International Journal of Revolution in Electrical and Electronic Engineering (IJREEE) ISSN (Online): 2350 –0220, ISSN (Print): 2350 –0212 Volume 2 Issue 3 Sep 2015

MAXIMUM POWER POINT TRACKING CONTROLLER FOR SOLAR-POWERED BATTERY CHARGER APPLICATIONS USING ANFIS BASED PREDICTION SOLAR RADIATION: A CASE STUDY Dr. R.Ramesh kumar Assistant professor, Dept of Automobile Engg, KSR college of Engineering, Tiruchengodu, India.

Abstract -The potential of solar energy on earth is enormous and the electrical power derived from solar PV module varies depending on many environmental factors. Many researchers have made an effort to position and track the maximum power point (MPPT) of the photovoltaic for different applications. This study, presents an adaptive neuro-fuzzy inference system (ANFIS) based modelling approach to predict the reference power of PV module from solar radiation and temperature, which is required for MPPT system. The photovoltaic(PV) module station at Coimbatore , with 11° N latitude and 77° E longitude, in India was considered for this study. The proposed ANFIS model is implemented with real meteorological data. The solar irradiance and temperature data for a period of Jan 2007 – Dec 2010 were utilised. Two years of data are used for training the ANFIS model and one year data for testing the model. The proposed model is capable of predicting reference power of 85W,17Vmax-PV module. The hybrid learning algorithm is used for training th e proposed ANFIS model. The statistical parameters such as root mean square error (RMSE) , mean bias error (MBE) and coefficient (R2 )are presented for the proposed model to validate the performance. The tracking algorithm integrated with a solar powered battery charging system has been successfully implemented on a low-cost PIC-microcontroller. The experimental results with a commercial solar array show that the proposed algorithm outperforms the conventional controller in terms of tracking speed and mitigation of fluctuation output power in steady state operation. The overall system efficiency is well above 90%. Keywords: Adaptive neuro-fuzzy inference system ,hourly global solar radiation, predictive model, root mean square error, mean bias error 1. INTRODUCTION Deployment of Conventional fossil fuel for many decades has led the globe to focus on the renewable energy. Effective harnessing and utilization of Solar energy gains viable importance for solar engineers, agriculturists and hydrologists in many applications. Recent researches on any PV application system focus on utilization technologies to achieve better stability, reliability, supervision and maintenance. Solar radiation data are the basic, very effective information resource for any researcher in solar energy. PV modules, which uses the solar radiations are manufactured under standard test condition (STC) such as an average irradiance spectrum at air mass (AM) 1.5, the normalised irradiance at 1kW/m2 and the cell temperature at 25°C[1] . Though the manufacturers' datasheets specify the normalised data, it is difficult to evaluate and analyze the performance of PV

20

ijreee

International Journal of Revolution in Electrical and Electronic Engineering (IJREEE) ISSN (Online): 2350 –0220, ISSN (Print): 2350 –0212 Volume 2 Issue 3 Sep 2015

module accurately in the field, because PV module temperature varies from 20°C to 40°C in a clear day. Different approaches were formulated by many investigators to estimate the hourly global solar radiation and its related parameters[2]-[5]. A multistage ANN was developed to predict the insolation of the next day[6]. Estimation of monthly mean hourly and daily global solar radiation was carried out using ANN with various meteorological parameters for India[7]. Hamdy et al.,[8] have used ANN model to determine solar radiation data in different spectrum bands from data of meteorology for Helwan (Egypt) monitoring station. Also many researchers have taken efforts to optimize PV systems to ensure accurate functioning of the PV station. Many techniques such as fuzzy logic (FL), artificial neural network (ANN) and neuro-fuzzy are used for different applications. ANFIS is used in many areas such as forecasting[9], classifying[10], controlling[11], recognition [12] and diagnosing [13]. In this study, an ANFIS model has been proposed and adapted to predict the reference power of PV module located at Coimbatore, India. The real meterological data such as hourly global solar irradiation and temperature are chosen for the analysis. The proposed ANFIS model comprises of Neural network (NN) which is used to adjust input and output parameters of membership function in the fuzzy logic controller (FLC). The hybrid learning algorithm is used for training this network. 2. DATA COLLECTION The real meterological data such as hourly global solar irradiation(kW/m2 ), temperature(ºC), and module power are used in this work. These data have been recorded by TNAU.meterological station located in Coimbatore(India) from 2009 to 2010. Table 1. shows the geographical features of Coimbatore in Tamilnadu, southern region of India. Fig 1. shows the sample meterological data used in the simulation analysis.

Table 1. Geographical features of Coimbatore in Tamilnadu Latitude Longitude Elevation Annual Relative humidity Annual Daily solar radiation Annual atm.pressure Earth temperature Annual Wind speed

21

°N °E m % kWh/m2/d kPa °C m/s

11 77 460 68.40 5.12 95.8 27.5 2.2

ijreee

International Journal of Revolution in Electrical and Electronic Engineering (IJREEE) ISSN (Online): 2350 –0220, ISSN (Print): 2350 –0212 Volume 2 Issue 3 Sep 2015

1400

32

Hourly Global Solar radiation

Temperature(1:26,1) 31

1200

Temperature ,degree Celcius

Hourly Global Solar radiation, W/m2

30 1000

800

600

400

29 28 27 26 25

200

24 0 6:00 6.30 7:00 7:30 8:00 8:30 9:00 9:30 10:00 10:30 11:00 11:30 12:00 12:30 1:00 1:30 2:00 2:30 3:00 3:30 4:00 4:30 5:00 5:30 6:00 6:30 7:00 7:30 Time , hr

23 6:00 6.30 7:00 7:30 8:00 8:30 9:00 9:30 10:00 10:30 11:00 11:30 12:00 12:30 1:00 1:30 2:00 2:30 3:00 3:30 4:00 4:30 5:00 5:30 6:00 6:30 7:00 Time , hr

(a) 38

34

Pre-Monsoon period Hourly Global Solar Irradince,W/m2

Temperature,

degree c elc ius

Pre-Monsoon period

Monsoon period

32 30 28 26 24 22 20

1500

Temperature

36

0 6 12 18 24 30 36 42 48 54 60 66 72 78 84 90 96102108114120126132138144150156162168174180186192198204210216222228234240246252258264270276282288294300306312318324330336342348354360 0 Time,hr

Monsoon period

Hourly Global Solar Irradiance

1000

500

0

0 6 12 18 24 30 36 42 48 54 60 66 72 78 84 90 96102108114120126132138144150156162168174180186192198204210216222228234240246252258264270276282288294300306312318324330336342348354360 12182430364206 Time,hr

(b) Fig 1.Solar Irradiance and Temperature data (a) 10 th August 2011 at coimbatore (b) Pre-monsoon and monsoon period. 3. ANFIS BASED PREDICTION The main goal of this work is to predict the PV module power from hourly global solar irradiation(kW/m2 ) and temperature(ºC) based on proposed ANFIS model. ANFIS has the neural network‟s capability to classify data in groups and find patterns and further develop a transparent fuzzy expert system Also, ANFIS has the capability to divide and adapt these groups to arrange a best membership functions resulting in clustering the data and deducing the output desired with minimum epochs. The learning mechanism fine-tunes the underlying fuzzy inference system. Using a given input/output data set, ANFIS constructs a fuzzy inference system (FIS) whose membership function parameters are tuned (adjusted) using either a back propagation algorithm alone, or in combination with a least squares type of method[14]. 3.1 ANFIS Architecture. In this study, ANFIS is utilised as a hybrid soft computing model comprising of a neuro-fuzzy system ,in which a fuzzy inference(high level reasoning capability) system can be trained by a neural network-learning(low level computational power) algorithm. Fig. 2. shows the ANFIS architecture

22

ijreee

International Journal of Revolution in Electrical and Electronic Engineering (IJREEE) ISSN (Online): 2350 –0220, ISSN (Print): 2350 –0212 Volume 2 Issue 3 Sep 2015

utilized in the proposed work. This system has two inputs x and y and one output, where its rule is expressed as : (1) where fi is output and pi , qi and ri are the consequent parameters of ith rule. A i and Bi are the linguistic labels represented by fuzzy sets whose membership function parameters are premise parameters. The so called firing strength or degree of fulfillment of a pair. Output of each node in every layer is denoted by where i specify the neuron number of next layer and l is the layer number. In the first layer called as fuzzifying layer. the linguistic labels are Ai and Bi. The output of the layer is the membership functions of these linguistic labels and are expressed as: ( )

(2)

( )

(3)

( )and where ( ) are membership functions that determine the degree to which the given x and y satisfy the quantifiers Ai and Bi.In the second layer, the firing strength for each rule quantifying the extent which any input data belongs to that rule is calculated. The output of the layer is the algebraic product of the input signals as can be given as: ( )

( ) ; i = 1,2

(4)

„„ ‟‟ denotes a fuzzy -norm operator which is a function that describes a superset of fuzzy intersection (AND) operators, including minimum or algebraic product. In this study algebraic product was used.In the third layer, called as the normalization layer every node calculates the ratio of the ith rule‟s firing strength to the sum of all rules‟ firing strengths. ̅

; i = 1,2.

(5)

In the fourth layer, the output of every node is ̅(

̅

)

(6)

The fifth layer computes the overall output as the summation of all incoming signals, which represents the results of wave height or wave period as can be given as: ∑

(7)

∑

The ANFIS learning algorithm employs two methods for updating membership function parameters [15-16]: 

A hybrid method consisting of back propagation for the parameters associated with the input membership functions, and least squares estimation for the parameters associated with the output membership functions.

23

ijreee

International Journal of Revolution in Electrical and Electronic Engineering (IJREEE) ISSN (Online): 2350 –0220, ISSN (Print): 2350 –0212 Volume 2 Issue 3 Sep 2015 

Back propagation for all parameters.

In order to improve the training efficiency, a hybrid learning algorithm is applied to justify the parameters of input and output membership functions. The only user-specified information is the number of membership functions for each input and the input–output training information. So, the output can be written as: ̅

̅

(8)

Fig 2. ANFIS architecture 3.2 Proposed Prediction Model A simple ANFIS prediction model chart is shown in Fig 3. The hourly global solar irradiance and temperature are the input data and PV module power is the output data .To train the proposed model, hybrid learning algorithm and sugeno fuzzy model has been used. Three gbell membership functions were assigned in the algorithm for each input. In the training and testing phase of the proposed model, a data set of 17,520 and 8,760 has been used respectively. During the training phase, the neural network in ANFIS model classifies the dataset in groups, discover patterns and further extend a transparent fuzzy expert system. Further, the network adapt these groups to arrange best membership function to deduce the desired output with minimum epochs. In the learning phase, ANFIS network model is trained effectively(better generalisation capability) to produce exact results for new testing data samples that are not present in the training data set. The iterative process(as in flowchart) of presenting an input-output pair continue until the error function reaches a predetermined value. To validate the proposed model, a statistical approach has been used to compare the predicted values and measured values.

24

ijreee

International Journal of Revolution in Electrical and Electronic Engineering (IJREEE) ISSN (Online): 2350 –0220, ISSN (Print): 2350 –0212 Volume 2 Issue 3 Sep 2015

INPUT

z

LEARNING TRAINING &

TESTING

PREDICTED OUTPUT

MEASURED OUTPUT

DATA

DATA

Fig 3. ANFIS prediction model chart 4. RESULTS AND DISCUSSIONS In this study, an effort has been made to design and develop an ANFIS model in MATLAB 7.5.0.3402(R2007) software version. The proposed ANFIS model was trained for three input membership functions. Fig 4 and Fig 5.shows the values of the proposed ANFIS model for a particular day(10th August 2011,on hourly basis) and Pre-Monsoon/Monsoon periods(Mar,May,Jun/Jul,Sept,Dec) respectively. It can be observed that the agreement between predicted and measured values are completely satisfactory. The performance of ANFIS models was evaluated in terms of RMSE, MBE and R2 . The formulae can be found in appendix.Presented results in the Table 2. clearly reveal that there is a negligible difference between calculated data and ANFIS model data. It can also be concluded that the proposed ANFIS model is an appropriate technique for exact prediction.

120 ANFIS Profile Measured Profile

Power in PV module,W

100

80

60

40

20

0 6:00 6.30 7:00 7:30 8:00 8:30 9:00 9:30 10:00 10:30 11:00 11:30 12:00 12:30 1:00 1:30 2:00 2:30 3:00 3:30 4:00 4:30 5:00 5:30 6:00 6:30 7:00 Time,hr

Fig 4. Comparision of power from PV module between ANFIS and Measured values at Coimbatore on 10 th August 2011

25

ijreee

International Journal of Revolution in Electrical and Electronic Engineering (IJREEE) ISSN (Online): 2350 –0220, ISSN (Print): 2350 –0212 Volume 2 Issue 3 Sep 2015

180

ANFIS Profile Measured Profile

160

Power in PV Module,Watts

Monsoon period

Pre-Monsoon period

140 120 100 80 60 40 20 0 -20

0 6 12 18 24 30 36 42 48 54 60 66 72 78 84 90 96102108114120126132138144150156162168174180186192198204210216222228234240246252258264270276282288294300306312318324330336342348354360 Time,hr

Fig 5. Comparision of power from PV module between ANFIS and Measured values during pre-monsoon and monsoon period

PV Power

Mean Vaue (Watts)

Measured Value

75.2410

Predicted Value

75.2198

Table 2. Statistical details. MBE RMSE

0.3210

0.0015

R2

0.9998

5. APPLICATION OF PROPOSED MODEL IN PHOTOVOLTAIC ENERGY CONVERSION SYSTEM (PECS). The application of proposed model in PECS, depicts the advantage of predicting the reference PV power from the solar irradiance and temperature for MPPT system. PECS consists of the PV module, DC-DC converter and the DC bus. The PECS includes the PV module, DC-DC converter and the DC bus as shown in Fig. 6. The DC-DC converter converts the PV voltage to the required DC bus voltage as shown in Fig 7. The reference power predicted from ANFIS model can be utilised by a current based MPPT controller system to determine the reference current as in Fig 8. The MPPT controller follows a simple perturb and observe tracking algorithm[17]. The MPPT algorithm can be found in the Appendix. A proportional-integral (PI) controller is employed to act upon the error with parallel proportional and integral responses in an attempt to drive the error to zero[18]. The Vpwm results from PI controller. In this way, the PWM control input to the driver circuit is adjusted , such that duty ratio (D) of boost converter varies between 0.0 and 1.0 and the process becomes almost linear because D is proportional to Vpwm. The consequent operation of prediction, design and analysis are pre requisites to continuew with this investigation[19-22]. Fig.9 shows the output voltage of the DC- DC boost converter ,supplied with 85 W,17Vmax -PV module.

26

ijreee

International Journal of Revolution in Electrical and Electronic Engineering (IJREEE) ISSN (Online): 2350 –0220, ISSN (Print): 2350 –0212 Volume 2 Issue 3 Sep 2015

Fig 6. Solar Photovoltaic sysytem

Fig 7. DC-DC converter model

Fig 8. ANFIS & MPPT model

50 DC bus voltage 45 40 35 30 25 20 15 10 5 0

0 6 121824303642485460667278849096102108114120126132138144150156162168174180186192 2082142202262322240 38224446225052225658226264226870227476228082228688229294239800330406331012331618332224332830333436342348354360 360

Fig 9. Output voltage of DC-DC boost converter. 6.CONCLUSIONS An ANFIS technique for modelling output power from PV module at a particular region has been presented. The performance parameters such as root mean square error (RMSE) and mean bias error (MBE) have been presented for the proposed model. The results prove that the ANFIS approach are more suitable to predict the output power from PV module for any season at any region, provided that samples of the input-output data from locations of all types of weather conditions are included in the

27

ijreee

International Journal of Revolution in Electrical and Electronic Engineering (IJREEE) ISSN (Online): 2350 –0220, ISSN (Print): 2350 –0212 Volume 2 Issue 3 Sep 2015

training process. The ANFIS model shows promising results for evaluating PV power in regions where a network of monitoring stations has not been setup. The application of ANFIS prediction model could enable designing a power conversion system with an efficient control to utilize the maximum possible solar energy at a particular region.

ACKNOWLEDGMENT We wholeheartedly thank the Department of science and Technology,India for the support of this work and TNAU for the database support. REFERENCES [1]. Angstrom, A., 1924.Solar and terrestrial radiation. Q.J.R.Meteorol.Soc., 50: 121-125. [2]. Rietveld, M.R., 1978. A new method for estimating the regression coefficients in the formula relating solar radiation to sunshine. Agric. Mete., 19: 243-252. [3]. Neuwirth, F., 1980.The estimation of global and sky radiation in Australia. Solar Energy, 24: 421-46. [4]. Gopinathan. K.K., 1988. A general formula for computing the coefficients of the correlation connecting global solar radiation to sunshine duration. Solar Energy, 41(6): 499-502. [5]. Hay, J.E., 1979. Calculation of monthly mean solar radiation for horizontal and inclined surfaces. Solar Energy, 23(4): 301-307. [6]. Alawi, S.M. and H.A. Ilinai, 1988. An ANN based approach for predicting global radiation in locations with no dir e c t measurement instrumentation. Renewable Energy, 14(1-4): 199-204. [7]. Kemmoku, Y., S. Orita, S. Nakagawa and T. Sakakibara, 1999. Daily insolation forecasting using a multi stage neural network. Solar Energy, 66(3): 193-199. [8]. Reddy, K.S., M. Ranjan, 2003. Solar resource estimation using artificial neural networks and comparison with other correlation models. Energy Conversion and Management, 37(2): 183-198. [9]. Hamdy, K.E., F.A. Faiz and S.E. Tarek, 2005. Estimation of solar radiation components incident on Helwan site using neural networks. Solar Energy, 79: 270-279. [10].Elmas, C., Ustun, O., & Sayan, H. H. (2008). A neuro-fuzzy controller for speed control of a permanent magnet synchronous motor drive. Expert Systems with Applications, 34, 1. [11].Sorousha, M., & Parisa, A. B. (2009). Intelligent scenario generator for business strategic planning by using ANFIS. Expert Systems with Applications, 36, 7729–7737. [12].Jang, J. (1993). ANFIS: Adaptive-network-based fuzzy inference system. IEEE Transactions on Systems, Man and Cybernetics, 23, 665–685. [13].Jang, J. S. R. (1991). Rule extraction using generalized neural networks. In Proceedings of the IFSA world congress (Vol. 4, pp. 82–86). [14].Avci, E., Hanbay. D., & Varol, A. (2007b). An expert discrete wavelet adaptive network based fuzzy inference system for digital modulation recognition. Expert Systems with Applications, 33(3), 582–589. [15].Aznarte, M. J. L., S/ánchez, J. M. B., Lugilde, D. N., Fernández, C. D. L., Guardia, C. D. Dl.,& Sánchez, F. A. (2007). Forecasting airborne pollen concentration time series with neural and neurofuzzy models. Expert Systems with Applications, 32(4), 1218–1225. [16].Wang, Y. M., Taha, M. S., & Elhag, T. M. S. (2008). An adaptive neuro-fuzzy inference system for bridge risk assessment. Expert Systems with Applications,34(3),3099–3106.

28

ijreee

International Journal of Revolution in Electrical and Electronic Engineering (IJREEE) ISSN (Online): 2350 –0220, ISSN (Print): 2350 –0212 Volume 2 Issue 3 Sep 2015 [17] V. Salas, E. Olias, A. Barrado, A. Làzaro,( 2005). “Review of the maximum power point tracking algorithms for stand-alone photovoltaic systems”, Solar Energy Materials &Solar Cells, vol. 89, issue 1, pp. 1-24. [18] Elena Villanueva, Pablo Correa, José Rodríguez, and Mario Pacas,(2009). Control of a singlephase cascaded h-bridge Multilevel Inverter for Grid-Connected Photovoltaic Systems. IEEE transactions on industrial electronics, vol. 56, no. 11.pp.4399-4406 [19] Sumithira, T.R., Nirmal kumar, A. and Ramesh Kumar, R. “An adaptive neuro-fuzzy inference system (ANFIS) based Prediction of Solar Radiation”, Journal of Applied. Sciences Research, Vol.8, No. 1, pp. 346-351, 2012 ; ISSN : 1816157X [20] Sumithira, T.R. and Nirmal kumar, A. “Prediction of monthly global solar radiation using adaptive neuro fuzzy inference system technique over the state of Tamilnadu (India) : a comparative study”, Applied Solar Energy, Vol. 48, No. 2, pp 140-145, 2012 ; ISSN : 0003701X [21] Sumithira, T.R. and Nirmal kumar, A. “Analysis on the design and development of off-grid solar photovoltaic uninterruptible power system using matrix converter topology : An experimental study”, International Review of Electrical Engineering, Vol. 8, No.1, pp 26-32, 2013. [22]Sumithira, T.R. and Nirmal kumar, A. “Elimination of Harmonics in Multilevel inverters connected to solar photovoltaic system using ANFIS : An experimental case study”, Journal of Applied Research Technology, Vol. 11, pp 124 -132, 2013. ISSN: 16656423

APPENDIX A. Statistical Formula:

∑

(

) ∑

√[∑

(

) ]

∑ ∑

(

) ∑

(

)

B.Mppt perturb & obsrve algorithm: function y = MPPtrackIref(P) global Pold; global Iref; global Increment; IrefH = 5.0; % upper limit for the reference current IrefL = 4.0; % lower limit for the reference current DeltaI = 0.05; % reference current increment if (P < Pold) Increment = -Increment; % change direction if P decreased end

29

ijreee

International Journal of Revolution in Electrical and Electronic Engineering (IJREEE) ISSN (Online): 2350 –0220, ISSN (Print): 2350 –0212 Volume 2 Issue 3 Sep 2015

% increment current reference Iref=Iref+Increment*DeltaI; % check for upper limit if (Iref > IrefH) Iref = IrefH; end % check for lower limit if (Iref < IrefL) Iref = IrefL; end % save power value Pold = P; % output current reference y = Iref;

30

ijreee