tracking maximum power point at any environmental ... Keyword- Solar Energy,
PV System, Maximum Power .... maximum power point (MPP) of PV cells. The.
International Journal of Emerging Technology and Advanced Engineering Website: www.ijetae.com (ISSN 2250-2459, ISO 9001:2008 Certified Journal, Volume 2, Issue 12, December 2012)
Maximum Power Point Tracking for PV System Using Advanced Neural Networks Technique A. M. Zaki, S. I. Amer, M. Mostafa Electronics Research Institute (ERI), Dokki, Cairo, Egypt. In order to overcome the undesired effects on the output PV power and draw its maximum power, it is possible to insert a DC/DC converter between the PV generator and the batteries, which can control the seeking of the MPP, beside including the typical functions assigned to controllers. These converters are normally named as maximum power point trackers (MPPTs) [1]. The cost of electricity from the solar array system is more expensive compared to electricity from the utility grid. For that reason, it is necessary to study carefully the efficiency of the entire solar system to design an efficient system to cover the load demands with lower cost. There are many external and internal influences which have an effect on the efficiency of the PV panel. A robust control using a PI regulator is used to track this maximum power point. The PI regulator used to control the boost DC/DC converter is synthesized by frequency synthesis using Bode method. For that, they have developed a transfer function of global mode using a small signal method [2]. An intelligent artificial technique to determine the maximum power point (MPP) based on artificial neural network is detailed. The approach is compared to perturb and observe (P&Q) method. The MPPT using artificial neural network proposed can reduce the noises and oscillations generated by classical methods and can be competitive against other MPPT algorithms [3]. Other researchers presented a method for the control of the PV system through the MPPT using Fuzzy Logic controller. This method succeeded to reduce the PV array area and increase their output, and used for control of MPPT for stand-alone PV system givning a minimum economic cost. Developed controller can be improved by changing the form of the functions of memberships as well as the number of subsets [4]. Also a neuro fuzzy controller (NFC) is proposed to track the MPP. It takes advantage in conjunction with the reasoning capability or fuzzy logical systems and the learning capability of neural networks. A gradient estimator based on a radial basis function neural network is developed to provide the reference information to the NFC. With a derived learning algorithm, the parameters of NFC are updated adaptively.
Abstract- Solar energy is clean, renewable and its decentralized character is appropriate well at the scattered state of the zones with low density of population. The cost of electricity from the solar array system is more expensive than the electricity from the utility grid. So, it is necessary to operate the PV system at maximum efficiency by tracking maximum power point at any environmental condition. In this work, the neural network back propagation algorithm is used to control the operation of the PV array in order to extract the maximum power. Two error functions are used. The first, the classic error function, and the second is a modified error function which takes into consideration the derivative of the error function also. The results obtained are compared and discussed. Keyword- Solar Energy, PV System, Maximum Power Point Tracking (MPPT), Neural Networks, Modified Error Function
I.
INTRODUCTION
The world is increasing experiencing a great need for additional energy resources so as to reduce dependency on conventional sources, and photovoltaic (PV) energy could be an answer to that need. Renewable energy becomes an essential source for many applications in the last four decades. It is difficult to supply electrical energy to small applications in remote areas from the utility grid or from small generators. Stand alone photovoltaic (PV) systems are the best solutions in many small electrical energy demand applications such as communication systems, water pumping and low power appliances in rural area. In addition, solar energy is clean, renewable and is used where it is and its decentralized character is appropriate well at the scattered state of the zones with low density of population. Consequently, it can contribute to the environmental protection and be regarded as an alternative with a future to conventional energies. There are two ways to generate electricity from sun; through photovoltaic (PV) and solar thermal systems. Generally, PV systems can be divided into three categories; stand-alone, grid-connection and hybrid systems. For places that are far from a conventional power generation system, stand-alone PV power supply system has been considered a good alternative.
58
International Journal of Emerging Technology and Advanced Engineering Website: www.ijetae.com (ISSN 2250-2459, ISO 9001:2008 Certified Journal, Volume 2, Issue 12, December 2012) The NFC is initialized using the expert knowledge from the traditional fuzzy control, which reduces the burden of the lengthy pre-learning with a derived learning algorithm , the parameters in the NFC are updated adaptively by observing the tracking errors. A radial basis function neural network (RBFNN) is designed to provide the NFC with gradient information, which reduces the complexity of the neural system [5]. The Adaptive Neuro-Fuzzy Inference System (ANFIS) has recently attracted the attention of researchers in various scientific and engineering areas. The ANFIS is designed as a combination of the surgeon fuzzy model and neural network. The fuzzy logic controller (FLC) utilizes the ANFIS output voltage to track the MPP, to acquire high efficiency with low fluctuation [6]. The modeling of a photovoltaic power supply PVPS-system using an ANFIS, was presented in [7]. For the modeling of the PVPS-system, it is required to find suitable models for its different components (ANFIS-PV-array, ANFISbattery and ANFIS-regulator) under variable climatic conditions. Test results provided that the ANFIS performed better than the neural networks. The results obtained from ANFIS can also be used for the prediction of the optimal configuration of PV systems, for the control of PV systems and for the prediction of the performance of the systems. Intelligent control technique using fuzzy logic control is associated to an MPPT controller in order to improve energy conversion efficiency and this fuzzy logic controller is then improved by using genetic algorithms (GA) [8]. In this paper, a MPPT for PV array algorithm based on Neural Networks technique is presented. Section 2 shows the PV cell equivalent circuit. Section 3 presents the PV array characteristics and section 4 presents the Artificial Neural Networks feed-forward algorithm. The error function used is shown in the classic form and also a modified error function. The Neural Network used in this work is presented in section 5. Section 6 shows the results and the discussion, while section 7 gives the conclusions of the work. II.
External sensors are used in many approaches to measure solar irradiation and ambient temperature to estimate the MPP as a function of data measured. A solar cell basically is a p-n semiconductor junction. When exposed to light, a dc current is generated which varies linearly with the solar irradiance. Figure (1) shows the equivalent circuit of the PV cell [9].
Fig.(1) Equivalent circuit of PV cell
The (I-V) characteristics of the PV model is described by the following equation: (
(
)
)
(1)
Where: I cell current (A) IL light generated current (A) ID diode saturation current (A) q charge of the electron = 1.6*10 -19 (coul) K Boltzman constant (j/k) T cell temperature (k) Rs cell series resistance (ohms) Rsh cell shunt resistance (ohms) V cell output voltage (V) III.
PV ARRAY CHARACTERISTICS
As it is crucial to operate the PV energy conversion systems at or near the maximum power point to increase the PV system efficiency, the study of the PV array characteristics became of great importance. The nonlinear nature of PV system is apparent from figure (2). In addition, the maximum power operating point varies with insolation level and temperature. Therefore the tracking control of the maximum power point is a complicated problem. Hence, the use of intelligent control techniques such as Artificial Neural Networks (ANNs) have gained great popularity to solve this problem [10].
PV CELL E QUIVALENT C IRCUIT
The maximum operating point of solar photovoltaic (PV) panels changes with environmental conditions. The maximum power point (MPP) of a PV system depends on cell temperature and solar irradiation, so it is necessary to continually track the MPP of the solar array. Many methods have been proposed to locate and track the maximum power point (MPP) of PV cells. The difficulties that face these methods are the rapid changes in solar radiation and the variety in cell temperature which affects the MPP setting.
59
International Journal of Emerging Technology and Advanced Engineering Website: www.ijetae.com (ISSN 2250-2459, ISO 9001:2008 Certified Journal, Volume 2, Issue 12, December 2012)
Fig. (2) Solar radiation & ambient temperature influence on the IV& P-V characteristic [11].
IV.
ARTIFICIAL NEURAL NETWORKS
Due to increasing need for artificial intelligence techniques, such as neural network, they have been recently widely applied on industrial electronics, and have a large perspective in intelligent control area that is evident by the publications in the literature. However, the literature in this area is hardly more in a last decade. Neural Network itself is a vast discipline in artificial intelligence, and the basic technology has advanced tremendously in recent years. Recently intelligent based schemes have been introduced [3-7]. Artificial Neural Networks try to mimic the biological brain neural networks into mathematical models. From two decades, artificial neural network captivates the attention of a large number of scientific communities. It has been advancing rapidly and its applications are expanding in different areas. A
The Feed-forward Neural Network Figure (3) illustrates an example of a feed-forward neural network (FFNN). The output of the neuron unit can be expressed as: ( ) ( ) (∑ ) Where: wi weight of connection Ɵ bias of the neuron unit N number of inputs to the neuron t time xi input to the neuron
60
(2)
International Journal of Emerging Technology and Advanced Engineering Website: www.ijetae.com (ISSN 2250-2459, ISO 9001:2008 Certified Journal, Volume 2, Issue 12, December 2012)
Fig.(4) Neural Network Configuration
The block diagram of the control system is presented in figure (5).
Fig. (3) Feed-forward neural network
The FFNNs have been applied successfully to solve some difficult and diverse problems by training them in a supervised manner with a highly popular algorithm known as the Back Propagation (BP) algorithm [12]. The BP training algorithm is an iterative gradient algorithm designed to minimize the mean square error between the actual output of a feed-forward net and the desired output. B
Fig.(5) Control System Block Diagram
Neuro-Controller with a Modified Error Function The back propagation algorithm is based on the steepest descent method. The algorithm is therefore stochastic in nature, i.e. it has a tendency to zigzag its way about the true direction to a minimum on the error surface. The error function used in the BP algorithm is as follows: Error = Vref – Vout (3) It suffers from a slow convergence property. To improve the neuro-controller performance in the on-line training mode, a modified error function was proposed [13]: Error = (Vref – Vout )- k1*(d Vout /dt) (4) Where k1 is constant. V.
According to the array temperature and the solar irradiation values, the maximum power point voltage Vmax is selected from the look-up table and input to the neural network as a reference voltage Vr. The output of the neural network is input to the solar array model to calculate the array voltage. The output is fedback as previous value to the input Vo-1. The system was tested using MATLAB M-file. VI.
RESULTS AND D ISCUSSION
The neural network control system presented in this work was tested using the solar array data shown in the following table and results obtained are presented in Table(1) .
NEURAL NETWORK USED CONFIGURATION
The neural network used in this work is shown in figure (4). It consists of three layers. Three inputs in the input layer, four nodes in the hidden layer and one output constitutes the output layer.
61
International Journal of Emerging Technology and Advanced Engineering Website: www.ijetae.com (ISSN 2250-2459, ISO 9001:2008 Certified Journal, Volume 2, Issue 12, December 2012) Table1. MSX-60 MODULE (SOLAREX COMPANY) Temp T
25oc
50 oc
75 oc
100 oc
125 oc
(G) Irrid 0
0
0
0
0
0
0.1
13
9.2
6
2.8
0.16
0.2
14
10
6.8
3.6
.36
0.3
14.6
10.9
7.4
4.3
1.6
0.4
15
11.3
7.8
4.5
1.8
0.5
15.2
11.6
8.2
4.8
2
0.6
15.5
11.8
8.43
5
2.32
0.7
15.6
12
8.6
5.3
2.4
0.8
15.9
12.2
8.7
5.5
2.6
0.9
16
12.4
8.9
5.6
2.76
1
16.12
12.5
9
5.7
2.8
Fig.(7) Reference Value at (T=25 , G= .2).
The results obtained in the three cases show that the neural network with modified error function shows better convergence rate. The output reaches the reference value in few iterations. While using the classic error function takes longer time to reach the reference maximum voltage target. So, the modified error function gives greatly better results during on-line control.
Fig.(8) Reference Value at (T=25 , G= 1) .
VII.
CONCLUSIONS
In this work, a neural network control system for Maximum Power Point Tracking (MPPT) for a PV array was presented. The back propagation neural network was used with the error calculated in its usual form and also with a modified error function which takes into consideration the derivative of the error. The two cases were tested and results were compared. Although both cases give good results, the modified error function resulted in faster convergence rate which give superiority in case of on-line control of the PV array. REFERENCES [1 ] V. Salas, E. Olias, A. Barrado and A. Lazaro, "Review of the maximum power point tracking algorithms for stand-alone photovoltaic systems", Solar Energy Materials & Solar Cells 90 (2006), pp. 1555-1578. [2 ] M. Salhi and R. El-Bachtiri, " Maximum Power Point Tracking Controller for PV systems using a PI regulator with Boost DC/DC converter", ICGST-ACSE journal ISSN 1687-4811,Vol.8, issue iii, January 2009, pp. 21-27.
Fig.(6) Reference Value at (T=50 , G= .2).
62
International Journal of Emerging Technology and Advanced Engineering Website: www.ijetae.com (ISSN 2250-2459, ISO 9001:2008 Certified Journal, Volume 2, Issue 12, December 2012) [3 ] M. Hatti, A. Mcharrar, M. Tiourci, " Novel approach of Maximum Power Point Tracking for Photovoltaic Module Neural Network Based", International Symposium on Environment Friendly Energies in Electrical Applications EFEEA'10, 2-4 November 2010, Ghardaia, Algeria, pp.1-6. [4 ] A. Mellit, M. Benghanme, A. Hadj and A. Guessoum, " Control of Stand-Alone Photovoltaic System Using Controller", Proc. Of Fourteenth Symposium on Improving Systems in Hot and Humid Climates, TX May 17-20, 2004. [5 ] LI Chun, ZHU Xin, SUI Sheng and HU Wan, "Maximum Power Point Tracking of a Photovoltaic Energy System Using Neural Fuzzy Techniques", J Shanghai Univ (Engl Ed), 2009, 13(1), pp.29-36. [6 ] Abdulaziz M, S. Aldobhani and Robert John, " Maximum Power Point Tracking of PV System Using ANFIS Prediction and Fuzzy Tracking", Procs. Of the Inter. Multi Conf. of engineers and Computer Scientists 2008, vol. II, IMECS 2008, 19-21 March, Hong Kong. [7 ] A. Mellit and S. A Kalogirou,"ANFIS-based Modeling for a Photovoltaic Power Supply System: A case Study", ELSEVIER, Renewable Energy 36 (2011), pp.250-258.
[8 ] C. Larbes, S.M. Ait Cheikh, T. Obeidi and A. Zerguerras, " Genetic Algorithms Optimized Fuzzy Logic Control for the Maximum Power Point Tracking in Photovoltaic System", Renewable Energy 34(2009) pp. 2093-2100. [9 ] M. Azab, " A New Maximum Power Point Tracking for Photovoltaic Systems", Inter. National Journal of Electrical and Electronics Engineering, Vol. 3, No. 11, 2003, pp. 702-705. [10 ] N. Patchara., S. Premrud. And Y. Sriuth., " Maximum Power Point Tracking using Adaptive Fuzzy Logic Control for GridConnected Photovoltaic System", Elsevier Publishing ltd, 2005. [11 ] S. Premrud. And N. Patana.," Solar-Array Modelling and Maximum Power Point Tracking Using Neural Networks", IEEE Bologna Power Tech Conference, June 23th-26, Bologna, Italy. 2003 [12 ] W. T. Miller, S. R. Suttin and P. J. Werbos, " Neural Network for Control", MIT Press, Cambridge, MA, 1990. [13 ] M. M. Salem, A. M. Zaki, O. A. Mahgoub, E. Abu-El-Zahab and O. P. Malik, "On-Line Trained Neuro-Controller with a Modified Error Function", Canadian Conference on Electrical and Computer Engineering, Habfax, May 7-10, 2000.
63