each phase Is calculated by the following formula: Among them, is the output voltage relative to the neutral point , is the DC bus voltage, , , are the switching states of the , , th phase, taking values 1, 0, or -1, corresponding to the output levels , 0, or , , , represent the A-phase, B-phase, and C-phase of the three-phase inverter respectively.
6. The model-free predictive current control method for a T-type three-level inverter according to claim 5, wherein The mapping relationship between the switch status and the output level is: Switch combination per phase The corresponding output level is determined by the conduction states of four groups of switching devices as follows: When and are turned on, the output level is ; When and are turned on, the output level is 0; When and are turned on, the output level is .
7. The model-free predictive current control method for a T-type three-level inverter according to claim 1, characterized in that, The multi-objective cost function is defined as: Among them, is the reference value of the phase current at time is the predicted current value of the phase current at time and are the upper and lower capacitor voltages on the DC side at time is the weight coefficient, used to adjust the priority of current tracking error and midpoint potential balance, , , represent the A-phase, B-phase, and C-phase of the three-phase inverter respectively.
8. The model-free predictive current control method for a T-type three-level inverter according to claim 7, characterized in that, The implementation method of the neutral point potential balance is: According to the dynamic equation of capacitor voltage Update the capacitor voltage, where is the capacitor voltage at time is the capacitor voltage at time is the capacitance value, is the capacitor index, is the sampling period, is the capacitor current at time, which is calculated by the following formula: Among them, is the DC bus current at a moment, is the phase switch state function, is the moment the phase current value.