Renewable Energy 34 (2009) Contents lists available at ScienceDirect. Renewable Energy. journal homepage:

Transcription

1 Renewable Energy 34 (2009) Contents lists available at ScienceDirect Renewable Energy journal homepage: Polar coordinated fuzzy controller based real-time maximum-power point control of photovoltaic system Syafaruddin a, *, Engin Karatepe b, Takashi Hiyama a a Department of Computer Science and Electrical Engineering of Kumamoto University, Kurokami, Kumamoto , Japan b Department of Electrical and Electronics Engineering of Ege University, Bornova-Izmir, Turkey article info abstract Article history: Received 28 April 2008 Accepted 25 April 2009 Available online 31 May 2009 Keywords: Photovoltaic Real-time simulator Maximum-power point Artificial neural network Fuzzy logic control It is crucial to improve the photovoltaic (PV) system efficiency and to develop the reliability of PV generation control systems. There are two ways to increase the efficiency of PV power generation system. The first is to develop materials offering high conversion efficiency at low cost. The second is to operate PV systems optimally. However, the PV system can be optimally operated only at a specific output voltage and its output power fluctuates under intermittent weather conditions. Moreover, it is very difficult to test the performance of a maximum-power point tracking (MPPT) controller under the same weather condition during the development process and also the field testing is costly and time consuming. This paper presents a novel real-time simulation technique of PV generation system by using dspace realtime interface system. The proposed system includes Artificial Neural Network (ANN) and fuzzy logic controller scheme using polar information. This type of fuzzy logic rules is implemented for the first time to operate the PV module at optimum operating point. ANN is utilized to determine the optimum operating voltage for monocrystalline silicon, thin-film cadmium telluride and triple junction amorphous silicon solar cells. The verification of availability and stability of the proposed system through the realtime simulator shows that the proposed system can respond accurately for different scenarios and different solar cell technologies. Ó 2009 Elsevier Ltd. All rights reserved. 1. Introduction In recent years, the increasing energy demand and environmental problems point out that the research and development activities are essential for distributed power generation systems. Among the distributed energy resources, photovoltaic systems generate electricity with no moving parts, operate quietly with no emissions and require low maintenance cost [1,2]. The worldwide annual growth rate in PV industry has averaged a little more than 30% in the last decade [3,4]. The current basic power delivery system is not adequate to handle the new demands. Due to growing energy demand, the generating electricity with photovoltaic solar cells continues to attract worldwide interests, most recently as a power source for distributed energy generation. The photovoltaic generators can be used in a wide range of applications from power supplies for satellite communications to large power stations feeding electricity into the grid [5,6]. Photovoltaic solar energy has been identified as rapid growth technology with a very large * Corresponding author. Tel.: þ ; fax: þ address: syafa@st.eecs.kumamoto-u.ac.jp (Syafaruddin). potential. However, it still remains a big challenge for the global community to maximize the use of photovoltaic systems. One of the barriers to the widespread use of photovoltaic is the high cost of module materials and encapsulation. The robust PV power conditioning system is another critical issue for efficient PV systems. The other constraints are the intermittency output characteristics of PV system due to solar irradiance, temperature, mismatched cells, partial shading and array configuration. The relationships between them have a non-linear characteristic and can t be easily expressed by a single analytical expression for each PV solar cell technology. In order to increase the power conversion efficiency of PV solar cells, PV system should be operated optimally. The state of the art techniques to track the maximum available output power of PV systems are called the maximum-power point tracking (MPPT). The MPPT controller works based on certain developed algorithms for identifying the maximum-power point (MPP) operation of PV system. The conventional MPPT control methods can be classified according to MPPT algorithms such as perturbation and observation (P&O), incremental conductance, fractional open-circuit voltage methods and so on [7,8]. The P&O and incremental conductance methods are widely used in PV applications due to the /$ see front matter Ó 2009 Elsevier Ltd. All rights reserved. doi: /j.renene

2 2598 Syafaruddin et al. / Renewable Energy 34 (2009) ease of implementation. Meanwhile, the incremental conductance has been proposed to improve the tracking accuracy by improving the P&O method. In general, the efficiency of both tracking techniques strongly depends on iteration step size. If the step size is increased for tracking speed up, the accuracy is decreased or vice versa. The fractional open-circuit voltage method is based on determining a constant value which gives the relationship between MPP voltage and open-circuit voltage to find the optimal operating point. On the other hand, relationship between the MPP voltage and irradiance might not be same for different solar cell technologies. It is impossible to determine the optimum voltage by using only one linear function of the open-circuit voltage [9]. Due to that reasons, the conventional MPPT algorithms might fail to determine the optimum voltage during fast changing irradiance conditions especially in fractional open-circuit voltage and P&O methods [9,10]. In addition, the popular algorithms such as P&O and incremental conductance continue oscillation around the optimum operating point which causes power losses [11]. Moreover, these algorithms may not give good results to the fast dynamic response [12] and also it is very difficult to get an analytical expression in determining the optimum operating voltage for different solar cell technologies under changing weather conditions. In order to solve this problem and to improve the stability of the MPPT controller, a novel MPPT algorithm based on ANN and fuzzy logic controller schemes using polar information on PV systems is proposed. The proposed method uses the advantages of neural networks to predict the MPP voltage in time as a reference voltage for fuzzy logic controller. This paper is inspired from the ANN methods which were developed in the previous works [13 15]. The benefits of using ANN are that there are no requirements for knowledge on internal system parameters, less computational effort and provide a compact solution for multivariable problems [16]. Since the optimum MPP voltage is directly obtained by using ANN, the proposed method does not need complex algorithms and advanced power electronic control units. Meanwhile, the fuzzy logic controller with polar information is used in a simple manner to establish the control signal in real-time for keeping the voltage of PV systems at optimum operating point. In this work, a novel fuzzy logic controller scheme using polar information is applied to the MPPT controller of PV system for the first time. This type of fuzzy logic rules has been used and achieved quite attractive results in the power system stabilizer (PSS) application [17]. The main advantage of the fuzzy logic controller is that the accurate description of the system to be controlled is not required and there are no wide parameter variations with respect to the standard regulators. The proposed method in this paper is also verified through the developed real-time simulator based dspace real-time interface system and tested for different solar cell technologies under different weather conditions. Testing with the real-time simulator will be able to overcome the field testing which is costly and time consuming for investigating the dynamics characteristics of PV system. With real-time simulation, not only performing analysis using real-time system parameters, what-if scenarios simply can be simulated by taking action using the online system model. For instance, using real-time data, the designer can simulate the impact of suddenly changing irradiation on a large scale PV system or different solar cell technologies without expense. A proposed system can be analyzed through the simulator easily during the development process. The real-time simulator can also allow increasing experience of the PV system behavior as well as actual system. The overall results confirm that the proposed technique is robust and insensitive to fast and slow variations in irradiation and temperature and achieves to enhance the dynamic behavior of the MPPT controller without retuning the gain parameters. 2. Configuration of proposed system The basic configuration of the proposed system is shown in Fig. 1. In the PV module block, the electrical outputs of PV modules, such as short-circuit current (I sc ), open-circuit voltage (V oc ), current at maximum-power point (I mp ) and voltage at maximum-power point (V mp ) are generated from the current voltage (I V) curve under varying irradiance level (E) in W/m 2 and cell temperature (T c ) Fig. 1. Basic configuration of proposed system.

3 Syafaruddin et al. / Renewable Energy 34 (2009) in degree Celsius. The V oc and T c are used as the input signals, whereas the optimum voltage (V * op ) is the output signal of the ANN. The neural network was trained once using some measured I V curves taking the V mp value as the output target. The optimum points were estimated by only reading the samples of open-circuit voltage and cell temperature very quickly without solving any nonlinear implicit equations that is necessary in conventional methods. The optimum voltage (V * op ) is used as reference voltage for the controller to generate the error signal (e). The error and integral error are used as the input signals for two-dimensional fuzzy logic controllers. A s and D r are the tuning parameters for the polar information based fuzzy logic controller. The control signal U c with time-delay is used to modify the V dc following V * op for every 0.01 s. The amount of power transferred to the load and reference MPP estimated power are denoted by P dc and P * dc, respectively The I V characteristics modeling of PV modules The PV module block is developed based on the Sandia s PV module model in [18,19] which can characterize module performance under various operational and environmental conditions. This model requires detailed input parameters that characterize the response of a given PV module to cell temperature, solar irradiance, incident angle, and air mass. Sandia s PV model can be used to generate the key points of I V curves such as short-circuit current, open-circuit voltage, and maximum-power point [20]. This model is based on empirical functions for several commercial PV modules. The following variables are defined to identify the electrical characteristics on the I V curve: E Irradiance level on module surface, W/m 2 E e Effective irradiance or Suns AM a Absolute air mass (dimensionless) AOI Angle of incident (degrees) T c, T o Cell and reference temperatures ( Celsius) I sco, I mpo Short-circuit current and current at MPP at STC V oco, V mpo Open-circuit voltage and voltage at MPP at STC a Isc, a Imp b Voco, b Vmpo Normalized temperature coefficients C 0 C 3 Empirically determined coefficients N s Number of cells in series STC E ¼ 1000 W/m 2, T c ¼ 25 C, AM a ¼ 1.5, AOI ¼ 0 By assuming the absolute Air Mass (AM a ) function is equal to 1.0 and combining the beam and diffuse components into a single component of irradiance, the short-circuit current is calculated as follows: E I sc ¼ I sco ð1 þ a E Isc ðt c T o ÞÞ (1) o Other electrical parameters such as MPP current (I mp ), opencircuit voltage (V oc ) and MPP voltage (V mp ) can be generated as follows: I mp ¼ I mpo C 0 E e þ C 1 E 2 e 1 þ aimp ðt c T o Þ (2) V oc ¼ V oco þ N s dðt c Þ lnðe e Þþb Voco E e ðt c T o Þ (3) V mp ¼ V mpo þ C 2 N s dðt c Þ lnðe e ÞþC 3 N s ðdðt c Þ lnðe e ÞÞ 2 þb Vmpo E e ðt c T o Þ (4) In this model, the concept of the effective irradiance (E e )is introduced due to the fact that the PV modules do not respond to all wavelengths of light contained in the solar spectrum. This parameter is obtained by: I sc E e ¼ (5) I sco ð1 þ a Isc ðt c T o ÞÞ Meanwhile, the thermal voltage per cell d(t c )in(3) and (4) is given by: dðt c Þ¼ nkðt c þ 275:15Þ q where n is the diode ideality factor of solar cell, k is the Boltzmann constant ( J/K) and q is the electric charge ( C). In this paper, three PV modules that consist of jointly connected different solar cell technologies are selected to compare the capability of proposed system. The first PV module is Siemens SM-55 PV module with monocrystalline silicon technology which can deliver high efficiency output power under reduced light condition by using pyramidal textured surface. The second one is First Solar FS-50 with thin-film CdTe technology that is recommended when high output voltage is desired. This module uses very thin layers of compound semiconductor material with low temperature coefficients which provides for cost effective and greater energy production. The last module is the USSC US-21 which is equipped with the latest new triple junction amorphous silicon solar cell technology. This module is composed of encapsulated polymer inside a rigid anodized aluminum frame. All coefficients in (1) (6) are given in Table 1 for SM-55, FS-50 and US-21 PV modules. The reason for selecting different PV modules is to show that the characteristic of MPP of each PV module might not be similar. The I V and P V curves are given in Fig. 2 for irradiance levels between Table 1 Specification of PV modules based on the Sandia National Laboratory (SNL) data. PV module specifications SM-55 FS-50 US-21 Materials c-si CdTe 3j a-si Vintage 1999(E) 2001(E) 1999(E) Area (m 2 ) Number of cells in series, N s Short-circuit current at STC, I sco (A) Open-circuit voltage at STC, V oco (V) Current at maximum-power point at STC, I mpo (A) Voltage at maximum-power point at STC, V mpo (V) Normalized temperature coefficient for short-circuit current, a Isc ( C 1 ) Normalized temperature coefficient for current at maximum-power point, a Imp ( C 1 ) Temperature coefficient at 1000 W/m for open-circuit voltage, b Voco (V/ C) Temperature coefficient at 1000 W/m for voltage at maximum-power point, b Vmpo (V/ C) Diode factor of solar cell, n Empirical coefficient relation between current at maximum-power point and irradiance, C 0 Empirical coefficient relating open-circuit voltage to irradiance, C 1 Empirical coefficient relating voltage at maximum-power point to irradiance, C 2 Empirical coefficient relating voltage at maximum-power point to irradiance, C 3 (V 1 ) (6)

4 2600 Syafaruddin et al. / Renewable Energy 34 (2009) Fig. 2. Current Voltage (I V) and Power Voltage (P V) characteristics of PV modules for E ¼ W/m 2 and T c ¼ 50 C.

5 Syafaruddin et al. / Renewable Energy 34 (2009) and 1000 W/m 2 with constant cell temperature of 50 C. Each PV module shows different characteristics with changing irradiation conditions especially around the MPP points. The MPP voltage of SM-55 PV module increases almost proportionally with irradiance levels, compared with other two module types. The MPP voltage of FS-50 PV module increases with irradiances between 100 W/m 2 and 400 W/m 2 and then starts decreasing at an irradiance level of 500 W/m 2. Meanwhile, the MPP voltage of US-21 starts decreasing gradually with irradiance level from 200 W/m 2. This kind of characteristic for different technologies might cause the conventional MPPT algorithms, especially in P&O and fractional open-circuit voltage methods, to fail for tracking the MPP of PV modules under the fast changing irradiation conditions. The ratio of MPP voltage to open-circuit voltage depends on solar cell parameters, but a commonly used value is 76%. This ratio shouldn t be used for all type of solar cell technology when tracking MPP of PV modules. It is another worth noting that the increasing irradiation causes to increase photocurrent, however, MPP voltage decreases with irradiation for US-21 and FS-50 PV module. Furthermore, perturbation observation is an unsuitable method with rapidly changing atmospheric conditions [21]. For these reasons, it must be noticed that a MPPT algorithm should overcome this kind of characteristics. On the other hand, the MPP current is proportional to light intensity that leads to the proportionality of the MPP power to the irradiance level for all PV module types. There is only difference in maximum level of output power at high irradiance levels as shown in Fig. 2. The SM-55 and FS-50 PV modules reach 50 W at irradiance level of 1000 W/m 2. On the other hand, the maximum output of US-21 is only about half of their power under the same irradiance level. The MPP voltage and power variations for different irradiance levels are summarized in Table 2. The irradiance level and cell temperature are unpredictably changing in daily practice of PV system. Any changes in these parameters are consequently affecting the shape of I V curve and the trajectory of MPP voltage and power. Therefore, intelligent techniques are a promising solution to solve the shifting of optimum points due to the environmental variations. The Artificial Neural Network (ANN) is accepted as a technology offering an alternative way to solve this kind of complex problems. The ANN is trained once by observing several combinations of MPP voltage as the function of open-circuit voltage and cell temperature and their optimum points are estimated very quickly without using complex algorithms and techniques. The output of ANN can be used as the voltage reference for the proposed MPPT controller ANN based estimated optimum voltage of PV modules Recently, the application of ANN has entered various engineering fields as an estimation method due to the high pattern recognition ability. As shown in Fig. 1, a three layer feed forward is Table 2 Voltage and power at maximum-power point of PV modules at T c ¼ 50 C. E (W/m 2 ) SM-55 FS-50 US-21 V mp (V) P mp (W) V mp (V) P mp (W) V mp (V) P mp (W) used: an input, a hidden and an output layer to determine the estimated optimum voltage (V * op ) of PV modules. The input layer consists of three nodes for the open-circuit voltage (V oc ), cell temperature (T c ) and a bias signal of 1.0. The open-circuit voltage of PV module can be easily measured by interrupting the normal operation of the system temporarily and storing the measured values [7,8]. Meanwhile, the cell temperature can be practically measured at the backside of PV panels. In the ANN process, the open-circuit voltage is derived from (3), while the cell temperature is arbitrarily determined in the range of C. The number of nodes of hidden layer depends on the minimum error obtained during the training process. The output layer provides the estimated optimum voltage. The successful implementation of ANN method is depending on the training process. During this process, the number of hidden layer and the connection weights between layers are determined following the total minimum error. Therefore, the first step in training process is how to set the functions between input and output through the hidden layers. Neural network training data should be selected to cover the entire region where the network is expected to operate. In the learning stage of the ANN, the generated data set for 228 operating conditions between C and W/m 2 is subdivided into a training set (200 sets) that well describes the entire problem domain and a test set (28 sets). In order to avoid the network losing its generalization ability, the training is stopped when the error on the test set begins to rise considerably. In the hidden layer, the sigmoid function is utilized for the input output characteristics of the nodes. For each node i in the hidden and output layers, the output O i (k) is given as follows: O i ðkþ ¼ 1 1 þ e I iðkþ The term I i (k) in(7) is the input signal to node i at the k-th sampling. The input I i (k) is given by the weighted sum of the input nodes as follows: I i ðkþ ¼ X j (7) w ij ðkþo j ðkþ (8) where w ij is the connection weight from node j to node i and O j (k)is the output from node j. During the training, the connection weights w ij are tuned recursively until the best fit is achieved for the input output patterns based on the minimum value of the sum of the squared errors [22]. The equation of the sum of the squared errors (SSE) is described as: SSE ¼ XN k ¼ 1 ðtðkþ OðkÞÞ 2 (9) where N is the total number of training patterns, t(k) and O(k) are the k-th output target and estimated values of MPP voltage power, respectively. For all the training patterns, the error function is evaluated and the connection weights w ij are updated to minimize the error in (9). In the training process, the learning rate and the momentum were specified to 0.2 and 0.85, respectively [13]. The number of hidden nodes is selected based on the total minimum errors during the training process. Fig. 3 illustrates the results of training process related to the total minimum errors and number of hidden nodes. For US-21 and FS-50 PV modules, the total minimum errors are achieved when the number of hidden nodes is 5. For the SM-55 PV module, it is achieved when the number of hidden nodes is 6. The crystalline silicon solar cells generally have higher fill factor than those of the others. In this study, the average

6 2602 Syafaruddin et al. / Renewable Energy 34 (2009) Fig. 3. Number of hidden nodes based on the total minimum errors. fill factor for SM-55 is , whereas and for FS-50 and US-21, respectively. Fill factor is a measure of the squareness of the I V curve. Therefore, more hidden nodes are required to accurately represent the characteristics of silicon solar cells Fuzzy logic controller scheme using polar information for MPP of PV modules In this section, a fuzzy logic control scheme with polar information is proposed for MPPT control of PV system. Similar fuzzy logic rules are already used in power system stabilizer (PSS) application [17]. Fuzzy logic controller has three important stages as rule base, fuzzification and defuzzification. As shown in Table 3, seven fuzzy levels are used for high accuracy, such as NB (negative big), NM (negative medium), NS (negative small), Z (zero), PS (positive small), PM (positive medium) and PB (positive big). In this table, the Z (zero) diagonal represents the switching line which divides the tables into two parts of control actions; negative signals for deceleration control and positive signals for acceleration control actions. The fuzzy rules assignment table which represents the rule base, then transformed into the phase plane as shown in Fig. 4. The sectors A in the first quadrant and B in the third quadrant are defined as the maximum control actions for deceleration and acceleration, respectively. Meanwhile, the control actions in sectors which are located in the second and fourth quadrant can be stated Table 3 Fuzzy rules assignment table. Fig. 4. Phase plane of fuzzy logic control with polar information. either as deceleration or acceleration. The coordinate of the point Z(k) in this figure is given by: ZðkÞ ¼½Int eðkþ; A s eðkþš (10) where Int_e(k) and e(k) are the state of integral of error and error respectively, A s is the scaling factor of the error. Normally, the input signals of fuzzy logic controller could be the error, derivative of error and integral of error. However, in this proposed method, the error and integral of error are selected to be the input signals for two-dimensional fuzzy logic controller. The error is defined as the deviation between the estimated optimum voltage (V * op ) and controlled dc side voltage (V dc ). The V dc itself is initially set very close to the MPP voltage (V mp ). By comparison with the PSS application in [17], the integral error is assumed equal to the speed deviation state, while the error is considered similar to the acceleration state. The control actions in the phase plane are transformed into two membership functions during the fuzzification stage. In this stage, the linguistic variables are obtained from numerical inputs based on membership function. In this case, there are two membership functions; angle and radius. The linguistic variables in the angle and radius membership function are the deceleration N(q(k)) and acceleration P(q(k)) control actions where their grades are shown in Figs. 5 and 6. Fig. 5. Angle membership function.

7 Syafaruddin et al. / Renewable Energy 34 (2009) Verification of the proposed system using developed real-time simulator The angle q(k) and the radius D(k) can be calculated by using state variables Int_e(k) and e(k) in Fig. 4 as follows: qðkþ ¼tan 1 As eðkþ Int eðkþ DðkÞ ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ðint eðkþþ 2 þða s eðkþþ 2 (11) (12) Meanwhile, the grade of the radius membership functions, denoted by G(D(k)) is specified as: ( DðkÞ GðDðkÞÞ ¼ D r for DðkÞ D r (13) 1:0 for DðkÞ D r where D r is the radius member. The selected tuning parameters of A s and D r were specified to 5.0 and 0.35, respectively. In the defuzzification stage, the linguistic variables are converted back into numerical variables as the fuzzy controller output based on the membership functions. The fuzzy control rules always bring the current state to the equilibrium point (O) to produce the desired control signal. The control (stabilizing) signal U c (k) is determined by the weighted averaging defuzzification algorithm as formulated in (14). U c ðkþ ¼ Fig. 6. Radius membership function. NðqðkÞÞ PðqðkÞÞ NðqðkÞÞ þ PðqðkÞÞ $GðDðkÞÞ$U max (14) The established control signal will be able to keep the dc side voltage (V dc ) of PV system to estimated optimum voltage (V * op ) through the deceleration and acceleration control actions. The main frame of developed simulator model consists of two personal computers (PCs) with Analog-Digital (AD) and Digital- Analog (DA) hardware under a real-time dspace platform. The rapid prototyping concept gives a user-friendly environment to develop some laboratory tests. Both computers may receive and send signals and exactly as if they are connected to the real system. The dspace provides complete solutions for electronic control unit software development. It is a powerful development tool for dedicated services in the field of function prototyping, target implementation and electronic control unit testing. Real-time control systems can be built using dspace and the control logic can be implemented. The dspace works on Matlab/Simulink platform which is a common engineering software and easy to understand and get experience without expense and time consuming. The setting of experimental real-time simulator is configured based on digital and analog simulation software environments. Initially, the overall system configuration is set for digital simulation by using Matlab/Simulink ver The digital simulation step is enough to be run in only one PC for ensuring that the model is perfectly running. Then, the I V characteristics model and the ANN/ fuzzy logic models are separated in PV modules PC and Controller PC, respectively. To obtain analog signal, both models are compiled using software dspace Real-Time Interface (RTI) ver The voltage signal margin about 10 V is used for Analog-Digital (AD) and Digital-Analog (DA) hardware. In Fig. 7, the PV modules based I V characteristics model is set up in the first PC, while the ANN and fuzzy controller properties are installed the second PC. The PV modules PC behaves as PV module hardware that can respond to produce V oc and I dc based on the variations in irradiance level (E) and cell temperature (T c ). The irradiance is measured by PhotoRecorder PHR-51, instead of using pyranometer. Normally, irradiance level can be monitored and measured by a pyranometer set. However, to reduce the cost of the system, the silicon solar cells are used instead of pyranometer to measure the incoming irradiance level. A silicon solar cell can be used as an irradiance sensor, because the short-circuit current is proportional to irradiance [23,24]. The PhotoRecorder PHR-51 is basically working following the silicon solar cell principal to measure the light intensity. To obtain the irradiance measurement in this equipment, a unit conversion is required to obtain the irradiance level in W/m 2, where 1 lx is equal to W/m 2. Fig. 7. Configuration of real-time simulator.

8 2604 Syafaruddin et al. / Renewable Energy 34 (2009) Meanwhile, the cell temperature is calculated and stored in this PC as the function of ambient temperature and wind speed. These measurements can be either directly transferred to the PV array computer using USB cable of RSAQ5R or through the AD converter. The real-time simulator works continuously by sending the signals of V oc, T c, V dc and I dc through the coaxial cable of RG-58c/u from the PV modules PC to the Controller PC units to produce the control signal (U c ) that is always fedback to correct the operating voltage of PV module at the optimum voltage. The V oc and T c signals are used for the ANN data processing, while V dc and I dc are used to produce the output controller. Of course, the best training process of ANN and selection of tuning parameters for the fuzzy logic controller are absolutely required for a precise design. The proposed real-time simulator configuration in this paper can be also called as virtual PV modules with the MPP trackers. This simulator can be utilized for tracking the optimum points of PV modules with different scenarios, even without the PV module hardware. The advantage of this method is no need for re-training the ANN pattern for the same technologies of PV modules. As a result, the simulator will be very useful for the research and development (R&D) of PV system and evaluation for the current existing PV system under special environmental conditions. In this study, several scenarios are implemented to evaluate the robustness of the proposed system for different types of PV module. Firstly, step, slow and fast changes in irradiance level from 300 W/m 2 to 800 W/m 2 at constant T c ¼ 45 C are applied to the SM-55 type PV module. The proposed controller is accurately tracking the V* op under the step change of irradiance as shown in Fig. 8a. The results also show that the proposed controller is Fig. 9. Daily basis of irradiance level (E) and cell temperature (T c ). suitable for slow and fast changing irradiance conditions as shown in Fig. 8b and c, respectively. In the second scenario, the proposed method is tested for three different PV solar cell technologies, SM-55 (monocrystalline Si), FS-50 (thin-film CdTe) and US-21 (triple junction amorphous Si), with given irradiance profiles and cell temperature (from 6am to 6pm) as shown in Fig. 9. This input is the typical cloudy day condition with much fluctuated irradiance levels and corresponded cell temperature. The clouds sometimes move quickly in real Fig. 8. The system responses to the step, slow and fast changes in irradiance level from 300 W/m 2 to 800 W/m 2 at T c ¼ 45 C for SM-55 PV module. Fig. 10. Voltage, power and control signal for SM-55 PV module.

9 Syafaruddin et al. / Renewable Energy 34 (2009) Fig. 11. Voltage, power and control signal for FS-50 PV module. application and in this respect the MPPT controller should give a good response. The scenario given in Fig. 9 is applied to the all PV modules to test the performance of proposed system. The real-time simulation results are given in Figs The results confirm that the proposed technique is robust and insensitive to changes in weather condition. The accuracy of the operating voltage is important since it represents MPP of PV module. This point is the key point of MPPT of PV system. The accuracy of the proposed system is evaluated by using two performance indexes. These are the average relative error (ARE) and the relative error (RE) over time to the optimum voltage. The indexes are given as follows: ARE ¼ 1 X jv op * V dcj 100% (15) K i V * op P jv op * V dcjdt i RE ¼ P V op * Dt 100% (16) i where K is the total number of sampling data. Results of the error measurements are presented in Table 4. As shown in Table 4, the proposed fuzzy controller can successfully track the optimum operating point. Fig. 12. Voltage, power and control signal for US-21 PV module. The performances of the proposed controller are also compared with the Proportional-Integral (PI) controller. To obtain the measurement in the PI controller, a simple PI controller of Matlab/ Simulink block is used to replace the fuzzy logic controller block, while the PV modules and ANN blocks are kept unchanged. In this case, the best tuning parameters for PI controller are specified with proportional gain (K p ) and integral gain (K i ) are 1.5 and 0.75, respectively. The reason for comparing our proposed controller is to show the robustness, stability and accuracy over the conventional PI controller because it is well-known that the PI controller is quite popular due to the ability to maintain an exact reference point. The results show that the PI controller has difficulty to find the best tuning parameters for minimizing the steady-state error, especially when the PV module type is changed. To compare the performance Table 4 Average relative error (ARE) and relative error (RE) of operating output voltage of PV modules under the MPPT processes. Operating conditions SM-55 FS-50 US-21 ARE (%) RE (%) ARE (%) RE (%) ARE (%) RE (%) Step change Slow change Fast change Daily condition

10 2606 Syafaruddin et al. / Renewable Energy 34 (2009) Table 5 The performance index (J) of conventional proportional-integral (PI) and fuzzy logic (FL) controllers. Operating conditions SM-55 FS-50 US-21 PI FL PI FL PI FL Step change Slow change Fast change Daily condition of the proposed fuzzy controller with the PI controller, the performance index is simply defined ratio of actual energy output of PV system to maximum possible output from the PV system over time and defined as Z P dc dt J ¼ Z (17) P dc * dt where P dc is the actual power produced by the PV array under the control of the MPPT, and P* dc is the true maximum power the module could produce under the given temperature and irradiance. Since irradiance and temperature are both function of time, powers are also time-varying. The performance index of controllers on SM-55, FS-50 and US-21 PV modules for different types of irradiance conditions are presented in Table 5. The performance index presents that the fuzzy logic controller has much improvement in the control performance over the conventional PI controller. The significant improvement can be obtained for FS-50 PV module with fuzzy logic controller for all types of operating conditions. 4. Conclusion In this paper, the polar coordinated fuzzy controller based realtime maximum-power point tracking control for different types of photovoltaic modules has been presented. The different behavior can be seen in the relationship between MPP voltage and irradiance variation for different solar cell technologies. The ANN is used to estimate the optimum voltage of PV module directly with the input of open-circuit voltage and cell temperature, instead of determining it by perturbing and observing. The simulation study of the proposed scheme has been realized using Matlab/Simulink. In order to validate its performance, the proposed system has been implemented in dspace software and digital signal processor card on personnel computer. For the first time in the PV system application, the polar coordinated fuzzy logic controller is used to keep the dc voltage output of PV module at its optimum operating point. The effectiveness and accuracy of the proposed method are demonstrated through the developed real-time simulator under different weather conditions and different types of solar cell technologies. References [1] Sukamongkol Y, Chungpaibulpatana S, Ongsaku W. A simulation model for predicting the performance of a solar photovoltaic system with alternating current loads. Renewable Energy 2002;27: [2] Abdullah AH, Ghoneim AA, Al-Hasan AY. Kuwaiti assessment of grid-connected photovoltaic systems in Kuwaiti climate. Renewable Energy 2002;26: [3] Wiser R, Bolinger M, Cappers P, Margolis R. Analyzing historical cost trends in California s market for customer-sited photovoltaics. Progress in Photovoltaic: Research and Applications 2007;15: [4] Denholm P, Margolis RM. Evaluating the limits of solar photovoltaics (PV) in traditional electric power systems. Energy Policy 2007;35: [5] Gabler H. Autonomous power supply with photovoltaics: photovoltaics for rural electrification reality and vision. Renewable Energy 1998;15: [6] Flood DJ. Advanced space photovoltaic technology: applications to telecommunication systems. In: International telecommunication energy conference proceedings; p [7] Esram T, Chapman PL. Comparison of photovoltaic array maximum power point tracking techniques. IEEE Transactions on Energy Conversion 2007;22: [8] Hohm DP, Ropp ME. Comparative study of maximum power point tracking algorithms. Progress in Photovoltaic: Research and Applications 2003;11: [9] Tafticht T, Agbossou K, Doumbia ML, Cheriti A. An improved maximum power point tracking method for photovoltaic system. Renewable Energy 2008;33: [10] Jain S, Agarwal V. New current control based MPPT technique for single stage grid connected PV systems. Energy Conversion and Management 2007;48: [11] Liu F, Duan S, Liu F, Liu B, Kang Y. A variable step size INC MPPT method for PV systems. IEEE Transactions on Industrial Electronics 2008;55: [12] Xiao W, Lind MGJ, Dunford WG, Capel A. Real-time identification of optimal operating points in photovoltaic power systems. IEEE Transactions on Industrial Electronics 2006;53: [13] Hiyama T, Kouzuma S, Imakubo T. Identification of optimal operating point of PV modules using neural network for real-time maximum power tracking control. IEEE Transactions on Energy Conversion 1995;10: [14] Hiyama T, Kitabayashi K. Neural network based estimation of maximum power generation from PV module using environmental information. IEEE Transactions on Energy Conversion 1997;12: [15] Hiyama T, Kouzuma S, Imakubo T, Ortmeyer TH. Evaluation of neural network based real-time maximum power tracking controller for PV system. IEEE Transactions on Energy Conversion 1995;10: [16] Karatepe E, Boztepe M, Colak M. Development of a suitable model for characterizing photovoltaic arrays with shaded solar cells. Solar Energy 2007;81: [17] Hiyama T. Fuzzy logic power system stabilizer using polar information. In: El- Hawary ME, editor. Electric power applications of fuzzy systems. New York: IEEE Press; p [18] King DL. Photovoltaic module and array performance characterization methods for all system operating conditions. In: Proceedings of NREL/SNL photovoltaics program review meeting; p [19] King DL. Sandia s PV module electrical performance model. Sandia National Laboratories: (version, 2000); September 5. [20] Karatepe E, Boztepe M, Colak M. Neural network based solar cell model. Energy Conversion and Management 2006;47: [21] Salas V, Olias E, Barrado A, Lazaro A. Review of the maximum power point tracking algorithms for stand-alone photovoltaic systems. Solar Energy Materials & Solar Cells 2006;90: [22] Wang GJ, Chen CC. A fast multilayer neural-network training algorithm based on the layer-b y-layer optimizing procedures. IEEE Transactions on Neural Networks 1996;7: [23] King DL, Boyson WE, Kratochvil JA. Photovoltaic array performance model. Sandia National Laboratories, SAND; August p [24] Driesse A, Harrison S, Jain P. Evaluating the effectiveness of maximum power point tracking methods in photovoltaic power systems using array performance models. Proceedings of Power Electronic Specialists Conference (PESC) 2007: