You are on page 1of 5

TSINGHUA SCIENCE AND TECHNOLOGY

ISSN 1007-0214 12/23 pp204-208


Volume 10, Number 2, April 2005

Artificial Neural Network Maximum Power Point Tracker


for Solar Electric Vehicle

Theodore Amissah OCRAN, CAO Junyi (曹军义)**,


CAO Binggang (曹秉刚), SUN Xinghua (孙兴华)

Research and Development Center for Electric Vehicle, Xi’an Jiaotong University, Xi’an 710049, China

Abstract: This paper proposes an artificial neural network maximum power point tracker (MPPT) for solar
electric vehicles. The MPPT is based on a highly efficient boost converter with insulated gate bipolar transis-
tor (IGBT) power switch. The reference voltage for MPPT is obtained by artificial neural network (ANN) with
gradient descent momentum algorithm. The tracking algorithm changes the duty-cycle of the converter so
that the PV-module voltage equals the voltage corresponding to the MPPT at any given insolation, tempera-
ture, and load conditions. For fast response, the system is implemented using digital signal processor (DSP).
The overall system stability is improved by including a proportional-integral-derivative (PID) controller, which
is also used to match the reference and battery voltage levels. The controller, based on the information sup-
plied by the ANN, generates the boost converter duty-cycle. The energy obtained is used to charge the lith-
ium ion battery stack for the solar vehicle. The experimental and simulation results show that the proposed
scheme is highly efficient.

Key words: artificial neural network; maximum power point tracker (MPPT); photovoltaic module; digital
signal processor; solar electric vehicle

(dc-dc or dc-ac converter) for load interface. Therefore,


Introduction the overall system cost could be reduced drastically by
using highly efficient power conditioners, such as the
Photovoltaic (PV) generation is gaining increased im-
maximum power point tracker (MPPT), to extract and
portance as a renewable source of energy. The undesir-
maintain the peak power from the PV module even
able rapid changes of solar power, which usually occur
when the above-mentioned unfavorable conditions
in a running vehicle, arise as a result of shade from
occur.
buildings, large trees, and clouds in the sky. Conven-
Various methods of maximum power tracking have
tional PV systems have difficulties in responding to
been considered in PV power applications[1-12]. Among
rapid variations due to shade. The main drawbacks of
the hill climbing methods[1-5], the perturb and observe
PV systems are that the initial installation cost is con-
(P&O) method tracks maximum power point (MPP) by
siderably high and the energy conversion efficiency
repeatedly increasing or decreasing the output voltage
(from 12% to 29%) is relatively low. Furthermore, in
at MPP of the PV module. This method requires calcu-
most cases, PV systems require a power conditioner
lation of dP/dV to determine the MPP[1,2,4]. Though the
Received: 2004-03-24 method is relatively simple to implement, it cannot
﹡﹡ To whom correspondence should be addressed. track the MPP when the irradiance changes rapidly.
E-mail: caojy@mailst.xjtu.edu.cn;
Also, the method oscillates around the MPP instead of
Tel: 86-29-87570252
Theodore Amissah OCRAN et al:Artificial Neural Network Maximum Power Point Tracker …… 205

directly tracking it. The incremental conductance


technique (ICT) is the most accurate[6,7] among the
other methods. This method gives good performance
under rapidly changing conditions. However, the com-
plex calculation of dI/dV and the complicated
algorithm require use of a digital signal processor
(DSP), which will usually increase the total system
Fig. 1 Equivalent circuit of a solar cell
cost.
The MPP tracking method using the short circuit
current of the PV module utilizes the fact that the op- ⎧ ⎫ V +I R
⎡ q
erating current at MPP of the PV module is linearly I o = I g − I sat ⎨exp ⎢ (Vo + I o Rs )⎤⎥ − 1⎬ − o o s
⎩ ⎣ AkT ⎦ ⎭ Rsh
proportional to the short circuit current of the PV mod-
ule[9]. Under rapidly changing atmospheric conditions, (1)
this method has a fast response speed of tracking the where I o and Vo are the output current and output
MPP, but the control circuit is complicated. The MPP voltage of the solar array; I g is the generated current
tracking method using open circuit voltage of the solar under a given insolation; I sat is the reverse saturation
panel[11] utilizes the fact that the operating voltage at current; q is the charge of an electron; k is the
MPP is almost linearly proportional to open circuit Boltzmann constant; A is the ideality factor for a P-
N junction; T is the array temperature; Rs and Rsh
voltage at MPP of the PV module (using 76% of open
are the intrinsic series and shunt resistances of the solar
circuit voltage as the MPP voltage). This method is
array.
very simple and cost-effective, but the reference volt-
The saturation current of the solar array varies with
age does not change between samplings. MPPTs using
temperature according to the following equation:
the fuzzy logic[13,14] and the artificial neural network
(ANN) [15] have been reported. These studies show that ⎡T ⎤
3
⎡ qE ⎛ 1 1 ⎞⎤
I sat = I o r ⎢ ⎥ exp ⎢ GO ⎜ − ⎟⎥ (2)
such modern control algorithms are capable of improv- ⎣ Tr ⎦ ⎢⎣ Bk ⎝ Tr T ⎠ ⎥⎦
ing the tracking performance as compared to λ
I g = ⎡⎣ I sc + K I (T − 25 ) ⎤⎦ (3)
the conventional methods. 100
In this paper, we propose an ANN MPPT for solar where Tr is the reference temperature; I or is the
electric vehicle PV systems. The tracking algorithm saturation current at Tr ; EGO is the band-gap en-
changes the duty ratio of the converter so that the PV ergy of the semiconductor used in the solar array; B is
module voltage equals the voltage corresponding to the also an ideality factor; I sc is the short circuit current
MPP at that atmospheric condition. This adjustment is at 25°C; K I is the short-circuit current temperature
carried out by using the back-propagation ANN. The coefficient and λ is the insolation in mW/cm².
reference voltage to the MPP is obtained by an offline Equations (1)-(3) are used in the development of
trained ANN. The controller generates the boost computer simulations for the solar array. The Matlab
converter duty-cycle. programming language is used. Figure 2 show the
simulated current-voltage and power-voltage curves
1 PV Module Characteristics for the solar array at different insolations and different
temperatures. These curves show that the output char-
The solar array characteristics significantly influence
acteristics of the solar array are nonlinear and greatly
the design of the converter and the control system, so
affected by the solar radiation, temperature, and load
the PV characteristics will be briefly reviewed here.
condition. Each curve has a maximum power point
The solar array is a nonlinear device and can be repre-
(Pmax), which is the optimal operating point for the ef-
sented as a current source model, as shown in Fig. 1.
ficient use of the solar array.
The traditional I-V characteristics of a solar array,
when neglecting the internal shunt resistance, are given 2 Artificial Neural Network
by the following equation:
ANN technology has been successfully applied to
206 Tsinghua Science and Technology, April 2005, 10(2): 204–208

j when neuron j is a hidden unit; e j ( n ) is the er-


ror signal at the output of neuron j for iteration n ;
and the set C includes all the neurons in the outer
layer of the network. The correction ∆w ji ( n ) to the
synaptic weight w ji ( n ) is given by
∆w ji ( n ) = α∆w ji ( n − 1) + ηδ j ( n ) yi ( n ) (5)
where α is the momentum constant; η is the learn-
ing-rate parameter of the back-propagation algorithm;
δ j (n) is the local gradient. The error signal at the out-
put is defined as
e j ( n) = d j ( n) − y j ( n) (6)
where d j ( n ) is the desired response or wanted tar-
gets and y j (n) is the output signal. Adjustment of the
weights for these layers is given by
w ji ( n + 1) = w ji ( n ) + α i w ji ( n − 1) + ηδ j ( n ) yi ( n )
(7)
From Eqs. (4) and (6), the mean squared error per-
formance index can be rewritten as

ε ( n ) = (Vref ( n ) − VA ( n ) ) .
1 2

2
The network training is performed repeatedly until the
performance index ε = (Vref − VA )
2
falls below a
specified value, ideally to zero. In other words ε → 0
(Vref − VA ) → 0, and then the connecting
2
implies
weights of the network are adjusted in such a way that
the array voltage VA is identically equal to the maxi-
mum power point voltage Vmp . At this stage the refer-
ence voltage Vref becomes equal to the maximum
power point voltage Vmp .
The configuration of the proposed three-layer feed-
forward neural network function approximator is
Fig. 2 Current-voltage and power-voltage curves for shown in Fig. 3. The neural network is used to obtain
the voltage of the maximum power Vmp ( n ) of the so-
the solar array at different insolations and different
temperatures (S is the solar radiation)
lar panel. The network has three layers: an input, a
solve very complex problems. Recently, its application hidden, and an output layer. The numbers of nodes are
in various fields is increasing rapidly[16,17]. two, four, and one in the input, the hidden, and output
The instantaneous sum of error squares or error en- layers, respectively. The reference-cell open circuit
ergy at iteration n is given by voltage Voc ( n ) and the time parameter T ( n ) are
1 supplied to the input layer of the neural network. These
ε (n) = ∑ e2j (n)
2 j∈C
(4) signals are directly passed to the nodes in the next hid-
den layer. The node in the output layer provides the
where neuron j lies in a layer to the right of neuron
identified maximum power point voltage Vmp ( n ) . The
i , and neuron k lies in a layer to the right of neuron nodes in the hidden layer get signals from the input
Theodore Amissah OCRAN et al:Artificial Neural Network Maximum Power Point Tracker …… 207

layer and send their output to the node in the output output layer mapping. During the training, the con-
layer. The sigmoid activation function is utilized in the necting weights are modified recursively until the best
layers of the network. The training program calculates fit is achieved for the input-output patterns in the train-
the connecting weights WI {1,1} with the bias b{1} ing data. The training of the net was accomplished
for the input to hidden layer mapping, the connecting off-line using Matlab.
weights WL {2,1} with bias b{2} for the hidden layer to

Fig. 3 Feed-forward neural network function approximator

η =0.1; momentum factor α =0.9; number of training


3 Experimental/Simulation Results iterations=10 000; error goal=0.000 001. The conver-
gence error for the training process is shown in Fig. 4.
A PV array used for the collection of the experimental
data is JDG-M-45 (Germany) type modules. The mod-
ule has a maximum power output of 45 W and a 20-V
open-circuit voltage at an irradiation of 1000 W/ m 2
and a 25°C temperature. The PV module specifica-
tions provided by the manufacturer in Table 1.
Table 1 Electrical specifications for the PV module
(JDG-M-45)
Maximum power, Pm (W) 45
Short circuit current, Isc (A) 2.93
Open circuit voltage, Voc (V) 20 Fig. 4 Convergence error for the neural network
training process
Voltage at max power point, Vmp (V) 17.1
Current at max power point, Imp (A) 2.64 The parameters used for the training of the ANN and
Size (mm×mm×mm) 971×441×28 the output values given by the ANN after the training
Mass (kg) 5 can be found in Table 2. Various sets of reference cell
The feed-forward back-propagation ANN, as shown open circuit voltage Voc and a time parameter T (as
in Fig. 3 was trained with values obtained from ex- shown in Table 2) are supplied as the input to the ANN.
perimental data of the reference cell. Gradient descent In order to validate the learning capability of the ANN,
algorithm was used in training as it improves the per- other sets of Voc different from the one in Table 2 were
formance of the ANN, reducing the total error by also supplied to the ANN, which gave out values of
changing the weights along its gradient. The training Vmp as expected. The software Matlab was used in the
parameters are as follows: learning rate parameter training of the ANN.

Table 2 The results of the trained data given by the ANN


Hour of the day
9:00 9:30 10:00 10:30 11:00 11:30 12:00 12:30 13:00 13:30 14:00
Time parameter −1 −0.9 −0.8 −0.7 −0.6 −0.5 −0.4 −0.3 −0.2 −0.1 0
Voc measured 0.555 0.519 0.516 0.529 0.517 0.514 0.516 0.510 0.511 0.509 0.508
Vmp measured 0.422 0.394 0.392 0.402 0.393 0.391 0.392 0.388 0.388 0.387 0.386
Vmp given by ANN 0.421 0.379 0.388 0.416 0.395 0.394 0.392 0.388 0.388 0.386 0.385
208 Tsinghua Science and Technology, April 2005, 10(2): 204–208

The weights to the hidden layer 1 from input 1 are as LA, USA, 2000: 115-1156.
follows: [6] Tse K K, Ho M T, Chung H S, Hui S Y. A novel maximum
⎡ 2.1706 −5.3865 −3.2326 −5.1766 ⎤ power point tracker for PV panels using switching fre-
WI {1,1} = ⎢ ⎥.
⎣109.8309 −27.9435 97.7250 −8.5640 ⎦
quency modulation. IEEE Trans. Power Electron., 2002,
17(6): 980-989.
The weights to the output layer 2 are WI {2,1} =
[7] Brambilla A, Gambarra M, Garutti A, Ronchi F. New ap-
[ −0.2039 0.0065 0.0561 −0.3029 ]. The bias to proach to photovoltaic arrays maximum power point track-
layer 1is b{1} = [−60.0887 12.996 − 54.4753 ing. In: Proc. IEEE Power Electron. Spec. Conf., Charles-
−1.5969]. The bias to layer 2 is b{2} = [−0.0397] . ton, SC, USA, 1999: 632-637.
[8] Mutoh N, Matuo T, Okada K, Sakai M. Prediction-data-
4 Conclusions based maximum power point tracking method for photo-
voltaic power generation systems. In: Proc. IEEE Power
An artificial neural network MPPT for charging the Electron. Spec. Conf., Caims, Austrilia, 2002: 1489-1494.
battery stack of a solar (hybrid) vehicle has been pro- [9] Noguchi T, Togashi S, Nakamoto R. Short-current pulse-
posed in this paper. An off-line ANN, trained using a based maximum power point tracking method for multiple
back-propagation with gradient descent momentum al- photovoltaic and converter module system. IEEE Trans.
gorithm, is utilized for online estimation of reference Ind. Electron., 2002, 49(1): 217-223.
voltage for the feed-forward loop. Experimental data is [10] Lee D Y, Noh H J, Hyun D S, Choy I. An improved
used for the offline training of the ANN, and software MPPT converter using current compensation method for
Matlab is used in the training of the net. The precision small scaled PV-applications. In: Proc. IEEE Applied
of the estimation has been verified by the graph of the Power Electron. Conf., Dalas, TX, USA, 2002: 540-545.
convergence error. The proposed method has several [11] Enslin J H R, Wolf M S, Snyman D B, Sweigers W. Inte-
advantages over the conventional methods, particularly grated photovoltaic maximum power point tracking con-
in that there is no need for voltage and current sensors, verter. IEEE Trans. Ind. Electron., 1997, 44(6): 769-773.
and in that it avoids a complex calculation of power. [12] Chen Y, Smedley K, Vacher F, Brouwer J. A new maxi-
The experimental and simulation results show that the mum power point tracking controller for photovoltaic
proposed scheme is highly efficient. power generation. In: Proc. IEEE Applied Power Electron.
Conf., Miami Beach, FL, USA, 2003: 56-62.
References
[13] Nafeh A, Fahmy F H, El-Zahab E M A. Evaluation of a
[1] Hua C, Lin J, Shen C. Implementation of a DSP-controlled proper controller performance for maximum power point
photovoltaic system with peak power tracking. IEEE Trans. tracking of a stand-alone PV system. International Journal
Ind. Electron., 1998, 45(1): 99-107. of Numerical Modelling, 2002, 15(4): 385-398.
[2] Koutroulis E, Kalaitzakis K, Voulgaris N C. Development [14] Veerachary M, Senjyu T S, Uezato K. Feedforward
of a microcontroller-based photovoltaic maximum power maximum power point tracking of PV systems using fuzzy
point tracking control system. IEEE Trans. Power Elec- controller. IEEE Trans. Aerosp. and Electron. Syst., 2002,
tron., 2001, 16(1): 46-54. 38(3): 969-981.
[3] Kuo Y C, Liang T J, Chen J F. Novel maximum power [15] Hiyama T, Kouzuma S, Ortmeyer T. Evaluation of neural
point tracking controller for photovoltaic energy conver- network based real time maximum power tracking control-
sion system. IEEE Trans. Ind. Electron., 2001, 48(3): 594- ler for PV system. IEEE Trans. on Energy Conv., 1995,
601. 10(3): 543-548.
[4] Sullivan C R, Powers M J. A high-efficiency maximum [16] Bose B K. Artificial neural network applications in power
power point tracker for photovoltaic arrays in a solar- electronics. In: Proc. IEEE Ind. Electron. Conf., Denver,
powered race vehicle. In: Proc. IEEE Power Electron. Spec. CO, USA, 2001: 1631-1638.
Conf., Seattle, WA, USA, 1993: 574-580. [17] Haykin S. Neural Networks—A Comprehensive Founda-
[5] Simoes M G, Franceschetti N N. A risc-microcontroller tion, 2nd Edition. New York: Prentice Hall Inc., 1999.
based photovoltaic system for illumination applications. In:
Proc. IEEE Applied Power Electron. Conf., New Orleans,