Performance of three-phase three-wire cascaded H-bridge multilevel inverter-based shunt active power filter

Musa Yusup Lada, Mohd Amran Mohd Radzi, Jasronita Jasni, Hashim Hizam, Auzani Jidin, Syahrul Hisham Mohamad 1,5,6 Power Electronic & Drive Laboratory, Center for Robotics and Industrial Automation, Faculty of Electrical Engineering, Universiti Teknikal Malaysia Melaka (UTeM), Malaysia Advanced Lightning, Power and Energy Research (ALPER) Centre, Faculty of Engineering, Universiti Putra Malaysia (UPM), Malaysia


INTRODUCTION
In the state of progression within the power electronic elements within the power system, applications of non-linear load are becoming more and more crucial. As those applications have exponentially increased, this phenomenon will generate a power quality issue in power system caused by the load current that drawing non-sinusoidal current form. The power quality distortion in power system can cause various problems such as high-power losses, extra current flow to neutral line, production of heat and implicit hazardous effect to the sensitive machine [1][2][3][4][5][6]. In the past decade, passive power filters such as series inductance, shunt capacitor and zigzag three-phase transformer are used to mitigate harmonic distortion at power system. However, those filters cannot work efficiently and furthermore they are unable to mitigate completely low frequency harmonics [7][8][9][21][22]. Hence, Active Power Filter (APF) has been introduced to mitigate low and high harmonics distortion in the power system. In conjunction to that, APF can also be used for reactive power compensation, load balancing, voltage regulation and voltage flicker compensation [1,3,5,6,10,24]. Multilevel Inverter (MLI) has gained much intention in high power system application due to its advantages such as produce common mode voltage, draw low distortion input current, operate in low and high switching frequency, and can reduce harmonic distortion using selective harmonic elimination technique. As the trend of increasing in research of renewable energy applications, researchers are progressively developing new topologies and algorithms in designing MLI in which the fruits of the research can also be applied in APF by replacing the conventional six-step three-phase inverter as shown in Figure  1(a). The advantage of using MLI compare to conventional six-step inverter is its capability to produce low harmonic output current which at the same time improving the performance of APF [5,[11][12][13][14][15]. Furthermore, MLI can produce high efficiency for fundamental frequency, is capable to work under high switching frequency, and can reduce voltage stresses across switches, which make it to even perform in medium and high voltage application [5,[11][12][13]15]. There are three common types of MLI namely Cascaded H-Bridge (CHB), Neutral Point Diode Clamped (NPC) and Flying Capacitor (FC) [11,12,14,15]. Among them, the CHB is the most popular due to its advantages such as less power devices requirement, Simple design and possible to produce output voltage level more than twice the number of DC source. Figure 1( Cdc2c S24a S11b S13b S12b S14b S11c S13c S12c S14c

CURRENT CONTROL ALGORITHMS
A current control algorithm is used to generate switching signal for controlling turn ON and OFF switching device. It can be grouped into two major control schemes namely Direct Current Control (DCC) and Indirect Control Current (ICC) algorithms. The DCC algorithm will calculate the current injection reference of APF and compare with the actual current injection of APF as stated in Equation 1. Meanwhile, for ICC algorithm the source current reference needs to be calculated first before comparing it to the actual current source as stated in Equation 2. In term of complexity of the algorithm, ICC is more desirable compared to DCC. ICC uses less mathematical calculation and practically less number of sensors compared to DCC [16][17][18][19].

RESEARCH METHODOLOGY
The effective operation of current control algorithm is depending on instantaneous power theory that used in this simulation. The instantaneous power theory or p-q theory was introduced by Akagi in 1983 and this method uses algebra transformation, which is also known as Clarke transform, to be used in three-phase voltage and current [4,9,10,12,16]. This theory is the most popular harmonic extraction algorithm, which is usually used for three-phase three-wire SAPF. The illustration of this algorithm is shown in Figure 2, where p is instantaneous total energy flow per unit of time and q is energy exchanged between the phases without transferring energy.
The modification of filtering for the instantaneous parameter is done by replacing the conventional Low Pass Filter (LPF) or High Pass Filter (HPF) with a new filtering technique called as Self Turning Filter (STF). This modification will improve the SAPF to work in more decent way of performing filtering in transient or steady state condition of the system [17,18,20,23,25]. The detailed principle of STF is described in Figure 3. In mathematical equation, STF can be elaborated in Equations 3 and 4, where x α and x β are the input signals, ̅̅̅ and ̅̅̅ are the output filtering signals, K is the selectivity parameter, and ω f is the fundamental pulsation. Since, the fundamental frequency is 50 Hz, the value of ω f is set to 100π rad/sec and the value of K is set to 100 in order to decrease the STF selectivity process by mean to extract the fundamental component from the voltage or current signal that was distorted without phase delay and amplitude changing. This is because the smaller value of K will increase the filter selectivity in STF.  Figure 4 shows the modification of DCC algorithm for CHB MLI based SAPF. The load current and source voltage are transformed into αβ axis as stated in Equations 6 and 7 respectively. Since the system only involves three-phase three-wire system, the zero sequence can be neglected.  STF is used to filter αβ axis for current and voltage to suppress harmonic component so that only fundamental component remains for computing instantaneous real power p and instantaneous reactive power q as stated in Equations 8 and 9.

Modification of direct current control algorithm
From Equations 8 and 9, the reference current αβ can be calculated as stated in Equation 10 before transforming it into active filter current reference as stated in Equation 11.
The DCC algorithm produces error current as stated early in Equation 1 and to be compared with the carrier signal. The comparison process used in this system is SPWM bipolar technique which is the error signal will be compared with four different signals that produce different PWM signal to create different levels of output voltage. The SPWM bipolar switching technique is shown in Figure 5.  Figure 6 shows the modification of ICC algorithm for CHB MLI based SAPF. Most of the equations used in DCC algorithm are used in ICC algorithm. The only difference of using ICC over DCC algorithm is it only uses single STF for filtering αβ voltage compare to DCC used two STF for filtering αβ voltage and current. Other than that, ICC used current source for error computing as stated in (1).

SIMULATION RESULT AND ANALYSIS
Simulation model for both algorithms is carried out using MATLAB/Simulink tools. Both of DCC and ICC simulation algorithms use same parameters as listed in Table 1. Figure 7 shows the non-linear voltage and current which are produced by combination of threephase rectifier and RL load as stated early in Table 1. Based on this result, both algorithms produce similar pattern of voltage and current waveforms due to use of same value of non-linear load. The voltage and current at non-linear load also remain the same pattern before and after the APF is connect to the system. The results prove that the APF does not affect or disturb the performance of voltage and current at the non-linear load while sudden injection of the APF current to the system.   Figure 8 shows the three-phase source current, load current and active filter current for both DCC and ICC algorithms. The waveform shows the condition for both of DCC and ICC algorithms before and after connected to the APF. The clear detail of this changing condition shown in Figure 9, where only show the source current, load current, active filter current and active filter voltage at phase A for both DCC and ICC algorithms. The used of STF in this simulation prove that under transient condition which between in transition before and after connected to the APF, immediately the load current been mitigated to produced less harmonics in line current. Line current start to change into the sinusoidal waveform after the APF connected to the system. Since the system used MLI, the active filter voltage shows the staircase seven-level voltage for both algorithm DCC and ICC. The total harmonic distortion load (THD) current spectrums at phase A for both DCC and ICC algorithms are shown in Figure 10. Since both DCC and ICC algorithms use the same value of non-linear load, the THDs of load current at all phases for both algorithms produce the same value of THD, which is 23.89%. Both DCC and ICC algorithms perform well in reducing the THD of load current so that the line current can produce less percentage value of THD and at the same time comply with 5% of THD in IEEE 519 Standard. Figure 11 shows the THD line current spectrums at phase A for both DCC and ICC algorithms. The THD value of line current at phase A for DCC algorithm is 1.00%, and meanwhile for ICC algorithm is 0.53%, which is less than DCC algorithm. The summary of THD values for load current and line current are listed in Table 2. Based on this result, ICC algorithm has a good performance in mitigation compared to DCC algorithm. In term of percentage reduction, DCC algorithm mitigates 95.8% of load current, and meanwhile ICC algorithm mitigates 97.8% of load current, which is higher than DCC algorithm.  The real powers at source and load are shown in Figure 12 for both DCC and ICC algorithms. The result shows that, real power load at source has less ripple percentage as compared to load after the APFs start to perform in the system. This condition also happens to the reactive power at source and load for both DCC and ICC algorithms, as shown in Figure 13.  Figure 14 shows the DC link capacitor voltages for each of capacitors connected to the CHB MLI. The PI controller will make sure that each of DC-link voltages are constantly at 400V and this will ensure that the CHB MLI work as APF by injecting the APF current to system. If the DC link voltages drop below than 400V the CHB MLI will not work as APF.

CONCLUSION
Based on the simulation result, both algorithms have performed well in mitigating the load current so that THD of line current reduces below than permissible value of IEEE 519 Standard, which is 5%. Modification of ICC and DCC algorithms by replacing the conventional low pass filter with STF shows that the line current waveform in each phase managed to immediately improve in transient and stead state condition, where the line current does not take long to become purely sinusoidal current and the waveform in constantly at the same amplitude. The key advantage of ICC algorithm with less calculation involved has made less THD of line current is obtained over DCC algorithm.