A model prediction-based multi-objective control method for three-level inverters

Abstract

The application discloses a kind of multi-objective control methods of three-level inverter based on model prediction using optimized pulse width modulation, current error, switching number, midpoint potential offset and the difference between selected switch state and optimized pulse width modulation switch state are weighted to obtain new optimization function, and the optimal vector considering multiple optimization targets is selected by rolling optimization, so as to adjust the switch state of three-level inverter in real time.The method of the patent combines the optimized pulse width modulation of three-level inverter with multi-objective model prediction algorithm, which can make the inverter output have good low-order harmonic characteristics, current fast tracking and midpoint potential balancing ability under dynamic at low switching frequency.

Description

Technical Field

[0001] This invention belongs to the field of multilevel inverter control, specifically relating to a model prediction multi-objective control method for three-level inverters using optimized pulse width modulation, which is mainly applicable to high-performance optimization control of three-level inverters at low switching frequencies. Background Technology

[0002] Under the same operating conditions, three-level inverters offer advantages over two-level inverters, including a higher sinusoidal output voltage and lower voltage tolerance for power devices, making them widely used in medium-voltage, high-power drive applications. As the power levels of drive systems continue to increase, the switching frequency of power devices is strictly limited to several hundred hertz to reduce switching losses. Traditional pulse width modulation techniques (such as SVPWM) suffer from dead-time effects and voltage pulse asymmetry at low switching frequencies, leading to increased low-order harmonic content in the current and decreased control performance.

[0003] Optimized pulse width modulation (PWM) can ensure good voltage or current harmonic characteristics at low switching frequencies and has been increasingly applied in fields such as rail transportation and ship propulsion in recent years. However, the optimized switching angle obtained based on steady-state solutions does not take into account the possible changes in the fundamental frequency, amplitude, and phase of high-performance speed control systems. Directly applying optimized PWM to high-performance dynamic control systems will cause PWM pulse turbulence, leading to system overcurrent.

[0004] Model predictive control (MPC), an advanced control strategy, first emerged in the late 1970s. With the development of digital signal processing technology, MPC has gradually demonstrated its superior performance in the field of power electronics. MPC combines a discretized mathematical model to predict the state of variables at the next moment using the state of variables at the previous moment. By optimizing the objective function, it integrates the reference value and the predicted value of the desired control quantity, evaluates each possible switching state, and selects the set of switching states that minimizes the optimization objective function as the final output switching sequence. MPC has advantages such as simple principle, simultaneous optimization of multiple control objectives, and rapid dynamic response. Existing optimized pulse width modulation techniques cannot simultaneously consider both steady-state and dynamic performance, and when applied to three-level inverters, they often struggle to simultaneously optimize issues such as current error, number of switching operations, low-order harmonics, and midpoint balance. Summary of the Invention

[0005] The purpose of this invention is to address the problems of insufficient dynamic characteristics of existing optimized pulse width modulation technology and the difficulty in coordinating the adjustment of low-order harmonic suppression and midpoint potential balance, and to provide a multi-objective control method for three-level inverters based on model prediction using optimized pulse width modulation.

[0006] To achieve the above objectives, the technical solution adopted by the present invention to solve its technical problem is: a multi-objective control method for a three-level inverter based on model prediction, comprising the following steps:

[0007] Step 1: Analyze the topology of the three-level inverter, and establish prediction models for current error, number of switches, and midpoint potential offset, respectively. Based on these models, combine optimized pulse width modulation switching states to obtain the final multi-objective control prediction model. Discretize the three-phase bridge arm current prediction model of the three-level inverter in the abc three-phase stationary coordinate system, resulting in the following form:

[0008]

[0009] In the formula, T is the sampling period, i(k) and e(k) represent the bridge arm current and equivalent load voltage in the k-th step, respectively, and L s R s The inductance and resistance of the equivalent load; u(k) = [u a (k)u b (k)u c (k)] T Let k be the inverter bridge arm output voltage at the k-th step, and update it according to the following form:

[0010]

[0011] Where, S(k)=[S a (k) S b (k) S c (k)] T Let U be the switching state of the inverter in the k-th phase, and the value range of the switching state of each phase is {-1, 0, 1}. dc (k) represents the DC bus voltage of the kth-stage three-level inverter;

[0012] Step 2: At the (k-1)th cycle, execute the optimal switching state S calculated in the previous cycle. opt (k-1), simultaneously sampling the equivalent load voltage, three-phase bridge arm current, and DC bus voltage at the (k-1)th step, to obtain the predicted current value i(k) at the kth step; then, based on i(k), iterate through the switch states S. n (k) Calculate the predicted current i corresponding to each switch state at time k+1. n (k+1), combined with the reference current value i * (k+1) Calculate the current error Δi(k+1), and select the optimal vector S between k and k+1 according to the optimization function. opt (k), and load at k time;

[0013] Step 3: Establish the switching frequency f of the three-level inverter between k and k+1 cycles. switch The relationship between (k+1) and the selected switch state is as follows: f switch (k+1)=|S a (k+1)-S a (k)|+|S b (k+1)-S b (k)|+|S c (k+1)-S c (k)|;

[0014] Step 4: Discretize the DC bus midpoint potential prediction model of the three-level inverter, resulting in the following form: Where C is the upper or lower half of the DC bus capacitance, here the midpoint potential offset ΔU is used. n (k) is defined as the voltage difference between the lower half DC bus capacitor and the upper half DC bus capacitor at the kth time.

[0015] Step 5, calculate the current error Δi(k+1) and the number of switches f. switch (k+1), Midpoint potential offset ΔU n (k+1) and the difference f between the selected switch state and the optimized pulse width modulation switch state opp (k+1) is used as the weight of the four parts of the optimization function J, and the optimization function is listed as follows: J=Δi(k+1)+λ dc |ΔU n (k+1)|+λ s f switch (k+1)+λ o f opp (k+1), where λ dc , λ s , λ o f represents the weighting factor for each optimization objective, with a value range of (0, 1), determined by the degree of importance attached to each optimization objective; opp (k+1)=|S n (k+1)-S opp (k+1)|, and S n (k+1) represents the switching state corresponding to the selected vector predicted by the model at time k+1, while S opp (k+1) represents the switching state corresponding to the optimized pulse width modulation at time k+1.

[0016] Step 6: Perform rolling optimization on the optimization function J to select the optimal vector that takes into account multiple optimization objectives, thereby adjusting the switching state of the three-level inverter in real time.

[0017] The multi-objective control method for model prediction of a three-level inverter, wherein the bridge arm side current prediction step in step 1 is as follows:

[0018]

[0019] In the three-level inverter model prediction multi-objective control method described above, step 6 involves traversing the 27 switching states of the inverter in each sampling period to find the optimal vector that minimizes the J value, and then adjusting the switching angle calculated by the optimized pulse width modulation method in real time.

[0020] The beneficial effects of this invention are: the multi-objective control method disclosed in this invention combines the optimized pulse width modulation of a three-level inverter with a multi-objective model prediction algorithm, which can enable the inverter output to not only have good low-order harmonic characteristics at low switching frequencies, but also ensure fast current tracking and neutral point potential balance under dynamic conditions. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the predictive multi-objective control principle of the model of the present invention;

[0022] Figure 2 The waveform of the three-phase current when the motor load is suddenly subjected to torque under the traditional optimized pulse width modulation method;

[0023] Figure 3 The three-phase current waveform under sudden load conditions using the method of this invention;

[0024] Figure 4 The waveforms of the upper and lower bus capacitors without multi-objective optimization;

[0025] Figure 5 The waveforms of the upper and lower bus capacitors after model-predicted multi-objective control are shown.

[0026] Figure 6 The output voltage waveform under sudden load conditions using the method of the present invention is shown. Detailed Implementation

[0027] To further illustrate the purpose and technical solution of the present invention, a more detailed description will be provided below in conjunction with the accompanying drawings and specific embodiments.

[0028] Reference Figure 1 As shown, this invention discloses a model-predictive multi-objective control method for a three-level inverter. It weights multiple objective functions, including current error, number of switch switching, midpoint potential offset, and the difference between the selected switch state and the optimized pulse width modulation switch state, to obtain a new optimization function. This function is then subjected to rolling optimization to select the optimal vector that balances multiple optimization objectives, thereby enabling real-time adjustment of the three-level inverter's switching states. Specifically, the method includes the following steps.

[0029] 1. Analyze the topology of a three-level inverter, and establish prediction models for current error, number of switch changes, and midpoint potential offset, respectively. Based on these models, combine optimized pulse width modulation switching states to obtain the final multi-objective control prediction model. Discretize the arm-side current prediction model in the abc three-phase stationary coordinate system, resulting in the following form:

[0030]

[0031] In the formula, T is the sampling period, i(k) and e(k) represent the three-phase bridge arm current and equivalent load voltage in the k-th phase, respectively, and L s R s The inductance and resistance of the equivalent load; u(k) = [u a (k)u b (k)u c (k)] T Let k be the inverter bridge arm output voltage at the k-th step, and update it according to the following form:

[0032]

[0033] Where, S(k)=[S a (k) S b (k) S c (k)] T Let U be the switching state of the inverter in the k-th phase, and the value range of the switching state of each phase is {-1, 0, 1}. dc (k) represents the DC bus voltage at the kth step.

[0034] 2. At the (k-1)th cycle, execute the optimal switching state S calculated in the previous cycle. opt (k-1), simultaneously sampling the equivalent load voltage, three-phase bridge arm current, and DC bus voltage at the (k-1)th step, to obtain the predicted current value i(k) at the kth step; then, based on i(k), iterate through the switch states S. n (k) Calculate the predicted current i corresponding to each switch state at time k+1. n (k+1), combined with the reference current value i * (k+1) Calculate the current error Δi(k+1), and select the optimal vector S between k and k+1 according to the optimization function. opt (k), and loaded at time k. Here, based on the current prediction model defined in step 1, the f operator is defined, and the bridge arm side current prediction steps are as follows:

[0035]

[0036] 3. Establish the number of three-level inverter switching f between k and k+1 cycles. switchThe relationship between (k+1) and the selected switch state is as follows: f switch (k+1)=|S a (k+1)-S a (k)|+|S b (k+1)-S b (k)|+|S c (k+1)-S c (k)|.

[0037] 4. The prediction model for the DC bus midpoint potential of a three-level inverter is discretized and has the following form: Where C is the upper or lower half bus capacitance, here the midpoint potential offset ΔU n (k) is defined as the voltage difference between the lower half bus capacitor and the upper half bus capacitor at the kth beat.

[0038] 5. The current error Δi(k+1) and the number of switches f are... switch (k+1), Midpoint potential offset ΔU n (k+1) and the difference f between the selected switch state and the optimized pulse width modulation switch state opp (k+1) are the four parts of the optimization function J, and the optimization function is listed as follows: J=Δi(k+1)+λ dc |ΔU n (k+1)|+λ s f switch (k+1)+λ o f opp (k+1), where λ dc , λ s , λ o f represents the weighting factor for each optimization objective, with a value range of (0, 1), determined by the degree of importance attached to each optimization objective; opp (k+1)=|S n (k+1)-S opp (k+1)|, and S n (k+1) represents the switching state corresponding to the selected vector predicted by the model at time k+1, while S opp (k+1) represents the switching state corresponding to the optimized pulse width modulation at time k+1.

[0039] Finally, in each sampling cycle, the 27 switching states of the inverter are traversed to find the optimal vector that minimizes the J value, and the switching angle calculated by the optimized pulse width modulation method is adjusted in real time.

[0040] Simulation verification was performed using a three-level inverter with a motor load. Figure 2 The three-phase current waveforms under a sudden torque increase in motor load using a traditional optimized pulse width modulation method. Figure 3To predict the three-phase current waveform corresponding to the multi-objective control method using the optimized pulse width modulation model proposed in this invention under the same operating conditions, the simulation results show that the traditional optimized pulse width modulation method has a slow dynamic response and takes 4 to 5 fundamental frequency cycles to enter a steady state, while the control method proposed in this invention can achieve current tracking within 1 fundamental frequency cycle and has a rapid dynamic response.

[0041] Figure 4 The waveforms of the upper and lower bus capacitors without multi-objective optimization are shown. Figure 5 The waveforms of the upper and lower bus capacitors after multi-objective control based on model prediction are shown. It can be seen that the amplitude of the bus capacitor voltage fluctuation after optimization has decreased from 100V to about 45V, and the midpoint potential offset has been significantly reduced.

[0042] Figure 6 When the motor load suddenly increases torque, the optimized pulse width modulation model proposed in this invention is used to predict the output voltage waveform corresponding to the multi-objective control method. As can be seen from the dashed line, the steady-state switching angle under dynamic conditions is constantly corrected to achieve rapid tracking of current and adjustment of midpoint potential, which causes the switching frequency to rise to a certain extent. Since the switching frequency has also undergone multi-objective optimization, the increase in switching frequency under dynamic conditions is controllable. After one fundamental cycle, the switching frequency returns to the steady state.

[0043] It is evident that the multi-objective control method proposed in this invention can ensure that the inverter output still has good current tracking and midpoint balancing capabilities at low switching frequencies.

[0044] This invention is not limited to the above-described preferred embodiments. Any person skilled in the art can derive other variations and improvements based on the inspiration of this invention. However, regardless of any changes made to its shape or structure, any technical solution that is the same as or similar to this application falls within the protection scope of this invention.

Claims

1. A multi-objective control method for a three-level inverter based on model prediction, characterized in that: Includes the following steps Step 1: Analyze the topology of the three-level inverter, establish prediction models for current error, number of switches, and midpoint potential offset, and based on these, combine optimized pulse width modulation switching states to obtain the final multi-objective control prediction model. Discretize the three-phase bridge arm current prediction model of the three-level inverter in the abc three-phase stationary coordinate system: In the formula, T is the sampling period, i(k) and e(k) represent the bridge arm current and equivalent load voltage in the k-th step, respectively, and L s R s The inductance and resistance of the equivalent load; u(k) = [u a (k) u b (k) u c (k)] T Let be the output voltage of the three-phase bridge arm in the k-th phase, and update it according to the following formula: Where, S(k)=[S a (k) S b (k) S c (k)] T This represents the switching state of the inverter in the k-th phase. The value range of the switching state for each phase is {-1, 0, 1}, U dc (k) represents the DC bus voltage of the kth-stage three-level inverter; Step 2: At the (k-1)th cycle, execute the switching state S calculated in the previous cycle. opt (k-1), simultaneously sampling the equivalent load voltage, three-phase bridge arm current, and DC bus voltage at the (k-1)th step, to obtain the predicted current value i(k) at the kth step; then, based on i(k), iterate through the switch states S. n (k) Calculate the predicted current i corresponding to each switch state at time k+1. n (k+1), combined with the reference current value i * (k+1) Calculate the current error Δi(k+1), and select the optimal vector S between k and k+1 according to the following optimization function. opt (k), and load at k time: J=Δi(k+1)+λ dc |D n (k+1)|+λ s f switch (k+1)+λ o f opp (k+1), In the formula λ dc , λ s , λ o The weighting factors for each optimization objective are in the range of (0, 1); Step 3: Establish the switching frequency f of the three-level inverter between k and k+1 cycles. switch The relationship between (k+1) and the selected switch state is as follows: f switch (k+1)=|S a (k+1)-S a (k)|+|S b (k+1)-S b (k)|+|S c (k+1)-S c (k)|; Step 4: Discretize the DC bus midpoint potential prediction model for the three-level inverter: Where C is the upper or lower half DC bus capacitance, and the midpoint potential offset ΔU n (k) is defined as the voltage difference between the lower half DC bus capacitor and the upper half DC bus capacitor at the kth time. Step 5, calculate the current error Δi(k+1) and the number of switches f. switch (k+1), Midpoint potential offset ΔU n (k+1) and the difference f between the selected switch state and the optimized pulse width modulation switch state opp (k+1) is used as a weighted average of the four parts of the optimization function J: f opp (k+1)=|S n (k+1)-S opp (k+1)|, and S n (k+1) represents the switching state corresponding to the selected vector predicted by the model at time k+1, S opp (k+1) represents the switching state corresponding to the optimized pulse width modulation at time k+1. Step 6: Perform rolling optimization on the optimization function J to select the optimal vector that takes into account multiple optimization objectives, thereby adjusting the switching state of the three-level inverter in real time.

2. The multi-objective control method for a three-level inverter based on model prediction according to claim 1, characterized in that, In step 1, the f operator is defined based on the three-phase bridge arm side current prediction model. The bridge arm side current prediction steps are as follows:

3. A multi-objective control method for a three-level inverter based on model prediction according to claim 1 or 2, characterized in that, In step 6, during each sampling period, the 27 switching states of the inverter are traversed to find the optimal vector that minimizes the J value, and the switching angle calculated by the optimized pulse width modulation method is adjusted in real time.

Images