Optimization of PV Systems Using ANN-PSO
Configuration Technique Under Different
Weather Conditions
Adedayo M. Farayola, Yanxia Sun, Ahmed Ali
Department of Electrical & Electronic Engineering Sciences, University of Johannesburg, South Africa Email: [email protected], [email protected], [email protected]
Abstract--- Conventional MPPT techniques like Perturb&observe perform ineffective under partial shading condition due to its inability to effectively track the global maximum power point (GMPP). Particle swarm optimization (PSO) technique is a recent meta-heuristic MPPT technique commonly used to extract maximum power from PV systems but takes time to iteratively locate the GMPP. This paper presents a novel use of hybrid ANN-PSO technique implemented using series-connected distributive MPPT configuration approach. The results of ANN-PSO distributive MPPT, PSO, and Perturb&observe (P&O) technique were compared with theoretical power values under different weather conditions. This work was done to determine the most efficient MPPT method that can be considered for MPPT task in PV systems under uniform irradiance and partial shading conditions. Obtained results show that ANN-PSO DMPPT configuration exhibited the best performance.
Keywords— MPPT; GMPPT; P&O; PSO; ANN-PSO; Partial shading, Photovoltaic systems; Genetic Algorithm.
I. INTRODUCTION
The use of photovoltaic (PV) systems to convert energy from sunlight into electrical energy has been of great importance these days due to some additional benefits solar energy systems offer compared to crude fossil fuel energy source [1, 2]. Apart from the cost of implementation, photovoltaic system exhibits a noise-free, carbon dioxide pollution-free to the atmosphere and with a rational energy-payback time (EPBT). The energy energy-payback time is the period of time (in years) required for a photovoltaic system to recompense for the total installation cost [3].
Despite the merits offered by solar energy systems, photovoltaic systems still suffer some limitations that greatly affect the performance of photovoltaic systems. For example, a PV system produce a low energy supply if operated without a good-working maximum power point tracking (MPPT) controller. MPPT is a tracking technique used to regulate and ensure that peak power is extracted from a varied power source such as PV energy or wind energy source [4]. Different techniques have been introduced and used to track the maximum power point in PV systems. However, most of these techniques underperform under partial shading weather
conditions with their inability to find the global maximum power point [5]. Good examples of such conventional (local MPPT) methods are the perturb&observe, incremental conductance, neural networks, fuzzy logic, and artificial neuro-fuzzy inference system (ANFIS) [6, 7].
Several heuristic-searching techniques such as particle swarm optimization (PSO), genetic algorithm (GA), ant bee colony (ABC) have been introduced to track the global maximum power point (GMPP). These metaheuristic MPPT techniques use minimization or maximization of objective function to achieve optimization tasks [8, 9]. Application of these metaheuristic machine learning techniques includes the optimization and tracking of GMPP in PV systems under partial shading and mismatch weather conditions [10]. However, most of these iterating techniques takes time to proffer optimum solutions.
The contribution of this paper is to introduce an innovative use of series-connected distributed MPPT configuration techniques for tracking the global maximum power point under uniform and partial shading conditions. Second contribution is a work done through comparison of the series-connected DMPPT, conventional Perturb&observe, and particle swarm optimization (PSO) technique results to determine the most appropriate MPPT technique that can be recommended for use in PV systems under partial shading weather conditions.
II. MPPT TECHNIQUES
The MPPT methods used in this paper are briefly discussed below:
a. Series-connected DMPPT
This is a type of configuration technique used to improve the performance of a complete PV systems under partial shading and mismatch conditions. To implement series-connected DMPPT configuration, each PV module is series-connected individually to a DC-DC converter and a fast tracking MPPT controller (e.g. ANN-PSO) is selected to confirm that maximum power is extracted from all the PV modules, thus
limiting the partial shading losses effect commonly experienced with PV arrays [11, 12]. The DC-DC converters outputs are then connected serially in order to sum-up the total output power from the converters [13]. In this paper, a hybrid combination of artificial neural network (ANN) and particle swarm optimization as ANN-PSO technique will be considered as the used MPPT method for the series-connected DMPPT configuration. The artificial neural network will be used to train while PSO will be used for MPPT optimization task.
b. Particle Swarm Optimization (PSO)
PSO is a heuristic-search technique inspired by the communal behaviour of bird flocks’ population as swarms. PSO optimizes complex problems using multivariable objective function [14]. PSO is similar to the genetic algorithm (GA) technique as both techniques are used for the minimization and maximization of objective functions in order to yield optimal results. However, PSO differs slightly from GA as fitness-proportional selection and genetic recombination character in GA algorithms is replaced with global best selection and the movement of the swarm particles themselves in PSO algorithm [15]. PSO works iteratively to achieve the best particles solution (Global best) that gives the optimum results [16, 17]. Equations (1) and (2) explain the working principle of PSO algorithm.
(1)
(2)
Where is the swarm size, is the particle velocity, is the current position of a particle, is the local best position, denotes the global best position, coefficients are random numbers between 0 and 1, is the inertia, and coefficients represent the learning factors respectively.
c. Perturb&observe (P&O) Technique
Perturb&observe technique is a hill-climbing, online MPPT technique that uses voltage perturbation to track MPP in PV systems [18, 19]. Conventional Perturb&observe works in such a way that if the MPP is towards the left, there will be a decrease in perturbation (V- DV) and if MPP is situated on the right, perturbation will shift toward the right direction (V+DV) [20]. V is the measured PV voltage and DV is the perturbation. Conventional P&O technique only tracks the local MPP during partial shading, thus resulting in power loss due to P&O inability to successfully track the GMPP under partial shading weather condition [21].
III. SIMULATION MODEL
To confirm the effectiveness of the above-mentioned MPPT techniques (series-connected DMPPT configuration, Perturb&observe, and PSO) for harvesting maximum power in
PV systems under partial shaded weather condition, an experiment was conducted using four-different weather conditions as case studies. The modelled complete PV system comprises of three similar Soltech 1STH-215-P modules, boosted DC-DC converter(s), MPPT controller(s) and a 20W load resistor. Table 1 displays the datasheet specifications for each of the used 1STH-215-P panels.
Table 1: PV and MCUK DC-DC converter specifications Solar Panel Specifications
PV Model 1STH-215-P
Standard Test Condition G = 1000 Wm-2 and T = 25 °C
Peak Voltage 29.0 Volts
Peak current 7.35 Amps
Peak power 213.15 Watts
No. of cell in series 60 cells
Short-circuit current 7.84 Amps
Open-circuit voltage 36.30 Volts
Table 2 displays the weather conditions that were considered to validate the effectiveness of the considered MPPT controllers. G1, G2, and G3 denotes the amount of light striking panel 1, panel 2, and panel 3 respectively under maintained room temperature (25°C) throughout the experiment. From the table, case I is the uniform condition, a condition where the insolation across panel 1, Panel 2 and Panel 3 are uniform. That is G1 = G2 = G3 = 1000 Wm-2. Case II is a weather condition where partial shading is experienced across panels 2 and 3. That is G1 is 1000 Wm-2, G2 is 700 Wm-2 and G3 = 900 Wm-2. Case III is a partial shaded condition where G1 = 700 Wm-2, G2 = 1000 Wm-2, and G3 = 550 Wm-2. While case IV is a condition where G1 = 450 Wm-2, G2 = 400 Wm-2 and G3 = 300 Wm-2.
Table 2: Considered weather conditions
Weather cases G1 (W/m2) G2 (W/m2) G3 (W/m2)
Case I (uniform) 1000 1000 1000
Case II 1000 700 900
Case III 700 1000 550
Case IV 450 400 300
Figure 1 shows the block diagram of a complete PV system designed using series-connected DMPPT configuration technique. This complete PV system comprises of three 1STH-215-P modules connected separately and uses three hybrid ANN-PSO MPPT controllers and three boost DC-DC converters. Each of the connected ANN-PSO controller senses both the irradiance (G) and temperature (T) across the solar panel and outputs a predicted current (Iref) as target and with a best cost error (minimum fitness value) of 5.6661e-3 after 1000 iterations.
The predicted currents ((Iref1, Iref2, Iref3) from the three PSO controllers were then compared with the measured currents (Ipv1, Ipv2, Ipv3) across PV1, PV2, and PV3 respectively as errors (e1, e2, e3). These errors were fine-tuned by the PI controllers and outputs duty cycles (d1, d2, d3) respectively. The duty cycles were then transmitted through pulse width modulators (PWM1, PWM2, and PWM3) as pulse signals that were used to initiate the Mosfet of the DC-DC converters. The output of the boost converters was connected in series.
Figure 2 presents the flowchart of the series-connected DMPPT configuration techniques implemented using ANN-PSO techniques. 1 1 1 2 2 * ( ) ( ) k k k k k k i i i i g i V + =w V +c r P -X +c r P -X 1 k k k i i i X + =X +P 1, 2,3,..., k= n k1 i V + 1 k i X + k i P k g P 1 2
( and )
r
r
w( and )
c
1c
2Fig. 1: series-connected DMPPT configuration
Fig. 2: Flowchart of the series-connected and parallel-connected DMPPT Figure 3 displays the graphical results for the training and optimization the system data using ANN-PSO technique after 1000 iterations.
Fig. 3: ANN-PSO training performance
For the PSO-GMPPT technique, Figure 4 presents the block diagram of a complete PV system designed using PSO technique while Figure 5 displays the flowchart of the used PSO-GMPPT algorithm. The PSO senses the irradiances (G1, G2, and G3) across the three PV modules and the temperature (T = T1= T2 = T3 = 25 °C) as inputs and outputs predicted current (Iref) as target. Iref was then compared with the measured current (Ipv) across the three serially-connected PV modules as error signal (e). The error signal was then minimized using ITAE (Integral of time and absolute error) criterion in order to find the optimal proportional-gain (Kp) and optimal proportional-integral (Ki) that gives the minimal fitness limit (1e-4). The boost DC-DC converter was modelled using a non-linear plant transfer function , where K is a constant coeffient and T is the sampled time.
Fig. 4: PSO PV1 PSO 1 + -Boost PI(s) PWM1 PV2 PSO 2 + -Boost PI(s) PWM2 PV3 PSO 3 + -Boost PI(s) PWM3 20W 1 G 2 G 3 G 1 T 2 T 3 T Iref1 Ipv1 e1 1 d 1 pulse Iref2 Ipv2 Ipv3 Iref3 e2 e3 2 d 3 d 2 pulse 3 pulse + + + + + + -+ + + 1 2 3 25 T T= = = = °T T C Start Algorithm Measure Measure PSO inputs
PSO predicted output currents
are passed to PID1, PID2, and PID3 controllers for tuning and
to obtain duty cycles are passed to PWM controllers at 50KHZ frequency as pulses To Mosfets of DC-DC Converters 1, ,1 2, 2, 3, 3 G T G T G T 1 1 2 2 3 3 (G T, ), (G T, ), and (G T, ) 1 2 3
(Iref ,Iref , and Iref )
1 2 3 (Ipv,Ipv ,Ipv ) 1 pv1 ref1 e =I -I 2 pv2 ref2 e =I -I 3 pv3 ref3 e =I -I 1 2 3 ( , , and ) Errors e e e 1 2 3 ( , , and ) Errors e e e 1 2 3 ( ,d d , and d) respectively 1 2 3 ( ,d d, and d) Output power 1 2 3 P=P+P+P Stop 0 50 100 Sample Index 150 200 250 2 4 6 8 Train Data Target Output 0 50 100 150 200 250 -4 -2 0 2 4 MSE = 0.67076, RMSE = 0.819 Error
Error Mean = -0.0047777, Error St.D. = 0.82084
-3 -2 -1 0 1 2 0 10 20 30 40 ( ) 1 sT Ke G s sT -é = ù ê + ú ë û PV1 PV2 + -Boost PI(s) plant PV3 20W 1 G 2 G 3 G 1 T 2 T 3 T Iref Ipv e d pulse + + + -- -+ PSO 1 2 3 25 T T T T= = = = °C PWM +
-ITAE
( ) e tFig. 5: PSO global MPPT algorithm
Table 3 displays the optimum proportional-gain values (Kp) and the optimum proportional integral values (Ki) that minimizes the ITAE error using PSO technique to track the global maximum power point under four different weather conditions.
Table 3: Determining optimum Kp and Ki using PSO Technique
Cases Optimum Kp Optimum KI Fitness limit
I -14.9777 -61.3619 1e-4
II -97.9008 -70.0026 1e-4
III 1.2720 7.0786 1e-4
IV -2.4635 3.2296 1e-4
Figure 6 presents the block diagram of a complete PV serially-connected (PV array) and modelled using Perturb&observe MPPT technique.
Fig. 6 Perturb&observe
Figure 7 displays the theoretical graphical results of a complete PV array systems, where under case 1, the theoretical global maximum power (GMP) is 638.8 W. Similarly, under case II weather condition, GMPP (theoretical) is 490.65 W. For case III, GMPP (theoretical) is 390.37 W, while case IV depicts a theoretical GMPP of 210.87 W respectively.
Fig. 7: Theoretical GMPP under different weather conditions The PV efficiency under different weather conditions was computed using equation 3,
(3)
where P1, P2, and P3 is the power harvested in panel 1, panel 2, and panel 3 respectively. While Ptotal is the total PV power and P (GMPP ref) = 638.8 W is the reference power across the three panels under standard test condition (STC), that is where G1 = G2 = G3 = 1000 Wm-2 and T1 = T2 = T3 = 25 °C.
IV. EXPERIMENTAL RESULTS
Figures (8-11) and Table 4 display the graphical results and the tabulated results of the considered MPPT techniques compared with theoretical values under four different weather conditions.
Figure 8 and Table 4 directly present the performance of the used MPPT techniques compared with theoretical power values under case 1 weather condition (uniform irradiance). Obtained results show that series-connected DMPPT configuration over-performed as 638.8 W power was harvested at the PV end and with an efficiency of 99.9999%, followed by PSO technique (638.7 W with a PV efficiency of 99.9843%). Perturb&observe (P&O) technique displayed the lowest performance as 607.5 W power could be extracted from the PV system and with an efficiency of 95.1002% using P&O technique.
Fig. 8: Used MPPT performances under case I weather condition For case II, similarly it can be observed from Figure 9 and Table 4 that series-connected DMPPT configuration displayed the best result as a total of 556.5 W PV power was extracted from the panels and with a PV efficiency of 87.1165% using configuration approach. On the other hand, P&O exhibited the lowest performance as 370.2 W PV power was harvested from the panels and with an efficiency of 57.9524% while PSO technique successfully tracked the global maximum power Start Algorithm
Sense G1,G2,G3, and T T=T1=T2=T3=298K
Measure PV current (Ipv) PSO inputs (G1,G2,G3, and T)
PSO predicted output currents (Iref)
Determine optimum Kp and Ki of the PI controller and output
duty cycle signal (d) d is transmitted through PWM controller at 50 kHZ
frequency as pulse signal To Mosfets of DC-DC
Converter Measure output power
Stop 0 ( ) ITAE t e t dt ¥ =
ò
(ITAE)=1e-4 Minimize Objective function If ? Re-iterate No YesError (e) =Ipv - Iref
PV1 PV2 P&O Boost PWM PV3 20W 1 G 2 G 3 G 1 T 2 T 3 T Ipv d pulse + + + -- -+ 1 2 3 25 T T T T= = = = °C Ipv Vpv Vpv + -20 40 60 80 100 120 PV Voltage (Volts) 0 100 200 300 400 500 600 700 P V P ow er (W at ts) case1 (Pmpp=638.81W) case 2 (Pmp=490.65W) case 3 (Pmp=390.37W) case 4 (Pmp=210.87W) 1 2 3 ( ref) *100% 638.8 total MPP P P P P efficiency P + + = =
point as power harvested using PSO equals the theoretical power (490.7 W) and with an efficiency of 76.8159%.
Fig. 9: Used MPPT performances under case I weather condition For case III, it can be seen from figure 10 and Table 4 that series-connected DMPPT configuration exhibited the best performance as a total of 471.8 W PV power was harvested from the panels and with an efficiency of 73.8572%. However, P&O displayed the lowest result as 305.9 W PV power could be harvested from the panels and with an efficiency of 57.9524% while PSO technique was used to harvest 372.9 W power and with an efficiency of 58.3751% respectively. Furthermore, the power extracted using configuration method exceeded the theoretical power when PV modules were serially connected as PV array.
Fig. 10: Used MPPT performances under case I weather condition For case IV weather condition, from Figure 11 and Table 4 results, series-connected DMPPT configuration produced the best result as 183.5 W power was extracted from the three panels and with an efficiency of 28.7257%, followed by PSO (175.5 W PV power and 27.4734% PV efficiency). While, Perturb&observe displayed the lowest performance as 104.7 W power could be extracted from the PV system and with an efficiency of 16.3901%.
Fig. 11: Used MPPT performances under case I weather condition
Table 4: Tabulated results of the considered MPPT Techniques compared with theoretical values under four-different weather conditions
Cases MPPT Techniques P 1 (W) (W) P2 (W) P3 P mpp (W) Eff. (%) 1 Series-DMPPT 213 213 213 638.8 99.9999 PSO 212.9 212.9 212.9 638.7 99.9843 P&O 202.5 202.5 202.5 607.5 95.1002 Theoretical - - - 638.8 Reference 2 Series-DMPPT 213.1 150.6 192.5 556.5 87.1165 PSO 173.8 146.3 170.6 490.7 76.8159 P&O 189.8 -3.85 184.6 370.2 57.9524 Theoretical - - - 490.7 76.8159 3 Series-DMPPT 147.8 206.6 117.8 471.8 73.8572 PSO 135.9 143 93.94 372.9 58.3751 P&O 134.0 175.6 -3.03 305.9 47.8867 Theoretical - - - 390.4 61.1115 4 Series-DMPPT 97.22 86.22 -0.03 183.5 28.7257 PSO 91.29 86.44 -1.65 175.5 27.4734 P&O 95.43 11.45 -1.65 104.7 16.3901 Theoretical - - - 210.9 33.0150 V. CONCLUSIONS
This paper presents a novel use of series-connected DMPPT configuration technique trained and optimized using ANN-PSO technique for MPPT tasks in PV systems under uniform irradiance and partial shaded weather conditions. Obtained results recommend that using configuration method can be used to extract more power from the PV system than using array method (connecting panels in series) as the power extracted using series-connected DMPPT configuration technique exceeded that of conventional PSO and Perturb&observe techniques that were implemented using PV array method in all the four weather cases. Findings also validate that Perturb&observe performs ineffectively under partial shading conditions as lesser power was extracted from the panels under such condition. While PSO technique can be used to track the global maximum power point in a PV array system.
Bibliography
[1] A. M. Farayola, Comparative study of different photovoltaic MPPT techniques under various weather conditions, Johannesburg, South Africa: University of Johannesburg, 2017. [2] J. Sampedro, I. Arto, and M. González-Eguino, "Implications of switching fossil fuel subsidies to solar: A case study for the european union," Sustainability, vol. 20, no. 50, pp. 1-12, 2018. [3] A. M. Farayola, A. N. Hasan, and A. Ali, "Use of MPPT techniques to reduce the energy pay-back time in PV systems," in 9th International Renewable Energy Congress (IREC), Hammamet, Tunisia, 2018.
[4] R. J. Prasanth and B. T. Sudhakar, "A comprehensive review on solar PV maximum power point tracking techniques," Renewable and Sustainable Energy (Elsevier), vol. 67, no. 3, pp. 826-847, 2017.
[5] A. H. EL-Din, S. S Mekhamer, and Hadi M.El-Helw, "Comparison of MPPT algorithms for photovoltaic systems under uniform irriadiance between PSO and P&O," International Journal of Engineering Technology and Management Research, vol. 4, no. 10, pp. 68-77, 2017.
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
Simulation time (sec) 0 100 200 300 400 500 P V P ow er (W at ts) Series-DMPPT PSO P&O
[6] B. Pakkiraiah and G. D. Sukumar, "Research survey on various MPPT performance issues to improve the solar PV system efficiency," Journal of Solar Energy, vol. 2016, pp. 1-20, 2016. [7] R. J. Prasanth and B. T. Sudhakar, "A comprehensive review on solar PV maximum power point tracking techniques," Renewable and Sustainable Energy, ELSEVIER, vol. 67, no. 3, pp. 826-847, 2017.
[8] "Metaheuristics: review and application," Journal of Experimental & Theoretical Artificial Intelligence, vol. 24, no. 4, pp. 503-526, 2013.
[9] L. Y. Chang, Y. N. Chung, K. H. Chao, and J. J. Kao, "Smart global maximum power point tracking controller of photovoltaic module arrays," Energies, vol. 11, no. 567, pp. 1-16, 2018. [10] P. Vasant, S. Z. Alparslan-Gok, and G. W. Weber, Handbook of
research on emergent applications of optimization algorithms (2 volumes), Hershey, Pennsylvania, USA: IGI Global, 2017. [11] A. M. Farayola, A. N. Hasan, A. Ali, and B. Twala, "Distributive
MPPT approach using ANFIS and perturb&observe techniques under uniform and partial shading conditions," in International Conference on Artificial Intelligence and Evolutionary Computations (ICAIECES 2017), India, Springerlink, 2017, pp. 27-37.
[12] S. QinStanton, T. CadyAlejandro, D. Dominguez, C. N. Pilawa-Podgurski, "A distributed approach to MPPT for PV sub-module differential power processing," in 5th Annual IEEE Energy Conversion Congress and Exposition (ECCE), Denver, Colorado, 2013.
[13] M. E. Eapen and D. E. Koshy, " Distributed PV power generation using PSO-based MPPT technique," International Journal of Innovative Research in Technology (IJIRT), vol. 2, no. 6, pp. 275-279, 2015.
[14] A. Lagrioui, G. Bennis, M. Karim, and M. Felja, "Comparison between the conventional PSO method and modified PSO method for PV system under shading," Wseas Transactions On Systems And Control, vol. 13, pp. 186-192, 2018.
[15] T. Iswari and A. M. S. Asih, "Comparing genetic algorithm and particle swarm optimization for solving capacitated vehicle routing problem," in International Conference on Industrial and System Engineering (IConISE) 2017 , Bali, Indonesia, 2017. [16] A. Badis, M. N. Mansouri, and A. Sakly, "PSO and GA-based
maximum power point tracking for partially shaded photovoltaic systems," in 7th International Renewable Energy Congress (IREC), Hammamet, Tunisia, 2016.
[17] J. P. Ram, T. S. Babu, and N.Rajasekar, "A comprehensive review on solar PV maximum power point tracking techniques," Renewable and Sustainable Energy Reviews (Elsevier), vol. 67, pp. 826-847, 2017.
[18] A. M. Farayola, A. N. Hasan, and A. Ali, "Comparison of Modified Incremental Conductance and Fuzzy Logic MPPT Algorithm Using Modified CUK Converter," in 8th IEEE International Renewable Energy Congress (IREC), Amman, Jordan, 2017.
[19] D. Sera, L. Mathe, T. Kerekes et al., "On the parturb-and-observe and incremental conductance MPPT methods for PV systems," IEEE Journal of Photovoltaics, vol. 3, no. 3, pp. 1070-1078, 2013.
[20] M. A. M. Ramli, S. Twaha, K. Ishaque, and Y. A. Al-Turki, "A review on maximum power point tracking for photovoltaic systems with and without shading conditions," Renewable and Sustainable Energy Reviews, vol. 67, pp. 144-159, 2017.
[21] A. Mohapatra, B. Nayak, P. Das, and K. Mohanty, "A review on MPPT techniques of PV system under partial shading condition," Renewable and Sustainable Energy Reviews, vol. 80, pp. 854-867, 2017.