Implementation of PV Based Multilevel Inverter to Improve Power Quality using Fuzzy PI and PSO PI Controllers

PV

Due to demand for industries and electronic devices, reactive linear loads and nonlinear loads are increasing. Nonlinear loads which uses switching converters require more reactive current and injects harmonics into grid source. This will affect other loads connected to source and also increases the line losses because of poor power factor. To find solutions for these Researchers are always researching on power quality issues such system imbalance, reactive current, load balancing, and harmonics distortion [1]. Initially passive filters are considered for harmonics reduction and for reactive power support. But passive filters are not dynamic and size cost and rating of filters should be increased with increase in power demand. Efficiency and performance of the system will be decreased with increase in amount of harmonics and amount of reactive current in the line [2]. STATCOM which can be connected in shunt at PCC, will improve transient stability and also provides reactive and active power demand and voltage support. STATCOM connected at distribution line works as active power filters by reducing harmonics and by improving power factor. By connecting renewable energy sources like PV at DC side of the compensator, required amount of reactive power can be injected or absorbed [3].
For high power rating distribution systems, compensators with two or three level inverters for photovoltaic systems are not suitable. The cascaded H Bridge converter for PVSTATCOM is new configuration used which improves dynamic compensation and can capable of overload. Comparing with H Bridge converters with cascaded diode clamped and flying capacitor multilevel inverters require less components [4]. Isolated DC sources are required for each H Bridge, which can be fulfilled by PV array strings. Due to independent DC links to each H Bridge, individual regulation of dc voltage is possible. Generating maximum power from each PV cell can be achieved by using buck boost converter at DC side of each H Bridge. Buck boost converter will regulate output voltage by transforming maximum and constant output power from PV arrays [5].
Synchronous reference frame theory (SRF) can be used to generate switching pulses for STATCOM inverter to compensate the reactive power and to reduce current harmonics [6]. Cascaded voltage and current strategy is proposed in [7] for interface converter. Sequence components control scheme proposed in [8] for CHB based STATCOM. Even when voltage of grid is imbalanced, the voltage of DC link can be adjusted by keeping an eye on the sequence components [9]. However main challenges associated with cascaded H bridge-based PV STATCOM are dynamic stability, balanced and regulated DC voltages and power mismatches of PV array. PV power Website: www.ijeer.forexjournal.co.in Implementation of PV Based Multilevel Inverter mismatches lead to unbalanced power from three phase CHB inverter which will make grid current to be unbalanced. This can be rectified by modifying modulation index using unbalanced factor of power in three phases [10].   Depending on the switching pulses applied to H Bridge, it generates three output voltage levels− , 0, + . For series connected modules, 2 + 1 voltage levels can be generated. As, for better harmonic compensation and for smooth current waveform, seven level is considered in this paper. Because of increased levels in voltage waveform, synthesized current is free of harmonics and requirement of more rated output filters can be reduced. Due to increased number of levels voltage stress on each switch can be reduced and hence efficiency will be improved compared to other topologies. For seven level converter 3 H Bridge modules should be connected in series for each phase and connected to PCC through filter. By applying KVL to one phase of the compensator Here is voltage output came by th module, is the filter inductance, is the resistance of the filter, is number of H bridges, is voltage grid and is current in grid.
Where the capacitance of filter is, is the voltage across individual capacitors and is current flowing into ℎ capacitor. Power on the both side of the converter are equivalent when switching losses are ignored.

░ 3. PROPOSED CONTROL METHOD
For Cascaded H Bridge seven level inverter, an advanced control method should be used to improve grid currents quality and to compensate reactive power. Current sine wave which is distorted due to harmonics injected by non-linear loads can be corrected or improved by injecting harmonics current in negative polarity with grid currents. Improvement of current wave also leads to reduction of distortions in voltage waveform. Reactive power burden on grid due to inductive loads can be reduced by injecting reactive current required by load into PCC by seven level H Bridge inverter. Hence, this will improve the power factor at source side and will reduce losses withstanding by other loads. These are the objectives to be consider while choosing a control method. Hence, an advanced SRF based control method is used and schematic is shown in figure 3.

SRF Control
DQ control or SRF control which is applied to H bridge reactive power compensation and inverter for harmonics is having two loops. One is DC outer voltage regulation loop and another one is current inner control loop. Outer voltage loop regulates average DC voltage of 9 H Bridge inverters and inner current loop is compensating current harmonics and reactive power.
Three phase voltages and current are transformed in x to dq0 components using parks transformation. Where is either voltage or current and is phase of grid voltage.

Implementation of PV Based Multilevel Inverter
Reference value for direct axis current is determined from PI controller which regulates average DC voltage of H bridges. Average of 9 H Bridges DC voltages is compared to voltage reference and PI controller is used to reduce tracking error and to improve dynamic behavior. In this paper, two algorithms fuzzy PID and PSO PID are considered for tuning of the PI controller. These two algorithms are explained in next sections.
can be set as zero and in case of reactive power compensation reactive current can be calculated using load reactive current requirement. and are compared to actual and using PI controllers which are tuned using trial and error method. Output of PI controllers are modulation indices , in DQ frame and these are transformed into three phase , , using inverse parks transformation.
In addition, with average DC voltage controlling, individual voltage control is required for each H Bridge to make PV module works at maximum power point. Consider phase A, two separate voltage loops with PI controller is used to generate modulation index proportions ( 2,3) . These values can be multiplied with modulation index of phase A( ) to generate two indices M 2, 3 for second and third H bridges. M 1 can be generated by subtracting the sum of M 2, 3 from . By using phase shifted SPWM (PSSPWM) pulses are generated for seven level output voltage [15][16][17][18]. Same process of pulse generation can be applied to phase B and phase C.

Modulation Correction
Due to atmospheric conditions PV arrays connected H bridges may not generate equal powers which lead to severe problems in H Bridge cascaded multilevel inverter. There is a possibility of unbalanced currents into the grid because of power mismatches in PV array connected to each H bridge module. To rectify this problem, modulation index can be corrected to make output phase voltage to be in proportional to the unbalanced power and thereby make currents to be balanced. Modification of modulation index imposes zero sequence voltage into each phase to make currents to be balanced.
Weighted ratio of unbalanced power is calculated as Where is the power input of phase ( = , , ), average power of three phases.
Then the modulation index correction to impose zero sequence voltage is given as Where , , are the modulation indices determined by current loop controller.
After the correction modulation index of each phase can be updated as

. PV ARRAY MODEL
Power generation of PV cell and its operation characteristics are affected by many environmental factors. Solar irradiance G, in W/m 2 , temperature t in O C are two main factors to be considered while defining of PV characteristics of cell. Characteristics I-V and P-V of PV system varies with the intensity of solar irradiation and cell temperature. To get required output power and voltage solar photovoltaic cells are connected in parallel and series referred ad PV modules.
Single diode model shown in figure 4 is mostly used to describe characteristics of PV cell. Parallel shunt resistance ℎ consider leakage currents of cell and series resistance consider shading effects of PV cell.
Where ℎ is photocurrent, is the saturation cell current, is an electron charge (1.6 10 −19 ), k is a Boltzmann's constant (1.38 10 −23 ⁄ ), is the cell's working temperature, is an ideal factor, ℎ is a shunt resistance and is a series resistance. The photocurrent of cell depends on solar insolation, working temperature of cell, given as Where is the reverse saturation current under standard sun illumination & temperature.
is the band gap energy of semiconductor used in the cell. The sun insolation and the cell's working temperature, which is stated as, determine the photo current Iph primarily.

International Journal of Electrical and Electronics Research (IJEER)
Open Access | Rapid and quality publishing Research Article | Volume 11, Issue 2 | Pages 378-388 | e-ISSN: 2347-470X Where T_ref is the reference temperature for the cell, G is the solar insolation in kW/m2, I_SC is the cell's short-circuit current at 25 o C and 1 kW/m2, K_I is the temperature coefficient of short circuit current, and T_ref is the cell's shortcircuit current. Typically, a PV cell produce very low voltage and power, When PV cells are arranged in series and parallel to provide the necessary voltage and power, the PV array's current and voltage equation becomes is the number of parallel PV cells and is the number of series PV cells.
Output I-V and P-V characteristics of PV array are shown in figure 5-8. Figure 5 and 6 showing the variation of current and power respectively under variable solar irradiance and constant temperature. Figure 7 and 8 showing the variation under variable temperature and constant irradiation. From these it is clear that variation of current and hence power with respect to voltage is nonlinear and varying with solar irradiation and temperature. Even during variations in atmospheric conditions, it is necessary to provide maximum power to the load and consequently, solar PV systems should adopt MPPT.

░ 5. FUZZY TUNED ADAPTIVE PI CONTROLLER
For controlling linear systems, conventional PI controller is simple and efficient. But for nonlinear systems such as PV based distribution systems in which control performance deteriorate with working conditions of the system normal PI controller is not effective. PI control with Variable parameters in proportional with working conditions can overcome this problem. Based on variable parameters fuzzy based PI controller can be designed. The detailed structure of the controller in figure 9.
The equation of the adaptive PID fuzzy is ( ) is the control action, , are proportional and integral gains. ( ) is the tracking error signal.
For fuzzy tuned PI controller input are error and the change of error , given as And the outputs of the fuzzy controller are variation in proportion and integral gain. Process of Fuzzy controller includes: fuzzification, rule base and defuzzification.

International Journal of Electrical and Electronics Research (IJEER)
Open Access | Rapid and quality publishing Research Article | Volume 11, Issue 2 | Pages 378-388 | e-ISSN: 2347-470X Website: www.ijeer.forexjournal.co.in Implementation of PV Based Multilevel Inverter    In defuzzification output of the fuzzy controller can be determined by weighted mean. Then from the fuzzy output proportional and integral gains are calculated as: is the initial value of the proportion gain, is the initial value of the integral gain. 1 2 are constants.

░ 6. PARTICLE SWARM OPTIMIZATION
Research communities and industries are increasing their attention on optimization algorithms in the past few years. To find the maximum or the minimum of a function with certain constraints optimization algorithm can be used. In area of designing a fitness function, multimodal optimization technique, parameter control methods computational intelligence which is successor of artificial intelligence is having major role. Compared with traditional optimization methods computational intelligence can find optimum values even in complicated optimization problems [19][20][21].

Implementation of PV Based Multilevel Inverter
One of the efficient optimization techniques in computational intelligence, which is based on swarm intelligence, is Particle Swarm Optimization (PSO). Animal behaviors and their social interaction such as bird flocking and fish schooling are the main motivation for this method. Competition and cooperation among the entire population of birds to find food is used to find optimal solution in PSO. Individual parameters of a swarm represent different possible set of the unknown values to be optimized. Population of some random solutions are set as initialization values of swarm. A multi-dimensional search space and particles flying around that space by adjusting their position and velocity can be defined in PSO systems. The main goal of PSO algorithm is to effectively search the solution space by particle swarming into the best-fitting solution obtained in earlier rounds with the goal of running into better solutions along the way and ultimately settling on a single minimal or maximum solution. Efficiency to find solution to the fitness function will define the performance of each particle to find solution.
In this paper, PSO algorithm is used to obtain optimal PI controller gains for a high-performance DC voltage regulation. Proportionality constant and Integral constant are tuned using possible controller setting of particles to reduce the error function as much as possible. Integral Time of Absolute Error is the error function employed here. In PSO algorithm, particles which are population of random solutions can be assigned to a system as initialization and each solution of these particles is also assigned a randomized velocity. This algorithm depends on information exchange between particles or random solutions. The trajectory of each particle into best solution for fitness function can be adjusted by itself. Trajectory of each particle can be modifying previous position attained by any nearby particle.
The performance of each particle can be evaluated by fitness function by observing whether the best fitting solution is achieved. During the process of algorithm, for every iteration each particle and fitness of the best individual particle improves and moves towards solution and hence to the end of the run [22][23][24][25].
At first in PSO algorithm swarm with no. of individuals called as particles are initialized. Every th particle in swarm holds the information of its position occupied , its moving velocity , the best position which is link with the best fitness value achieved by a particle far and the global best position which is link with best fitness value among all the particles .
In our application positions of particles represents the gains of PI controller (Kp, Ki). The position of each particle defines fitness of a particle. Particle closure to the solution will have higher fitness value then the particle which far away. For every iteration every particle position and velocity are updated to get better fitness by using the equation: For = 1,2, … , where is the number of particles, = (1,2, … , ) as is number of dimensions, is the iteration number, is the inertia weight, 1 and 2 are 2 random numbers distributed uniformly in the range [0,1], 1 and 2 are the acceleration factors. 1 is cognitive acceleration constant, forces the particle to move into the position to get the more fitness and 2 is social acceleration constant forces the particle towards the particle that currently has highest fitness.
Velocity of particle should be bounded between chosen limits < < and particle position is also bounded as < < .The personal best of each particle can be updated as Then the global best of swarm can be updated as: is a function to evaluate fitness value of given position.
This process of PSO can be repeated iteratively until maximum iterations are reached or no improvement over a number of iterations. PSO algorithm Flow chart is shown in figure 12.
To  These indices will define the minimization of error signal given as the input to PI controller. In this work ITAE is used as fitness function for PSO to tune the PI gains.

░ 7. SIMULATION RESULTS
To exhibit the performance of proposed control strategy the system shown in figure 13 is simulated using MATLAB/SIMULINK using tools simscape, Fuzzy logic and PSO. System Parameters used in simulation are given in  figure  14. Seven level CHB multilevel converter is shown in figure 15.

░ 8. EXPERIMENTAL RESULTS
The proposed method is done in simulation and also an experimental prototype is implemented to confirm the simulation results which is as shown in figure 18, control signals are shown in figure 19 and figure 20 and output voltage in figure 21. Sinusoidal pulse width modulation is used to generate control signals for switching power switches. The output voltage of inverter is successfully produced and filtered to get sine wave which results lesser total harmonic distortion and results meets with simulation results. To produce modulation indices for H Bridge switching, synchronised reference frame theory is modified. To rectify power mismatches in PV arrays, which leads to unbalanced grid currents, modulation indices are modified by adding a value called modulation compensation. H bridges' average DC voltage is managed by a PI controller, and controller gains are fine-tuned using the trial and error, fuzzy, and PSO methods. Simulation results are presented by using these three tuning methods. In dc voltage with PSO based PI controller tracking error, peak overshoot and percentage of ripples are very less compared to other two methods. The results of inverter output are experimentally confirmed and hence CHB seven level inverter-based PV STATCOM with PSO tuned PI controller is efficiently improves the power quality by increasing quality of current waveform at source side of the line.