Current Reconstruction Method of NPC Three-Level Converter Based on Model Predictive Control

By adopting a current reconstruction method based on model prediction control in the NPC three-level converter, using virtual voltage vectors to avoid CUAs, the problem of current reconstruction in traditional methods is solved, and efficient current reconstruction and neutral points are achieved Balanced, improving control performance.

CN114499249BActive Publication Date: 2024-11-12ZHENGZHOU UNIVERSITY OF LIGHT INDUSTRY
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210073134.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-01-21
Publication Date
2024-11-12
Estimated Expiration
2042-01-21

AI Technical Summary

Technical Problem

In the case of three-phase load balancing, the traditional multi-vector control method has a short vector duration or insufficient measured current, resulting in unmeasurable current zones (CUAs), making current reconstruction difficult in NPC three-level converters, and More sensors are needed.

Method used

Using a model prediction control method, the voltage vector is divided into three groups, and the virtual voltage vector is designed. By synthesizing different groups of voltage vectors, the adverse effects of CUAs are avoided, and the optimal vector is selected to reconstruct the three-phase current, so as to achieve no alternation. Reconstruction of current in the case of current sensor.

Benefits of technology

The current reconstruction of a single DC line current sensor is realized, eliminating the influence of CUAs, improving control performance, maintaining DC voltage utilization, and achieving neutral point voltage balance without weight coefficients.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114499249B_ABST
    Figure CN114499249B_ABST
Patent Text Reader

Abstract

The present invention proposes a current reconstruction method for an NPC three-level converter based on model predictive control to solve the CUAs problem of the traditional method. The steps are as follows: first, collect DC current and grid voltage; second, sample twice in the sub-vector action period according to the optimal virtual vector of the previous cycle and calculate the three-phase current; third, the three-phase current and grid voltage are transformed by Clark to obtain their αβ Component, and at the same time obtain the grid angle through the phase-locked loop, set the reference current, and calculate the reference voltage; then, construct a virtual voltage vector control set according to the switching state of the three-level NPC grid-connected converter; finally, construct a cost function, calculate the minimum value of the cost function according to the reference voltage and the virtual voltage vector, and use the virtual voltage vector corresponding to the minimum value to control the current prediction control of the next cycle. The present invention solves the problem of reduced DC voltage utilization in conventional MPC current reconstruction schemes and can provide accurate reconstruction current.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of power electronics, and in particular to a current reconstruction method of an NPC three-level converter based on model predictive control. Background Art

[0002] Compared with two-level converters, NPC three-level converters are widely used in grid-connected converter systems due to their low switching frequency, small output current harmonic components, and low switching stress. In most grid-connected converter control strategies, it is very important to obtain accurate three-phase current. Direct measurement and current reconstruction are the two main methods to obtain three-phase current. In the direct measurement method, even when the three-phase load is balanced, the measured three-phase current may be unbalanced due to the different sensitivities of the AC current sensors. Compared with current reconstruction, direct measurement requires more sensors. Therefore, current reconstruction has attracted much attention due to its small number of sensors, low cost, small size, and ability to ensure continuous operation in the event of current sensor failure. Summary of the invention

[0003] In the case of three-phase load balancing, to reconstruct the three-phase current from the DC line current sensor, it is necessary to use a voltage vector that can measure two-phase currents in one cycle. However, the traditional multi-vector control method faces the problem of current unmeasurable areas (CUAs) due to short vector duration or insufficient number of measurable currents. The present invention proposes a current reconstruction method for an NPC three-level converter based on model predictive control, which divides the voltage vectors into three groups according to the measurable phase currents; by synthesizing appropriate voltage vectors of different groups, a virtual voltage vector that can balance the neutral point voltage and avoid the adverse effects of CUAs is designed; by selecting the optimal vector from the virtual voltage vector, the three-phase current is reconstructed without an AC current sensor.

[0004] The technical solution of the present invention is achieved in this way:

[0005] A current reconstruction method for an NPC three-level converter based on model predictive control, the steps of which are as follows:

[0006] Step 1: Collect the DC line current i of the three-level NPC grid-connected converter through the current sensor dc , and according to the DC line current i dc The two-phase currents obtained by sampling the optimal virtual vector of the previous cycle are used to calculate the three-phase current i a 、i b 、i c ;

[0007] Step 2: construct a virtual voltage vector control set according to the switching state of the three-level NPC grid-connected converter;

[0008] Step 3: Collect the grid voltage e of the three-level NPC grid-connected converter through the voltage sensor a 、e b 、e c ;

[0009] Step 4: Three-phase current i a 、i b 、i c The αβ components i of the three-phase current are obtained by Clark transformation α 、i β ; Grid voltage e a 、e b 、e c The αβ component of the voltage is obtained by Clark transformation α 、e β ;

[0010] Step 5: Set the grid voltage e a 、e b 、e c The input phase-locked loop PLL obtains the grid angle and sets the reference current according to the grid angle;

[0011] Step 6: According to the reference current and the αβ component i of the three-phase current α 、i β and the αβ component of the voltage e α 、e β Calculate the reference voltage;

[0012] Step 7: construct a cost function, calculate the minimum value of the cost function according to the reference voltage and the virtual voltage vector in the virtual voltage vector control set, and use the virtual voltage vector corresponding to the minimum value of the cost function as the optimal voltage vector;

[0013] Step 8: Apply the switch state corresponding to the optimal voltage vector to the three-level NPC grid-connected converter to reconstruct the three-phase current at the next moment, and return to step 3 to predict the next control cycle.

[0014] Preferably, the DC line current i dc With the three-phase current i a 、i b 、i c The relationship between them is:

[0015] i dc =S Pa *i a +S Pb *i b +S Pc *i c ;

[0016] Among them, S Pa , S Pb and S Pc is the coefficient of the corresponding phase which is 1 in P state and 0 otherwise;

[0017]

[0018] Preferably, the virtual voltage vector control set is {V V1 ,V V2 ,…,V V36}, where the corresponding relationship between the virtual voltage vector and the sub-vector is shown in Table 1:

[0019] Table 1 Correspondence between virtual voltage vector and sub-vector

[0020] Virtual Voltage Vector Sub-vector <![CDATA[V V1 -V V3 ]]> PPN, PNN, NNN, PPP <![CDATA[V V4 -V V6 ]]> PPN,NPN,NNN,PPP <![CDATA[V V7 -V V9 ]]> NPN, NPP, NNN, PPP <![CDATA[V V10 -V V12 ]]> NPP, NNP, NNN, PPP <![CDATA[V V13 -V V15 ]]> NNP, PNP, NNN, PPP <![CDATA[V V16 -V V18 ]]> PNP, PNN, NNN, PPP <![CDATA[V V19 -V V21 ]]> PPN,PNN <![CDATA[V V22 -V V24 ]]> PPN,NPN <![CDATA[V V25 -V V27 ]]> NPN,NPP <![CDATA[V V28 -V V30 ]]> NPP,NNP <![CDATA[V V31 -V V33 ]]> NNP,PNP <![CDATA[V V34 -V V36 ]]> PNP,PNN

[0021] The sub-vector is related to the switch state. The switch state corresponding to PPN is (S Pa S Pb S Pc )=(110), the switch state corresponding to PNN is (S Pa S Pb S Pc )=(100), the switch state corresponding to NNN is (S Pa S Pb S Pc )=(000), the switch state corresponding to PPP is (S Pa S Pb S Pc )=(111), the switch state corresponding to NPN is (S Pa S Pb S Pc )=(010), the switch state corresponding to NPP is (S Pa S Pb S Pc )=(011), the switch state corresponding to NPP is (S Pa S Pb S Pc )=(011), the switch state corresponding to NNP is (S Pa S Pb S Pc )=(001), the switch state corresponding to PNP is (S Pa S Pb S Pc )=(101).

[0022] Preferably, the reference voltage is calculated as follows:

[0023]

[0024] in, is the α-axis component of the reference voltage, is the β-axis component of the reference voltage, i αref is the α-axis component of the reference current, i βref is the β-axis component of the reference current, T s is the sampling period, L is the filter inductance, and R is the filter resistance.

[0025] Preferably, the cost function is expressed as:

[0026]

[0027] Among them, G is the cost function, u α is the α-axis component of the virtual voltage vector, u β is the β-axis component of the virtual voltage vector.

[0028] Preferably, the method of applying the switch state corresponding to the optimal voltage vector to the three-phase current of the three-level NPC grid-connected converter to reconstruct the three-phase current at the next moment is: the optimal voltage vector corresponds to a group of switch states, and two sampling points are respectively selected from the group of switch states to collect two current values ​​of the a-phase, b-phase, and c-phase currents; according to i a +i b +i c =0 principle, the third current value can be obtained, that is, the three-phase current is obtained; wherein, the two sampling points correspond to different switching states, and each switching state corresponds to a long vector.

[0029] Preferably, the synthesis method of the virtual voltage vector is:

[0030] Calculate the duty cycle of the sub-vector:

[0031]

[0032]

[0033] Among them, G1 and G2 are V VL The cost function value of the active sub-vector, d L1 d L2 V VL Duty cycle of the active sub-vector; d S1 and d S2 V VS The duty cycle of the active sub-vector;

[0034] The expression of the virtual voltage vector is:

[0035]

[0036] Among them, V VL_αβ is the αβ component of the large virtual voltage vector, V VS_αβ is the αβ component of the small virtual voltage vector, V 1αβ and V 2αβ Both represent the αβ components of the long vector in the sub-vector.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] 1) The virtual voltage vector is combined with the MPC method to achieve midpoint voltage balancing and current reconstruction of the NPC three-level grid-connected converter using a single DC line current sensor.

[0039] 2) In order to eliminate the influence of the two CUAs, a suitable position and a new vector synthesis combination are selected in the space vector diagram to design the virtual voltage vector; a large voltage vector with zero neutral current is used to synthesize the virtual voltage vector, which can achieve neutral point voltage balance without a weight coefficient.

[0040] 3) Compared with the current reconstruction based on conventional MPC, the method of the present invention does not reduce the DC voltage utilization rate and improves the control performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0042] Figure 1 This is the topological structure diagram of the NPC three-level grid-connected converter.

[0043] Figure 2 Available vectors for the NPC three-level converter; (a) a-phase sensor failure. (b) b-phase sensor failure.

[0044] Figure 3 Arrangement of single DC line current sensor for three cases.

[0045] Figure 4 The spatial voltage vector diagram shows the CUAs in three cases, including (a) case 1, (b) case 2, and (c) case 3.

[0046] Figure 5 Spatial virtual voltage vector diagram for the NPC three-level converter.

[0047] Figure 6The synthesis process of the virtual voltage vector of sector I, where (a) is a large virtual voltage vector and (b) is a small virtual voltage vector.

[0048] Figure 7 The influence of different vectors on the neutral point potential; among them, (a) the original voltage vector, (b) the virtual voltage vector.

[0049] Figure 8 It is the control principle diagram of the present invention.

[0050] Fig. 9 Comparison results of current waveforms generated by conventional current reconstruction methods and the method proposed in the present invention; among them, (a) conventional current reconstruction based on conventional MPC, (b) conventional current reconstruction based on conventional SVPWM, and (c) the method proposed in the present invention.

[0051] Fig.10 Comparison results of current waveform and error; among them, (a) measured current and reconstructed current, (b) error between measured current and reconstructed current.

[0052] Fig.11 are the capacitor voltage and output current of the NPC three-level converter when the reference current changes suddenly.

[0053] Fig.12 It is the FFT analysis result of the output current after the reference current changes. DETAILED DESCRIPTION

[0054] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0055] Conventional current reconstruction of MPC: When an AC current sensor fails, it is usually Figure 1 The structure shown uses the conventional MPC-based current reconstruction method to reconstruct the three-phase current. Taking the A-phase sensor failure as an example, the A-phase current can be obtained by the B-phase current i b , DC line current i dc and phase voltage state P(U dc / 2)、O(0)、N(-U dc / 2)get.

[0056] Since the DC sensor is located between capacitor C1 and the first switching device S of each phase i1 The DC line current i dcThe expression can be derived as:

[0057]

[0058] i dc =S Pa *i a +S Pb *i b +S Pc *i c (2)

[0059] where i a 、i b 、i c They are phase currents a, b and c respectively.

[0060] According to a +i b +i c =0,i a It can be expressed as:

[0061]

[0062] Therefore, the current of the fault phase can be reconstructed using equation (3). However, in equation (3), S Pa Cannot be equal to S Pc Similar limitations apply when the phase B sensor fails. These limitations reduce the number of available voltage vectors for current reconstruction. Figure 2 The available vectors under different sensor faults are given. The shaded area is the voltage vector area after considering vector synthesis, but there are still some white areas remaining, indicating that the DC voltage utilization is reduced.

[0063] Based on the above analysis, it can be seen that: 1) The reduction of available vectors leads to a decrease in the control performance of the current reconstruction method based on traditional MPC. 2) Since the current reconstruction strategy based on traditional MPC lacks some vectors, the DC voltage utilization rate is reduced.

[0064] Conventional current reconstruction based on SVPWM: The current reconstruction method based on conventional SVPWM can reconstruct the three-phase current without the need for AC current sensors, solving more challenging conditions. Figure 1 As shown in the sensor placement scheme in , it can be seen that each active vector can determine a phase current according to equation (2). To achieve current reconstruction, at least two phase currents need to be obtained under three-phase balance conditions. This is also the reason why multi-vector control methods such as SVPWM are widely used in current reconstruction.

[0065] However, conventional SVPWM current reconstruction has CUAs. CUAs are related to the placement of the DC link current sensor. There are three possible placements such as Figure 3shown.

[0066] In these three cases, there are two reasons for CUAs, namely, too short vector duration and insufficient number of vectors of measurable phase currents. Based on these two reasons, CUAs can be divided into two categories, namely, CUAs1 and CUAs2.

[0067] In SVPWM, the duty cycle of the sub-vector is different in different cycles. However, in order to achieve accurate measurement, the measurement time of the current sensor needs to be at least greater than the minimum allowed duty cycle d min The corresponding time. CUAs 1 is composed of duty cycle less than d min The area composed of sub-vectors. Minimum duty cycle d min It is expressed as:

[0068]

[0069] Where T s is the sampling period, t dead To avoid the dead time of bridge arm short circuit fault, t sett for i dc The stabilization time, t AD The time required for sampling and A / D conversion.

[0070] In the conventional SVPWM strategy, the reference voltage V ref The three-phase currents cannot be reconstructed in the triangular area formed by the NTVs if the measurable currents of the NTVs are in the same phase. CUAs 2 are composed of these triangular areas.

[0071] The CUAs in three cases are as follows Figure 4 shown and marked with two different colors.

[0072] Through the analysis of CUAs, it can be seen that: 1) The current reconstruction in the three single DC line current sensor arrangements all has CUAs problems. 2) CUAs in the three cases all contain CUAs 1, and only cases 1 and 3 contain CUAs 2. Compared with CUAs 1, CUAs 2 has a greater impact on current reconstruction.

[0073] In order to avoid the adverse effects of CUAs in current reconstruction, for situation 1, the present invention proposes a current reconstruction method for an NPC three-level converter based on model predictive control. Due to the requirements of single DC line current sampling technology, multiple vectors need to be output within one cycle. Therefore, virtual vectors that avoid CUAs can be pre-designed and combined into the control set of the finite control set model predictive control (FCS-MPC) method to achieve current reconstruction. The number of different voltage vectors output by FCS-MPC is fixed. This means that when the angle and size of the virtual vector are set reasonably, all possible output vectors are not affected by CUAs. Therefore, the model predictive control current reconstruction method based on virtual vectors can effectively solve the problem of CUAs.

[0074] In MPC, the optimal voltage vector is usually selected by a cost function related to the predicted current. Assume that the sampling period T s is very small, and the inductor voltage can be calculated using the Euler forward formula. Then, according to Figure 1 , predict the current i α p and i β p The αβ components can be expressed as:

[0075]

[0076] Where u, i, e represent the converter output voltage, output current, and grid voltage respectively. The subscript αβ represents their αβ components. L and R are the filter inductance and resistance respectively.

[0077] Then the cost function g of conventional MPC can be expressed as:

[0078]

[0079] where i αref and i βref is the αβ component of the reference current.

[0080] In addition, the cost function in equation (6) can be further improved to direct virtual vector control, and the optimal vector is selected according to the gradient theory. The derivation process is as follows:

[0081]

[0082] The second-order derivative 2(Ts / L) can be obtained 2 Greater than 0. Therefore, when the value of equation (7) is 0, the reference voltage vector u α opt 、u βopt :

[0083]

[0084] Then the reference current i αref ,i βref It can be written as:

[0085]

[0086] According to equations (5), (6) and (9), the cost function of the present invention can be expressed as:

[0087]

[0088] In equation (10), the variable is only the voltage component. Therefore, the cost function value can be obtained by directly comparing the virtual voltage vector and the reference voltage.

[0089] Figure 8 The control principle diagram of the method of the present invention is as follows:

[0090] Step 1: Collect the DC line current i of the three-level NPC grid-connected converter through the current sensor dc , and according to the DC line current i dc The two-phase currents obtained by sampling the optimal virtual vector of the previous cycle are used to calculate the three-phase current i a 、i b 、i c ;

[0091] DC line current i dc With the three-phase current i a 、i b 、i c The relationship between them is:

[0092] i dc =S Pa *i a +S Pb *i b +S Pc *i c (11)

[0093] Among them, S a is the switching state of phase a, S b is the b-phase switch state, S c It is the switching state of phase c;

[0094]

[0095] Step 2: construct a virtual voltage vector control set according to the switching state of the three-level NPC grid-connected converter;

[0096] The virtual voltage vector control set is {V V1 ,V V2 ,…,V V36},like Figure 5 As shown, V V1 -V V18 is a small virtual vector, V V19 -V V36 is a large virtual vector. Referring to the distribution of the original voltage vector, three evenly distributed virtual vector angles are designed in each sector. VS Located inside the inner hexagon, the large virtual vector V VL Located on the outer hexagon.

[0097] Take the small virtual vector V in sector I VS and the large virtual vector V VL For example, the virtual vector synthesis process is as follows Figure 6 shown.

[0098] According to formula (10), Figure 6 The duty cycle of each sub-vector in can be expressed as:

[0099]

[0100] Among them, G1 and G2 are V VL The cost function value of the active sub-vector, d L1 d L2 V VL Duty cycle of the active sub-vector.

[0101] Each small virtual vector V VS Corresponding to a large virtual vector V with twice the length VS . In V VS During the synthesis process, the zero vector duty cycle is 0.5. Therefore, the remaining 0.5 can be expressed as:

[0102]

[0103] Among them, d S1 and d S2 V VS The duty cycle of the active sub-vector.

[0104] When the duty cycle is greater than d min When , the virtual vector can reconstruct the current. In the present invention, d min is 0.05, and all virtual vectors meet this requirement.

[0105] The expression of the virtual voltage vector is:

[0106]

[0107] Among them, V VL_αβ is the αβ component of the large virtual voltage vector, V VS_αβ is the αβ component of the small virtual voltage vector, V 1αβ and V 2αβ Both represent the αβ components of the long vector in the sub-vector.

[0108] The corresponding relationship between the virtual voltage vector and the sub-vector is shown in Table 1:

[0109] Table 1 Correspondence between virtual voltage vector and sub-vector

[0110] Virtual Voltage Vector Sub-vector <![CDATA[V V1 -V V3 ]]> PPN, PNN, NNN, PPP <![CDATA[V V4 -V V6 ]]> PPN,NPN,NNN,PPP <![CDATA[V V7 -V V9 ]]> NPN, NPP, NNN, PPP <![CDATA[V V10 -V V12 ]]> NPP, NNP, NNN, PPP <![CDATA[V V13 -V V15 ]]> NNP, PNP, NNN, PPP <![CDATA[V V16 -V V18 ]]> PNP, PNN, NNN, PPP <![CDATA[V V19 -V V21 ]]> PPN,PNN <![CDATA[V V22 -V V24 ]]> PPN,NPN <![CDATA[V V25 -V V27 ]]> NPN,NPP <![CDATA[V V28 -V V30 ]]> NPP,NNP <![CDATA[V V31 -V V33 ]]> NNP,PNP <![CDATA[V V34 -V V36 ]]> PNP,PNN

[0111] The sub-vector is related to the switch state. The switch state corresponding to PPN is (S Pa S Pb S Pc )=(110), the switch state corresponding to PNN is (S Pa S Pb S Pc )=(100), the switch state corresponding to NNN is (S Pa S Pb S Pc )=(000), the switch state corresponding to PPP is (S Pa S Pb S Pc )=(111), the switch state corresponding to NPN is (S Pa S Pb S Pc )=(010), the switch state corresponding to NPP is (S Pa S Pb S Pc )=(011), the switch state corresponding to NPP is (S Pa S Pb S Pc )=(011), the switch state corresponding to NNP is (S Pa S Pb S Pc )=(001), the switch state corresponding to PNP is (S Pa S Pb S Pc )=(101).

[0112] According to Table 1, each small virtual vector is composed of two long vectors and a zero vector (corresponding to two switching states NNN and PPP), and the large virtual vector is composed of only two long vectors. By synthesizing the virtual vectors for control according to the selected sub-vectors, CUAs 2 will be eliminated. In addition, since the duty cycle of the long vector in the sub-vector is greater than d min , so CUAs 1 will not affect the current reconstruction.

[0113] Figure 6 The synthesis method shown also solves the midpoint balancing problem of conventional MPC during current reconstruction. In conventional MPC, a neutral point voltage balancing term with a weight coefficient is added to the cost function to balance the neutral point voltage. A small primitive vector corresponds to two switching states with the same vector output voltage but opposite neutral point current. Therefore, these switching states are important for the balance of the neutral point. However, since the phase current cannot be detected, half of the switching states of the small primitive vector cannot be used, which seriously affects the performance of the neutral point balancing term of conventional MPC.

[0114] from Figure 6 and Figure 4 As can be seen from case 1 in , the measurable currents of the non-zero sub-vectors are different. In addition, all sub-vectors do not generate neutral point current, so the virtual vector can achieve neutral point balance in current reconstruction. Figure 7 The effect of the voltage vector and the virtual vector on the neutral point potential is shown.

[0115] from Figure 7 It can be seen that when the proposed virtual vector acts, the neutral point current i O = 0. Therefore, the method of the present invention does not need to add a neutral point balance term to the cost function, which means that the method of the present invention is not affected by the change of the optimal weight coefficient with different operating conditions.

[0116] According to the virtual vectors analyzed above, the influence of the two CUAs can be eliminated: 1) For CUAs1, the designed virtual vectors all satisfy d min , that is, the duration is sufficient for the sensor to measure accurately. 2) For CUAs2, the voltage vectors in two different measurable current groups are selected as non-zero subvectors of the virtual vector. Therefore, all virtual vectors are not affected by CUAs2. From the reconstruction process of the above virtual voltage vectors, it can be seen that they have the inherent characteristics of balanced neutral point voltage and are all located in the measurable area.

[0117] Step 3: Collect the grid voltage e of the three-level NPC grid-connected converter through the voltage sensor a 、e b 、e c ;

[0118] Step 4: Three-phase current i a 、i b 、i c The αβ components i of the three-phase current are obtained by Clark transformation α 、i β ; Grid voltage e a 、e b 、e c The αβ component of the voltage is obtained by Clark transformation α 、e β ;

[0119] Step 5: Set the grid voltage e a 、e b 、e c The input phase-locked loop PLL obtains the grid angle and sets the reference current according to the grid angle;

[0120] Step 6: According to the reference current and the αβ component i of the three-phase current α 、i β and the αβ component of the voltage e α 、e β Calculate the reference voltage;

[0121] The reference voltage is calculated as follows:

[0122]

[0123] in, is the α-axis component of the reference voltage, is the β-axis component of the reference voltage, i αref is the α-axis component of the reference current, i βref is the β-axis component of the reference current, T s is the sampling period, L is the filter inductance, and R is the filter resistance.

[0124] Step 7: construct a cost function, calculate the minimum value of the cost function according to the reference voltage and the virtual voltage vector in the virtual voltage vector control set, and use the virtual voltage vector corresponding to the minimum value of the cost function as the optimal voltage vector;

[0125] The expression of the cost function is:

[0126]

[0127] Among them, G is the cost function, u α is the α-axis component of the virtual voltage vector, u β is the β-axis component of the virtual voltage vector.

[0128] Step 8: Apply the switch state corresponding to the optimal voltage vector to the three-level NPC grid-connected converter to reconstruct the three-phase current at the next moment, and return to step 3 to predict the next control cycle. The optimal voltage vector corresponds to a set of switch states, and two sampling points are selected from this set of switch states to obtain two current values ​​of phase a, phase b, and phase c; according to i a +i b +i c =0 principle, the third current value can be obtained, that is, the three-phase current is obtained; wherein, the two sampling points correspond to different switching states, and each switching state corresponds to a long vector.

[0129] Based on the method proposed in the present invention, it can be seen that: 1) This method can only output the designed virtual vector, eliminating the influence of CUAs. 2) Since no neutral point current is generated in one cycle, the proposed method can achieve neutral point balance without a weight coefficient. 3) Compared with conventional MPC, this method has more available vectors, which helps to improve tracking performance. In addition, the DC voltage utilization rate is not reduced.

[0130] In order to compare the conventional current reconstruction with the proposed method, a comparative experiment was carried out using the b-phase sensor failure as an example. Fig. 9 shown.

[0131] exist Fig. 9 In (a), when the sector number is 1 or 6, the output current has obvious errors, which is consistent with Figure 2 (b) is consistent with the analysis. Fig. 9 In (b), when the reference vector is located at CUAs, the reconstructed i reca 、i recb 、i recc There is also a significant error in the current. Fig. 9 In (c), the output current can accurately track the reference current. The results show that the proposed method does not reduce the DC voltage utilization and CUAs.

[0132] exist Fig.10 In (a), a error 、b error 、c error It is the percentage of the error between the three-phase actual current and the three-phase reconstructed current in the reference current. Fig.10 (b) is a comparison of the actual current and the reconstructed current waveform under different reference currents. The peak value of the reference current is 10A in the first 0.03 seconds and 15A thereafter.

[0133] from Fig.10It can be seen that when the reference current is 10A and 15A, the reconstructed current waveform is consistent with the measured current waveform, and the maximum error is about 3%. The results show that this method can effectively reconstruct the three-phase current under different reference currents and has high current reconstruction accuracy.

[0134] In order to evaluate the dynamic performance of the proposed MPC method, Fig.11 The capacitor voltage and output current of the NPC three-level converter are shown when the reference current changes suddenly. The reference current changes from 10A to 15A in 0.03 seconds, and the output current can track the reference current in a short time. In addition, when the proposed virtual vector is applied, the neutral point current is equal to 0A in each cycle. Therefore, the proposed MPC method can balance the neutral point even without using the weight coefficient.

[0135] Fig.12 This is the FFT analysis of the output current after the reference current changes. The THD value is 1.83%. Fig.11 and Fig.12 It can be seen that the proposed MPC method can quickly reconstruct the current when the reference value fluctuates and output accurate current.

[0136] The present invention proposes an MPC current reconstruction strategy based on virtual vector control for current reconstruction of an NPC three-level converter with a single DC line current sensor. The CUAs of three sensor positions are analyzed and divided into two types. In order to avoid the influence of these two CUAs on current reconstruction, a virtual vector is designed. Theoretical analysis and experimental results show that this method has the following advantages: 1) Compared with the conventional SVPWM current reconstruction method, the proposed virtual vector eliminates the adverse effects of CUAs. 2) In current reconstruction, the use of vectors without measurable phase currents is limited. In order to overcome its influence on neutral point balance, a virtual vector without neutral point current is designed. Therefore, this method does not require an additional neutral point balance method. 3) Compared with the current reconstruction based on conventional MPC, this method does not reduce the DC voltage utilization and improves the control performance. Therefore, the proposed current reconstruction method solves the problems of reduced DC voltage utilization and CUAs. In addition, the method can also provide accurate reconstructed current and output current with low THD.

[0137] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A current reconstruction method for NPC three-level converter based on model predictive control, It is characterized in that The steps are as follows: Step 1: Collect the DC line current i of the three-level NPC grid-connected converter through the current sensor dc , and according to the DC line current i dc The two-phase currents obtained by sampling the optimal virtual vector of the previous cycle are used to calculate the three-phase current i a 、i b 、i c ; Step 2: construct a virtual voltage vector control set according to the switching state of the three-level NPC grid-connected converter; Step 3: Collect the grid voltage e of the three-level NPC grid-connected converter through the voltage sensor a 、e b 、e c ; Step 4: Three-phase current i a 、i b 、i c The αβ components i of the three-phase current are obtained by Clark transformation α 、i β ; Grid voltage e a 、e b 、e c The αβ component of the voltage is obtained by Clark transformation α 、e β ; Step 5: Set the grid voltage e a 、e b 、e c The input phase-locked loop PLL obtains the grid angle and sets the reference current according to the grid angle; Step 6: According to the reference current and the αβ component i of the three-phase current α 、i β and the αβ component of the voltage e α 、e β Calculate the reference voltage; The reference voltage is calculated as follows: in, is the α-axis component of the reference voltage, is the β-axis component of the reference voltage, i αref is the α-axis component of the reference current, i βref is the β-axis component of the reference current, T s is the sampling period, L is the filter inductance, and R is the filter resistance; Step 7: construct a cost function, calculate the minimum value of the cost function according to the reference voltage and the virtual voltage vector in the virtual voltage vector control set, and use the virtual voltage vector corresponding to the minimum value of the cost function as the optimal voltage vector; The expression of the cost function is: Among them, G is the cost function, u α is the α-axis component of the virtual voltage vector, u β is the β-axis component of the virtual voltage vector; Step 8: Apply the switch state corresponding to the optimal voltage vector to the three-level NPC grid-connected converter to reconstruct the three-phase current at the next moment, and return to step 3 to predict the next control cycle; The method of applying the switch state corresponding to the optimal voltage vector to the three-level NPC grid-connected converter to reconstruct the three-phase current at the next moment is as follows: the optimal voltage vector corresponds to a set of switch states, and two sampling points are selected from this set of switch states to collect two current values ​​of phase a, phase b, and phase c; according to i a +i b +i c =0, the third current value is obtained, that is, the three-phase current is obtained; wherein, the two sampling points correspond to different switching states, and each switching state corresponds to a long vector.

2. The current reconstruction method of the NPC three-level converter based on model predictive control according to claim 1, It is characterized in that DC line current i dc With the three-phase current i a 、i b 、i c The relationship between them is: i dc =S Pa *i a +S Pb *i b +S Pc *i c ; Among them, S Pa , S Pb and S Pc is the coefficient of the corresponding phase which is 1 in P state and 0 otherwise; 3. The current reconstruction method of the NPC three-level converter based on model predictive control according to claim 2, It is characterized in that The virtual voltage vector control set is {V V1 ,V V2 ,…,V V36 }, where the corresponding relationship between the virtual voltage vector and the sub-vector is shown in Table 1: Table 1 Correspondence between virtual voltage vector and sub-vector The sub-vector is related to the switch state. The switch state corresponding to PPN is (S Pa S Pb S Pc )=(110), the switch state corresponding to PNN is (S Pa S Pb S Pc )=(100), the switch state corresponding to NNN is (S Pa S Pb S Pc )=(000), the switch state corresponding to PPP is (S Pa S Pb S Pc )=(111), the switch state corresponding to NPN is (S Pa S Pb S Pc )=(010), the switch state corresponding to NPP is (S Pa S Pb S Pc )=(011), the switch state corresponding to NPP is (S Pa S Pb S Pc )=(011), the switch state corresponding to NNP is (S Pa S Pb S Pc )=(001), the switch state corresponding to PNP is (S Pa S Pb S Pc )=(101).

4. The current reconstruction method of the NPC three-level converter based on model predictive control according to claim 3, It is characterized in that The synthesis method of the virtual voltage vector is: Calculate the duty cycle of the sub-vector: Among them, G 1 , G 2 V VL The cost function value of the active sub-vector, d L1 d L2 V VL Duty cycle of the active sub-vector; d S1 and d S2 V VS The duty cycle of the active sub-vector; The expression of the virtual voltage vector is: Among them, V VL_αβ is the αβ component of the large virtual voltage vector, V VS_αβ is the αβ component of the small virtual voltage vector, V 1αβ and V 2αβ Both represent the αβ components of the long vector in the sub-vector.