Research on three-level PWM Rectifier without grid voltage sensor

Traditional three-level PWM rectification technology usually uses voltage sensors to sample the grid voltage, and then obtains the angle information of the voltage vector through a phase-locked loop, thereby achieving the d/q axis control function of voltage and current. However, the use of voltage sensors not only increases system costs, but also increases system volume and debugging difficulty. In addition, when the voltage sensor is damaged, it can cause system shutdown and affect system reliability. After analyzing the three-level control method of PWM rectifiers, a sensorless control strategy for grid voltage based on sliding mode observer method is proposed. By using the sliding mode observer method to estimate the amplitude and phase of the power grid voltage at the voltage and current values of each PWM cycle, a phase-locked loop is used to determine the position of its voltage vector. The simulation results all demonstrate that the proposed method can accurately and effectively estimate the voltage vector angle of the power grid, and ensure that the output of the rectification system meets the standard voltage waveform.


Introduction
PWM rectifiers, due to their advantages of bidirectional energy flow and achieving unit power factor, have significantly reduced requirements for filtering devices compared to uncontrolled and semi controlled rectifiers, and have a larger adjustment range.Therefore, they have been widely used in emerging industries such as new energy vehicles, energy storage converters, and electric aircraft [1][2][3].
Traditional PWM rectifiers are usually two-level.In scenarios with low power requirements, twolevel rectifiers have the advantages of simple structure, convenient control, and low device cost.However, in areas such as energy storage systems with high power requirements, two-level rectifiers require higher switching frequencies, which can lead to excessive device switching losses and affect system efficiency.Compared with three-level rectifiers, they have lower switching frequency, higher efficiency, and can produce high-quality voltage and current waveforms, which have been widely used in energy storage, wind power, and other fields [4].At present, the three-level topology mainly includes NPC-I topology, T-type topology, and ANPC topology.The NPC-I topology was proposed by Japanese scholar A Nabael [5], and each phase is composed of four power switches and two clamping diodes, making a single switch withstand a DC voltage that is generally the same as the DC bus voltage.Under the same power switch, the system has a higher withstand voltage level.However, the NPC-I topology has the problem of uneven losses, Due to the excessively high junction temperature of a single power switch, it affects the design of system modules.Therefore, people have replaced the clamping diode with a power switch based on the NPC-I topology [6][7], adding a neutral point circuit to flexibly control the selection of the commutation circuit for the switch transistor, making the switch loss relatively balanced, and thus improving the output power of the module to a certain extent.Due to the need for more power devices in NPC-I topology and ANPC topology, in order to reduce costs, T-type three-level topology is often used in systems with lower voltage levels [8][9].On the basis of two-level, two IGBTs in reverse series are used to connect the output terminal to the neutral point, thereby introducing zero level.Compared to NPC-I and ANPC topology, T-type topology can reduce equipment volume and cost, but its voltage withstand level is consistent with that of two-level; This article intends to take the NPC-I converter as the research object.
However, regardless of the topology structure and control method adopted, voltage sensors and current sensors are required to obtain three-phase AC voltage, three-phase AC current, and DC voltage for control.At least six sensors are required at this time.However, excessive sensors will inevitably reduce the reliability of the system.When one sensor fails, it will inevitably lead to the failure of the entire control system, thereby affecting the reliability of the rectification system, Therefore, literature [9] proposes to apply three-phase current reconstruction technology based on single current sampling to PWM rectifiers.However, current reconstruction is affected by dead zones and there is a certain reconstruction blind zone, which will reduce the control accuracy of the current.Moreover, this method can only reduce one current sensor and has limited cost impact.Therefore, literature [10][11] proposes a control strategy without grid voltage to reduce three grid voltage sensors, thereby reducing the cost of PWM rectifiers, and improve the reliability of the rectifier system.Voltage control without power grid is usually divided into virtual flux method and power grid voltage observer method.The traditional virtual flux estimation method uses a pure integration link to estimate the flux, but the use of a pure integrator inevitably introduces initial value and DC bias issues.Therefore, reference [10] proposes using three first-order low-pass filters to replace the pure integration link.However, the low-pass filter potential will inevitably introduce phase bias, so reference [11] proposes an orthogonal signal generator based on second-order generalized integrator to improve integration accuracy and eliminate phase error.
At present, the application of power grid voltage observer method in three-level PWM rectifiers is relatively limited.Therefore, this paper proposes a power grid voltage observation control method based on sliding mode observer.Based on theoretical analysis of the sliding mode observer and three-level SVPWM control principles, the feasibility of the proposed method is verified through simulation.

Control strategy of PWM rectifier
This article intends to use id=0 control, with the voltage given as the outer loop and the d-q axis current as the inner loop.The sliding mode observer method is used to estimate the power grid voltage, and the position of its voltage vector angle is determined using a phase-locked loop.
The d-q axis voltage balance equality for PWM rectifiers is: According to equality (1), there is a coupling term related to velocity in the d-q axis voltage equality, and the existence of the coupling term will increase the difficulty of voltage and current dual closedloop control.To reduce the difficulty of control, decouple equality (1) with current feedforward.Based on the current PI controller, the control equations for vd and vq are as follows: In the equality(2), iP K and iI K represent the coefficients of proportion and integration in the current inner loop PI controller; represent the given values of current d i and q i .Substitute equality (1) into equality (2) to obtain the current balance equality of the PWM rectifier after feedforward decoupling control: According to equality (3), it can be seen that the decoupling control of the d-axis and q-axis of the PWM rectifier system can be achieved, which can achieve separate control of the d-axis and q-axis.Combined with the control process of the voltage outer loop, the control strategy of the decoupled dual closed-loop control system is shown in Figure 1:

Three level SVPWM algorithm
The SVPWM control algorithm is used for the three-level rectifier of I-NPC topology, which decomposes the three-level space vector map into six large sectors.After determining the large sector where the reference voltage vector is located, the voltage vector is decomposed to obtain the two-level vector.Then, the two-level control method used to determine the sector and calculate the vector action time, thereby achieving three-level control.
Among them, t1, t2, and t3 are the action times of V1, V7, and V13, respectively.Translate Vref along V1, and the equality after translation is: (5) The two-level vector synthesis equality can be obtained as: By translating this coordinate, three-level SVPWM modulation can be achieved within the first sector using a two-level SVPWM modulation strategy, and other sectors can also be achieved similarly.

Control strategy of power grid voltage observer based on sliding mode observer
The acquisition of grid voltage vector angle can be transformed into the acquisition of grid voltage information.This article adopts the method of constructing a sliding mode observer to observe the voltage of the power grid.The error between the actual current value and the observed value in the stationary coordinate system is selected as the sliding mode surface s.When the estimated value is equal to the actual current value, the accurate voltage value of the power grid can be obtained.The traditional sliding mode observer is constructed as shown in equality (7):      (7) In the equality (7), z  and z  is α-β axis sliding mode control function is expressed as equality (8): sgn( ) sgn( ) In the equality, k is the sliding mode gain coefficient, and sgn() is the sign function.At , the condition for the existence of sliding mode was met.When there was a deviation between the estimated current and the actual current, that is, when the system deviated from the sliding mode surface, the control function z  and z  would take effect, causing the system to continuously approach the sliding mode surface, reach the sliding mode within a certain time, and stabilize near the equilibrium point.z  and z  including the voltage fundamental wave of the power grid in and, if a low-pass filter is used to filter out higher-order harmonics, the estimated power grid voltage can be obtained.Subsequently, the phase angle of the grid voltage vector can be obtained through a normalized phase-locked loop or tangent function.In summary, the schematic diagram of the sliding mode observer is shown in Figure 3:

Simulation verification
To verify the effectiveness of the three-level SVPWM modulation strategy and the control strategy without grid voltage sensors, a simulation model was built for simulation experiments.The simulation parameters are shown in Table 1.Under the normal control strategy of using voltage sensors to obtain grid voltage information, the rectifier system is started with load at 0s, and its output waveform is shown in Figure 4.The steady-state value is 750 V, the dynamic response speed is less than 0.1s, and the overshoot is small.The peak values of the voltage ripple are all below 0.05 V, which meets the technical requirements.The waveforms of phase A voltage and phase A current are shown in Figure 5. Due to the system being in rectification mode, the power factor is -1, and the phase difference between phase voltage and phase current is 180°.The phase current waveform can present a good sinusoidal waveform with less harmonic content, avoiding any impact on the reliability of the power grid.The observed voltage waveform of the power grid is shown in Figure 6, which shows that it can present a sinusoidal waveform.Fourier analysis shows that its total harmonic distortion rate is 5.01%, mainly consisting of third and fifth harmonics.The high-frequency harmonics are relatively small, indicating that the chattering of the sliding mode observer has a relatively small impact on its observation effect.Under the control strategy of using a sliding mode observer to obtain grid voltage information, the rectifier system is started with load at 0 s, and its output voltage waveform is shown in Figure 7.The visible voltage waveform is consistent with the dynamic response speed of the grid voltage sensor, with a small overshoot.The peak values of the voltage ripple slightly increase, but both are below 0.5 V, which still meets the technical specifications.The power grid voltage vector angle observed by its sliding mode observer and the actual angle waveform are shown in Figure 8, indicating that the estimated power grid voltage vector angle by its sliding mode observer can quickly follow the actual angle with a small error of about 9.6°.To further test the dynamic performance of the system, the given DC voltage was reduced from 750 V to 600 V 0.25 s after the rectifier system reached steady state.The output DC voltage waveform is shown in Figure 9, indicating that the system can quickly adjust after the given voltage change.The dynamic response time is less than 0.1 s, and it enters steady state after 0.1 s, verifying that the rectifier system has good dynamic performance.

Conclusion
This paper combines the PWM rectifier model to decouple the voltage equation.Based on the threelevel SVPWM modulation strategy, a sliding mode observer is used to estimate the vector angle of the power grid voltage.On the one hand, it solves the disadvantage of traditional PWM rectifiers requiring grid voltage sensors.On the other hand, through the three-level SVPWM modulation strategy, the harmonic content of the AC side current is reduced, greatly improving the efficiency of the rectification system, and resulting in a significant improvement in output voltage compared to uncontrolled rectification and indirect current control.The effectiveness of the proposed grid voltage observer method in the three-level modulation strategy is verified in simulation.

Figure 1 .
Figure 1.Diagram of voltage and current dual closed loop decoupling control strategy.

Figure 2 .
Schematic diagram of three-level space vector decomposition.

Figure 2 (
Figure2(a) shows the six large sectors divided.This control strategy equates the three-level reference voltage vector to the sum of a small vector in a three-level basic vector and a two-level voltage vector.Taking the reference voltage vector in the first largest sector as an example, the three-level voltage reference vector Vref is transformed into the sum of the small vector V1 and the two-level voltage reference vector V'ref in the first largest sector through coordinate translation, as shown in Figure2 (b).According to the principles of vector synthesis and volt second balance, Vref can be synthesized by V1,

Figure 5 .
Figure 5. Voltage and current waveform of the power grid.

Figure 8 .
Figure 8. Grid voltage vector angle and observation angle.

Figure 9 .
Figure 9. DC voltage waveform after given voltage change.