Method for suppressing neutral point potential fluctuation of three-level inverter in rural micro-grid

Abstract

The present application relates to a kind of rural micro-grid three-level inverter midpoint potential fluctuation suppression method. Including: S1, the upper capacitor and the lower capacitor of the series connection of the DC side of inverter are respectively collected voltage, and the correction amount is calculated according to the voltage difference of the upper capacitor and the lower capacitor voltage collected;S2, three-phase phase current is collected in inverter AC side, and current polarity is judged;S3, the correction direction of S1 correction amount is determined after current polarity judged by S2;S4, calculate small vector time allocation factor, and get positive small vector action time and negative small vector action time by small vector time allocation factor respectively;S5, generate SVPWM switching sequence according to the positive small vector action time and negative small vector action time obtained by S4, and export PWM pulse;S6, adjust midpoint current, so that the voltage difference of upper capacitor and lower capacitor tends to balance.The present application can realize DC midpoint potential balance by changing the direction and size of average midpoint current in a switching cycle.

Description

Technical Field

[0001] This invention relates to a method for suppressing potential fluctuations, and more particularly to a method for suppressing potential fluctuations at the midpoint of a three-level inverter in a rural microgrid. Background Technology

[0002] In recent years, rural residential photovoltaic (PV) systems, distributed PV systems, and agricultural electricity consumption have developed rapidly, and rural power distribution networks are gradually transforming from traditional unidirectional power supply networks to active distribution networks with distributed power source access. However, rural power grids still suffer from shortcomings such as weak grid connections in some areas, low voltage, and frequent power outages, necessitating further improvements in the carrying capacity of distributed renewable energy. In rural low-voltage distribution areas, the increase in distributed PV access capacity may cause problems such as reverse power flow, reverse overload of distribution transformers, high voltage at the grid connection point, and voltage fluctuations. With large-scale integration of distributed PV in rural areas, reverse overload of distribution transformers and excessively high voltage at the distributed PV access point are particularly prominent safety issues; the randomness, volatility, and intermittency of PV output also affect voltage fluctuations in the distribution network.

[0003] In rural microgrids and distributed photovoltaic (PV) systems, interface inverters not only convert DC to AC power but also directly impact the microgrid's output power quality, operating efficiency, and control stability. Traditional two-level inverters are simple in structure, but in medium- and high-voltage applications, they are prone to problems such as high voltage stress on components, excessive output harmonics, and large filter size. Three-level inverters, by increasing the number of output levels, can reduce the amplitude of single voltage transitions, lower the voltage stress on switching devices, and improve output waveform quality, thus possessing high value in engineering applications.

[0004] For distributed photovoltaic (PV) interface converters in rural microgrids, inverters often operate under conditions of load fluctuations, changes in PV output, and variations in grid connection voltage. To ensure stable system operation, inverters need to possess good output performance. Lower harmonic content, less device voltage stress, and strong dynamic regulation capability are crucial for improving the operational stability of rural microgrids. Among various three-level topologies, the diode-neutral-clamped inverter, also known as the NPC (non-polarizing inverter), is widely studied due to its mature topology, abundant control methods, and broad engineering applications. Compared to two-level inverters, each arm of the NPC three-level inverter can output positive, zero, and negative levels. The combination of the three phases forms a richer space voltage vector, providing conditions for improving modulation performance.

[0005] Space Vector Pulse Width Modulation (SVPWM) is a commonly used inverter modulation method. This method equates the three-phase reference voltage to a rotating reference vector in the α-β plane, and then synthesizes the target output voltage by time-weighting adjacent basic vectors within one switching cycle. SVPWM features high DC bus voltage utilization, good output waveform quality, and ease of digital implementation, making it suitable for three-level inverter control.

[0006] The NPC three-level inverter also suffers from DC-side midpoint potential deviation during operation. If the voltages of the upper and lower capacitors are unbalanced for a long period, it may lead to output voltage distortion, uneven voltage stress on components, and decreased system stability.

[0007] In rural microgrids, distributed photovoltaic (PV) output is significantly affected by factors such as sunlight intensity, cloud cover, and temperature, exhibiting a degree of randomness and volatility. Large-scale PV integration can easily cause voltage fluctuations in the distribution network and may also lead to power quality issues such as flicker. Rural areas have diverse load types, with significant variations in residential electricity consumption, agricultural irrigation, and livestock processing loads, resulting in frequent load fluctuations. When both PV output and load change simultaneously, the AC current of the inverter changes accordingly, and the average value of the midpoint current in the NPC three-level inverter also changes. This means that the NPC three-level inverter also experiences DC-side midpoint potential offset during operation. This, in turn, causes uneven charging and discharging of the upper and lower capacitors on the DC side, leading to midpoint potential fluctuations. If the voltage imbalance between the upper and lower capacitors persists for a long period, it may result in output voltage distortion, uneven voltage stress on components, and decreased system stability.

[0008] Therefore, how to utilize the redundant small vectors in a three-level inverter to adjust the midpoint current without changing the target output voltage, in order to achieve the balance of the midpoint potential, is a technical problem that urgently needs to be solved. Summary of the Invention

[0009] To address the aforementioned technical problems, this invention provides a method for suppressing midpoint potential fluctuations in a three-level inverter for rural microgrids. The purpose is to adjust the direction and magnitude of the average midpoint current through SVPWM. The average midpoint current further controls the charging and discharging process of the upper and lower capacitors, causing the voltage on the side with higher voltage to gradually decrease and the voltage on the side with lower voltage to gradually increase, ultimately driving the voltage of the upper capacitor and the voltage of the lower capacitor to become consistent, thereby achieving DC midpoint potential balance.

[0010] To achieve the above objectives, the present invention provides a method for suppressing midpoint potential fluctuations in a three-level inverter for rural microgrids, comprising: S1. Collect the voltage of the upper and lower capacitors connected in series on the DC side of the inverter, and calculate the correction amount based on the voltage difference between the upper and lower capacitors. S2. Collect the three-phase current on the AC side of the inverter and determine the current polarity; S3. Determine the correction direction of the correction amount in S1 based on the current polarity determined by S2. S4. Calculate the small vector time allocation factor, and obtain the action time of the positive small vector and the action time of the negative small vector through the small vector time allocation factor; S5. Based on the positive small vector action time and negative small vector action time obtained in S4, generate the SVPWM switching sequence and output PWM pulses; S6. Adjust the midpoint current to balance the voltage difference between the upper and lower capacitors.

[0011] S1 includes the following steps: The collected upper capacitor voltage and lower capacitor voltage are subtracted after a unit delay to keep the voltage feedback quantity synchronized with the SVPWM calculation cycle. The difference between the voltage of the upper capacitor and the voltage of the lower capacitor is the midpoint voltage balance error, which is used to obtain the PI correction value through a parallel PI controller. The PI correction limit is used to ensure that the correction amount sent to the timing calculation module is within the allowable range, preventing the midpoint balance correction from being too strong and causing the action time of a certain vector to be negative, exceeding the switching cycle, or destroying the PWM sequence.

[0012] The input error formula for the parallel PI controller of S12 is: e np [k]=U c2 [k-1]-U c1 [k-1]=ΔU[k-1]; In the formula e np [k] represents the midpoint voltage balance error in the k-th control cycle; ΔU[k-1] represents the midpoint voltage deviation in the previous sampling cycle; therefore, when e np When e > 0, the voltage of the lower capacitor is higher than that of the upper capacitor; when e np When e < 0, the voltage of the upper capacitor is higher than that of the lower capacitor; when e np When =0, the upper and lower capacitors reach an ideal balance; A parallel PI controller includes a proportional element and an integral element. The proportional element directly adjusts rapidly based on the current error, and its output formula is: u p (t)=K p e np (t); In the formula, K p U is the proportionality coefficient. p The proportional control is a variable that can immediately change the controller output according to the instantaneous voltage deviation, which is used to improve the recovery speed of the midpoint voltage. The larger the deviation, the larger the proportional control and the stronger the correction of the action time of the redundant small vector. The integration step first multiplies the error signal by the integration coefficient K. iThe error is then accumulated using a continuous integrator of 1 / s, and the original integral output formula is as follows: x i (t)= i e np (τ)dτ; x i (t) represents the original output of the integral stage at time t; K i e is the integral coefficient, used to determine the rate of error accumulation and the strength of integral adjustment; np (τ) represents the midpoint voltage balance error at time τ during the integration process; τ is the integral variable; t is the current control time. The integral element is used to accumulate the persistent midpoint voltage error and eliminate the steady-state deviation that may remain when only proportional control is used. Therefore, the formula for the integral part is: u i (t)=sat[x i (t),u i,min ,u i,max ]; Among them, u i,min and u i,max These are the lower and upper limits of the integral output, set to -0.5 and 0.5 respectively, to limit the continuous accumulation of the integral caused by long-term deviations. ; The formula for the final PI correction is to add the proportional adjustment to the integral adjustment after the amplitude limit is: u PI (t)=sat[u p (t)+u i (t),u min ,u max ].

[0013] The S2 signal is acquired by the three-phase voltage and current measurement module in the main circuit and output in the form of Iabc. This signal represents the instantaneous phase current formula of phases A, B, and C: Iabc=[ia ib ic]ᵀ; In the formula, ia, ib, and ic are the instantaneous currents of phases A, B, and C, respectively. After the Iabc signal enters the three-level space vector pulse width modulation module, it first goes through a unit delay stage to keep synchronized with the current SVPWM calculation cycle, and then is split into ia, ib, and ic, and input into the vector action time calculation module. The three-phase current is not directly used to calculate the midpoint voltage error, but is used to determine whether the midpoint current flows into or out of the DC midpoint when the current redundant switch is in operation, thereby determining the correction direction of the redundant small vector action time.

[0014] After specifying the positive direction of the midpoint current, the instantaneous midpoint current can be expressed as: inp = δa ia + δb ib + δc ic; In the formula, δa, δb and δc are the connection state coefficients of the bridge arm midpoint. For any phase x∈{a,b,c}, the values ​​can be expressed as follows: δx=1, phase x is in state O; δx=0, phase x is in state P or N.

[0015] S3 is the correction amount after determining the current direction based on the current polarity: dcorr=u PI (t)·dir; In the formula, dir is the direction factor, and its sign is determined by ia, ib, ic, and the current large sector, small sector, and redundant switch status; the sign of dcorr is used to determine which type of redundant small vector should have its duration increased, and its absolute value reflects the degree of time allocation correction.

[0016] The correction amount, after considering the current direction, is processed in the time calculation module and converted into a time allocation factor f ranging from 0 to 1; If the total duration of a pair of P-type and N-type redundant small vectors is T0, then the durations of the two redundant states are as follows:

[0017] T P =f·T 0， T N =(1-f)·T 0； In the formula, T P and T N These represent the durations of the P-type and N-type redundant small vectors, respectively. Since the two redundant states correspond to the same space voltage vector position, T is maintained. P With T N The sum is T0, which allows us to change only the duration of the two midpoint current paths without significantly altering the vector's contribution to the average output voltage on the AC side; f is obtained by shifting the output of the midpoint balancing PI controller, i.e., f = u. PI (t)+0.5, u PI (t) represents the time allocation offset; simultaneously, based on the large sector, small region, and phase current polarity, it is determined which value should be allocated to T. P Or T N Therefore, f=0.5 indicates that the two redundant states are allocated time equally; f>0.5 indicates that the P-type redundant state occupies more time; f<0.5 indicates that the N-type redundant state occupies more time.

[0018] The S5 module includes a time calculation module that receives the local reference vector angle, large sector number, small sector auxiliary quantity, switching cycle, DC bus voltage, and three-phase current to perform two types of calculations: first, it calculates the total basic vector time required for the reference voltage based on the volt-second balance; second, it writes the midpoint balance correction into the time allocation between redundant small vectors. Therefore, the volt-second balance formula in SVPWM is: V ref T s =ΣV j T j , ΣT j =Ts,T j ≥0; In the formula V ref As the reference voltage vector, T s For the switching period, V j For the j-th fundamental voltage vector participating in the reference vector synthesis, T j The first equation states that the sum of the volt-second effects generated by each basic voltage vector within one switching cycle should be equal to the volt-second value of the reference voltage vector, thus ensuring that the average output voltage of the inverter is consistent with the reference voltage; the second equation states that the sum of the effects of each vector should be equal to one complete switching cycle; T j ≥0 indicates that the duration of action of each vector cannot be negative; The time calculation module outputs T P T N T2 and T3 represent the duration of action of different basic vectors or redundant switching states within a switching cycle; where T2 and T3 are the durations of action of two effective vectors, T... P and T N Allocate two redundancy distributions for the remaining time, satisfying the following: T P +T N +T2+T3=T s ; Midpoint balance control is achieved by adjusting T P With T N The proportion changes the duration of the redundant state. The above time amount needs to be further converted into time boundary and P / O / N state to finally generate IGBT gate pulse; SVPWM employs a symmetrical sequence of switching states, which is represented as follows: S1→S2→S3→S4→S3→S2→S1; Midpoint balance control is achieved by changing the time boundary corresponding to redundant states, making the pulse of one state wider and the pulse of another state narrower. When the local time crosses a certain cumulative time boundary, the system immediately switches to the next P / O / N combination. Since adjacent states are usually arranged according to the principle of changing only one bridge arm state, the number of switching devices involved in a single switch can be reduced. This section outputs the state of the three-phase bridge arm: P, O, or N. Each phase arm of the NPC three-level inverter contains four power switching devices. The pulse generator module converts the state of each phase into four 0 / 1 gate signals according to a fixed mapping relationship:

[0019] P→[1, 1, 0, 0]

[0020] O→[0, 1, 1, 0]

[0021] N→[0, 0, 1, 1] When the three-phase state is POO, phase A outputs [1,1,0,0], and phases B and C output [0,1,1,0]. When the time crosses a certain boundary within the cycle, it switches to the next P / O / N combination, and the gate pulse flips accordingly. Thus, the corrected T... P T N Ultimately, this manifests as a change in the gate pulse width of the corresponding redundant state; T P Increasing the holding time of the P-type redundant state corresponds to a wider gate pulse T. N Increasing the holding time of the N-type redundant state increases the corresponding gate pulse width. Thus, the PI controller first changes the duration of the redundant state, and then indirectly changes the gate pulse width through the time boundary and the switching state table.

[0022] The midpoint current changes with the switching state during one switching cycle, and the average midpoint current during the switching cycle is: i np,avg =(1 / T s ) dτ; In the formula, i np (τ) is the instantaneous midpoint current, i np,avg The average midpoint current over one switching cycle; the integral term represents the net charge flowing into or out of the DC midpoint during that cycle, divided by T. s The corresponding average current is then obtained, which represents the net charge flowing into or out of the midpoint during one switching cycle. If the influence of a pair of P-type and N-type redundant small vectors on the average midpoint current is i, then the midpoint currents during the period of their action are respectively i np , P and i np , N Then we have: i np,avg PN =[i np,P TP+inp , N TN] / T s； For a pair of typical redundant small vectors, under the same phase current conditions, the midpoint currents they generate are usually similar in magnitude and opposite in direction; if i approximately satisfies np,N =-i np,P ,for: i np,avg PN =(T0 / T s (2f-1)i np , P； T0 / T s This represents the proportion of time occupied by this pair of redundant small vectors in the entire switching cycle, 2f-1 indicates whether the time allocation of the two redundant states is balanced, and to which side it is biased, i np,P This represents the midpoint current generated during the operation of the P-type redundancy state; As can be seen from the above formula, when f=0.5, the duration of the P-type and N-type redundant states is equal, and their effects on the average midpoint current cancel each other out. When f>0.5, the average midpoint current is closer to the direction generated by the P-type redundant state. When f<0.5, the average midpoint current is closer to the direction generated by the N-type redundant state. Therefore, changing the time allocation factor f is to adjust the direction and amplitude of the average midpoint current within one switching cycle.

[0023] The change in voltage deviation within one switching cycle can be approximately expressed as: ΔU[k+1]=ΔU[k]+(T s / C)i np,avg [k]; This formula shows that the direction of the average midpoint current determines the direction of the voltage difference between the upper and lower capacitors, while the magnitude of the average midpoint current determines the rate of change of the voltage difference; under a specified positive direction, when i np,avg When i > 0, the net charge flowing into the DC midpoint increases, causing the voltage difference between the upper and lower capacitors to increase; when i np , avg When <0, the net charge at the DC midpoint decreases, which reduces the voltage difference between the upper and lower capacitors; Midpoint potential balance control selects a suitable redundant small vector based on the current voltage deviation and adjusts its duration, thereby changing the direction and magnitude of the average midpoint current within a switching cycle. The average midpoint current further controls the charging and discharging process of the upper and lower capacitors, causing the voltage on the side with higher voltage to gradually decrease and the voltage on the side with lower voltage to gradually increase, ultimately driving the voltage of the upper capacitor and the voltage of the lower capacitor to become consistent, thus achieving DC midpoint potential balance.

[0024] Advantages and effects of this invention: This invention can change the direction and magnitude of the average midpoint current within a switching cycle by selecting a suitable redundant small vector based on the voltage deviation between the upper and lower capacitor voltages and adjusting its duration. The average midpoint current further controls the charging and discharging process of the upper and lower capacitors, causing the voltage on the higher side to gradually decrease and the voltage on the lower side to gradually increase, ultimately pushing the upper and lower capacitor voltages to converge and achieving DC midpoint potential balance. This method, while maintaining a relatively constant total duration of the redundant small vector, adjusts the direction and magnitude of the average midpoint current within a switching cycle by changing the corresponding SVPWM pulse width, thereby controlling the charging and discharging process of the upper and lower capacitors. Compared with fixed-time allocation methods, this invention avoids erroneous adjustments caused by changes in current direction, improving the accuracy and dynamic response capability of midpoint potential balance control; simultaneously, it eliminates the need for additional power devices and voltage balancing circuits, minimizing the impact on the original SVPWM basic structure and AC side reference voltage synthesis, and offers advantages such as clear calculation process, simple implementation, and ease of digital controller deployment. Attached Figure Description

[0025] Figure 1 This is the main circuit topology diagram of an NPC-type three-level inverter.

[0026] Figure 2 This is a flowchart of the present invention.

[0027] Figure 3 It is a simulation model of a three-level inverter.

[0028] Figure 4 It is the internal structure of a three-level SVPWM modulation module.

[0029] Figure 5 This is the output voltage waveform of phase A without LC filtering.

[0030] Figure 6 This is a steady-state waveform diagram of the three-phase output phase voltage under SVPWM modulation.

[0031] Figure 7 This is a waveform diagram of the three-phase output line voltage without LC filtering.

[0032] Figure 8 This is a steady-state waveform diagram of the three-phase output line voltage under SVPWM modulation.

[0033] Figure 9 This is a waveform diagram of the three-phase output current.

[0034] Figure 10 This is the waveform of the capacitor voltage difference without the addition of midpoint balance control.

[0035] Figure 11 This is a waveform diagram of the capacitor voltage difference when midpoint balance control is added. Detailed Implementation

[0036] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0037] like Figure 1 As shown, this invention is a method for suppressing midpoint potential fluctuations in NPC-type three-level inverters. In an inverter topology such as... Figure 1 As shown: The DC side consists of two capacitors connected in series, with the midpoint denoted as point O; each phase arm consists of four power switching devices and two midpoint clamping diodes. The function of the clamping diodes is to limit the voltage across the switching devices in the off state to near the voltage of a single DC-side capacitor. If the total DC bus voltage is... And the voltages of the upper and lower capacitors are respectively and Then, under ideal equilibrium conditions, we have At this point, the output of a single-phase bridge arm can be selected. 0 and Three levels.

[0038] If the four switching devices of a certain phase bridge arm are labeled S1, S2, S3, and S4 from top to bottom, the typical conduction combinations corresponding to the positive, zero, and negative levels of the bridge arm are shown in Table 1. To avoid shoot-through, S1 and S3, and S2 and S4 are usually complementary. The zero-level state of a three-level bridge arm will cause the phase current to flow through the DC side midpoint, thus directly affecting the midpoint potential.

[0039] Table 1 shows the typical output states of the NPC three-level bridge arm.

[0040] .

[0041] Each phase of a three-level inverter has three output states: P, O, and N. The three phases combined form a total of 27 switching states. Different space voltage vectors are divided into four types according to their amplitude and influence on the neutral point potential: large vector, medium vector, small vector, and zero vector.

[0042] The large vector has 6 switching states, corresponding to 6 vector positions. When the large vector is in operation, the three-phase output terminals are only connected to the positive or negative terminal of the DC bus, and do not pass through the midpoint. Therefore, it will not cause changes in the midpoint current and has virtually no effect on the midpoint potential.

[0043] The medium vector has six switching states, corresponding to six vector positions. When the medium vector is in operation, one of the three phases is connected to the DC side midpoint, thus generating a midpoint current and affecting the midpoint potential. However, compared to the small vector, the medium vector has a weaker effect on regulating the midpoint potential and is generally not used as the primary means of midpoint balance control.

[0044] The small vector has 12 switching states, corresponding to 6 vector positions. Small vectors are the most important type of vector in three-level SVPWM because each small vector position typically corresponds to two redundant switching states. For example, (POO) and (ONN) can produce the same output space vector. Based on the composition of the switching states, small vectors can be further divided into P-type small vectors and N-type small vectors. The former consists of P and O states, while the latter consists of O and N states. Since these two types of redundant small vectors have opposite effects on the midpoint current, this invention utilizes them to achieve midpoint potential balance control.

[0045] The zero vector has three switching states: (PPP), (OOO), and (NNN). The spatial vector positions corresponding to these three states are all located at the origin, therefore their magnitude is 0, and they are usually denoted as . .

[0046] The following details how to use small vectors to suppress midpoint potential fluctuations in three-level inverters for rural microgrids, including the following steps:

[0047] S1. Collect the voltage of the upper and lower capacitors connected in series on the DC side of the inverter, and calculate the correction amount based on the voltage difference between the collected upper and lower capacitors; the details are as follows:

[0048] The collected upper capacitor voltage and lower capacitor voltage are subtracted after a unit delay to keep the voltage feedback quantity synchronized with the SVPWM calculation cycle.

[0049] The difference between the voltage of the upper capacitor and the voltage of the lower capacitor is the midpoint voltage balance error, which is used to obtain the PI correction value through a parallel PI controller; the input error formula of the parallel PI controller is:

[0050] e np [k]=U c2 [k-1]-U c1 [k-1]=ΔU[k-1];

[0051] In the formula e np [k] represents the midpoint voltage balance error in the k-th control cycle; ΔU[k-1] represents the midpoint voltage deviation in the previous sampling cycle; therefore, when e np When e > 0, the voltage of the lower capacitor is higher than that of the upper capacitor; when e np When e < 0, the voltage of the upper capacitor is higher than that of the lower capacitor; when e npWhen =0, the upper and lower capacitors reach an ideal balance;

[0052] To reduce deviation, a parallel PI controller is used, comprising a proportional element and an integral element. The proportional element directly adjusts rapidly based on the current error, and its output formula is:

[0053] u p (t)=K p e np (t);

[0054] In the formula, K p U is the proportionality coefficient. p The proportional control is a variable that can immediately change the controller output according to the instantaneous voltage deviation, which is used to improve the recovery speed of the midpoint voltage. The larger the deviation, the larger the proportional control and the stronger the correction of the action time of the redundant small vector.

[0055] The integration step first multiplies the error signal by the integration coefficient K. i The error is then accumulated using a continuous integrator of 1 / s, and the original integral output formula is as follows:

[0056] x i (t)= i e np (τ)dτ;

[0057] x i (t) represents the original output of the integral stage at time t; K i e is the integral coefficient, used to determine the rate of error accumulation and the strength of integral adjustment; np (τ) represents the midpoint voltage balance error at time τ during the integration process; τ is the integral variable; t is the current control time. The integral element is used to accumulate the persistent midpoint voltage error and eliminate the steady-state deviation that may remain when only proportional control is used.

[0058] Therefore, the formula for the integral part is:

[0059] u i (t)=sat[x i (t),u i,min ,u i,max ];

[0060] Among them, u i,min and u i,max These are the lower and upper limits of the integral output, set to -0.5 and 0.5 respectively, to limit the continuous accumulation of the integral caused by long-term deviations.

[0061] ;

[0062] The formula for the final PI correction is to add the proportional adjustment to the integral adjustment after the amplitude limit is:

[0063] u PI (t)=sat[u p (t)+u i (t),u min ,u max ];

[0064] The PI correction limit is used to ensure that the correction amount sent to the timing calculation module is within the allowable range, preventing the midpoint balance correction from being too strong and causing the action time of a certain vector to be negative, exceeding the switching cycle, or destroying the PWM sequence.

[0065] S2. Collect the three-phase current on the AC side of the inverter and determine the current polarity;

[0066] The voltage and current are acquired by the three-phase voltage and current measurement module in the main circuit and output in the form of Iabc. This signal represents the instantaneous phase current formula of phases A, B, and C:

[0067] Iabc=[ia ib ic]ᵀ;

[0068] In the formula, ia, ib, and ic are the instantaneous currents of phases A, B, and C, respectively. After the Iabc signal enters the three-level space vector pulse width modulation module, it first goes through a unit delay stage to keep synchronized with the current SVPWM calculation cycle, and then is split into ia, ib, and ic, and input into the vector action time calculation module. The three-phase current is not directly used to calculate the midpoint voltage error, but is used to determine whether the midpoint current flows into or out of the DC midpoint when the current redundant switch is in operation, thereby determining the correction direction of the redundant small vector action time.

[0069] After defining the positive direction of the midpoint current, the instantaneous midpoint current can be expressed as:

[0070] inp = δa ia + δb ib + δc ic;

[0071] In the formula, δa, δb and δc are the connection state coefficients of the bridge arm midpoint. For any phase x∈{a,b,c}, the values ​​can be expressed as follows: δx=1, phase x is in state O; δx=0, phase x is in state P or N.

[0072] S3. Determine the correction amount u for S1 based on the current polarity determined in S2. PI (t); u PI (t) represents the correction direction of the correction amount output by the midpoint voltage balance PI controller at time t. In the simulation, it corresponds to BusBalancePIOut.

[0073] The correction amount only reflects the correction strength required for the midpoint voltage deviation. Whether the actual duration of the P-type or N-type redundant small vector should be increased needs to be determined in conjunction with the current three-phase current direction, sector position, and redundant vector status.

[0074] Therefore, the time calculation module constructs a direction factor dir based on the current polarity and obtains a correction amount that takes the current direction into account.

[0075] dcorr=u PI (t)·dir;

[0076] In the formula, dir is the direction factor, whose sign is determined by ia, ib, ic, and the current large sector, small sector, and redundant switch status; the sign of dcorr is used to determine which type of redundant small vector should have its duration increased, and its absolute value reflects the degree of correction in time allocation. dcorr is processed in the time calculation module and converted into a time allocation factor f ranging from 0 to 1. The time calculation module first calculates T2 and T3 based on the reference vector position, where T2 and T3 are the durations of the two effective vectors. P and T N This is achieved by assigning two types of redundant switches.

[0077] S4. Calculate the small vector time allocation factor, and obtain the action time of the positive small vector and the action time of the negative small vector through the small vector time allocation factor;

[0078] The correction amount, after considering the current direction, is processed in the time calculation module and converted into a time allocation factor f in the range of 0 to 1.

[0079] If the total duration of a pair of P-type and N-type redundant small vectors is T0, then the durations of the two redundant states are as follows:

[0080] T P =f·T 0，

[0081] T N =(1-f)·T 0；

[0082] In the formula, T P and T N These represent the durations of the P-type and N-type redundant small vectors, respectively. Since the two redundant states correspond to the same space voltage vector position, T is maintained. P With T N The sum is T0, which allows us to change only the duration of the two midpoint current paths without significantly altering the vector's contribution to the average output voltage on the AC side; f is obtained by shifting the output of the midpoint balancing PI controller, i.e., f = u. PI (t)+0.5, u PI(t) represents the time allocation offset; simultaneously, based on the large sector, small region, and phase current polarity, it is determined which value should be allocated to T. P Or T N Therefore, f=0.5 indicates that the two redundant states are allocated time equally; f>0.5 indicates that the P-type redundant state occupies more time; f<0.5 indicates that the N-type redundant state occupies more time.

[0083] S5. Based on the positive small vector action time and negative small vector action time obtained in S4, generate the SVPWM switching sequence and output PWM pulses;

[0084] The time calculation module receives the local reference vector angle, large sector number, small sector auxiliary quantity, switching cycle, DC bus voltage, and three-phase current to perform two types of calculations: first, it calculates the total basic vector time required for the reference voltage based on the volt-second balance; second, it writes the midpoint balance correction into the time allocation between redundant small vectors. The reference vector angle and large sector number are determined by the reference voltage u. α u β After angle calculation and large sector judgment, y is obtained from small sector judgment, and T... s The given value is 1 / 6000, calculated by the internal input time calculation module of the three-level space vector pulse width modulation.

[0085] So, the volt-second balance formula in SVPWM is:

[0086] V ref T s =ΣV j T j , ΣT j =Ts,T j ≥0;

[0087] In the formula V ref As the reference voltage vector, T s For the switching period, V j For the j-th fundamental voltage vector participating in the reference vector synthesis, T j The first equation states that the sum of the volt-second effects generated by each basic voltage vector within one switching cycle should be equal to the volt-second value of the reference voltage vector, thus ensuring that the average output voltage of the inverter is consistent with the reference voltage; the second equation states that the sum of the effects of each vector should be equal to one complete switching cycle; T j ≥0 indicates that the duration of action of each vector cannot be negative;

[0088] The time calculation module outputs T P T N T2 and T3 represent the duration of different basic vectors or redundant switching states within a switching cycle, where T2 and T3 are the durations of two effective vectors, and T...P and T N Allocate two redundancy distributions for the remaining time, satisfying the following:

[0089] T P +T N +T2+T3=Ts;

[0090] Midpoint balance control changes the duration of the redundant state by adjusting the ratio of TP to TN. The above time amount needs to be further converted into time boundary and P / O / N state to finally generate IGBT gate pulse;

[0091] To reduce the number of switching operations and improve the output harmonic characteristics, the SVPWM of this invention employs a symmetrical switching state sequence. Therefore, firstly, according to the preset symmetrical sequence, T... P T N The total action time corresponding to T2 and T3 is divided into several consecutive time intervals. If a certain switching state needs to appear symmetrically in the first half and the second half of the cycle, its total action time is usually divided into two parts and placed at the two ends of the sequence respectively.

[0092] Suppose a switching cycle is ultimately divided into m time periods, and the duration of the r-th time period is... Then all time periods should satisfy:

[0093] ;

[0094] Based on this, the module calculates the cumulative time boundaries for each time period:

[0095] b0=0、b r = and b m =T s ;

[0096] Therefore, 0 to b1 correspond to the first switch state, b1 to b2 correspond to the second switch state, and so on.

[0097] Then divide the current total running time t by the switching period T. s The remainder after division is tc, which indicates the current position within the current switching cycle. Then, the local time tc is compared with the cumulative time boundary. When tc is within the r-th time interval, the r-th switching state S is output. r :

[0098] S(tc)=S r When b r-1 ≤tc r ;

[0099] The symmetric state sequence is represented as:

[0100] ​S1→S2→S3→S4→S3→S2→S1;

[0101] Midpoint balance control is achieved by changing the time boundary corresponding to redundant states, making the pulse of one state wider and the pulse of another state narrower.

[0102] When the local time crosses a certain cumulative time boundary, the system immediately switches to the next P / O / N combination. Since adjacent states are usually arranged according to the principle of changing only one bridge arm state, the number of switching devices involved in a single switch can be reduced.

[0103] This section outputs the state of the three-phase bridge arm: P, O, or N. Each phase arm of the NPC three-level inverter contains four power switching devices. The pulse generator module converts the state of each phase into four 0 / 1 gate signals according to a fixed mapping relationship:

[0104] P→[1, 1, 0, 0]

[0105] O→[0, 1, 1, 0]

[0106] N→[0, 0, 1, 1]

[0107] When the three-phase state is POO, phase A outputs [1,1,0,0], and phases B and C output [0,1,1,0]. When the time crosses a certain boundary within the cycle, it switches to the next P / O / N combination, and the gate pulse flips accordingly. Thus, the corrected T... P T N Ultimately, this manifests as a change in the gate pulse width of the corresponding redundant state; T P Increasing this value increases the P-type redundant state hold time, corresponding to a wider gate pulse T. N Increasing the holding time of the N-type redundant state increases the corresponding gate pulse width. Thus, the PI controller first changes the duration of the redundant state, and then indirectly changes the gate pulse width through the time boundary and the switching state table.

[0108] S6. Adjust the midpoint current to balance the voltage difference between the upper and lower capacitors.

[0109] The midpoint current changes with the switching state during a switching cycle. The average midpoint current during the switching cycle is:

[0110] i np,avg =(1 / T s ) dτ;

[0111] In the formula, i np (τ) is the instantaneous midpoint current, i np,avg The average midpoint current over one switching cycle; the integral term represents the net charge flowing into or out of the DC midpoint during that cycle, divided by T.s The corresponding average current is then obtained, which represents the net charge flowing into or out of the midpoint during one switching cycle.

[0112] If the influence of a pair of P-type and N-type redundant small vectors on the average midpoint current is i, then the midpoint currents during the period of their action are respectively i np , P and i np , N Then we have:

[0113] i np,avg PN =[i np,P T P +i np , N T N ] / T s

[0114] For a pair of typical redundant small vectors, under the same phase current conditions, the midpoint currents they generate are usually similar in magnitude and opposite in direction; if i approximately satisfies np,N =-i np,P ,for:

[0115] i np,avg PN =(T0 / T s (2f-1)i np , P；

[0116] T0 / T s This represents the proportion of time occupied by this pair of redundant small vectors in the entire switching cycle, 2f-1 indicates whether the time allocation of the two redundant states is balanced, and to which side it is biased, i np,P This represents the midpoint current generated during the operation of the P-type redundancy state;

[0117] As can be seen from the above formula, when f=0.5, the duration of the P-type and N-type redundant states is equal, and their effects on the average midpoint current cancel each other out. When f>0.5, the average midpoint current is closer to the direction generated by the P-type redundant state. When f<0.5, the average midpoint current is closer to the direction generated by the N-type redundant state. Therefore, changing the time allocation factor f is to adjust the direction and amplitude of the average midpoint current within one switching cycle.

[0118] The average midpoint current reflects the net current flowing into or out of the DC midpoint during one switching cycle. When the average midpoint current is not zero, a net charge will accumulate or be released at the DC midpoint, causing the upper and lower voltage-dividing capacitors to have different charging and discharging states. Since the total DC bus voltage can be approximated as constant during a relatively short switching cycle, when the voltage of one capacitor increases, the voltage of the other capacitor will decrease accordingly, thus changing the voltage difference between the upper and lower capacitors.

[0119] make i np >0 indicates that current flows from the AC side to the DC midpoint, and the midpoint voltage deviation is defined as the lower capacitor voltage minus the upper capacitor voltage:

[0120] ΔU=Uc2-Uc 1；

[0121] Where Uc1 and Uc2 represent the voltages of the upper and lower capacitors, respectively. When the parameters of the upper and lower capacitors are the same, both being C, the change in voltage deviation within one switching cycle can be approximated as:

[0122] ΔU[k+1]=ΔU[k]+(T s / C)i np,avg [k];

[0123] This formula shows that the direction of the average midpoint current determines the direction of the voltage difference between the upper and lower capacitors, while the magnitude of the average midpoint current determines the rate of change of the voltage difference; under a specified positive direction, when i np,avg When the value is greater than 0, the net charge flowing into the DC midpoint increases, causing the voltage difference between the upper and lower capacitors to increase; when... inp,avg When the value is less than 0, the net charge at the DC midpoint decreases, which reduces the voltage difference between the upper and lower capacitors.

[0124] Midpoint potential balance control selects a suitable redundant small vector based on the current voltage deviation and adjusts its duration, thereby changing the direction and magnitude of the average midpoint current within a switching cycle. The average midpoint current further controls the charging and discharging process of the upper and lower capacitors, causing the voltage on the side with higher voltage to gradually decrease and the voltage on the side with lower voltage to gradually increase, ultimately driving the voltage of the upper capacitor and the voltage of the lower capacitor to become consistent, thus achieving DC midpoint potential balance.

[0125] To illustrate how the midpoint voltage balance correction affects the time allocation of the redundant small vector, under the simulation condition with midpoint balance control, an analysis is performed on a switching cycle corresponding to t*=0.02837s in the transient process. The model's switching frequency is 6000Hz, therefore the switching cycle is:

[0126] T s =1 / 6000=166.67μs;

[0127] At that moment, the actual output of the midpoint balancing PI controller is:

[0128] u PI =0.4788;

[0129] Based on the calculation relationship of the time allocation factor:

[0130] f=u PI +0.5;

[0131] We can obtain:

[0132] f = 0.4788 + 0.5 = 0.9788;

[0133] At the selected representative time t*=0.02837s, the large sector number and small region number are respectively:

[0134] N=6, y=6;

[0135] Based on the time calculation relationship between the 6th major sector and the 6th minor sector in the time calc module, the time of action of the two effective vectors is determined by the local reference vector angle θ. 60 The normalized amplitude k and the switching period Ts are determined together. At this moment, the local reference vector angle and the normalized amplitude are calculated in real time based on the amplitude and phase angle of the reference voltage vector, and are approximated as follows:

[0136] θ 60 ≈0.795642 rad;

[0137] k≈0.717654;

[0138] When N=6 and y=6, the effective vector action time formulas used by the time calc module are as follows:

[0139] T2=T s [2ksin(θ 60 )-1];

[0140] T3=2kT s sin(π / 3-θ 60 );

[0141] T s =1.6667×10 -4 s, k=0.717654 and θ 60 Substituting 0.795642 rad into the formula for calculating T2, we get:

[0142] T2 = 1.6667 × 10 -4 ×[2×0.717654×sin(0.795642)-1]=4.210μs;

[0143] Similarly, substituting the above parameters into the formula for calculating T3, we get:

[0144] T3=2×0.717654×1.66666667×10 -4 ×sin(π / 3-0.795642)=59.544μs;

[0145] Therefore, within this representative switching cycle, the durations of the two effective vectors are as follows:

[0146] T2 = 4.210 μs;

[0147] T3 = 59.544 μs;

[0148] Therefore, the total remaining time available for allocation between the two redundant switching implementations within this switching cycle is:

[0149] T0 = ​​T s -T2-T 3；

[0150] T0=166.67-4.210-59.544=102.916μs;

[0151] Based on the determination of the sector, sub-region, and three-phase current polarity of the reference vector at that moment, the time calculation module allocates a larger time share fT0 to T. N The smaller time share (1-f)T0 is allocated to T. P ,Right now:

[0152] T N =fT0=0.9788×102.916=100.734μs;

[0153] T P =(1-f)T0=(1-0.9788)×102.916=2.181μs;

[0154] Ultimately, the durations of each action satisfy the following:

[0155] T P +T N +T2+T3=2.181+100.734+4.210+59.544=166.67μs=T s ;

[0156] The above results show that, after adding the midpoint balance correction, the sum of the action times of each vector is still equal to one complete switching cycle, indicating that the time calculation results satisfy the SVPWM action time constraint. Meanwhile, T P With T NThe significant difference indicates that the midpoint balance controller has redistributed the duration of the two redundant switching states based on the capacitor voltage deviation and the phase current direction.

[0157] T N As the holding time of the corresponding state increases, the gate pulse width increases accordingly; T P As the holding time of the corresponding state decreases, its gate pulse width decreases accordingly. By changing the duration of the two redundant states, the proportion of different midpoint current paths within a switching cycle can be adjusted, thereby changing the direction and magnitude of the average midpoint current.

[0158] At the selected representative moment, the three-phase currents are as follows:

[0159] i a =-21.6401A、i b =20.2402A and i c =1.3999A;

[0160] The sum of the three-phase currents is approximately zero, i. a +i b +i c ≈0. Based on the state mapping representing the current moment, the P-type redundant state is POO, and the N-type redundant state is ONN. For the POO state, phases B and C are connected to the DC midpoint, therefore the midpoint current is:

[0161] i np,P =i b +i c =20.2402 + 1.3999 = 21.6401A;

[0162] In the ONN state, only phase A is connected to the DC neutral point, therefore the neutral point current is:

[0163] i np,N =i a =-21.6401A;

[0164] The ratio of the duration of action of the two states is:

[0165] T N / T P =100.734 / 2.181≈46.19;

[0166] That is, the duration of the N-type midpoint current path is approximately 46.19 times that of the P-type path. Therefore, although the instantaneous midpoint current magnitudes corresponding to the two states are approximately the same, the positive and negative midpoint charges within a switching cycle no longer cancel each other out due to the significant difference in duration.

[0167] The amount of charge transferred to the midpoint by a switching state can be expressed as Q = iT. The amount of midpoint charge generated by a P-type redundant state is:

[0168] Q P =i np,P T P ;

[0169] Q P =21.6401×2.181×10 -6 =4.7197×10 -5 C=47.20μC;

[0170] The midpoint charge generated by the N-type redundant state is:

[0171] Q N =i np,N T N ;

[0172] Q N =-21.6401×100.734×10 -6 =-2.1799×10 -3 C = -2179.89 μC;

[0173] Therefore, the net midpoint charge generated by this redundant vector pair within one switching cycle is:

[0174] Q np =Q P +Q N ;

[0175] Q np =47.20-2179.89=-2132.69μC=-2.1327mC;

[0176] The negative sign indicates that, under the current direction convention adopted in this paper, the negative direction midpoint current path dominates within this cycle. Therefore, considering only this redundant vector pair, the average midpoint current within one switching cycle is:

[0177] i np,avg PN =Q NP / T s ;

[0178] i np,avg PN =(-2.1327×10 -3 ) / (166.67×10 -6 )≈-12.80A;

[0179] If the durations of the two redundant states are exactly equal, then due to i np,N ≈-i np,PThe charges generated by the two will approximately cancel each other out, and the average midpoint current will approach zero. The current controller significantly extends the N-type state and shortens the P-type state, giving the average midpoint current a clear negative direction, thereby providing an action for the directional regulation of the capacitor voltage difference.

[0180] Figure 3 The parameters of the upper and lower voltage divider capacitors in the model are as follows:

[0181] C1 = 1.36 × 10 -5 F, C2 = 6.5 × 10 -3 F;

[0182] The initial voltages of the capacitors are U C1 (0) = 200V, U C2 (0) = 300V, and the midpoint voltage difference is defined as:

[0183] ΔU=U C2 -U C1 ;

[0184] Under the condition that the total DC bus voltage remains approximately constant during a single switching cycle, the net charge Q at the midpoint... np The resulting voltage difference change can be approximated as:

[0185] Δ(ΔU)=2Q np / (C1+C2);

[0186] in:

[0187] C1 + C2 = 1.36 × 10 -5 +6.5×10 -3 =6.5136×10 -3 F;

[0188] Substituting the net midpoint charge of the representative period, we get:

[0189] Δ(ΔU)=[2×(-2.1327×10 -3 )] / (6.5136×10 -3 )≈-0.655V;

[0190] Correspondingly, the theoretical change in voltage between the upper and lower capacitors is approximately:

[0191] ΔU C1 =-Q np / (C1+C2)≈+0.327V;

[0192] ΔU C2 =Q np / (C1+C5)≈-0.327V;

[0193] Therefore, in the current UC2 >U C1 Under these conditions, the representative cycle causes the upper capacitor voltage to rise and the lower capacitor voltage to fall, thus reducing the midpoint voltage difference ΔU. According to this theoretical estimate, the voltage difference decreases from approximately 100V to approximately 99.35V.

[0194] The theoretical correction of approximately 0.655V within a single switching cycle is not significant relative to the initial voltage difference of 100V, representing about 0.655%. This is normal because the SVPWM midpoint balancing control does not cause a large abrupt change in capacitor voltage within a single cycle. Instead, it gradually accumulates and converges the voltage difference through controlled small-amplitude charging and discharging adjustments within each switching cycle. The smaller single-cycle correction also helps avoid current surges and midpoint voltage oscillations caused by over-straining.

[0195] The model's switching frequency is 6000Hz, corresponding to a switching period of:

[0196] T s =1 / 6000=166.67μs;

[0197] To avoid misjudgments caused by the midpoint voltage difference briefly crossing the set range due to switching ripple, this paper uses the midpoint voltage difference entering the ±5V range and remaining there continuously for 20ms as the criterion for reaching a stable equilibrium state. First, the midpoint voltage difference is calculated based on the voltages of the upper and lower capacitors recorded in the simulation. Then, it is checked point by point whether the midpoint voltage difference enters the ±5V range, and the first continuous interval with a duration of not less than 20ms is searched. The start time of this continuous interval is defined as the time when the midpoint voltage enters the stable equilibrium range.

[0198] After adding midpoint balancing control, the time required for the midpoint voltage difference to enter the ±5V range and remain continuously for 20ms is 0.040216s, and the corresponding number of switching cycles is approximately:

[0199] N on =0.040216 / (166.67×10 -6 )≈241 cycles;

[0200] Without midpoint balance control, achieving the same conditions would take 0.367028 seconds, corresponding to approximately:

[0201] N off =0.367028 / (166.67×10 -6 )≈2202 cycles;

[0202] Therefore, it can be seen that after adding control, the number of switching cycles required to reach the equilibrium range is reduced by approximately 1961, and the convergence speed is increased by approximately:

[0203] 0.367028 / 0.040216≈9.13 times;

[0204] Based on a rough equivalent reduction from a 100V voltage difference to a 5V range, the average correction amount for the entire process after adding control is approximately:

[0205] (100-5) / 241≈0.394V / cycle;

[0206] This value is on the same order of magnitude as the 0.655V estimated for the representative period. The larger correction for the representative period is because the PI output is close to a more throttling state at that moment; as the voltage difference gradually decreases, the PI output and T... P T N The differences between them will also decrease accordingly.

[0207] The following uses a MATLAB / Simulink simulation model to examine the effect of this invention on suppressing midpoint potential.

[0208] A simulation model of an NPC three-level inverter was built on the MATLAB / Simulink platform. The DC side of the main circuit consists of an ideal DC voltage source and two series-connected voltage divider capacitors, with the common point of the two capacitors being the neutral point O. The inverter side adopts a three-phase NPC three-level bridge structure, and the AC side is configured with an LC filter network to filter out the high-frequency switching components introduced by SVPWM modulation. A DC side voltage measurement module was included in the model to collect the total bus voltage, the upper capacitor voltage, and the lower capacitor voltage, and calculate the capacitor voltage difference to provide feedback for the neutral point potential balance control.

[0209] like Figure 3 and 4 As shown, the control system mainly includes voltage and current sampling, coordinate transformation, PI regulation, SVPWM modulation calculation, and midpoint balance compensation. The main controller generates reference voltage components in the α-β coordinate system based on the AC side voltage and three-phase current, and then sends them to the three_level_svpwm module. Within the SVPWM module, the control algorithm sequentially completes sector determination, action time allocation, redundancy vector correction, and PWM pulse output.

[0210] To verify the model's output capability and the effectiveness of the midpoint balance control, this invention sets up two operating conditions: steady-state output waveform verification and midpoint potential balance comparison. The steady-state output waveform verification mainly observes the phase voltage, line voltage, and output current; the midpoint potential balance comparison involves setting different initial voltages for the upper and lower capacitors and running the system in two modes: without balance control and with balance control.

[0211] The parameters are shown in Table 2.

[0212]

[0213] The simulation model can be explained from the following main aspects.

[0214] from Figure 5 The output voltage of phase A without LC filtering can be observed. The phase voltage switches frequently between +250V, 0V, and -250V, exhibiting clear three-level PWM stepped wave characteristics. The pulse width varies with the reference sinusoidal quantity, indicating that the SVPWM module can adjust the basic vector's duration based on the reference vector.

[0215] from Figure 6 The steady-state waveform of the three-phase output phase voltage under SVPWM modulation after LC filtering shows that the voltage amplitudes of phases A, B, and C are basically the same, the phase difference of the three phases is about 120°, and the waveform is close to sine. This indicates that the filter network can effectively suppress high-frequency switching components, and the established main circuit and control system can form a stable three-phase AC output.

[0216] from Figure 7 The waveform of the three-phase output line voltage without LC filtering shows that, due to the three phase voltage levels, the composite line voltage exhibits five stepped levels: +500V, +250V, 0V, -250V, and -500V. Compared to a two-level inverter, the three-level structure has finer line voltage steps and smaller single voltage jump amplitude, which helps reduce harmonics and device stress.

[0217] from Figure 8 The steady-state waveform of the three-phase output line voltage under SVPWM modulation shows that the amplitudes of the three sets of line voltages are basically the same after LC filtering, and the periodicity and symmetry are good, indicating that the system has good three-phase balanced output capability during steady-state operation.

[0218] from Figure 9 The three-phase output current waveforms show that there is a transient adjustment process in the initial startup phase, followed by a gradual entry into a steady state. In the steady-state phase, the amplitudes of the three-phase currents are basically the same, with a phase difference of approximately 120°, and the overall waveform is close to a sine wave, retaining only a small high-frequency ripple. This result further demonstrates that the model can achieve relatively stable three-phase current output under SVPWM control.

[0219] To verify the effect of the midpoint potential balance control, the initial voltage of the upper DC capacitor was set to 200V and the initial voltage of the lower DC capacitor was set to 300V, so that the initial voltage difference between the upper and lower capacitors was approximately 100V.

[0220] Figure 10 The waveform of the capacitor voltage difference without midpoint balancing control is shown. The voltage difference gradually decreases over time and approaches zero, indicating that the system has a certain natural equilibrium tendency under symmetrical operating conditions, but its decay rate is slow.

[0221] Figure 11This is a waveform of the capacitor voltage difference when midpoint balance control is applied. Figure 10 In contrast, the capacitor voltage difference rapidly approaches zero within a short period of time, followed by only small high-frequency fluctuations, indicating that the PI balance control can actively regulate the charging and discharging process of the upper and lower capacitors and accelerate the recovery speed of the midpoint potential.

[0222] Table 3 compares the effects of midpoint balance control before and after.

[0223]

[0224] It can be seen that, using an NPC-type three-level inverter, a main circuit, control system, and midpoint potential balance control model were established on the MATLAB / Simulink platform. Simulation analysis was conducted on the output voltage, current, and capacitor voltage difference under SVPWM control. The model can output symmetrical and stable three-phase voltage and current, verifying that the established model can meet the basic power quality requirements of rural microgrids for inverters, laying the foundation for its stable operation in remote and weak grid environments.

[0225] The invention demonstrates that the established three-level inverter model can output symmetrical and stable three-phase phase voltage, line voltage, and current, indicating that the main circuit topology, SVPWM modulation module, and filtering stage settings can achieve the suppression effect.

[0226] Three-level SVPWM can achieve reference voltage vector synthesis through sector judgment, action time calculation, and symmetrical switching sequence generation. This also conforms to the output characteristics of a three-level inverter. Under the same initial voltage deviation, without balancing control, the capacitor voltage difference converges slowly, making it difficult for the inverter to adapt to the actual operating conditions of rural microgrid photovoltaic processing frequency fluctuations and drastic load changes, posing a risk of voltage imbalance. With the addition of PI-type midpoint balancing control, the capacitor voltage difference can converge to near zero in a shorter time, indicating that redundant small vector allocation combined with PI regulation can effectively suppress midpoint potential deviation.

Claims

1. A method for suppressing midpoint potential fluctuations in a three-level inverter for a rural microgrid, characterized in that... include: S1. Collect the voltage of the upper and lower capacitors connected in series on the DC side of the inverter, and calculate the correction amount based on the voltage difference between the upper and lower capacitors. S2. Collect the three-phase current on the AC side of the inverter and determine the current polarity; S3. Determine the correction direction of the correction amount in S1 based on the current polarity determined by S2. S4. Calculate the small vector time allocation factor, and obtain the action time of the positive small vector and the action time of the negative small vector through the small vector time allocation factor; S5. Based on the positive small vector action time and negative small vector action time obtained in S4, generate the SVPWM switching sequence and output PWM pulses; S6. Adjust the midpoint current to balance the voltage difference between the upper and lower capacitors.

2. The method for suppressing midpoint potential fluctuations in a three-level inverter for a rural microgrid according to claim 1, characterized in that... S1 includes the following steps: The collected upper capacitor voltage and lower capacitor voltage are subtracted after a unit delay to keep the voltage feedback quantity synchronized with the SVPWM calculation cycle. The difference between the voltage of the upper capacitor and the voltage of the lower capacitor is the midpoint voltage balance error, which is used to obtain the PI correction value through a parallel PI controller. The PI correction limit is used to ensure that the correction amount sent to the timing calculation module is within the allowable range, preventing the midpoint balance correction from being too strong and causing the action time of a certain vector to be negative, exceeding the switching cycle, or destroying the PWM sequence.

3. The method for suppressing midpoint potential fluctuations in a three-level inverter for a rural microgrid according to claim 2, characterized in that... The input error formula for the parallel PI controller is: e np [k]=U c2 [k-1]-U c1 [k-1]=ΔU[k-1]; In the formula e np [k] represents the midpoint voltage balance error in the k-th control cycle; ΔU[k-1] represents the midpoint voltage deviation in the previous sampling cycle; therefore, when e np When e > 0, the voltage of the lower capacitor is higher than that of the upper capacitor; when e np When e < 0, the voltage of the upper capacitor is higher than that of the lower capacitor; when e np When =0, the upper and lower capacitors reach an ideal balance; To reduce deviation, a parallel PI controller is used, comprising a proportional element and an integral element. The proportional element directly adjusts rapidly based on the current error, and its output formula is: u p (t)=K p e np (t); In the formula, K p U is the proportionality coefficient. p The proportional control is a variable that can immediately change the controller output according to the instantaneous voltage deviation, which is used to improve the recovery speed of the midpoint voltage. The larger the deviation, the larger the proportional control and the stronger the correction of the action time of the redundant small vector. The integration step first multiplies the error signal by the integration coefficient K. i The error is then accumulated using a continuous integrator of 1 / s, and the original integral output formula is as follows: x i (t)= i e np (τ)dτ; x i (t) represents the original output of the integral stage at time t; K i e is the integral coefficient, used to determine the rate of error accumulation and the strength of integral adjustment; np (τ) represents the midpoint voltage balance error at time τ during the integration process; τ is the integral variable; t is the current control time. The integral element is used to accumulate the persistent midpoint voltage error and eliminate the steady-state deviation that may remain when only proportional control is used. Therefore, the formula for the integral part is: u i (t)=sat[x i (t),u i,min ,u i,max ]; Among them, u i,min and u i,max These are the lower and upper limits of the integral output, set to -0.5 and 0.5 respectively, to limit the continuous accumulation of the integral caused by long-term deviations. ； The formula for the final PI correction is to add the proportional adjustment to the integral adjustment after the amplitude limit is: u PI (t)=sat[u p (t)+u i (t),u min ,u max ]。 4. The method for suppressing midpoint potential fluctuations in a three-level inverter for a rural microgrid according to claim 1, characterized in that... The S2 signal is acquired by the three-phase voltage and current measurement module in the main circuit and output in the form of Iabc. This signal represents the instantaneous phase current formula of phases A, B, and C: Iabc=[ia ib ic]ᵀ; In the formula, ia, ib, and ic are the instantaneous currents of phases A, B, and C, respectively. After the Iabc signal enters the three-level space vector pulse width modulation module, it first goes through a unit delay stage to keep synchronized with the current SVPWM calculation cycle, and then is split into ia, ib, and ic, and input into the vector action time calculation module. The three-phase current is not directly used to calculate the midpoint voltage error, but is used to determine whether the midpoint current flows into or out of the DC midpoint when the current redundant switch is in operation, thereby determining the correction direction of the redundant small vector action time.

5. The method for suppressing midpoint potential fluctuations in a three-level inverter for a rural microgrid according to claim 4, characterized in that... After defining the positive direction of the midpoint current, the instantaneous midpoint current can be expressed as: inp = δa ia + δb ib + δc ic; In the formula, δa, δb and δc are the connection state coefficients of the bridge arm midpoint. For any phase x∈{a,b,c}, the values ​​can be expressed as follows: δx=1, phase x is in state O; δx=0, phase x is in state P or N.

6. The method for suppressing midpoint potential fluctuations in a three-level inverter for a rural microgrid according to claim 1, characterized in that... S3 is the correction amount after determining the current direction based on the current polarity: dcorr=u PI (t)·dir; In the formula, dir is the direction factor, and its sign is determined by ia, ib, ic, and the current large sector, small sector, and redundant switch status; the sign of dcorr is used to determine which type of redundant small vector should have its duration increased, and its absolute value reflects the degree of time allocation correction.

7. The method for suppressing midpoint potential fluctuations in a three-level inverter for a rural microgrid according to claim 6, characterized in that... The correction amount, after considering the current direction, is processed in the time calculation module and converted into a time allocation factor f ranging from 0 to 1; If the total duration of a pair of P-type and N-type redundant small vectors is T0, then the durations of the two redundant states are as follows: T P =f·T 0， T N =(1-f)·T 0； In the formula, T P and T N These represent the durations of the P-type and N-type redundant small vectors, respectively. Since the two redundant states correspond to the same space voltage vector position, T is maintained. P With T N The sum of these two is T0, which allows us to change the duration of the two midpoint current paths without significantly altering the vector's contribution to the average output voltage on the AC side. f is obtained by shifting the output of the midpoint balancing PI controller, i.e., f = u PI (t)+0.5, u PI (t) represents the time allocation offset in the model; simultaneously, based on the large sector, small region, and phase current polarity, it is determined which time fT0 should be allocated to T. P Or T N Therefore, f=0.5 indicates that the two redundant states are allocated time equally; f>0.5 indicates that the P-type redundant state occupies more time; f<0.5 indicates that the N-type redundant state occupies more time.

8. The method for suppressing midpoint potential fluctuations in a three-level inverter for a rural microgrid according to claim 1, characterized in that... The S5 module includes a time calculation module that receives the local reference vector angle, large sector number, small sector auxiliary quantity, switching cycle, DC bus voltage, and three-phase current to perform two types of calculations: first, it calculates the total basic vector time required for the reference voltage based on the volt-second balance; second, it writes the midpoint balance correction into the time allocation between redundant small vectors. Therefore, the volt-second balance formula in SVPWM is: V ref T s =ΣV j T j ,ΣT j =Ts,T j ≥0; In the formula V ref As the reference voltage vector, T s For the switching period, V j For the j-th fundamental voltage vector participating in the reference vector synthesis, T j Its duration of action within one switching cycle; The first equation indicates that the sum of the volt-second effects generated by each basic voltage vector within one switching cycle should be equal to the volt-second value of the reference voltage vector, thereby ensuring that the average output voltage of the inverter is consistent with the reference voltage. The second equation states that the sum of the action times of each vector should equal one complete switching cycle; T j ≥0 indicates that the duration of action of each vector cannot be negative; The time calculation module outputs T P T N T2 and T3 represent the duration of action of different basic vectors or redundant switching states within a switching cycle; where T2 and T3 are the durations of action of two effective vectors, T... P and T N Allocate two redundancy distributions for the remaining time, satisfying the following: T P +T N +T2+T3=T s； Midpoint balance control is achieved by adjusting T P With T N The proportion changes the duration of the redundant state. The above time amount needs to be further converted into time boundary and P / O / N state to finally generate IGBT gate pulse; SVPWM employs a symmetrical sequence of switching states, which is represented as follows: S1→S2→S3→S4→S3→S2→S1; Midpoint balance control is achieved by changing the time boundary corresponding to redundant states, making the pulse of one state wider and the pulse of another state narrower. When the local time crosses a certain cumulative time boundary, the system immediately switches to the next P / O / N combination. Since adjacent states are usually arranged according to the principle of changing only one bridge arm state, the number of switching devices involved in a single switch can be reduced. This section outputs the state of the three-phase bridge arm: P, O, or N. Each phase arm of the NPC three-level inverter contains four power switching devices. The pulse generator module converts the state of each phase into four 0 / 1 gate signals according to a fixed mapping relationship: P→[1，1，0，0] O→[0，1，1，0] N→[0，0，1，1] When the three-phase state is POO, phase A outputs [1,1,0,0], and phases B and C output [0,1,1,0]. When the time crosses a certain boundary within the cycle, it switches to the next P / O / N combination, and the gate pulse flips accordingly. Thus, the corrected T... P T N Ultimately, this manifests as a change in the gate pulse width of the corresponding redundant state; T P Increasing the holding time of the P-type redundant state corresponds to a wider gate pulse T. N Increasing this value increases the N-type redundant state holding time, resulting in a wider gate pulse. Therefore, it can be seen that the PI controller first changes the duration of the redundant state, and then indirectly changes the gate pulse width through the time boundary and the switching state table.

9. The method for suppressing midpoint potential fluctuations in a three-level inverter for a rural microgrid according to claim 1, characterized in that... The midpoint current changes with the switching state during one switching cycle, and the average midpoint current during the switching cycle is: i np,avg =(1 / T s ) dτ; In the formula, i np (τ) is the instantaneous midpoint current, i np,avg The average midpoint current over one switching cycle; the integral term represents the net charge flowing into or out of the DC midpoint during that cycle, divided by T. s The corresponding average current is then obtained, which represents the net charge flowing into or out of the midpoint during one switching cycle. If the influence of a pair of P-type and N-type redundant small vectors on the average midpoint current is i, then the midpoint currents during the period of their action are respectively i np , P and i np , N Then we have: i np,avg PN =[i np,P T P +i np , N T N ] / T s； For a pair of typical redundant small vectors, under the same phase current conditions, the midpoint currents they generate are usually similar in magnitude and opposite in direction; if i approximately satisfies np,N =-i np,P ,for: i np,avg PN =(T0 / T s )(2f-1)i np , P； T0 / T s This represents the proportion of time occupied by this pair of redundant small vectors in the entire switching cycle, 2f-1 indicates whether the time allocation of the two redundant states is balanced, and to which side it is biased, i np,P This represents the midpoint current generated during the operation of the P-type redundancy state; As can be seen from the above formula, when f=0.5, the duration of the P-type and N-type redundant states is equal, and their effects on the average midpoint current cancel each other out. When f>0.5, the average midpoint current is closer to the direction generated by the P-type redundant state. When f<0.5, the average midpoint current is closer to the direction generated by the N-type redundant state. Therefore, changing the time allocation factor f is to adjust the direction and amplitude of the average midpoint current within one switching cycle.

10. The method for suppressing midpoint potential fluctuations in a three-level inverter for a rural microgrid according to claim 9, characterized in that... The change in voltage deviation within one switching cycle can be approximately expressed as: ΔU[k+1]=ΔU[k]+(T s / C)i np,avg [k]; This formula shows that the direction of the average midpoint current determines the direction of the voltage difference between the upper and lower capacitors, while the magnitude of the average midpoint current determines the rate of change of the voltage difference; under a specified positive direction, when i np,avg When i > 0, the net charge flowing into the DC midpoint increases, causing the voltage difference between the upper and lower capacitors to increase; when i np , avg When <0, the net charge at the DC midpoint decreases, which reduces the voltage difference between the upper and lower capacitors; Midpoint potential balance control selects a suitable redundant small vector based on the current voltage deviation and adjusts its duration, thereby changing the direction and magnitude of the average midpoint current within a switching cycle. The average midpoint current further controls the charging and discharging process of the upper and lower capacitors, causing the voltage on the side with higher voltage to gradually decrease and the voltage on the side with lower voltage to gradually increase, ultimately driving the voltage of the upper capacitor and the voltage of the lower capacitor to become consistent, thus achieving DC midpoint potential balance.

Images