An improved model predictive control method for three-level NPC inverter based on junction temperature

Abstract

The present invention discloses an improved model predictive control method for a three-level NPC inverter guided by junction temperature, comprising the following steps: measuring the load current and capacitor voltage values ​​at the current sampling moment, and measuring the switching state of each phase at the current sampling moment; using the load current measurement value at the current sampling moment to predict the reference voltage vector at the next moment, and determining the sector in which the reference voltage vector is located based on the load current measurement value; and performing the following iterative calculation on seven voltage vectors in the candidate vector set corresponding to the sector in which the reference voltage vector is located to obtain the optimized vector at the current sampling moment. By applying the improved model predictive current control strategy to the resistive-inductive load of the three-level NPC inverter, the present invention ensures the midpoint voltage balance effect and other control performance while reducing and balancing the junction temperature of the inverter power components and improving the inverter life.

Description

Technical Field

[0001] The present invention belongs to the technical field of three-level NPC inverter modulation, and in particular relates to an improved model predictive control method for a junction temperature-guided three-level NPC inverter. Background Art

[0002] Three-level neutral-point clamped (NPC) inverters are widely used in rail transit traction drives, wind power generation, shield tunneling, and other fields due to their low switching device withstand voltage, low output voltage harmonics, and mature control technology. However, the operating environment of inverters is becoming increasingly harsh, leading to increasingly stringent requirements for three-level system-level reliability.

[0003] Engineering practice shows that the vast majority of three-level inverter failures are caused by power devices. During the switching process of a three-level inverter, instantaneous energy changes and distribution are highly likely to become unbalanced, resulting in localized excessive energy in the power devices and causing their failure. Insulated-gate bipolar transistors (IGBTs), primarily used as power devices in high-power converters, are significantly affected by variable operating conditions and are relatively fragile. Studies have shown that the lifetime of an IGBT module is largely determined by the junction temperature variation and average junction temperature. The lifetime of an IGBT is related to the number of temperature cycles at different junction temperature variations and average junction temperatures. Therefore, the lifetime of an IGBT can be expressed in terms of the junction temperature fluctuation and average junction temperature. Therefore, reducing the average junction temperature and junction temperature fluctuation of the IGBT is crucial for improving the lifetime of the IGBT module, and thus the inverter lifespan. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to address the deficiencies of the existing technology and provide an improved model predictive control method for a three-level NPC inverter guided by junction temperature, so as to coordinately control the output current, midpoint potential balance and junction temperature of the power device of the resistive-inductive load of the three-level NPC inverter.

[0005] In order to solve the above technical problems, the present invention includes:

[0006] An improved model predictive control method for a junction temperature-guided three-level NPC inverter includes the following steps:

[0007] Step A1: Measure the load current measurement value i(k) and capacitor voltage value U at the current sampling time k c1 (k) and U c2 (k), measure the switching state S of each phase at the current sampling time k A 、S B and S c ;

[0008] Step A2: Use the load current measurement value i(k) at the current sampling time k to predict the reference voltage vector υ*(k+1) at the next time k+1, and determine the sector in which the reference voltage vector υ*(k+1) is located based on the reference voltage vector υ*(k+1):

[0009] The sectors are divided into 12 sectors at intervals of 30° in the counterclockwise direction, starting from the α-axis in the α-β coordinate system. The sectors are sequentially named sectors 1 to 12. The candidate vector set corresponding to each sector has seven voltage vectors.

[0010] Then, the reference voltage vector υ*(k+1) is decomposed in the α-β coordinate system, and the angle between the reference voltage vector υ*(k+1) and the α axis is calculated to determine the sector in which the reference voltage vector υ*(k+1) is located;

[0011] Step A3: In the candidate vector set corresponding to the sector where the reference voltage vector υ*(k+1) determined in step A2 is located, the following iterative calculation is performed on the seven voltage vectors in sequence to obtain the optimized vector at the current sampling time k:

[0012] A3-1: First, use the load current measurement value i(k) at the current time k and each voltage vector to predict the load current value i at the next time k+1 p (k+1);

[0013] A3-2: Then use the capacitor voltage value U at the current moment k c1 (k), U c2 (k) and the switching state S of each phase A 、S B and S C Predict the midpoint potential ΔU at the next moment k+1 c (k+1);

[0014] A3-3: In the cost function, dynamically determine whether the power device is in the switching state or the conducting state at the next moment relative to the previous moment, and add a factor to reduce the corresponding power loss in the corresponding state. This results in a final cost function that comprehensively considers current control, midpoint potential balance control, and power device loss reduction.

[0015] A3-4: The load current value i at the next moment k+1 predicted in step A3-1 is p (k+1) and the midpoint potential ΔU at the next moment k+1 predicted in step A-2 c Substitute (k+1) into the final cost function to calculate the g value corresponding to each voltage vector;

[0016] A3-5: Compare and determine the minimum value among all g values, and use the voltage vector corresponding to the minimum g value as the optimized vector υ at the current sampling time k opt .

[0017] Furthermore, in step A2, the reference voltage vector υ*(k+1) at the next moment k+1 is predicted according to the following formula:

[0018]

[0019] In the formula, R represents the three-phase load resistance, L represents the three-phase load inductance, T s represents the sampling time, i*(k+1) represents the reference current value at the next moment k+1;

[0020] Furthermore, in step A3-1, the load current value i at the next moment k+1 is calculated using the following formula: p (k+1) makes predictions:

[0021]

[0022] Where υ(k) represents the voltage vector.

[0023] Furthermore, in step A3-2, the midpoint potential ΔU at the next moment k+1 is calculated using the following formula: c (k+1) makes predictions:

[0024]

[0025] Where, T s represents the sampling time, C represents the capacitance value;

[0026] ΔU c (k) represents the voltage difference between the two capacitors at the current moment k:

[0027] ΔU c (k)=U c1 (k)-U c2 (k)

[0028] i0(k+1) represents the predicted value of the midpoint current at the next moment k+1:

[0029] i0(k+1)=(1|S A |)i a +(1|S B |)i b +(1|S C |)i c

[0030] Where i a 、i b and i cIndicates the three-phase load current measurement value.

[0031] Furthermore, in step A3-3, the final cost function is:

[0032]

[0033] Where, represents the real part of the reference current, represents the imaginary part of the reference current, represents the real part of the predicted current, represents the imaginary part of the predicted current, Indicates the weight coefficient for controlling the midpoint potential balance, ΔU c Represents the voltage difference between the two capacitors, g A 、g B and g C They represent the loss cost functions of phase A, phase B, and phase C respectively, and the three loss cost functions have the same form;

[0034] The loss cost function of phase A is:

[0035] g A =2λ con P con +(i a >0)(4(S a (k)==-1)(S a (i)==1)+3(S a (k)==-1)(S a (i)==0)+(S a (k)==1)(S a (i)==0))λ sw P sw +2(i a >0)((S a (k)==0)(S a (i)==1)+(S a (k)==1)(S a (i)==-1)+(s a (k)==0)(S a (i)==-1))λ sw P sw +(i a <0)(4(S a (k)==1)(S a (i)==-1)+3(S a (k)==1)(S a (i)==0)+(S a (k)==-1)(S a (i)==0))λ swP sw +2(i a <0)((S a (k)==0)(S a (i)==1)+(S a (k)==-1)(S a (i)==1)+(S a (k)==0)(S a (i)==-1))λ sw P sw

[0036] Where, P con Represents the conduction loss of the power device, λ con Represents the weight coefficient of the conduction loss factor, P sw Represents the switching loss of the power device, λ sw Represents the weight coefficient of the switching loss factor; S a (k) represents the switching state of phase A at time k; S a (i) represents the switching state of phase A corresponding to the i-th voltage vector.

[0037] The beneficial effects of the present invention are:

[0038] The present invention is guided by reducing the junction temperature of IGBTs / Diodes and aims to increase the life of inverters. Based on the model predictive current control (MPCC) method, it reduces the system workload by dividing sectors while dynamically adding loss factors to the cost function. In this way, while ensuring the current control and midpoint potential balance effects, it minimizes and balances the junction temperature of power devices, increases their life, and ultimately increases the life of the inverter. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 This is the topology diagram of the three-level NPC inverter;

[0040] Figure 2 It is the control flow chart of the traditional MPCC;

[0041] Figure 3 It is the voltage vector and partition diagram of the present invention;

[0042] Figure 4 This is a schematic diagram of the Foster thermal network model of the inverter power device;

[0043] Figure 5 This is a schematic diagram of the switching loss generated by the inverter power device IGBT during the turn-on and turn-off process;

[0044] Figure 6 This is the A-phase current flow diagram when ia>0;

[0045] Figure 7 This is the A-phase current flow diagram when ia < 0;

[0046] Figure 8 It is a control flow chart of the improved MPCC of the present invention;

[0047] Figure 9 This is a comparison chart of the simulation results of the MPCC method of the present invention and the traditional three-level NPC inverter;

[0048] Figure 10 This is a statistical result diagram of IA1 rain flow counting of the present invention (MPCC-II). DETAILED DESCRIPTION

[0049] For the convenience of understanding the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood by those skilled in the art that the embodiments are only provided to help understand the present invention and should not be regarded as specific limitations of the present invention.

[0050] like Figure 1 As shown in the topology of the three-level NPC inverter with resistive inductive load, each phase has 4 IGBTs (I X1 , I X2 , I X3 , I X4 ), 4 freewheeling diodes (D X1 , D X2 , D X3 , D X4 ) and two clamping diodes (D X5 , D X6 ), wherein X represents A, B and C phases.

[0051] In order to output the three levels of P, O, and N, two tubes in each phase must be turned on at the same time, so the switching function of each phase is defined as follows:

[0052]

[0053] The spatial voltage vector u of the three-phase voltage synthesis and the voltage u of each phase X They are expressed as formula (2) and formula (3) respectively:

[0054]

[0055]

[0056] Substituting equation (3) into equation (2), the calculation formula of the basic voltage vector is obtained as equation (4):

[0057]

[0058] Substituting the 27 switch combinations into the equation, we obtain 27 basic voltage vectors and their classifications as shown in Table 1.

[0059] Table 1 Basic voltage vectors of three-level NPC inverter

[0060]

[0061] like Figure 2 The figure shows the control flow of a conventional MPCC for resistive and inductive loads. The conventional MPCC requires 27 iterations of calculations when selecting the voltage vector.

[0062] according to Figure 1 The current dynamic equation of the three-level NPC inverter resistive-inductive load can be listed as:

[0063]

[0064] In formula (5), R represents the three-phase load resistance, L represents the three-phase load inductance, v represents the three-phase load voltage, and i represents the three-phase load current.

[0065] Using the forward Euler approximation, the derivative of the load current can be expressed as:

[0066]

[0067] In formula (6), T s Indicates the sampling time.

[0068] Substituting equation (6) into equation (5), we get the load current prediction value at time (k+1):

[0069]

[0070] In formula (7), i p (k+1) represents the current predicted value at the next moment, i(k) represents the current measured value at the current moment, and υ(k) represents the voltage vector.

[0071] Secondly, since the three-level NPC inverter needs to consider the midpoint potential balance control, the factor of controlling the midpoint potential is added to the cost function and a weight coefficient is assigned. The voltage difference between the two capacitors at the next moment is directly given as follows:

[0072]

[0073] In formula (8), C represents the capacitance value;

[0074] ΔU C (k) represents the voltage difference between the two capacitors at the current moment k:

[0075] ΔU c (k)=U c1(k)-U c2 (k)

[0076] i0(k+1) represents the predicted value of the midpoint current at the next moment k+1:

[0077] i0(k+1)=(1-|S A |)i a +(1-|S B |)i b +(1-|S C |)i c

[0078] Where i a , ib and i c Indicates the three-phase load current measurement value.

[0079] To sum up, the cost function at this time can be expressed as:

[0080]

[0081] In formula (9), represents the real part of the predicted current, represents the imaginary part of the predicted current, represents the real part of the reference current, represents the imaginary part of the reference current, Represents the weight coefficient of the midpoint potential.

[0082] In order to coordinately control the output current, midpoint potential balance and junction temperature of the power device of the resistive-inductive load of the three-level NPC inverter, the present invention is guided by reducing the junction temperature of the IGBT / Diode and aims to improve the life of the inverter. Based on the MPCC (Model Predictive Current Control) method, the system workload is reduced by dividing the sectors, and the loss factor is dynamically added to the cost function. Therefore, while ensuring the current control and midpoint potential balance effects, the junction temperature of the power device is minimized and balanced, its life is improved, and ultimately the life of the inverter is improved.

[0083] First, in order to reduce the 27 iterations of calculation required by traditional MPCC when screening voltage vectors, Figure 3 As shown, in the α-β coordinate system, the present invention divides the voltage vector sector into 12 sectors at intervals of 30° in the counterclockwise direction, starting from the α axis. The sectors are sequentially named sectors 1 to 12. A reduced set of candidate vectors is obtained based on the sector where the reference voltage vector is located.

[0084] In each sector, seven voltage vectors are involved in the iterative calculation: three zero vectors, two small vectors, one medium vector, and one large vector. Table 2 lists each sector and its corresponding candidate vector set. This reduces the number of cost function iterations from 27 to 7, significantly reducing the computational effort.

[0085] Table 2 Sectors and corresponding candidate vector sets

[0086] Sector division Alternative vector Sector 1 ONN POO PNN PON PPP OOO NNN Sector 2 OON PPO PPN PON PPP OOO NNN Sector 3 OON PPO PPN OPN PPP OOO NNN Sector 4 NON OPO NPN OPN PPP OOO NNN Sector 5 NON OPO NPN NPO PPP OOO NNN Sector 6 NOO OPP NPP NPO PPP OOO NNN Sector 7 NOO OPP NPP NOP PPP OOO NNN Sector 8 NNO OOP NNP NOP PPP OOO NNN Sector 9 NNO OOP NNP ONP PPP OOO NNN Sector 10 ONO POP PNP ONP PPP OOO NNN Sector 11 ONO POP PNP PNO PPP OOO NNN Sector 12 ONN POO PNN PNO PPP OOO NNN

[0087] In order to determine which sector the voltage vector is running to and thus update the candidate range of the voltage vector set, the current prediction formula under resistive and inductive loads is the same as formula (7). According to the idea of ​​current deadbeat, at the next moment, the current needs to follow the given value, so:

[0088] i p (k+1)=i*(k+1) (10)

[0089] Combining equations (7) and (10), the calculation formula for the reference voltage vector at the next moment is:

[0090]

[0091] In formula (11), υ*(k+1) represents the reference voltage value at the next moment.

[0092] υ*(k+1) represents the reference current value at the next moment k+1. The reference current is a set standard sinusoidal current, and the control goal is to make the load output current follow this reference current. After setting up the three-phase sinusoidal current module, the current value is updated at each sampling point. The updated reference current and system parameters are used to predict the reference voltage at the next moment.

[0093] The reference voltage vector is decomposed in the α-β coordinate system, and the angle between the reference voltage vector and the α axis is calculated. Then, the sector where the reference voltage vector is located is determined, and a reduced set of candidate vectors is obtained.

[0094] Secondly, in order to reduce the junction temperature of the power devices of the three-level NPC inverter and improve the life of the inverter, the junction temperature of the power devices T j The calculation formula (12) and Figure 4It can be seen from the Foster thermal network model of the power device shown that the junction temperature is related to power loss, thermal impedance and ambient temperature. Therefore, the present invention reduces the loss by adding a factor that considers the IGBT\Diode power loss into the cost function, thereby reducing the junction temperature. Among them, the power loss of the IGBT is mainly divided into two parts: switching loss and conduction loss, and its calculation is based on the table lookup method. During the turn-on and turn-off process of the power device IGBT, the voltage and current waveforms overlap in a very small period of time, so switching loss will occur, such as Figure 5 As shown in Figure 2, the loss of the anti-parallel diode during the turn-on process is very small and can generally be ignored. However, reverse recovery loss occurs during the turn-off process. Due to the initial saturation voltage drop and on-resistance, the power device will generate conduction loss during the conduction process.

[0095] Power device junction temperature T j The calculation formula is:

[0096] T j =P loss (Z j-c +Z c-h +Z h-a )+T a (12)

[0097] In formula (12), P loss Represents power loss, including switching loss P sw and conduction loss P con , Z j-c , Z c-h and Z h-a They represent the equivalent thermal impedance from junction to case, from case to heat sink, and from heat sink to environment, respectively. a Indicates the ambient temperature.

[0098] Switching loss P sw The calculation formula is:

[0099]

[0100] In formula (13), t on / off Indicates the switching time, υ ce Indicates the IGBT collector-emitter instantaneous voltage, i c Indicates the instantaneous current of the IGBT collector.

[0101] Conduction loss P con The calculation formula is:

[0102] P con =υ ce (t)(υ f (t))×i c (t)(i f(t)) (14)

[0103] In formula (14), υ f It represents the voltage drop across the diode when it is conducting, i f Represents the instantaneous current flowing through the diode.

[0104] Formula (13) and Formula (14) are simplified to two power loss factors shown in Formula (15):

[0105]

[0106] In formula (15), i a，b，c Represents the three-phase current, and R0 represents the on-resistance of the power device when it is turned on.

[0107] When these two power loss factors are added directly to the cost function, because the cost function includes control terms such as current control and midpoint potential, this will directly reduce the effectiveness of current and midpoint potential control, thus losing sight of the main goal. Therefore, the present invention dynamically determines the state of the power device at the next moment relative to the previous moment (switching state or conducting state) in the cost function, and then adds a factor to reduce the corresponding power loss in the corresponding state, thereby improving the system's ability to achieve the above-mentioned three-phase control.

[0108] The three-level NPC inverter has three phases. Phase A is used as an example for explanation below. Due to the same structure, the other phases are similar to phase A.

[0109] Assuming the current flowing out of the inverter and into the resistive-inductive load is in the positive direction, when the predicted current value at the next moment calculated by equation (1) is positive, the losses of each transistor in phase A under different switching states and the corresponding cost function values ​​are analyzed as follows. It should be noted that the diode, whether it is an anti-parallel freewheeling diode or a midpoint clamp diode, has no turn-on loss compared to the IGBT.

[0110] (1) When the current of phase A is positive

[0111] ①If Sa(k)=0 and Sa(k+1)=1, then Figure 6 As shown, Figure 6 (a) means Sa=0 (ia>0), Figure 6(b) indicates Sa = 1 (ia > 0), indicating that at the previous moment DA5 and IA2 were in the on state, while the other transistors were in the off state; at the next moment IA1 and IA2 were in the on state, while the other transistors were in the off state. It can be clearly seen that DA5 switches from the on state to the off state during this switching process, which will incur turn-off loss; IA2's on state remains unchanged before and after the switching, and still incurs conduction loss; IA1 switches from the off state to the on state, which will incur turn-on loss + conduction loss. In summary, three transistors are involved in this switching process, and the losses generated are as follows:

[0112] IA1: Turn-on loss + conduction loss

[0113] IA2: Conduction loss

[0114] DA5: Turn-off loss

[0115] Then, the cost function for loss of phase A is:

[0116] g A =λ sw P sw +λ con P con +λ con P con +λ sw P sw

[0117] =2λ sw P sw +2λ con P con (16)

[0118] In formula (16), λ sw Represents the weight coefficient of the switching loss factor, λ con Indicates the weight coefficient of the conduction loss factor. Two weight coefficients λ con and λ sw It is adjusted according to actual simulation results and experience.

[0119] The following analysis of power device losses caused by the change in the phase A switch state when the phase A current is greater than 0 and the corresponding cost function are consistent with the above, and only the conclusion is given.

[0120] ② If Sa(k) = 1 and Sa(k+1) = 0, the tubes involved and the losses generated are:

[0121] IA1: Turn-off loss

[0122] IA2: Conduction loss

[0123] DA5: Conduction loss

[0124] Then, the cost function for loss of phase A is:

[0125] g A =λ sw P sw +2λ con P con (17)

[0126] ③If Sa(k)=1 and Sa(k+1)=-1, the tubes involved and the losses generated are:

[0127] IA1: Turn-off loss

[0128] IA2: Turn-off loss

[0129] DA3: Conduction loss

[0130] DA4: Conduction loss

[0131] Then, the cost function for loss of phase A is:

[0132] g A =2λ sw P sw +2λ con P con (18)

[0133] ④ If Sa(k)=-1 and Sa(k+1)=1, the tubes involved and the losses generated are:

[0134] DA3: Turn-off loss

[0135] DA4: Turn-off loss

[0136] IA1: Turn-on loss + conduction loss

[0137] IA2: Turn-on loss + conduction loss

[0138] Then, the cost function for loss of phase A is:

[0139] g A =4λ s wP sw +2λ con P con (19)

[0140] ⑤ If Sa(k) = 0 and Sa(k+1) = -1, the tubes involved and the losses generated are:

[0141] IA2: Turn-off loss

[0142] DA5; Turn-off loss

[0143] DA3: Conduction loss

[0144] DA4: Conduction loss

[0145] Then, the cost function for loss of phase A is:

[0146] g A =2λ sw P sw +2λ con P con (20)

[0147] ⑥ If Sa(k) = -1 and Sa(k+1) = 0, the tubes involved and the losses generated are:

[0148] IA2: Turn-on loss + conduction loss

[0149] DA5: Conduction loss

[0150] DA3: Turn-off loss

[0151] DA4: Turn-off loss

[0152] Then, the cost function for loss of phase A is:

[0153] g A =3λ sw P sw +2λ con P con (twenty one)

[0154] ⑦ If Sa(k) = 1 and Sa(k+1) = 1, the tubes involved and the losses generated are:

[0155] IAl: conduction loss

[0156] IA2: Conduction loss

[0157] Then, the cost function for loss of phase A is:

[0158] g A =2λ con P con (twenty two)

[0159] ⑧If Sa(k)=0 and Sa(k+1)=0, the tubes involved and the losses generated are:

[0160] IA2: Conduction loss

[0161] DA5: Conduction loss

[0162] Then, the cost function for loss of phase A is:

[0163] gA =2λ con P con (twenty three)

[0164] ⑨ If Sa(k) = -1 and Sa(k+1) = -1, the pipes involved and the losses generated are:

[0165] DA3: Conduction loss

[0166] DA4: Conduction loss

[0167] Then, the cost function for loss of phase A is:

[0168] g A =2λ con P con (twenty four)

[0169] (2) When the current of phase A is negative:

[0170] ①If Sa(k)=0 and Sa(k+1)=1, then Figure 7 As shown, Figure 7 (a) means Sa=0 (ia<0), Figure 7 (b) indicates Sa = 1 (ia < 0), indicating that at the previous moment, IA3 and DA6 were on, while the other transistors were off. At the next moment, DA1 and DA2 were on, while the other transistors were off. It is clear that during this switching process, DA6 and IA3 switch from on to off, incurring turn-off losses; while DA1 and DA2 switch from off to on, incurring conduction losses.

[0171] In summary, four tubes are involved in this switching process, and the losses generated are as follows:

[0172] DA1: Conduction loss

[0173] DA2: Conduction loss

[0174] IA3: Turn-off loss

[0175] DA6: Turn-off loss

[0176] Then, the cost function for loss of phase A is:

[0177] g A =2λ sw P sw +2λ con P con (25)

[0178] The following analysis of power device losses caused by the change in the phase A switch state when the phase A current is less than 0 and the corresponding cost function are consistent with the above, so only the conclusion is given.

[0179] ② If Sa(k) = 1 and Sa(k+1) = 0, the tubes involved and the losses generated are:

[0180] DA1: Turn-off loss

[0181] DA2: Turn-off loss

[0182] IA3: Turn-on loss + conduction loss

[0183] DA6: Conduction loss

[0184] Then, the cost function for loss of phase A is:

[0185] g A =3λ sw P sw +2λ con P con (26)

[0186] ③If Sa(k)=1 and Sa(k+1)=-1, the tubes involved and the losses generated are:

[0187] DA1: Turn-off loss

[0188] DA2: Turn-off loss

[0189] IA3: Turn-on loss + conduction loss

[0190] IA4: Turn-on loss + conduction loss

[0191] Then, the cost function for loss of phase A is:

[0192] g A =4λ sw , P sw +2λ con P con (27)

[0193] ④ If Sa(k)=-1 and Sa(k+1)=1, the tubes involved and the losses generated are:

[0194] DA1: Conduction loss

[0195] DA2: Conduction loss

[0196] IA3: Turn-off loss

[0197] IA4: Turn-off loss

[0198] Then, the cost function for loss of phase A is:

[0199] g A =2λ sw P sw +2λ con P con (28)

[0200] ⑤ If Sa(k) = 0 and Sa(k+1) = -1, the tubes involved and the losses generated are:

[0201] IA3: Conduction loss

[0202] DA6; Turn-off loss

[0203] IA4: Turn-on loss + conduction loss

[0204] Then, the cost function for loss of phase A is:

[0205] g A =2λ sw P sw +2λ con P con (29)

[0206] ⑥ If Sa(k) = -1 and Sa(k+1) = 0, the tubes involved and the losses generated are:

[0207] IA3: Conduction loss

[0208] IA4: Turn-off loss

[0209] DA6: Conduction loss

[0210] Then, the cost function for loss of phase A is:

[0211] g A =λ sw P sw +2λ con P con (30)

[0212] ⑦ If Sa(k) = 1 and Sa(k+1) = 1, the tubes involved and the losses generated are:

[0213] DA1: Conduction loss

[0214] DA2: Conduction loss

[0215] Then, the cost function for loss of phase A is:

[0216] g A =2λ con P con(31)

[0217] ⑧If Sa(k)=0 and Sa(k+1)=0, the tubes involved and the losses generated are:

[0218] IA3: Conduction loss

[0219] DA6: Conduction loss

[0220] Then, the cost function for loss of phase A is:

[0221] g A =2λ con P con (32)

[0222] ⑨ If Sa(k) = -1 and Sa(k+1) = -1, the pipes involved and the losses generated are:

[0223] IA3: Conduction loss

[0224] IA4: Conduction loss

[0225] Then, the cost function for loss of phase A is:

[0226] g A =2λ con P con (33)

[0227] In summary, the loss cost function of phase A can be expressed as:

[0228] g A =2λ con P con +(i.>0)(4(S a (k)==-1)(S a (i)==1)+3(S 。 (k)==-1)(S a (i)==0)+(S a (k)==1)(S a (i)==0))λ sw P sw +2(i a >0)((S a (k)==0)(S a (i)==1)+(S a (k)==1)(S a (i)==-1)+(S a (k)==0)(S a (i)==-1))λ sw P sw +(i a<0)(4(Sa(k)==1)(S a (i)==-1)+3(S a (k)==1)(S a (i)==0)+(S a (k)==-1)(S a (i)==0))λ sw P sw +2(i a <0)((S a (k)==0)(S a (i)==1)+(S a (k)==-1)(S a (i)==1)+(S a (k)==0)(S a (i)==-1))λ sw P sw (34)

[0229] Since the analysis of phases B and C is the same as that of phase A, they will not be expanded.

[0230] Finally, the cost function obtained by integrating current control, midpoint potential balance control and reducing power device losses is:

[0231]

[0232] In formula (35), The weight coefficient of the control midpoint potential balance expressed by g needs to be adjusted according to the simulation results; B and g C The loss cost functions of phases B and C are the same as those of phase A.

[0233] For the final cost function calculated by formula (35), since the improved MPCC considers more control variables than the traditional MPCC, in order to avoid a significant impact on the current following effect, which in turn causes a large harmonic distortion of the load current, the current error of the cost function of the traditional MPCC is replaced by the square of the current error. Secondly, it should be noted that, firstly, the power losses of the three phases obtained through simulation are not exactly the same, so the power loss weight coefficients of the three phases need to be fine-tuned based on the set value according to the actual situation to achieve the best effect; secondly, the losses of IGBTs and diodes are also inconsistent, and their weight coefficients also need to be fine-tuned based on the actual situation to ultimately achieve a good effect.

[0234] like Figure 8 As shown, the present invention provides an improved model predictive control method for a three-level NPC inverter guided by junction temperature, comprising the following steps:

[0235] Step A1: Measure the load current measurement value i(k) and capacitor voltage value U at the current sampling time k c1 (k) and U c2 (k), measure the switching state S of each phase at the current sampling time k A 、S B and S C ;

[0236] Step A2: Use the load current measurement value i(k) at the current sampling time k and predict the reference voltage vector U*(k+1) at the next time k+1 according to formula (11), and determine the sector in which it is located based on the reference voltage vector U*(k+1):

[0237] Step A3: In the candidate vector set corresponding to the sector where the reference voltage vector U*(k+1) determined in step A2 is located, the following iterative calculation is performed on the seven voltage vectors in sequence to obtain the optimized vector at the current sampling time k:

[0238] A3-1: First, use the load current measurement value i(k) at the current time k and each voltage vector and predict the load current value i at the next time k+1 according to formula (7) p (k+1);

[0239] A3-2: Then use the capacitor voltage value U at the current moment k c1 (k), U c2 (k) and the switching state S of each phase A 、S B and S C And according to formula (8), the midpoint potential ΔU at the next moment k+1 is predicted c (k+1);

[0240] A3-3: In the cost function, dynamically determine whether the power device is in the switching state or the conducting state at the next moment k+1 relative to the previous moment k (comparison S A，B，C (k) and S A，B，C (k+1)), and adding the factor of reducing the corresponding power loss in the corresponding state, the final cost function that comprehensively considers current control, midpoint potential balance control and reduction of power device loss is obtained;

[0241] A3-4: The load current value i at the next moment k+1 predicted in step A3-1 is p (k+1) and the midpoint potential ΔU at the next moment k+1 predicted in step A-2 c Substitute (k+1) into the final cost function to calculate the g value corresponding to each voltage vector;

[0242] A3-5: Compare and determine the minimum value among all values, and use the voltage vector corresponding to the minimum g value as the optimized vector υ at the current sampling time k opt .

[0243] The present invention applies an improved model prediction current control strategy to the resistive-inductive load of a three-level NPC inverter, thereby ensuring the midpoint voltage balance effect and other control performances, while reducing and balancing the junction temperature of the inverter power devices and improving the inverter life.

[0244] Finally, the inverter lifespan under this method is calculated using the rain flow counting method and the life prediction model. The effectiveness of this method is verified by comparing it with the traditional three-level NPC inverter MPCC method in MATLAB / SIMULINK.

[0245] Figure 9 This is a comparison of the simulation results of the two strategies. Figure 9 (a) is the comparison of the junction temperature of the A-phase IGBT, Figure 9 (b) is the comparison of diode junction temperature of phase A. Figure 9 (c) is the control performance comparison.

[0246] Figure 9 In this paper, MPCC-I refers to the traditional MPCC strategy, and MPCC-II refers to the improved MPCC strategy proposed in this invention. Figure 9 As shown in (a) and (b), the junction temperatures of the IGBT and diode of MPCC-I are not balanced. For example, the maximum junction temperature of the IGBT is 39.1°C, while the minimum junction temperature is only 33°C, and the maximum junction temperature fluctuation is 4.9°C; the maximum junction temperature of the diode is 28.8°C, while the minimum junction temperature is only 24.2°C, and the maximum junction temperature fluctuation is 3.2°C. Such a junction temperature distribution will cause the power device with the highest junction temperature and the largest junction temperature fluctuation to fatigue and damage first, while other power devices can still work. That is, the life of the entire inverter will depend on the power device with the largest junction temperature and the largest junction temperature fluctuation, resulting in serious waste. Figure 9 From the comparison of (a) and (b), it can be seen that the MPCC-II strategy proposed in the present invention balances and reduces the junction temperature of the IGBT and the diode, and reduces their junction temperature fluctuation. At this time, the maximum junction temperature of the IGBT is 37.4°C, the minimum junction temperature is 32.8°C, and the maximum junction temperature fluctuation is 4.6°C; the maximum junction temperature of the diode is 27.5°C, the minimum junction temperature is 24.0°C, and the maximum junction temperature fluctuation is 2.3°C.

[0247] like Figure 9(c) Comparison of the control performance of the two strategies shown. In the MPCC-I strategy, the maximum fluctuation of the midpoint potential (and the voltage difference between the two capacitors) is 0.4V; the three-phase current has good sinusoidality, and the total harmonic distortion is 4.2%, which is lower than 5%; the phase voltage of phase A has a good waveform and is consistent with the theory. In the MPCC-II strategy, the maximum fluctuation of the midpoint potential is 0.4V; the three-phase current; the three-phase current has good sinusoidality, and the total harmonic distortion is 4.6%, which is lower than 5%; the phase voltage of phase A has a good waveform and is consistent with the theory. Therefore, except that the three-phase current harmonic distortion of the MPCC-II strategy is slightly higher than that of the MPCC-I strategy, the rest of the performance is the same.

[0248] according to Figure 9 As a result, since MPCC-II lowers the junction temperature compared to MPCC-I, it is preliminarily judged that MPCC-II optimizes the inverter life. In order to make the conclusion more convincing, this paper uses the rain flow counting method to count the number of junction temperature cycles of power devices. Due to space limitations, only the statistical results of MPCC-II IA1 rain flow counting are given. Figure 10 Finally, the Lesit life prediction model calculates the life of each phase power device and compares the converter life of the two methods. The comparison results are shown in Table 4.

[0249] Table 4 Lifespan comparison

[0250] Power device life Life of Phase A power device Life of Phase B power device Life of Phase C power devices Inverter life MPCC-I 3.1e53 3.0e53 3.2e53 3.0e53 MPCC-II 3.6e55 2.99e55 2.46e55 2.46e55

[0251] It can be seen from Table 4 that the total life of the three phases under the MPCC-II strategy is greater than the total life of the three phases under the MPCC-I strategy, that is, the life of the three-level NPC inverter based on the MPCC-II strategy is greater than the life of the three-level NPC inverter based on the MPCC-I strategy.

Claims

1. An improved model predictive control method for a junction temperature-guided three-level NPC inverter, characterized in that: The method comprises the following steps: Step A1: Measure the load current measurement value i(k) and capacitor voltage value U at the current sampling time k c1 (k) and U c2 (k), measure the switching state S of each phase at the current sampling time k A 、S B and S C ; Step A2: Use the load current measurement value i(k) at the current sampling time k to predict the reference voltage vector v at the next time k+1 * (k+1), and according to the reference voltage vector v * (k+1) determines the sector it is in: The sectors are divided into 12 sectors at intervals of 30° in the counterclockwise direction, starting from the α-axis in the α-β coordinate system. The sectors are sequentially named sectors 1 to 12. The candidate vector set corresponding to each sector has seven voltage vectors. Then, the reference voltage vector v * (k+1) is decomposed in the α-β coordinate system by calculating the reference voltage vector v * The angle between (k+1) and the α axis is used to determine the reference voltage vector v * The sector where (k+1) is located; Step A3: The reference voltage vector v determined in step A2 is * In the candidate vector set corresponding to the sector where (k+1) is located, the following iterative calculation is performed on the seven voltage vectors in sequence to obtain the optimized vector at the current sampling time k: A3-1: First, use the load current measurement value i(k) at the current time k and each voltage vector to predict the load current value i at the next time k+1 p (k+1); A3-2: Then use the capacitor voltage value U at the current moment k c1 (k), U c2 (k) and the switching state S of each phase A 、S B and S C Predict the midpoint potential ΔU at the next moment k+1 c (k+1); A3-3: In the cost function, dynamically determine whether the power device will be in the switching state or the conducting state at the next moment relative to the current moment, and add a factor to reduce the corresponding power loss in the corresponding state. This results in a final cost function that comprehensively considers current control, midpoint potential balance control, and power device loss reduction. A3-4: The load current value i at the next moment k+1 predicted in step A3-1 is p (k+1) and the midpoint potential ΔU at the next moment k+1 predicted in step A-2 c Substitute (k+1) into the final cost function to calculate the g value corresponding to each voltage vector; A3-5: Compare and determine the minimum value among all g values, and use the voltage vector corresponding to the minimum g value as the optimized vector v at the current sampling time k opt ; In step A3-3, the final cost function is: Where, represents the real part of the reference current, represents the imaginary part of the reference current, represents the real part of the predicted current, represents the imaginary part of the predicted current, Indicates the weight coefficient for controlling the midpoint potential balance, ΔU c Represents the voltage difference between the two capacitors, g A 、g B and g C They represent the loss cost functions of phase A, phase B, and phase C respectively, and the three loss cost functions have the same form; The loss cost function of phase A is: g A =2λ con P con +(i a >0)(4(S a (k)==-1)(S a (i)==1)+3(S a (k)==-1)(S a (i)==0)+(S a (k)==1)(S a (i)==0))λ sw P sw +2(i a >0)((S a (k)==0)(S a (i)==1)+(S a (k)==1)(S a (i)==-1)+(S a (k)==0)(S a (i)==-1))λ sw P sw +(i a <0)(4(S a (k)==1)(S a (i)==-1)+3(S a (k)==1)(S a (i)==0)+(S a (k)==-1)(S a (i)==0))λ sw P sw +2(i a <0)((S a (k)==0)(S a (i)==1)+(S a (k)==-1)(S a (i)==1)+(S a (k)==0)(S a (i)==—1))λ sw P sw Where, P con Represents the conduction loss of the power device, λ con Represents the weight coefficient of the conduction loss factor, P sw Represents the switching loss of the power device, λ sw Represents the weight coefficient of the switching loss factor; S a (k) represents the switching state of phase A at time k; S a (i) represents the switching state of phase A corresponding to the i-th voltage vector.

2. The improved model predictive control method for a junction temperature-guided three-level NPC inverter according to claim 1, characterized in that: In step A2, the reference voltage vector v at the next moment k+1 is predicted according to the following formula: * (k+1): In the formula, R represents the three-phase load resistance, L represents the three-phase load inductance, T s represents the sampling time, i * (k+1) represents the reference current value at the next moment k+1.

3. The improved model predictive control method for a junction temperature-guided three-level NPC inverter according to claim 1, characterized in that: In step A3-1, the load current value i at the next moment k+1 is calculated using the following formula: p (k+1) makes predictions: Where v(k) represents the voltage vector.

4. The improved model predictive control method for a junction temperature-guided three-level NPC inverter according to claim 1, characterized in that: In step A3-2, the midpoint potential ΔU at the next moment k+1 is calculated using the following formula: c (k+1) makes predictions: Where, T s represents the sampling time, C represents the capacitance value; ΔU c (k) represents the voltage difference between the two capacitors at the current moment k: ΔU c (k)=U c1 (to c2 (k) i0(k+1) represents the predicted value of the midpoint current at the next moment k+1: i0(k+1)=(1-|S A |)i a +(1-|S B |)i b +(1-|S C |)i c Where i a 、i b and i c Indicates the three-phase load current measurement value.

Images