Four-level flying capacitor converter based on model predictive control algorithm

Abstract

The invention relates to a four-level flying capacitor converter (4L-SFC) based on a model predictive control algorithm, the converter defines a single target cost function to control an output current, and capacitor voltage balance is realized by reconstructing a phase voltage level selected by a 4L-SFC inverter. According to the method, the voltage of the flying capacitor is balanced by reconstructing the phase voltage level selected by the 4L-SFC converter, so that the calculated amount is relatively low; compared with a traditional FCS-MPC, a plurality of switch states are applied in the converter, and therefore current distortion is reduced.

Description

Technical Field

[0001] The invention relates to a four-level converter, which is applied to a renewable energy power generation system to improve the performance and efficiency of the renewable energy power generation system. Background Art

[0002] Over the past few decades, driven by the advancement of digital signal processors (DSPs), the finite control set model predictive control (FCS-MPC) algorithm has been widely applied to power electronic converters. The FCS-MPC algorithm consists of two main phases: prediction and optimization. Given a limited set of available switching states in a power electronic converter, the FCS-MPC algorithm uses a mathematical model of the system to predict all possible values ​​for each controlled variable. Therefore, at each sampling period, a switching state is selected based on minimizing a cost function. The FCS-MPC algorithm has many advantages. For example, it is easy to understand and has fast dynamic response. Furthermore, it does not require any PI controller. Capacitor voltage balance is a key control objective in multilevel converters. The FCS-MPC algorithm uses feasible switching states and selects appropriate weighting factors to minimize capacitor voltage errors. Despite its advantages and ability to control capacitor voltages in multilevel converters, the FCS-MPC method still has some drawbacks that affect system operation. These include high computational complexity, inappropriate selection of weighting factors, and the system's harmonic content. Summary of the Invention

[0003] The purpose of the present invention is to invent a method that can further improve the shortcomings of the FCS-MPC algorithm and overcome the limitation problem of capacitor voltage balance caused by simplifying the flying capacitor multi-level (FCM) converter topology components.

[0004] In order to achieve the above object, the technical solution adopted by the present invention is:

[0005] A four-level flying capacitor converter (4L-SFC) based on a model predictive control algorithm is characterized in that a single bridge arm of the converter includes a power electronic switch tube S p1 , power electronic switch tube S p2 , power electronic switching tube Power electronic switching tube and capacitor C; this converter provides four-level output voltage while reducing the number of capacitors from three to one; to avoid unnecessary harmonic distortion caused by unequal voltage levels, C should be kept at V dc The power electronic switch tube S p1 The drain of the power electronic switch S is connected to the converter input point. p1 The source of the power electronic switch S p2 The drain of the power electronic switch tube S p2The source of the power electronic switch tube The drain of the power electronic switch is connected to The source of the power electronic switch tube The drain of the power electronic switch is connected to The source of the capacitor C is connected to the input terminal of the converter, and one end of the capacitor C is connected to the power electronic switch S p2 The other end of the capacitor C is connected to the drain of the power electronic switch The source is connected.

[0006] Furthermore, a discrete-time (DT) model of the system is obtained, and model predictive control is applied to the power electronic converter. In order to eliminate calculations related to voltage prediction, a single-objective cost function is defined in the model predictive control process to evaluate and obtain the optimal voltage vector of the inverter. After selecting the appropriate switching state to ensure the load current required by the inverter output, the voltage of the flying capacitor is balanced by reconstructing the selected voltage phase level.

[0007] Compared with the prior art, the present invention has the following advantages: it defines a single-objective cost function to control the output current and balances the flying capacitor voltage by reconstructing the phase voltage level selected by the 4L-SFC converter. Therefore, the present invention has a low computational complexity; and compared with the traditional FCS-MPC, the present invention applies multiple switching states in the converter, thereby reducing current distortion. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1 The switching states of the single-phase 4L-SFC converter of the present invention are: (a) 3, (b) 2, (c) 1, (d) 0;

[0009] Figure 2 The process of reconstructing the phase voltage level of the single-phase 4L-SFC converter of the present invention. DETAILED DESCRIPTION

[0010] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0011] from Figure 1 As can be seen from the figure, the four-level flying capacitor (4L-SFC) converter based on the model predictive control algorithm of the present invention provides a four-level output voltage while reducing the number of capacitors from three to one; each bridge arm of the converter includes a power electronic switch S p1 , power electronic switch S p2 , power electronic switches Power electronic switches and capacitor C; the power electronic switch tube S p1The drain of the power electronic switch S is connected to the converter input point. p1 The source of the power electronic switch S p2 The drain of the power electronic switch tube S p2 The source of the power electronic switch tube The drain of the power electronic switch is connected to The source of the power electronic switch tube The drain of the power electronic switch is connected to The source of the capacitor C is connected to the input terminal of the converter, and one end of the capacitor C is connected to the S p2 The other end of the capacitor C is connected to the drain of To avoid unnecessary harmonic distortion caused by unequal voltage levels, the voltage of the capacitor C should be kept at V dc .

[0012] Table 1 shows the four different switching states of a single phase of a 4L-SFC converter. It is clear from Table 1 that, without any redundancy, each phase voltage level is generated by only one switching state of each bridge arm. Using the single-phase schematic diagram of a 4L-SFC converter, the operating mode and current path of each level are shown as follows: Figure 1 shown.

[0013] Table 1 4L-SFC converter operating modes

[0014]

[0015] Applying model predictive control to power electronic converters requires a discrete-time (DT) model of the system. The model of a voltage source inverter connected to an RL load is represented as follows:

[0016]

[0017] Where I, V, R and L are load current, inverter voltage, load resistance and load inductance respectively. In this stage, a single objective cost function is defined to control the load current at the inverter output. Assume that the actual current is equal to the reference current of the previous cycle, that is, i(k+1)=i * (k+1). By using the Euler forward discretization method we get:

[0018]

[0019] The inverter output reference voltage of the next sampling period is:

[0020]

[0021] Among them, V αβ* (k+1), i αβ * (k+1) are the output voltage and output current in the αβ reference frame at time k+1, i αβ (k), T s is the output current measurement at time k and the sampling period. Given the inverter reference voltage vector, only one sector in the spatial diagram is selected and its corresponding voltage vector is considered. As a result, the number of states evaluated in the predictive control is reduced from 64 to 10. In addition, to eliminate the computation associated with the current prediction, the classic cost function is replaced by a cost function consisting of the inverter voltage vector error, defined as:

[0022] g=(V αβ * (k+1)-V αβ i (k+1)) 2 ,i=1,...,10 (4)

[0023] V αβ i (k+1) is the voltage vector containing ten different values ​​in the selected sector. By using the single-objective cost function shown in (4), the calculations related to capacitor voltage prediction can be omitted to reduce the amount of computation. In addition, there is no need to heuristically select the weighting factors.

[0024] The next step is to select the appropriate redundant state corresponding to the optimal voltage vector. As can be seen from sector I of the 4L-SFC converter space vector diagram, the voltage vector is a linear combination of several phase voltages. For example, V5 is a linear combination of two states. The selection of redundant states is based on the reduction of switching frequency, which means that if a switching state with a higher voltage level is selected in the first sector, the voltage level of phase A remains fixed at V during the entire sector. dc Repeating this strategy in the third and fifth sectors, the voltage levels of phase B and phase C are maintained at V dc Similarly, the state with lower voltage is selected in the even sectors of the space vector.

[0025] After selecting the appropriate switching states to ensure the inverter outputs the required load current, the second stage is to balance the voltages across the flying capacitors. In previous studies, virtual vectors have been assigned an average capacitor current of zero. These vectors are formed by linear combinations of the actual voltage vectors, which allows the capacitor voltages to remain constant under ideal conditions. However, this approach also introduces some problems, such as the need for a lookup table (LUT) to store the vectors and increased computational effort. In the concept of reconstructing the phase voltage levels, instead of combining the actual voltage vectors, the phase voltage levels are linearly combined. Since the actual voltage vectors do not change, the number of states in the FCS-MPC method does not increase.

[0026] The voltage level reconstruction data is derived from the available switching states of the inverter and the effect of each switching state on the flying capacitor voltage. P ∈{1,2,3,4}, then That is, the phase level I of the voltage vector s P The phase reconstruction voltage level. Figure 1 In the 4L-SFC converter shown, states 3 and 0 are not affected by the level reconstruction process because the flying capacitor does not cause voltage drift in these states. Unlike 3 and 0, states 1 and 2 enter the reconstruction process because they change the capacitor voltage from its nominal value. Each of these voltage levels is reconstructed in a manner that includes three levels to obtain zero average capacitor current within the sampling interval and achieve equilibrium. The process is as follows Figure 2 shown.

[0027] The reconstructed finite control set model predictive control (RFCS-MPC) method selects a state at each sampling time interval by minimizing the cost function, and only one voltage level enters the phase voltage level reconstruction. Therefore, the duty cycle and optimal vector of the reconstructed energy level are equal to the sampling period:

[0028]

[0029] is the duty cycle of the reconstructed voltage, and the duty cycle of each voltage level of the phase p of the space vector s is obtained as follows:

[0030]

[0031] in is the phase level I of the voltage vector s p The duty cycle of phase levels 1 and 2 of each voltage vector is the same according to (6). Therefore, the average capacitor voltage change in each sampling interval is zero due to the opposite effect of each level on the capacitor voltage. In addition, the average value of the reconstructed voltage level is equal to the primary voltage level, which will make the system balanced.

[0032] The converter described above defines a single-objective cost function to control the output current. Capacitor voltage balancing is achieved by reconstructing the phase voltage levels selected by the 4L-SFC inverter, resulting in a low computational effort. Compared to conventional FCS-MPC, this approach applies multiple switching states in the converter, thereby reducing current distortion. Furthermore, despite the use of a model predictive control approach based on virtual voltage vectors, the number of feasible voltage vectors in RFCS-MPC does not increase because the phase levels are linearly combined. Simulation results show that the current spectra of this approach tend to be concentrated near the sampling frequency.

[0033] The above embodiments are intended only to illustrate the technical concepts and features of the present invention. Their purpose is to enable those skilled in the art to understand the contents of the present invention and implement them accordingly. They are not intended to limit the scope of protection of the present invention. Any equivalent changes or modifications made in accordance with the spirit of the present invention are intended to be covered by the scope of protection of the present invention.

Claims

1. A four-level capacitor converter based on a model predictive control algorithm, characterized in that: A single bridge arm of the converter includes a power electronic switch tube , power electronic switching tube , power electronic switching tube , power electronic switching tube and a capacitor C; The converter provides a four-level output voltage while reducing the number of capacitors from three to one; the power electronic switch tube The drain of the power electronic switch is connected to the input point of the converter. The source of the power electronic switch tube The drain of the power electronic switch is connected to The source of the power electronic switch tube The drain of the power electronic switch is connected to The source of the power electronic switch tube The drain of the power electronic switch is connected to The source of the capacitor C is connected to the input terminal of the converter, and one end of the capacitor C is connected to the The drain of To avoid unnecessary harmonic distortion caused by unequal voltage levels, the voltage of the capacitor C should be kept at V dc .

2. The four-level capacitor converter based on the model predictive control algorithm according to claim 1, characterized in that: The model To apply predictive control to the converter, it is necessary to obtain a discrete-time (DT) model of the system. The model of the voltage source inverter connected to the RL load is expressed as follows: （1） Among them, I, V, R and L are load current, inverter voltage, load resistance and load inductance respectively. In this stage, a single objective cost function is defined to control the load current at the inverter output. It is assumed that the actual current is equal to the reference current of the previous cycle, that is, , by using the Euler forward discretization method: (2) The inverter output reference voltage of the next sampling period is: （3） in, , are the output voltage and output current in the αβ reference frame at time k+1, , is the output current measurement value and sampling period at time k; given the inverter reference voltage vector, only one sector in the spatial diagram is selected and its corresponding voltage vector is considered; in addition, to eliminate the calculation related to the current prediction, the classic cost function is replaced by a cost function composed of the inverter voltage vector error, which is defined as: （4） The voltage vector contains ten different values ​​in the selected sector; The next step is to select the appropriate redundancy state corresponding to the optimal voltage vector. The selection of the redundancy state is based on the reduction of the switching frequency. After selecting the appropriate switching state to ensure the inverter outputs the required load current, the voltage level is reconstructed to balance the flying capacitor voltage; The reconstructed finite control set model predictive control (RFCS-MPC) method selects a state at each sampling time interval by minimizing the cost function. Only one voltage level enters the phase voltage level reconstruction. Therefore, the duty cycle and optimal vector of the reconstructed energy level are equal to the sampling period: （5） is the duty cycle of the reconstructed voltage, and the duty cycle of each voltage level of the phase p of the space vector s is obtained as follows: （6） in is the phase level of the voltage vector s According to (6), the duty cycle of phase levels 1 and 2 of each voltage vector is the same. Therefore, due to the inverse effect of each level on the capacitor voltage, the average capacitor voltage change in each sampling interval is zero; in addition, the average value of the reconstructed voltage level is equal to the primary voltage level, which will make the system balanced.

Images