A hybrid model predictive control method for eight-switch three-phase inverter
By adopting the hybrid model predictive control (H-MPC) method in the eight-switch tube three-phase inverter, vector reconstruction and PD controller are used, the output current ripple and switching frequency in the traditional FCS-MPC method are solved. As well as the complex problem of NP voltage balancing, better steady-state performance and computational efficiency are achieved.
Patent Information
- Application Number
- CN202411114701.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-14
- Publication Date
- 2024-11-29
- Estimated Expiration
- 2044-08-14
AI Technical Summary
The traditional finite control set model predictive control (FCS-MPC) has problems with large output current ripple, uncertain switching frequency and complex weight coefficient adjustment in the eight-switch tube three-phase inverter, making it difficult to effectively balance the neutral point voltage.
A hybrid model predictive control (H-MPC) method is proposed to eliminate modeling errors in the cost function through vector reconstruction, and combine a proportional differential (PD) controller to adjust the vector action time in the "NP voltage adjustable sector" , realize mixed output and adopt a multi-step optimization strategy to reduce computing time.
The H-MPC method can significantly improve the steady-state performance of the three-phase inverter, suppress the three-phase output imbalance, maintain a constant switching frequency, and effectively balance the NP voltage, reducing the calculation time by about 65.8%.
Smart Images

Figure CN118646286B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of power electronics, and specifically relates to a hybrid model predictive control method for an eight-switch three-phase inverter. Background Art
[0002] Compared with traditional two-level inverters, three-phase three-level voltage source inverters (3P-3L VSIs) have the advantages of less voltage stress, less output current harmonics, and higher efficiency. They are one of the key devices for medium voltage and high power applications in industrial scenarios. However, due to the introduction of more power semiconductor devices in 3P-3L VSIs, their failure rate has also increased exponentially.
[0003] In order to improve the stability of 3P-3L VSIs to meet the stringent requirements of modern power electronic systems for cost, safety, and reliability, fault-tolerant technologies have been widely studied in recent years. Generally speaking, there are many types of faults in power electronic inverters, among which the faults of power semiconductor devices are the most common. The fault types of power semiconductor devices can be divided into short-circuit faults and open-circuit faults. Short-circuit faults can cause serious damage to the entire system and require the inverter to be shut down immediately. In contrast, open-circuit faults are mainly caused by thermal cycles, resulting in lifting and cracking of solder joints within the module. These faults usually do not cause serious safety issues, but can lead to increased harmonic content in the converter output current and NP voltage fluctuations, especially for three-level inverters with a neutral point structure.
[0004] In response to open-circuit failures of power semiconductor devices, academia and industry have proposed several effective strategies. First, redundant circuit structures can be added to the inverter to intervene when a fault occurs; second, a fault-tolerant control algorithm can be designed. However, this method usually requires the design of at least two modulation modes to work in a hybrid manner, which increases the complexity of the algorithm; in contrast, reconstructing the post-fault topology to obtain a new fault-tolerant configuration is relatively economical and easy to implement, and has attracted widespread attention from scholars.
[0005] Eight-switch three-phase inverters (ESTPIs) are the classic reconstruction topology of 3P-3L VSIs. The fault phase is directly connected to the neutral point, thus forcing the fault phase to change to the "O" state. For ESTPIs, the direct connection of the fault phase to the neutral point will interfere with the neutral point current, resulting in NP voltage fluctuations. Unbalanced NP voltage can cause serious problems. For example, long-term deviation of the NP voltage from the reference value may cause aging or even breakdown of the electrolytic capacitor; in addition, unbalanced NP voltage will also lead to unbalanced three-phase output. Therefore, how to effectively control ESTPIs is a topic worthy of in-depth study. Generally speaking, ESTPIs can be controlled using space vector pulse width modulation (SVPWM). Some people combine PI controllers with SVPWM to achieve NP voltage balance by adjusting the zero-sequence voltage; some people design moving average filters (MAFs) to extract the DC component in the NP voltage as compensation for the vector action time to suppress three-phase imbalance. However, these SVPWM-based methods involve complex trigonometric function calculations and are relatively complex to implement.
[0006] On the other hand, finite control set model predictive control (FCS-MPC) shows significant advantages in solving such multi-objective optimization problems. First, FCS-MPC establishes a discrete mathematical model of the system, then establishes the state prediction equation, and designs a multi-objective cost function. Finally, the control input that minimizes the cost function is obtained through rolling optimization. However, traditional FCS-MPC usually only acts on one voltage vector in each switching cycle, resulting in large output ripple and variable switching frequency, which complicates the filter design. Using multiple voltage vectors in one control cycle is a feasible solution. Some people use three voltage vectors in one switching cycle, and the duration of each vector is inversely proportional to its corresponding cost value. Some people use visualization methods to verify that this method can indeed reduce tracking errors.
[0007] However, in order to achieve multi-objective control, it is time-consuming to design appropriate weight factors of the cost function. In order to eliminate the weight coefficient, some people achieve NP voltage balance by adjusting the action time of redundant small voltage vectors; some people propose sequential model predictive control based on hierarchical optimization. However, these methods are not applicable because ESTPIs have no redundant small voltage vectors and hierarchical optimization will lead to excessive tracking errors. Therefore, in order to achieve NP voltage balance in ESTPIs, some people still use weighting factors in the cost function; some people design virtual zero vectors to adjust the neutral point current, but the virtual zero vector will increase the switching loss. In summary, among the existing methods, there is no MPC method for ESTPIs that can effectively eliminate the weight factor.
[0008] Therefore, the traditional FCS-MPC only acts on one voltage vector in one switching cycle, resulting in an uncertain switching frequency and large output current ripple, as well as complex weight factor adjustment problems. Summary of the invention
[0009] The objectives of the present invention are achieved through the following technical solutions.
[0010] The present invention proposes a hybrid model predictive control (H-MPC) with vector reconstruction. First, the voltage vector is reconstructed under unbalanced NP voltage, the modeling error term in the cost function is eliminated, a better steady-state current output is obtained, and the three-phase imbalance is effectively suppressed; then, in order to eliminate the weight coefficient, a proportional differential (PD) controller is connected in parallel in the "NP voltage adjustable sector" to adjust the vector action time and realize hybrid output. Finally, a multi-step optimization strategy is proposed to reduce the calculation time.
[0011] Specifically, the present invention proposes a hybrid model predictive control method for an eight-switch three-phase inverter, comprising:
[0012] Obtaining an upper capacitor voltage and a lower capacitor voltage of an eight-switch tube three-phase inverter, and performing vector reconstruction according to a preset table and the upper capacitor voltage and the lower capacitor voltage to obtain nine reconstructed voltage vectors;
[0013] Determine a large triangular sub-sector and a small triangular sector according to two large voltage vectors, evaluate the cost value of each voltage vector located at the midpoint of the small triangular sector, and determine the small triangular sector where the voltage vector that minimizes the cost function is located;
[0014] In each small triangle sector, the action time of three preset voltage vectors in one switching cycle is calculated;
[0015] If the small triangle sector is an NP voltage adjustable sector, the PD controller is further used to adjust the vector action time, otherwise no adjustment is performed;
[0016] The duty cycle is calculated according to the vector action time and is used to control the eight-switch three-phase inverter in the next cycle.
[0017] Furthermore, the voltage vector with the minimum cost function is obtained by Lagrangian extrapolation.
[0018] Furthermore, the vector reconstruction includes: calculating accurate values of all voltage vectors as inputs of model predictive control.
[0019] Furthermore, each of the small triangular sectors includes two small voltage vectors. After determining the small triangular sector where the voltage vector that minimizes the cost function is located, the method further includes:
[0020] The neutral point currents corresponding to the two small voltage vectors are analyzed, and the small triangle sector is divided into an NP voltage adjustable sector and an NP voltage non-adjustable sector.
[0021] Furthermore, the analysis of the neutral point currents corresponding to the two small voltage vectors, dividing these small triangle sectors into NP voltage adjustable sectors and NP voltage non-adjustable sectors, includes:
[0022] When the neutral point currents corresponding to the two small voltage vectors in the triangular small sector are in opposite directions, the triangular small sector is considered to be an NP voltage adjustable sector, otherwise it is an NP voltage non-adjustable sector.
[0023] Furthermore, time compensation is performed in the NP voltage adjustable sector.
[0024] Further, the time compensation is controlled by a PD controller.
[0025] The advantages of the present invention are: 1. The vector reconstruction method based on NP voltage proposed in the present invention enables H-MPC to have better steady-state performance, and can effectively suppress the imbalance of three-phase output, and can maintain a constant switching frequency; 2. The parallel mode of PD controller and model predictive control proposed in the present invention eliminates the weight coefficient in the cost function, so that the reference current tracking is decoupled from the NP voltage balance, thereby effectively balancing the NP voltage; 3. The multi-step optimization strategy proposed in the present invention significantly simplifies the calculation, and the calculation time is reduced by about 65.8% compared with before optimization. BRIEF DESCRIPTION OF THE DRAWINGS
[0026] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Moreover, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:
[0027] Figure 1 The circuit structure diagram of ESTPIs according to the embodiment of the present invention is shown, wherein (a) ESTPI-A; (b) ESTPI-B; (c) ESTPI-C; (d) is the output LC filter and linear resistance circuit.
[0028] Figure 2 1 shows a voltage vector distribution diagram according to an embodiment of the present invention. Wherein, (a) ESTPI-A; (b) ESTPI-B; (c) ESTPI-C;
[0029] Figure 3 A schematic diagram of a multi-step optimization strategy proposed according to the present invention is shown.
[0030] Figure 4The voltage vector diagram of the ESTPI-A proposed according to the present invention under different NP voltage conditions is shown. (a) ζ=0 (b) ζ<0 (c) ζ>0.
[0031] Figure 5 A schematic diagram showing the relationship between the three-phase current directions and sectors according to the present invention is shown.
[0032] Figure 6 FIG. 4 shows a closed-loop control block diagram of ζ according to the present invention.
[0033] Figure 7 The overall system control block diagram according to the present invention is shown.
[0034] Figure 8 A schematic diagram showing the time complexity of different algorithms according to the present invention is shown.
[0035] Fig. 9 The steady-state experimental waveforms under experimental conditions 1 and 2 according to the present invention are shown. (a) T-MPC; (b) T-SVPWM; (c) H-MPC.
[0036] Fig.10 The FFT analysis diagrams of steady-state experiments under experimental conditions 1 and 2 according to the present invention are shown. (a) T-MPC; (b) T-SVPWM; (c) H-MPC.
[0037] Fig.11 The dynamic experimental waveforms under experimental conditions 3 and 4 according to the present invention are shown. (a) T-MPC; (b) T-SVPWM; (c) H-MPC.
[0038] Fig.12 The schematic diagram of the three-phase output imbalance suppression capability under NP voltage imbalance according to the present invention is shown. (a) T-MPC; (b) T-SVPWM; (c) H-MPC.
[0039] Fig.13 A schematic diagram showing the comparison of the imbalance suppression capability of the vector reconstruction method according to the present invention is shown: (a) H-MPC without vector reconstruction; (b) H-MPC with vector reconstruction. DETAILED DESCRIPTION
[0040] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.
[0041] Terminology explanation:
[0042] Eight-switch three-phase inverters (ESTPIs): When a single-phase open-circuit fault occurs in a three-phase three-level inverter, the faulty phase can be directly connected to the neutral point to reconstruct it into an eight-switch three-phase inverter. In this case, the faulty phase is forced to output the "O" state. Depending on the faulty phase, there are a total of 3 possible eight-switch three-phase inverter topologies. Since one phase is directly connected to the neutral point, how to effectively balance the neutral point voltage and reduce the output current ripple is a difficult problem in the control of the eight-switch three-phase inverter.
[0043] Finite Control Set Model Predictive Control (FCS-MPC): FCS-MPC first establishes a discrete mathematical model of the system, then calculates the control input that minimizes the cost function through rolling optimization, and acts on the next control cycle. FCS-MPC can simply handle multi-objective optimization problems. For power electronic inverters, the control input is limited, so the cost value of each control input can be calculated. However, the complex calculation process and cumbersome weight coefficient adjustment limit the development of FCS-MPC.
[0044] Vector reconstruction technology: Vector reconstruction technology means that when the neutral point (NP) voltage is unbalanced, traditional modeling ignores the offset of the voltage vector, which causes large output current ripple and three-phase imbalance. At this time, the voltage vector can be reconstructed according to the real NP voltage to eliminate the modeling error. Vector reconstruction technology can effectively improve the steady-state performance of the system and maintain three-phase balance even when the NP voltage is unbalanced.
[0045] The specific embodiments of the present invention are as follows:
[0046] A. System Model
[0047] The circuit structures of three ESTPIs with LC filters are as follows: Figure 1 As shown, is the DC side voltage; and are the upper capacitor and lower capacitor voltages respectively; , and The current flowing through the upper capacitor is And the lower capacitor The current at the neutral point O; , , , LC filter that constitutes the output; is the filter inductor current; and are the filter capacitor voltage and output current respectively; , and They are linear resistance loads for the three phases A, B, and C respectively.
[0048] These three structures are essentially equivalent, and without loss of generality, ESTPI-A is used as an example for modeling and analysis.
[0049] For each arm of a fault-free T-type 3P-3L VSI, there are three possible output states, defined as “P”, “O”, and “N”, where “P” represents the switching device and On at the same time; "O" indicates a switching device and Both are turned on, "N" indicates the switch device and Since ESTPIs directly connect the fault phase to the neutral point, there are only 9 possible voltage vectors for each ESTPI, such as Figure 2 shown.
[0050] For ESTPI-A, the switching functions of phases B and C are Defined as:
[0051]
[0052] Assuming that the NP voltage has been effectively balanced, the output voltage relative to the neutral point "O" , and It can be expressed as:
[0053] [ VxO VxD VxD ] = [ 0 0 0 0 V dc / 2 0 0 0 V dc / 2 ] [ 0 Sb Sc ]
[0054] Furthermore, the output voltage relative to the load neutral point "N" , and It can be expressed as:
[0055] [ V N VxD VxD ] = 1 3 [ 2 − 1 − 1 − 1 2 − 1 − 1 − 1 2 ] [ VxO VxD VxD ] ,
[0056] The resultant voltage vector can be expressed as:
[0057]
[0058] in
[0059] According to Kirchhoff's law, assuming the three phases are symmetrical, then Figure 1 The circuit equation in (a) can be derived as follows:
[0060] [ V aN V Bn V cN ] = L f d dt [ i fa i fb i fc ] + R L [ i fa i fb i fc ] + [ V fa V fb V fc ] .
[0061] Assume the inductor current is the state variable. The model in (5) can be transformed into a two-phase stationary coordinate system through Clarke transformation and expressed as:
[0062] L f d dt [ i f α i f β ] = [ V α V β ] − R L [ i f α i f β ] − [ V f α V f β ]
[0063] When the sampling period When the time is small enough, the forward Euler approximation is used to discretize the continuous model. This gives the recursive expression of the state variables at time (k+1) and (k). The discrete state space model of formula (6) is:
[0064]
[0065] It is worth noting that all the above models are based on the balanced NP voltage. In practical applications, the NP voltage needs to be analyzed. Figure 1 As shown, the current and voltage relationship of the upper and lower capacitors on the DC side can be expressed as:
[0066]
[0067] NP voltage balance and neutral point current related, It can be expressed as:
[0068]
[0069] Similarly, formula (8) can also be discretized by forward Euler approximation. Thus, the difference between the upper and lower capacitor voltages at time (k+1) can be obtained:
[0070]
[0071] in .
[0072] B. Traditional FCS-MPC
[0073] In order to achieve accurate reference current tracking and NP voltage balance, the cost function of FCS-MPC is It is expressed as:
[0074]
[0075] Considering the inherent one-step delay compensation of FCS-MPC and the reduced prediction accuracy caused by multi-step prediction (p=2), the weight coefficient matrix , prediction matrix and the reference matrix It is expressed as:
[0076]
[0077] Y c ( k + i | k ) = [ i f α ( k + i | k ) i f β ( k + i | k ) Δ VPN ( k + i | k ) ] Τ ,
[0078] R ( k + i ) = [ i f α * ( k + i ) i f β * ( k + i ) 0 ] Τ .
[0079] in is the weight factor of NP voltage balance, and the prediction matrix is obtained by formulas (7) and (10). and The reference value of the inductor current can be calculated using the Lagrange interpolation method. For a sinusoidal reference, the order "n" should not be less than 2. Therefore, the reference value of the inductor current in the αβ system can be expressed as:
[0080]
[0081] Each voltage vector corresponds to a specific cost value. Therefore, through rolling optimization, the voltage vector that minimizes the cost function can be determined and applied to the next control cycle. The optimization process can be expressed as:
[0082] (16)
[0083] C. Reduce current ripple
[0084] Traditional FCS-MPC usually only acts on one voltage vector in one switching cycle, which will lead to large output current ripple, which is undesirable in modern power systems. To solve this problem, the present invention uses modulation, which can effectively reduce the output ripple.
[0085] Therefore, the space vector is first divided into 8 triangular small sectors, and 3 voltage vectors are applied in one switching cycle. , and The final synthesized voltage vector It can be expressed as:
[0086]
[0087] Among them, the action time of the three voltage vectors , and The tracking cost function is inversely proportional to the corresponding tracking cost value. It can be defined as:
[0088]
[0089] So, , and It can be expressed as:
[0090]
[0091] The above modulation method can effectively reduce the output current ripple, but it also brings the problem of excessive calculation, so the algorithm needs to be further optimized.
[0092] D. Multi-step optimization method
[0093] In order to achieve rapid optimization, the present invention proposes a multi-step optimization method.
[0094] like Figure 2 As shown in (a), the entire space sector can be regarded as a symmetrical rhombus. First, the large triangle sub-sector is determined according to the two large voltage vectors ((0, 1, -1) and (0, -1, 1)), as shown in Figure 3 As shown. Subsequently, in order to further determine the small triangular sector, the cost value of each voltage vector located at the midpoint of the small triangular sector is evaluated. Thus, the sector where the voltage vector that minimizes the cost function is located is determined.
[0095] The above optimization method is also applicable to ESTPI-B and ESTPI-C. It is worth noting that the weighting term of NP voltage is not considered in the optimization process because it will be eliminated in the subsequent analysis.
[0096] Table I. Accurate voltage vector based on NP voltage
[0097]
[0098] E. Reconstruction of voltage vector under unbalanced NP voltage
[0099] For ESTPIs, since one phase is directly connected to the neutral point, the fluctuation of NP voltage is inevitable and always forms a dynamic balance near the balance point. However, when the NP voltage fluctuates greatly, it will cause obvious output ripple or even three-phase imbalance.
[0100] To analyze this problem, we will Defined as:
[0101]
[0102] when , it is necessary to rewrite equation (2) according to the actual value of the NP voltage to obtain an accurate voltage vector, as shown in Table I. The voltage vector of ESTPI-A under different NP voltage conditions is shown in Figure 4 As shown. It can be seen that when When ≠0, the voltage vector deviates from its original position, resulting in modeling error.
[0103] Therefore, substituting the voltage vector into equation (11) The actual voltage vector The relationship can be expressed as:
[0104] [ V α ( k + 1 ) V β ( k + 1 ) ] = [ V ′ α ( k + 1 ) V ′ β ( k + 1 ) ] + E u ( k + 1 )
[0105] in, E u = [ E u α E u β ] T represents the modeling error caused by NP voltage imbalance, and .
[0106] Substituting formula (21) into formula (7) we can get the current prediction error:
[0107]
[0108] in, i ′ αβ ( k + 2 | k ) = [ i ′ α ( k + 2 | k ) i ′ β ( k + 2 | k ) ] T The actual voltage vector The predicted value of the inductor current is obtained.
[0109] Substituting formula (22) into formula (11), we get the inaccurate cost caused by the inaccurate voltage vector: for:
[0110]
[0111] It can be seen from equations (21), (22) and (23) that the unbalanced NP voltage superimposes a disturbance term on the correct cost function. This disturbance affects the optimization process and causes the controller to output the wrong control quantity, which is the root cause of output current distortion and three-phase imbalance.
[0112] To eliminate this effect, one method is to adjust the action time of the voltage vector to correct the voltage vector offset caused by the unbalanced NP voltage. However, this method is complicated and not suitable for model predictive control. Therefore, the present invention proposes a more intuitive method to directly calculate the exact values of all voltage vectors according to Table I and use them as the input of model predictive control. This method eliminates the disturbance term in formula (23) and ensures that the correct control quantity is obtained.
[0113] The voltage vector reconstruction of ESTPI-A can be expressed as:
[0114] [ V ′ 0 V ′ 1 V ′ 2 V ′ 3 V ′ 4 V ′ 5 V ′ 6 V ′ 7 V ′ 8 ] = [ 0 0 0 2 / 3 0 1 / 3 + j 3 / 3 0 − 1 / 3 + j 3 / 3 − 2 / 3 0 − 1 / 3 − j 3 / 3 0 1 / 3 − j 3 / 3 0 − 1 / 3 + j 3 / 3 1 / 3 + j 3 / 3 − 1 / 3 − j 3 / 3 1 / 3 − j 3 / 3 ] [ V p V n ]
[0115] Formula (24) can also be expanded to ESTPI-B and ESTPI-C as needed.
[0116] F. NP voltage balance
[0117] For ESTPIs, since there is no redundant small voltage vector, it is not possible to achieve NP voltage balance by adjusting the action time of the redundant small voltage vector like 3P-3L VSIs. To solve this problem, a feasible alternative is to control the charging and discharging of the upper and lower capacitors by adjusting the neutral point current.
[0118] Table II. Neutral point current corresponding to the voltage vector of ESTPI-A
[0119]
[0120] Table II summarizes the neutral point currents corresponding to the nine voltage vectors of ESTPI-A. Since each small triangular sector designed by the present invention contains two small voltage vectors, these small triangular sectors can be divided into "NP voltage adjustable sectors" and "NP voltage non-adjustable sectors" by analyzing the neutral point currents corresponding to the two small voltage vectors. Specifically, when the neutral point currents corresponding to the two small voltage vectors in the triangular small sector are in opposite directions, it is considered to be an "NP voltage adjustable sector"; otherwise, it is an "NP voltage non-adjustable sector".
[0121] In steady state, it is assumed that the inductor current accurately tracks the reference current. Figure 5The relationship between the three-phase current direction and the space vector is given. Combined with Table II, it can be seen that only small triangle sectors 2, 5, 7, and 8 are "NP voltage adjustable sectors", while small triangle sectors 1, 3, 4, and 6 are "NP voltage non-adjustable sectors".
[0122] Therefore, time compensation is only required in the “NP voltage adjustable sector”, for example, when , the action time of the two small voltage vectors are and , and the corresponding neutral point currents are and ; The time compensation is t, and the time after compensation can be expressed as:
[0123]
[0124] Then, the equivalent neutral point current It can be expressed as:
[0125]
[0126] Assuming the bus voltage is constant, and equation (26) is further transformed into the s domain:
[0127]
[0128] It can be seen from equation (27) that the imbalance degree ζ can be adjusted by controlling the compensation time. In order to simplify the control algorithm, the adjustment process of ζ can be regarded as position-type PID control. In order to eliminate the over-adjustment caused by integration, a PD controller can be used instead. The closed-loop control block diagram is shown in Figure 6 shown.
[0129] In addition, in order to avoid excessive NP voltage fluctuations, the reference value of ζ for sector 2 can be slightly greater than 0, because sectors 3 and 4 always reduce ζ. Similarly, the reference value of ζ for sector 5 can be slightly less than 0.
[0130] G. System control block diagram
[0131] Based on the above analysis, the complete control block diagram of H-MPC proposed in the present invention is as follows: Figure 7 shown.
[0132] The present invention proposes a hybrid model predictive control method for an eight-switch three-phase inverter, comprising:
[0133] Obtaining an upper capacitor voltage and a lower capacitor voltage of an eight-switch tube three-phase inverter, and performing vector reconstruction according to a preset table and the upper capacitor voltage and the lower capacitor voltage to obtain nine reconstructed voltage vectors;
[0134] Determine a large triangular sub-sector and a small triangular sector according to two large voltage vectors, evaluate the cost value of each voltage vector located at the midpoint of the small triangular sector, and determine the small triangular sector where the voltage vector that minimizes the cost function is located;
[0135] In each small triangle sector, the action time of three preset voltage vectors in one switching cycle is calculated;
[0136] If the small triangle sector is an NP voltage adjustable sector, the PD controller is further used to adjust the vector action time; otherwise, no adjustment is performed.
[0137] The duty cycle is calculated according to the action time and is used to control the eight-switch three-phase inverter in the next cycle.
[0138] Furthermore, the voltage vector with the minimum cost function is obtained by Lagrangian extrapolation.
[0139] Furthermore, the vector reconstruction includes: calculating accurate values of all voltage vectors as inputs of model predictive control.
[0140] Furthermore, each of the small triangular sectors includes two small voltage vectors. After determining the small triangular sector where the voltage vector that minimizes the cost function is located, the method further includes:
[0141] The neutral point currents corresponding to the two small voltage vectors are analyzed, and the small triangle sector is divided into an NP voltage adjustable sector and an NP voltage non-adjustable sector.
[0142] Furthermore, the analysis of the neutral point currents corresponding to the two small voltage vectors, dividing these small triangle sectors into NP voltage adjustable sectors and NP voltage non-adjustable sectors, includes:
[0143] When the neutral point currents corresponding to the two small voltage vectors in the triangular small sector are in opposite directions, the triangular small sector is considered to be an NP voltage adjustable sector, otherwise it is an NP voltage non-adjustable sector.
[0144] Furthermore, time compensation is performed in the NP voltage adjustable sector.
[0145] Further, the time compensation is controlled by a PD controller.
[0146] In order to verify the effectiveness of the H-MPC proposed in this paper, an experimental platform of ESTPI-A based on DSP-TMS320F28374S was built. ) and output current ( ) is sampled by the VAC current sensor, the load voltage ( ) and NP voltage ( ) is sampled by the LEM voltage sensor. DSP-TMS320F28374S is responsible for the implementation of sampling and control algorithms. The experimental parameters are detailed in Table III.
[0147] Table III. System parameters
[0148]
[0149] At the same time, the H-MPC proposed in the present invention is compared with the following two algorithms, as shown below:
[0150] Traditional MPC: T-MPC, the cost function is designed as formula (11), and a weight factor of 0.15 is used to achieve NP voltage balance. The optimal voltage vector is obtained by evaluating 9 voltage vectors.
[0151] Traditional SVPWM: T-SVPWM adopts a closed-loop control strategy with dq current decoupling, where k p = 0.8, k i =0.02, and at the same time, NP voltage balance is achieved through zero-sequence voltage injection.
[0152] H-MPC: The proposed H-MPC, where k p =0.6, k d =0.3
[0153] The experimental conditions are set up as follows:
[0154] Experimental condition 1: Set the output current amplitude to 3A, the frequency to 50Hz, and the load to a linear resistance load ( ).
[0155] Experimental condition 2: Set the output current amplitude to 5A, the frequency to 50Hz, and the load to a linear resistance load ( ).
[0156] Experimental condition 3: Set the output current frequency to 50Hz, the amplitude jumps from 3A to 5A, and the load is a linear resistance load ( ).
[0157] Experimental condition 4: Set the output current frequency to 50Hz, the amplitude jumps from 5A to 3A, and the load is a linear resistance load ( ).
[0158] Experimental condition 5: Set the output current amplitude to 3A, the frequency to 50Hz, and the load to a linear resistance load ( ), injecting a 40V DC bias into the NP voltage.
[0159] Experimental condition 6: Set the output current amplitude to 3A, the frequency to 50Hz, and the load to a linear resistance load ( ), the inductance value and the parasitic resistance value vary by ±50% respectively, and the robustness of the algorithm is compared.
[0160] The experimental effects of the present invention are as follows:
[0161] A. Calculation time
[0162] The operation time of the three methods is compared on a DSP-TMS320F28374S with a crystal frequency of 200 MHz and a sampling period of 62.5 μs.
[0163] Among them, the execution time of tasks such as A / D conversion and system protection is about 6.22μs, and the rest is the execution time of the three methods. In addition, in order to highlight the superiority of the proposed multi-step optimization method, experiments are also conducted using a non-optimization method, which calculates the cost value of each small sector.
[0164] The experimental results are as follows Figure 8 As shown. Among them, the calculation time of T-SVPWM is 11.1μs, the calculation time of T-MPC is 11.9μs, and the calculation time of H-MPC is 17μs. Due to the introduction of modulation and cost function calculation, the calculation time of the MPC-based method is longer than that of T-SVPWM. It is worth noting that the multi-step optimization method proposed in the present invention reduces the calculation time by about 65.8% compared with the non-optimized method.
[0165] Experimental results show that the proposed multi-step optimization method significantly reduces the computational complexity.
[0166] B. Steady-state performance evaluation
[0167] Fig. 9 The steady-state experimental waveforms of the three methods under experimental conditions 1 and 2 are given. The experimental results show that the three methods can effectively track the reference current. In terms of NP voltage fluctuation, the fluctuation amplitude of T-MPC and H-MPC is slightly smaller than that of T-SVPWM.
[0168] also, Fig.10 The harmonic spectrum of the output current is given, which is generated by MATLAB / Simulink, and the experimental data is collected by an oscilloscope. Under experimental condition 1, the total harmonic distortion (THD) of T-MPC, TSVPWM and H-MPC are 11.2%, 4.3% and 3.52%, respectively. Under experimental condition 2, the THD of the three methods are 6.9%, 2.98% and 2.71%, respectively. The experimental results show that H-MPC has the lowest THD under different reference current amplitudes, indicating that the vector reconstruction strategy proposed in this invention effectively reduces the impact of NP voltage fluctuations.
[0169] In addition, both T-SVPWM and H-MPC maintain a constant switching frequency. The high harmonics of the output current are mainly concentrated at the sampling frequency and its multiples, such as 16kHz and 32kHz. However, the high harmonics of T-MPC are scattered in a range of frequencies, complicating the filter design.
[0170] C. Dynamic performance evaluation experiment
[0171] In this section, the dynamic performance of the three methods is evaluated under experimental conditions 3 and 4 respectively. The experimental results are shown in Fig.11 As shown, the following conclusions can be drawn:
[0172] 1) When the reference current amplitude changes, all three methods can quickly track the given reference current. However, under experimental condition 3, due to the effect of the PI controller, T-SVPWM has an overshoot phenomenon, which shows that the dynamic performance of T-MPC and H-MPC is slightly better than that of T-SVPWM.
[0173] 2) In terms of NP voltage fluctuation, the NP voltage fluctuation of T-MPC and H-MPC is about 8 V under experimental conditions 3 and 4. In contrast, the NP voltage fluctuation of T-SVPWM under condition 3 is about 12 V. This shows that T-MPC and H-MPC perform better in suppressing NP voltage fluctuation.
[0174] D. Three-phase imbalance suppression experiment
[0175] In order to further evaluate the ability of the three methods to suppress output current imbalance when the NP voltage is unbalanced, experiments were conducted under experimental condition 5.
[0176] The experimental results are as follows Fig.12 As shown. It can be seen that when the NP voltage is unbalanced, both T-MPC and T-SVPWM have three-phase imbalance and output current distortion. In contrast, the H-MPC proposed in the present invention maintains better three-phase balance and sinusoidal current output.
[0177] In addition, in order to verify the effectiveness of the proposed vector reconstruction method, the experimental performance of H-MPC with and without vector reconstruction is compared. Fig.13 As shown in Figure 2, it can be seen that H-MPC with vector reconstruction has better output performance.
[0178] E. Model parameter mismatch experiment
[0179] In practical applications, model parameter mismatch is inevitable due to various interferences. Under experimental condition 6, the parameter sensitivity of the two MPC methods is evaluated by changing the true values of the inductance and parasitic resistance from -50% to 50%, where the change rates and are defined as:
[0180]
[0181] in, and are the true values of inductance and parasitic resistance, respectively. and are the values of the inductance and resistance actually used respectively, and the rest are the same as experimental condition 2.
[0182] The experimental results are shown in Table IV. It can be observed that even in the case of 50% parameter mismatch, the H-MPC proposed in the present invention can maintain a THD of less than 5%. Compared with T-MPC, this method has better robustness.
[0183] Table IV. Parameter mismatch analysis
[0184]
[0185] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.
Claims
1. A hybrid model predictive control method for an eight-switch three-phase inverter, characterized in that: include: Obtaining an upper capacitor voltage and a lower capacitor voltage of an eight-switch tube three-phase inverter, and performing vector reconstruction according to a preset table and the upper capacitor voltage and the lower capacitor voltage to obtain nine reconstructed voltage vectors; Determine a large triangular sub-sector and a small triangular sector according to two large voltage vectors, evaluate the cost value of each voltage vector located at the midpoint of the small triangular sector, and determine the small triangular sector where the voltage vector that minimizes the cost function is located; the voltage vector with the minimum cost function is obtained by Lagrangian extrapolation; In each small triangle sector, the action time of three preset voltage vectors in one switching cycle is calculated; If the small triangle sector is an NP voltage adjustable sector, the PD controller is further used to adjust the vector action time, otherwise no adjustment is performed; A duty cycle is calculated according to the vector action time, and is used to control the eight-switch three-phase inverter in the next cycle; The vector reconstruction includes: calculating the precise values of all voltage vectors as inputs of model predictive control.
2. The method according to claim 1, characterized in that Each of the small triangular sectors includes two small voltage vectors. After determining the small triangular sector where the voltage vector that minimizes the cost function is located, the method further includes: The neutral point currents corresponding to the two small voltage vectors are analyzed, and the small triangle sector is divided into an NP voltage adjustable sector and an NP voltage non-adjustable sector.
3. The method according to claim 2, characterized in that The analysis of the neutral point current corresponding to the two small voltage vectors divides these small triangle sectors into NP voltage adjustable sectors and NP voltage non-adjustable sectors, including: When the neutral point currents corresponding to the two small voltage vectors in the triangular small sector are in opposite directions, the triangular small sector is considered to be an NP voltage adjustable sector, otherwise it is an NP voltage non-adjustable sector.
4. The method according to claim 2 or 3, characterized in that: Further including: Time compensation is performed in the NP voltage adjustable sector.
5. The method according to claim 4, characterized in that The time compensation is controlled by a PD controller.
6. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: The processor runs the computer program to implement the method according to any one of claims 1 to 5.
7. A computer-readable storage medium having a computer program stored thereon, characterized in that: The program is executed by a processor to implement the method according to any one of claims 1 to 5.
Citation Information
Patent Citations
Vector sector positioning method and local optimization model prediction control method and device
CN111614282A
Model prediction voltage control method of T-type three-phase three-level inverter
CN117526675A
