A dummy peak elimination based MPPT technique for PV generation under partial shading condition

Correspondence Boni Satya Varun Sai, Department of Electrical Engineering, Jadavpur University, 188, Raja S.C. Mullick Road, Jadavpur, Kolkata 700032, India. Email: satyavarunsai1234@gmail.com Abstract The increased energy demand with recent advancement in technology has led the global energy producers to move towards solar energy. Partial shading is a phenomenon that can severely affect the production of the solar power in the photovoltaic system. With partial shading situation, the cascaded solar panels show numerous dummy peaks both in power– voltage (P–V) and current–voltage (I–V) curves, resulting in difficulty of extracting maximum power. A dummy peak elimination based MPPT technique is introduced in proposed work for accurate tracking of maximum power. The proposed dummy peak elimination based MPPT estimates the dummy peaks without imposing any high computational burden for the processor. The proposed technique has better dynamic response and is more compatible for real time application compared to existing optimization-based techniques. The proposed dummy peak elimination based MPPT is tested on a practical system with 3S configuration having three 249 W PV panels for different shading patterns, which displays acceptable outputs with good dynamic performance and better efficiency.


INTRODUCTION
Currently, the expanded interest of energy has prompted huge exploitation of fossil fuels. This leads to significant issues like exhaustion of fossil fuels, unsteady energy cost along with extreme natural contamination. This has lead analysts around the world towards environmentally friendly generation systems. Photovoltaic (PV) energy is one of the most focused and well-established renewable energy sources among all others due to its hefty availability and relatively simpler control strategies [1]. With high establishment cost till date, the activity of solar modules under MPPT condition is essential for greatest utilization of accessible power. The adaptability in development and precision in delivering performance characteristics of solar cell encourages the single diode model as the most utilized one for photovoltaic cell construction [2]. Because, the power generated by single solar panel is lower, a noteworthy number of PV panels are to be associated either in series or parallel configurations etc., for any PV power generation system. In etc., prevent the availability of uniform irradiance everywhere throughout all the panels. Since, uniform irradiation results in a single peak for the overall configuration, tracking of maximum power is easier [3,4]. On the other hand, the partial shading (PS) situation leads to multiple peaks both in the power-voltage (P-V) and current-voltage (I-V) characteristics of the cascaded PV panels, resulting in difficulty to track maximum power. Boost converter is chosen commonly as an intermediate power conditioner in PV system due to its simple structure, cost compatibility and flexible control schemes with noninvertible output voltage [5,6]. Under PS case, the conventional MPPT schemes results in lesser efficiency. The mostly used traditional MPPT technique is Perturb and Observe (P&O) algorithm, which has a limitation in identifying global maximum point under PS condition. Similarly, other conventional MPPT schemes like hill climbing (HC) method, incremental conductance (IC) method etc., are also inefficient in tracking power at global maximum under PS condition [7]. For effective tracking of global maximum under shading case, numerous optimization-based algorithms are introduced to improve efficiency, reduction of cost and complexity [8,9]. Execution of these optimization-algorithms is subjected to numerous adjustable parameters for example, population size, the number of iterations, acceleration and tuning off parameters etc., [10]. Under PS condition, global maximum tracking based on artificial vision is presented recently, in which web cam is utilized to recognize the shading pattern. Even though this method shows improved characteristics, the implementation cost can be high due to high end hardware requirement with higher computational compatibility [11]. Beside this, it is also mentioned that, low-cost camera can be used for tracking purpose, but it may lead to precision error. Maximum power point scanning (MPPS) technique is the one wherein group selection optimizer along with P&O is utilized for tracking global maximum, where the research to be carried out in controlled voltage source for improved performance [12]. Fusion firefly algorithm (FFA) is able to show the improved MPPT characteristics with moderate difficulty in the algorithm [13]. Using optical isolation mechanism, PS MPPT can be successfully implemented with low cost and fast tracking [14]. An artificial neural network-based system is also developed for shading pattern detection to generate particular reference voltage to track global peak power, but precision initialization routine is necessary for successful hardware implementation [15].
Particle swarm optimization (PSO) and cuckoo search (CS) algorithms are the most normally employed optimization techniques in tracking global maximum under shading environment [16,17]. The CS search algorithm displays better performance characteristics compared to PSO [18,19]. In [20], an enhanced adaptive P&O (EA-P&O) MPPT is proposed where the scanning is done to measure the region of convergence for global maximum. This work avoids the inability of P&O in detecting MPP under PS scenario. But the scanning takes considerable time, resulting in inferior dynamic response of system. In [21], an accurate MPPT to detect PS occurrence is determined in which, based on the shading pattern detection, it is observed whether the system is under PS condition or uniform irradiation condition. If the PS condition occurs PSO comes into picture, while for uniform irradiation P&O is considered. This leads to high complexity with poor dynamic response of the system for practical implementation under PS condition. Moreover, recent studies reveal that, MPPT techniques with low dynamic response employed in the PV systems are one of the major reasons behind the introduction of interharmonics into the system. So, a MPPT scheme with good tracking efficiency, easy real time implementation and good dynamics is strongly required [22]. However, all these schemes discussed have the major difficulty of high computational requirement for the processor with low dynamic performance, which restricts these to be efficiently used for MPPT schemes under PS condition.
In this paper, a dummy peak elimination (DPE) based MPPT scheme is proposed which removes the disadvantages of the existing schemes. In the proposed scheme, all the peaks including the peak very close to global maxima for the combined P-V characteristic under PS condition are calculated based on the measured irradiance on each panel. The proposed algorithm uses P&O where the dummy peaks are eliminated and the peak nearest to the required global maxima becomes the initial search point for P&O. The initial estimation of the peak near the global maxima avoids the tendency of P&O to stick to any of the local peaks under PS condition. The proposed scheme uses simple control structure compared to any of the comparable existing schemes as it uses P&O along with dummy peak elimination algorithm. The proposed MPPT has better dynamic properties compared to existing schemes as the initial point is very near to the global maxima. Moreover, it is highly suitable for real time implementation with existing drive compatible hardware. The proposed algorithm is practically studied for 3S configuration under different shading patterns to justify the proposed theoretical concept.
The proposed dummy peak elimination based MPPT technique can be implemented for series configuration with more number of parallel strings with different irradiations, but the approach is different, which will make the understanding of situation complex and can be a different work. Moreover, there are few irradiation sensors based MPPT schemes are available in research platform primarily handling series connected schemes. Out of which, the recent publication [23], deals with series configuration. By keeping the research demand and developing improves methodologies, the present research article detailed more about series connected modules. The main advantageous features of the proposed DPE MPPT algorithm compared to available MPPT techniques are, (i) accurate tracking; (ii) good dynamic properties with low search time; (iii) suitability for real time implementation with hardware having moderate computational facility; (iv) economic viability with no requirement of high-end processing units including scanners etc.
The present research work is articulated as solar module model in Section 2. The proposed peak detection technique in Section 3, the dummy peak elimination based MPPT technique in Section 4, simulation results and analysis in Section 5, experimental results and discussion in Section 6 and conclusion in Section 7.

SOLAR MODULE MODEL
Among the available designs for PV modelling, single diode model is prevalent due to its simple construction and precision in generating performance characteristics at par with practical PV cell [2]. Figure 1 represents the PV cell design using single diode model where I p is photo current, d is diode, R se is series resistance, R pa is parallel resistance, I so solar panel output current and V so is solar panel voltage. The relation between I so and V so from Figure 1 can be formulated as, where A is curve fitting factor, k is Boltzmann constant, q is electron charge, I o is saturation current of diode and T is temperature. It is established that, short circuit current (I s ) and open of solar cell are the functions of irradiation and temperature which can be represented as [1], where is temperature coefficient of V op and a is irradiation correction factor of V op . Table 1 represents the manufacturer data for different parameters of the PV panel used in the proposed work for simulation and experimentation. These parameters are utilized to develop single diode model in MATLAB/Simulink environment. Using single diode modelling, the output characteristics developed for the solar panel for different irradiation levels are shown in Figure 2(a,b). By using these curves, maximum power (P m ), maxim power point (MPP) voltages and currents V m and I m , open circuit voltage (V op ) and short circuit current (I s ) for

THE PROPOSED PEAK DETECTION TECHNIQUE
The partial shading scenario in the proposed work mainly focuses on the method of [23], where the panel is experiencing uniform irradiance throughout the module. The practical consideration of partial shading and real time situations are explained in this section.

Dummy peak detection using P-V and I-V relationship between individual and cascaded panel
The relation between I-V and P-V characteristics between individual and cascaded panels can be established by considering different shading patterns of individual panels under 3S configurations as shown in Figure 3. The different combinations of the shading patterns can result in three peaks, two peaks or single peak in the combined P-V characteristic of the 3S configuration. For the first case Pa 1 , the irradiation levels of individual panels considered are 1000, 800 and 600 W/m 2 respectively, which can result in three peaks in the overall P-V curve. Similarly, for the case Pa 2 , the considered irradiation levels are 1000, 200 and 200 W/m 2 where two peaks will appear. On the other hand, for Pa 3 , the irradiations for all the three panels are kept at 1000 W/m 2 and the overall P-V curve will show a single peak. For the considered shading patterns, the simulated P-V and I-V characteristics plots are displayed for the 3S configuration in   It can be observed from Figure 4(a) that, P-V and I-V curves display three peaks. The three panels are at irradiation levels of 1000, 800 and 600 W/m 2 respectively. The first peak occurs at V 1 = 28.22 V, P 1 = 235.78 W and I 1 = 8.36 A, the second peak occurs at V 2 = 60.64 V, P 2 = 415.04 W and I 2 = 6.84 A and the third peak occurs at V 3 = 95.01 V, P 3 = 492.67 W and I 3 = 5.19 A. From Table 2, V m and I m for 1000 W/m 2 are 30.01 V and 8.3 A, for 800 W/m 2 the values are 29.97 V and 6.69 A and for 600 W/m 2 they are 30.44 V and 4.95 A, respectively. As per the proposed methodology the peaks P 1 and P 2 can be considered as dummy peaks. So, by intently noticing these quantities, the relations for voltages and currents at different power peaks for cascaded and individual panels are developed as, Similarly, Here, V 1 , V 2 and V 3 are voltages and I 1 , I 2 and I 3 are the currents at the three peaks respectively. The numbers at the suffixes of Equations (4)-(9) denotes the irradiation levels.
In the same manner, from Figure 4(b), the P-V curve displays two peaks. In this, the first peak is at V 1 = 28.20 V, P 1 = 235.69 W and I 1 = 8.36 A. The second peak appears at V 2 = 59.19 V, which is close to sum of V m at 1000 and 200 W/m 2 and I 2 = 1.68 A which is close to I m at 200 W/m 2 as observed in Table 2. Thus, the power at second peak is, P 2 = V 2 × I 2 = 99.44 W. The third peak is at V 3 = 94.94 V, P 3 = 158.37 W and I 3 = 1.67 A. As the concept of dummy peak is developed on the basis of the local peak which appears on the combined P-V plot, the peaks P 2 and P 3 can be designated as dummy peaks. The dummy peak can fall on the same slope and may not be visible in the combined P-V characteristic of cascaded configuration. On the other hand, the dummy peaks can be also visible in the combined P-V plot. In Figure 4(b), the second and the third peaks are dummy peaks. The second peak at V 2 = 59.19 V falls on the same slope with third peak at lower voltage, but at nearly same current of 1.68 A. On the other hand, here first and third peaks are visible, of which third one is a dummy peak. In this case also, the relation for voltages and currents at peaks for individual and cascaded panels can be developed as, Similarly, Similarly, from Figure 4(c), the combined P-V curve displays single visible peak. Here, both the two dummy peaks appear on the same slope as that of the third peak. The first peak is at  Table 2. The power at this peak is P 1 = V 1 × I 1 = 249.1 W. The second peak occurs at V 2 = 60.02 V which is double to that of the V m at 1000 W/m 2 , I 2 = 8.3 A which is I m at 1000 W/m 2 and P 2 = V 2 × I 2 = 498.2 W. The third peak occurs at V 3 = 90.25 V, P 3 = 747.23 W and I 3 = 8.28 A. So, by intently noticing these quantities, the relationships for voltages and currents at peaks are developed as, Similarly, In the Equations (10)-(21), the subscript numbers in voltage and current symbols indicate the irradiation level.  By employing Equations (4)-(9), V 1 = 30.01 V, V 2 = 59.98 V, V 3 = 90.42 V, I 1 = 8.3 A, I 2 = 6.69 A and I 3 = 4.95 A. Therefore, power at first peak is P 1 = V 1 × I 1 = 249.1 W, similarly power at second peak is P 2 = V 2 × I 2 = 401.27 W and power at third peak is P 3 = V 3 × I 3 = 447.58 W. From this, it can be detected that P 1 and P 2 are dummy peaks and the maximum power is near third peak at V 3 = 90.42 V. So, if P&O is allowed to track from the voltage V 3 = 90.42 V, the maximum power can be reached with very low search period.

Proposed global peak detection
Similarly, for Pa 2 , the individual panels V m and I m values from Table 2  Therefore, approximate power at first peak is P 1 = V 1 × I 1 = 249.1 W, at second peak is P 2 = V 2 × I 2 = 99.44 W and at third peak is P 3 = V 3 × I 3 = 148.46 W. From this, it can be observed that the second and third peaks are dummy and maximum power is near to voltage 30.01 V.
In the same manner for Pa 3 , the individual panels V m and I m values from Table 2  Therefore, power at first peak is P 1 = V 1 × I 1 = 249.1 W, at second peak is P 2 = V 2 × I 2 = 498.2 W, at third peak is P 3 = V 3 × I 3 = 747.3 W. From this, it can be observed that first two are dummy peaks and maximum power is near to voltage 90.03 V. Table 3 represents the comparison between actual V m from model and calculated V m through the proposed technique. Inspection of Table 3 shows close conformity of V m obtained through proposed calculation with those from actual model.

Practical implementation of proposed system
It is assumed earlier that solar panels experience uniform shading throughout the panel as considered in [23]. The practical implementation of the same can be carried out as shown in Figure 5, where partial shading within a panel is considered. In the proposed work, 3S and 4S configurations are considered where one of the panels is experiencing the shading. The effective irradiation for any panel can be calculated after averaging the irradiations obtained by adjacent sensors to the panel.
For the proposed zig-zag type of sensor placement as shown in Figure 5, the irradiation for individual panels can be calculated for 3S configuration as, Similarly, for four panels, the irradiation for each panel is calculated as, Thus, in this case, the required number of irradiation sensors is N + 1, with N being the number of panels. However, accuracy can be further improved with increased number of sensors effective per panel. Since the area of single panel is less, the considered configuration of N + 1 sensor will give FIGURE 6 P-V curve for 50-50% 3S configuration fairly accurate results. The accuracy can be further improved with a larger number of series connected panels. In the partial shading pattern three cases considered, where in one case 50% is shaded and 50% unshaded, in second case it is considered that 33.7% is unshaded and 66.7% shaded and for third cases it is 33.7% is shaded and 66.7% unshaded. For the case of, partial shading within the panel for 3S-configuration, of 50-50% shading with irradiation levels of G 11 = 150 W/m 2 , G 12 = 250 W/m 2 , G 21 = 350 W/m 2 and G 22 = 750 W/m 2 , P-V curve is represented in Figure 6.
In general, the irradiance sensed by the sensors are almost same except for the shaded panel. For the considered practical cases, the maximum power point voltage (V m ) and maximum power (P m ) for partially shaded and unshaded configurations are represented in Tables 4 and 5. It can be observed that, calculated V m with proposed averaging technique for partial shading within the panel almost nears the actual value.

Proposed system
The general schematic structure for the proposed work is represented in Figure 7, where the boost converter is utilized in progressing PV panel voltage (V so ) in the degree of system requirement. The PV modules with 3S configuration are under the influence of irradiation levels of G 1 , G 2 and G 3 at temperature (T). The boost converter circuit constraints such as, inductance (L P ) and capacitance (C P , C o ) are formulated according to the system condition.

Flow chart for proposed system
The schematic diagram for proposed DPE MPPT method is signified in Figure 8. To explain the overall theme of proposed  system, a pattern can be considered in which irradiation levels for the three panels under 3S configuration are G 1 , G 2 and G 3 respectively. The overall flowchart can be explained with following steps.
Step 1: Sensing parameters For the proposed system, irradiation and temperature sensors are required for sensing the operating irradiation levels of individual panels and temperature of array for finding pattern change and dummy peak of the combinations under PS condition. Voltage and current sensors for the overall configuration are employed for tracking power using conventional P&O MPPT.
Step 2: Dummy peak detection In this step, based on the proposed concept developed in the preceding section, the sensed irradiation levels are arranged in decreasing order as G 1 ≥ G 2 ≥ G 3 . By using the available irradiation levels and array temperature, I s and V op are calculated using Equations (2) and (3). It can be observed from Table 2 that the ratio of V m to V op is almost same for all considered irradiation levels. It can also be observed that the ratio of I m to I s are very close for all studied irradiation cases. The two ratios of V m to V op and I m to I s can be defined as K 1 and K 2 where, Thus, based on the calculated I s and V op for each panel through Equations (2) and (3), the individual panel voltages and currents V m and I m corresponding to maximum power for measured irradiation can be calculated as Using the calculated values of V m and I m through Equations (31)-(36) for individual panels with measured shading patterns, the voltage and current values at every peak for the considered 3S configuration can be calculated as, Using Equations (37)-(42), the peak powers calculated as, Comparing Equations (43)-(45), the maximum power can be known and the respective voltage for that power can be taken as reference (V p ) for P&O MPPT algorithm for tracking global maximum power. The process is same for any possible pattern.

Loop operation
As shown in Figure 8, the proposed algorithm can be split into three separate loops. The first loop consists of standard P&O algorithm, the second loop is for pattern detection and the third loop is for dummy peak elimination. The maximum power tracking can be done initially and if there is a change in shading pattern, the same can be detected by employing loop 2 and it can finally activate loop 3 for dummy peak elimination by maximum peak point detection. The respective maximum peak point is used to set the reference voltage (V p ) from where the P&O starts its perturbation to track maximum power.
In proposed system, P&O with PI controller is employed to obtain the required dynamic characteristics with low settling time and zero steady state error of the system. Moreover, the initial point is estimated near to the global maximum, the time required for tracking maximum power is reduced. PI controller gain values are obtained using trial and error method [24].

SIMULATION RESULTS AND ANALYSIS
The simulation is carried out for similar situations given in [23], where each solar panel experiences a different irradiation with uniform isolation over panel surface. For the proposed system, simulation is carried out in MATLAB Simulink atmosphere for three following cases, (i) Single pattern; (ii) Single step change in patterns; (iii) Multiple step change in patterns.

Single pattern
The term single pattern corresponds to any of the fixed pattern of shading for example, Pa 1 , Pa 2 or Pa 3 as indicated in Figure 3. The respective PV panel power obtained for each considered patterns by employing the proposed DPE MPPT. The simulation results for maximum power tracking for each of the patterns Pa 1 , Pa 2 and Pa 3 are shown in Figure 9(a-c) correspondingly, showing that the proposed MPPT can successfully track the maximum power for all the considered patterns.

Single step change in patterns
In the second case of simulations with considered shading patterns in Figure 3, a single step change between the patterns is applied and the results are observed. The step changes between the patterns considered are Pa 3 -Pa 2 , Pa 1 -Pa 2 and Pa 3 -Pa 1 . However, any other possible step changes also can be applied for test purpose. The step change is applied at time t = 0.5 s in the simulation for all the cases, where the loop 2 activates loop 3 in finding the global peak after eliminating the dummy peaks, which helps the P&O in tracking global maximum power. The results are displayed in Figure 10 (a-c) respectively for the three step changes applied. From Figure 10, it can be observed that the proposed DPE MPPT scheme successfully tracks the PV panel power under the considered varying pattern condition with single step change. The convergence time is compared with [20] and shown in Table 7. Moreover, from Figure 10, it is noticed that the convergence time for the DPE MPPT method is below 10 ms compared to much larger time in [15] and 250 ms for [20]. Thus, it can be concluded that the proposed DPE MPPT works faster than the existing MPPT schemes. In proposed MPPT scheme the tracking is smoother with considerable reduction of searching period for pattern to pattern variations.

Multiple step change in patterns
In the next stage of simulation with the proposed system, an arbitrary multiple step change in patterns is applied for example, Pa 2 -Pa 3 -Pa 1 -Pa 3 -Pa 2 to test the dynamic performance. In this case, any other possible pattern can be applied for testing purpose. The step change is applied at each 0.5 s time interval. The tracked power for considered step changes is shown in Figure 11. From this, it can be observed that, with the multiple variations in patterns, the proposed MPPT can perform smooth tracking with good dynamic behaviour.

EXPERIMENTAL RESULTS AND DISCUSSION
The experimental verification is performed with three solar panels connected in 3S configuration along with anti-parallel diode protection as shown in Figure 12. In the proposed scheme, artificial insolation is created using incandescent lamps in the test room, where each solar panel is employed with incandescent lamp as shown in Figure 12. Moreover, the lamps can be switched on and off instantly and by this way the solar panels tend to operate with sudden changes in irradiation.
The system can supply generated power to load through boost converter. The PV panel current and voltage are driven as input to PIC 18F452 controller using sensors. LM35 is used for sensing temperature in which the temperature is converted to equivalent voltage. Pyranometer is used for sensing irradiance of each panel, as an equivalent voltage with precision. These input signals are passed through suitable low pass filters to remove the  Figure 12.
The red line indicates the flow of inputs given to PIC controller which is programmed using a programmer with PC interface. The output pulses generated by PIC are in the range of 5 V with low driving capability with no isolation. Thus, in order to enhance the voltage to 12 V, the driver circuit with TLP250H with isolation is employed. The pulses are given to gate of the power MOSFET K2611 of the boost converter.
For practical validation of the proposed concept, initially a step change in shading patterns for example, Pa 3 -Pa 2 , Pa 1 -Pa 2 and Pa 3 -Pa 1 are applied. The proposed MPPT has been employed and the results are displayed in Figure 13 (a-c). Similarly, for the case of, partial shading within the panel for 3Sconfiguration, of 50-50% shading, with irradiation levels of G 11 = 150 W/m 2 , G 12 = 250 W/m 2 , G 21 = 350 W/m 2 and G 22 = 750 W/m 2 , the tracked maximum power 143.29 W is shown in Figure 14, where the actual power is 144.62 W and same is represented in Table 6. From this, it can be observed that the proposed MPPT can efficiently track the maximum power with good dynamic behaviour and low convergence time.
The proposed dummy peak elimination based MPPT technique, is based on perturb and observe algorithm for tracking. It is known fact that, P&O efficiency depends on steady state oscillation and drift due to change in irradiation. Since, the initial point is estimated near to the global maximum, the time required for tracking maximum power is reduced. In proposed system, PI controller is employed to make the steady state error near to zero. Moreover, the P&O is assisted with irradiation sensors, making it drift free with improved efficiency. So,  the P&O shows the similar tracking properties as of [25], with a major advantage of tracking global maximum under partial shading condition. For the considered scenarios the experimentally obtained efficiency for different irradiations on panel and for partial shading with in panel case is represented in Table 6. Table 7 represents the comparison between the existing EA-P&O [20], the recent method described in [15] and with proposed DPE based MPPT. Even though the proposed MPPT requires irradiation sensors unlike the other two, it has the major