IET Renewable Power Generation

Research Article

Intelligent wind power smoothing control with ISSN 1752-1416

Received on 21st September 2015
Revised 22nd August 2016
BESS Accepted on 21st October 2016
E-First on 17th January 2017
doi: 10.1049/iet-rpg.2015.0427
www.ietdl.org

Faa-Jeng Lin1 , Hsuang-Chang Chiang2, Jin-Kuan Chang1, Yung-Ruei Chang3

1Department of Electrical Engineering, National Central University, Chungli 320, Taiwan
2Department of Electrical Engineering, National United University, Miaoli 360, Taiwan
3Engineering Technology and Facility Operation Division, Institute of Nuclear Energy Research, Taoyuan 335, Taiwan

E-mail: linfj@ee.ncu.edu.tw

Abstract: An intelligent wind power smoothing control using recurrent fuzzy neural network (RFNN) is proposed in this study.
First, the modeling of wind power generator and the designed battery energy storage system (BESS) are introduced. The BESS
is consisted of a bidirectional interleaved DC/DC converter and a 3-arm 3-level inverter. Then, the network structure of the
RFNN and its online learning algorithms are described in detail. Moreover, actual wind data is adopted as the input to the
designed wind power generator model. Furthermore, the three-phase output currents of the wind power generator are converted
to dq-axis current components. The resulted q-axis current is the input of the RFNN power smoothing control and the output is a
gentle wind power curve to achieve the effect of wind power smoothing. The difference of the actual wind power and smoothed
power is supplied by the BESS. The minimum energy capacity of the BESS with a small fluctuation of the grid power can be
achieved by the RFNN power smoothing control. A digital signal processor (DSP) based BESS is built using two
TMS320F28335. From the experimental results of various wind variation sceneries, the effectiveness of the proposed intelligent
wind power smoothing control is verified.

1Introduction superconducting magnetic energy storage [6, 7], super capacitor

In recent years, global warming caused by emissions of greenhouse
gases has hugely attracted the attention of governments and general
public. Therefore, the development of renewable energy resources
(RERs) to cope with the increasing demand for electricity and to
reduce the emissions of greenhouse gases has become an important
topic especially for the most cost-effective wind energy [1]. Thus,
many large-scale wind power plants have been built for electrical
power generation. However, wind power in many places though
rich but are vulnerable to seasonal and weather influence resulted
in the variation of wind energy output. Since the wind speed is
random in nature, the wind power generation system should be
designed to track the maximum power point of the wind turbines
[2]. In addition, if a large number of wind turbines or wind farms
are directly connected to the power system, the need to maintain
the power and load balancing in terms of the power system stability
at all times and to ensure a smooth supply of electricity owing to an
instantaneous change of wind power becomes a very important
issue [1, 2]. On the other hand, RERs such as wind energy cannot
directly replace the existing centralised electric energy
technologies. The centralised electric power plants are by far well
established, while the new developed RESs are not sufficiently
developed to meet the total energy demand. Therefore, it is rational
to infuse RERs into existing grids gradually by using of the smart
grid and power electronics technologies, and transform the power
system over time [3].
There are several methods that can be used to gentle wind
power fluctuations on the grid. The simplest and cheapest solution
involves the use of the pitch control of the blades [4]. However, the
rapid change in the blade pitch is likely to cause torque pulsation of
the wind turbine. Since the blade usually has a large inertia, the
torque pulsation will have a direct effect on the gear box resulted in
susceptible damage on the shaft and gearbox of the wind turbine.
With the development of new energy storage technology, the use of
energy storage systems to alleviate fluctuations of wind power on
the grid has been taken gradually in many wind farms. Currently,
the most commonly adopted energy storage systems include hydro
pump storage, compressed air energy storage, flywheel storage [5],
storage [8, 9] and battery energy storage [10] etc. Moreover, the
use of battery energy storage system (BESS) to reduce the
volatility of the output power of wind farm with various power
smoothing control methods is an important research topic [1013].
Most of the research using a first-order low-pass filter (FOLPF) to
track the waveform of the wind farm output power, while the
difference between the original power waveform and the smoothed
power waveform is compensated by the BESS [10]. However, the
performance of the FOLPF depends entirely on its time constant.
Using a large time constant results in a better smoothing of the
output power of the wind farm but large battery capacity is
required. Conversely, when the time constant is small, the required
battery capacity decreases; however, the resulted smoothing effect
could be unacceptable. Therefore, the compromise between the
smoothing effect of the power curve and the capacity of the battery
is necessary; it can achieve the savings in the energy storage
capacity while allowing the power curve within an acceptable
range to improve the power quality. Several methods have been
proposed to adjust the smoothing effect of the power curve.
Examples are the adjusting of the time constant of a resilient low-
pass filter using two-time-scale coordination [11], a dual-layer
control strategy consisting of a fluctuation mitigation control [12]
and application of the concept of automatic segmentation moving
average method [13]. These control methods can enable the BESS
effectively managing the impact of output power variation of the
wind farm. In addition, a combined problem formulation for
activereactive optimal power flow (A-R-OPF) in distribution
networks with embedded wind generation and battery storage has
been proposed in [14]. Moreover, the A-R-OPF method has been
extended in [15] by developing a flexible battery management
system which is accomplished by optimising the hours of charge
and discharge periods of battery storage systems for each day.
Although sizing the battery systems for wind generators in power
systems should consider many aspects such as storage life, state of
charge (SOC), reactive power provision, curtailment processes and
electricity market requirements [14, 15], since this paper focuses
on the development of a BESS with intelligent wind power
smoothing control using recurrent fuzzy neural network (RFNN),

IET Renew. Power Gener., 2017, Vol. 11 Iss. 2, pp. 398-407 398
The Institution of Engineering and Technology 2016

Fig. 1 Characteristics of wind power generator and operating modes
only the minimum energy capacity of the battery to achieve the
requirement of the smoothing control and the SOC of the battery
are considered in this paper.
Several intelligent control methods such as FNN have been
used as wind power smoothing control methods [5, 16]. In [5], an
adaptive linear neuro-based control with a flexible learning rate
was proposed to mitigate the power fluctuation of a wind farm. In
[16], the use of fuzzy control rules to adjust the parameters of
Kalman filter was proposed to smooth the output power fluctuation
of a wind power generation system. Although the FNN possesses
the advantages of the capability of fuzzy reasoning in handling
uncertain information and the capability of artificial NNs in
learning from processes, the major drawback of the existing FNN
is the limitation of the application domain to static problem due to
their inherent feedforward network structure. On the other hand,
since the recurrent neuron has an internal feedback loop to capture
the dynamic response of a system without external feedback
through delays, the RFNN has superior capabilities than the FNN
[17, 18].
In this paper, an RFNN is adopted for the development of wind
power smoothing control. Actual wind data from the wind farm
built in Penghu archipelago, Taiwan, is adopted as the input signal
of the designed wind power generator model in this paper. Then,
the three-phase output currents of the wind power generator are
converted to dq-axis current components where the q-axis current
represents the active power. The resulted q-axis current is the input
of the RFNN power smoothing control. By using the excellent
approximation and online training abilities of the RFNN with a
grid power change rate limitation, a smoothed power with very
limited time delay can be obtained. Moreover, the difference of the
actual wind power and smoothed power is supplied by the BESS.
Therefore, using the RFNN power smoothing control, a minimum
energy capacity of the battery can be achieved with acceptable
power quality of the grid, i.e. the grid active power fluctuation
limit per minute is set to be 10% of the rated power [19]. This
paper is organised as follows. Section 2 illustrates the modelling of
wind power generator and BESS. Section 3 describes the RFNN
smoothing architecture including the network structure of the
RFNN and its online learning algorithms. Section 4 provides some
experimental results and conclusions are given in Section 5.
2System architecture
2.1 Modelling of wind power generator
A wind turbine extracts kinetic energy from the swept area of its
blades. The power in the airflow is given by
1 transmission between the battery terminal and the DC bus and the
Pair = AU 3 (1)
2 control of DC bus voltage vdc. The second stage is a three-level
IET Renew. Power Gener., 2017, Vol. 11 Iss. 2, pp. 398-407
The Institution of Engineering and Technology 2016
where Pair is the power from the wind; is the air density in
kilograms per cubic metres; A is the exposed area in square metres;
and U is the wind speed in metres per second. Although (1) gives
the power available in the wind, the power transferred to the wind
turbine rotor is reduced by the power coefficient Cp as follows:
Pm
Cp = (2)
Pair
1
Pm = CpPair = C AU 3 (3)
2 p
where Cp is the wind turbine power coefficient and Pm is the output
power of the wind turbine. A maximum value of Cp is defined by
the Betz limit, which states that a turbine can never extract more
than 59.3% of the power from an air stream. In reality, wind
turbine rotors have maximum Cp values in the range 2545%. The
developed aerodynamic torque can then be calculated as
Pm
Tm = (4)
m
where T m is the torque provided by the wind turbine and m is the
speed of the wind turbine. Moreover, the dynamic equation of the
generator is in the following:
de
Tm = J + Be + T e (5)
dt
where T e is the electromagnetic torque of the generator; J is the
moment of inertia of the generator; B is the friction coefficient of
the generator; and e is the speed of the generator. Assume the
generator is direct driven by the wind turbine, then
1
(T m T e) = e (6)
JS + B
Thus, the available output power of the generator can be
represented as
mT m = eT e = Pwind (7)
where is the efficiency of the generator.
In this paper, the model of wind power generator with 2kW is
established using power sim (PSIM) with setting as 0.95 and Cp
setting as 0.4 as shown in Fig. 1. Three operating modes of the
wind power generator are: (i) maximum power point tracking
mode: when Pwind < Pmax, the wind power generator is operated at
maximum power point; (ii) constant power mode: when the wind
speed is too high and the output power exceeds the maximum
output power (Pmax), Pwind is required to maintain at Pmax; and (iii)
brake mode: when the wind speed exceeds the maximum set speed,
the wind turbine must be operated in brake mode to avoid
damaging the wind turbine mechanical structure. Moreover, to
verify the effectiveness of the PSIM model of the wind power
generator, actual wind data from the wind farm built in Penghu
archipelago, Taiwan, is adopted as the input signal of the designed
wind power generator model as shown in Fig. 2.
2.2 Battery energy storage system
In this paper, a BESS with two-stage circuit architecture is
developed as a wind power power smoothing control as shown in
Fig. 3. The first stage is a bidirectional interleaved DC-to-DC
converter [20, 21] and is responsible for the two-way power
neutral-point clamped (NPC) inverter [21] and is responsible for
supplying the power difference of the actual wind power and
399
Fig. 2 PSIM simulation of wind power generator for
(a) Profile of actual wind speed, (b) Output power of wind turbine generator

smoothed power will be supplied by the BESS. The BESS and the and integral (PI) controller to produce the battery charge current
wind power generation emulator are connected to the grid as command ibat. Then, the current command ibat is divided by three
shown in Fig. 3. Moreover, in Fig. 3, iga, igb, igc are the three-phase for the three arms of the interleaved DC/DC converter, and the
output currents of the wind power generator; ioa, iob, ioa are the differences between the current command and the sensed phase
three-phase output currents of the DC/AC inverter; vdc
and vdc are current ibata, ibatb, ibatc are regulated via individual PI controller to

the DC bus voltage command and DC bus voltage; vd1 and vd2 are generate the control signal commands vbata , vbatb , vbatc for PWM.

the half high voltage and half low voltage of the DC bus voltage; When the current ibat command is positive, the circuit operates in
ibat, ibata, ibatb and ibatc are the input current command and converter boost mode, and the battery is discharged. Conversely, when the
input currents of the of the interleaved DC/DC converter; vbata
, current ibat command is negative, the circuit operates in buck mode,

vbatb, vbatc are the pulse-width modulation (PWM) voltage and the battery is charged. For the three-level NPC inverter, it is
controlled by the dq0-axis current control. The d-axis current
commands of the interleaved DC/DC converter; va , vb, vc are the
control is responsible for the reactive power control by using the
PWM voltage commands of the DC/AC inverter; va, vb, vc are the reactive power current command ido and is set to be zero. The q-
three-phase voltages of the grid; vq, vd, v0 are the dq0-axis PWM axis current control is responsible for the active power control by
voltage commands of the DC/AC inverter; iqg and iqg are the q-axis using the active power current command iqo, which is the
output current command and q-axis output current of the wind difference between the output of the RFNN iqg and the active
power generator, i.e. the active power component; idg is the d-axis power component of the wind power generator iqg. In addition, the
output current of the wind power generator, i.e. the reactive power 0-axis current control is responsible for the balance of the upper
component; ido, iqo, i0o and ido, iqo, i0o are the dq0-axis output half DC bus voltage vd1 and the lower half DC bus voltage vd2 by
current commands and dq0-axis output currents of the DC/AC using a PI controller to generate the 0-axis current command i0o.
inverter; i is the synchronous angle obtained from Phase- Locked
The individual difference between the current command ido, iqo, i0o
Loop (PLL). The axis conversion between the ioa, iob, ioa and ido,
and the current ido, iqo, i0o is regulated via three PI controllers to
iqo, i0o and vq, vd, v0 and va , vb, vc are shown in the following
generate the respective dq0-axis voltage command vq, vd, v0. Then
equations:
vq, vd, v0 is converted to voltage command va , vb, vc in the abc-axis
2 2 for sinusoidal PWM by using the synchronous angle i.
cos i cos i cos i +
ido 3 3 ioa
iqo 2 2 2 iob 3RFNN power smoothing control
= sin i sin i sin i + (8)
3 3 3
i0o ioc Owing to the excellent approximation and online training abilities
1 1 1 of the RFNN, an RFNN is adopted in this paper for the wind power
2 2 2 smoothing control. Using the RFNN power smoothing control and
considering the constraint of grid power change rate, a smoothed
cos i sin i 1 power with very limited time delay can be obtained. Since the
va vq
2 2 difference of the actual wind power and smoothed power is
vb = cos i 3 sin i
3
1 v
d (9) supplied by the BESS, thus, the minimum energy capacity of the
battery can be achieved with acceptable power quality of the grid.
vc 2 2
cos i + sin i + 1 v 0 The network structure of the RFNN and its online learning
3 3 algorithms are introduced in this section.
For the bidirectional interleaved DCDC converter, the
operating modes include boost mode and buck mode. The former 3.1 Recurrent FNN
corresponds to the battery discharge state, and the latter A four-layer RFNN as shown in Fig. 4, which comprises the input
corresponds to the battery charge state. Average current control (the i layer), membership (the j layer), rule (the k layer), and output
method is adopted where the difference of the DC bus voltage layer (the o layer), is adopted to implement the RFNN power

command vdc and DC bus voltage vdc is regulated by a proportional

400 IET Renew. Power Gener., 2017, Vol. 11 Iss. 2, pp. 398-407
The Institution of Engineering and Technology 2016
Fig. 3 BESS with intelligent wind power smoother

smoothing control in this paper. The signal propagation and the Layer 4 output layer: The single node o in this layer is labelled
basic function in each layer of the RFNN are introduced as follows: with , which computes the overall output as the summation of all
input signals
Layer 1 input layer: For every node i in this layer, the net input
and the net output are represented as net4o(N) = wkoxk(N)
4 4

k (13)
neti1(N) = x1i y4o(N) = f 4onet4o(N) = net4o(N), o=1
(10)
yi1 = f i1 neti1(N) = net1i (N), i = 1, 2
where the connecting weight w4ko is the output action strength of the
1 1
where x = e(t); x = e(t); N denotes the number of iterations.
1 2 oth output associated with the kth rule; xk4 is determined by the
Layer 2 membership layer: In this layer, each node performs a selected membership function and 0 xk4 1; and y4o is the RFNN
membership function and the Gaussian function is adopted as the output.
membership function. For the jth node
2 3.2 Online parameter training
2
x2i mi j
net j(N) = 2 To train the RFNN effectively, an online parameter training
i j (11) methodology, which is derived by using the gradient descent
y2j(N) = f net2j(N) = exp net2j(N) , j = 1, 2, , n method, is proposed. Not only the connecting weights between rule
layer and output layer are adjusted online but also the recurrent
weights and the membership functions. This training scheme will
where mi j and i j are the mean and the standard deviation,
increase the learning capability of the RFNN. First, define an
respectively, of the Gaussian function in the jth term of the ith energy function E as the following formula:
input linguistic variable xi2 to the node of layer 2, and n is the total
number of the linguistic variables with respect to the input nodes. 1 1
E= (i i )2 = e2 (14)
Layer 3 rule layer: Each node k in this layer is denoted by , 2 qg qg 2
which multiplies the input signals and outputs the result of product.
For the kth rule node in which e is the difference between iqg and iqg. The detailed
learning algorithms are described below.
netk3 (N) = w jkx j(N)wkyk(N 1)
3 3 3

j (12) Layer 4 output layer: The error term to be propagated is given

yk3 (N) = f 3k netk3 (N) = netk3 (N), k = 1, 2, , l 4
E E e iqg yo
4o = = (15)
where x3j are the jth input to the node of layer 3; w3jk are the weights net4o e iqg y4o net4o
between the membership layer and the rule layer and are assumed
to be unity; wk are recurrent weights for the units in the rule layer; and the weight is updated by the amount
i
l = n/i are number of rules with complete rule connection if each
input node has the same linguistic variables.

IET Renew. Power Gener., 2017, Vol. 11 Iss. 2, pp. 398-407 401
The Institution of Engineering and Technology 2016
 The mean and standard deviation of the hidden layer are updated as
follows:

mi j(N + 1) = mi j(N) + mi j (23)

i j(N + 1) = i j(N) + i j (24)

3.3 RFNN power smoothing control

In the power smoothing control, first, the output three-phase
currents of the wind power generator are converted to dq-axis
current components. The resulted q-axis current iqg represents the
active current component of the wind power. The difference
between the output of the RFNN iqg and iqg, i.e. e(t), and its
derivative e(t) are the two inputs of the RFNN power smoothing
control. After that, the output signal iqg is generated through the
feedforward structure of the RFNN following by the online
parameters learning of the network. This cycle is repeated every
10ms. Since the learning process is to minimise the difference
between iqg and iqg, i.e. e(t), the curve of the output of RFNN iqg
will be quite similar to the curve of iqg with limited time delay. On
the other hand, though the RFNN possesses excellent
Fig. 4 Structure of four-layer RFNN approximation ability but it is not perfect, and the non-perfect iqg
4 4 results in a more gentle wind power curve to achieve the effect of
E E yo neto wind power smoothing. Therefore, the RFNN wind power
w4ko = 4 = 4 4 4 = o4 xk4 (16)
wko yo neto wko smoothing scheme not only can effectively reduce the amount of
Layer 3 rule layer: Since the weights in this layer are unity, only battery capacity but also takes the power quality of the grid into
the error term in the following needs to be calculated and account.
propagated: In Fig. 3, the difference between iqg and iqg is also indicated as
iqo. When iqo is negative indicating that the output power of the

4 4 3
E E yo neto yk wind power generator is greater than the smoothing power, the
k3 = 3 = = 4ow4ko (17)
netk yo neto yk netk3
4 4 3
three-level NPC inverter will absorb excess power resulted in the
rising of the DC bus voltage vdc. Therefore, the current ibat
The update law of the recurrent weight wk also can be obtained command will be negative. The bidirectional interleaved DCDC
by the following equation: converter will operate in buck mode, and the battery will be
charged. On the other hand, when iqo is positive indicating that the
4 4 3 3
E E yo neto yk netk output power of the wind power generator is smaller than the
wk = = 4 smoothing power, the three-level NPC inverter will compensate
wk yo neto yk netk wk
4 3 3
(18) insufficient power resulted in the decreasing of the DC bus voltage
= 3k x3j(N)w3jk yk3 (N 1) vdc. Then, the current ibat command will be positive. The
j
bidirectional interleaved DCDC converter will operate in boost
mode and the battery will be discharged.
The recurrent weights are updated as follows:

wk(N + 1) = wk(N) + k (19) 3.4 Comparison of smoothing methods

The main objective of the research is to determine the necessary
Layer 2 membership layer: The error term is computed as capacity of battery for use in conjunction with a wind power
follows: generator that allows the output of combined wind power generator
and BESS to meet the connected grid requirements, i.e. the
4 4 3 3 2
E E yo neto yk netk y j fluctuation of injected power to the grid should be kept below some
2j = = percentage of the rated power to maintain the grid power quality
net2j net net4o y3k netk3 y2j net2j (20) [19]. To verify the effectiveness of the proposed control strategy,
= kyk(N)
3 3
simulation analyses were performed using the wind power
generation model and MATLAB/SIMULINK. Actual 24h wind
The update laws of mi j and i j are data from the wind farm built in Penghu archipelago, Taiwan, is
adopted as the input signal of the designed wind power generator
4 4 3 3 2 2 model.
E E yo neto yk netk y j net j The comparison of smoothing methods is shown in Figs. 5ad
mi j = = 4
mi j yo net4o yk3 netk3 y2j net2j mi j including the average method, the moving average method [22],
(21) the low-pass filter method [11] and the RFNN method. In Figs. 5a
2
2(x2i mi j) d, Pwind is the actual wind power and Pout in the smoothed power
= j
(i j)2 which can be calculated from iqg. The required energy capacity of
4 4 3 3 2 2
the battery could be described by the following equation:
E E yo neto yk netk y j net j
i j = = 4
i j yo neto4 y3k netk3 y2j net2j i j i=a+N

(22) EBESS = Max ( Pwind(ti) Pout(ti) )t ,

2(xi2 mi j)
2
i=a (25)
2
= j Pwind > Pout or Pout > Pwind
(i j)3

402 IET Renew. Power Gener., 2017, Vol. 11 Iss. 2, pp. 398-407
The Institution of Engineering and Technology 2016
where a and a+N are the start and stop time of continuous
charging or discharging; t is the sampling interval which is 1s.
As shown in Figs. 5ad, when the condition Pwind > Pout changes
to Pwind < Pout, the start time a is determined and N is reset to zero.
Then, N is counted up every 1s until the condition Pwind > Pout
reaches again. The resulted energy capacities using (25) of the
various methods as shown in Figs. 5ad are shown in Table 1
where the average method requires the largest energy capacity of
the battery and the RFNN method required the smallest energy
capacity of the battery. The required power capacities of the battery
of various smoothing methods are also given in Table 1 using the
following equation:
PBESS = Max{ Pwind(ti) Pout(ti) } (26)
From the data shown in Table 1, the smallest power capacity of
the battery still can be obtained by the proposed intelligent
smoothing method using RFNN. Moreover, the responses from 13
to 16h, which are shown in Figs. 5ad, are enlarged and shown in
Figs. 5eh. From the simulated results shown in Figs. 5eh, the
time delay phenomena of the smoothed curves of the moving
average and FOLPF methods are very obvious. Only a smoothed
power curve with very limited time delay and with the minimum
energy capacity of the battery can be obtained by the proposed
intelligent smoothing method using RFNN. Furthermore, in this
paper, the SOC estimation using Coulomb counting method [23] is
adopted for the management of the active power. In addition, in the
energy management of leadacid battery, the ordinary minimum
and maximum SOC is 10 and 90%. For the comparison of various
smoothing methods and considering the allowed minimum SOC, a
BESS consisting of 30 leadacid batteries with 12V/12Ah is
adopted in this paper. The resulted SOC responses of the average
method, the moving average method, the low-pass filter method
and the RFNN method are shown in Figs. 5il. From the simulated
results, the SOC of the average method reaches nearly its lowest
limit 10%. On the other hand, there are still plenty energy of
capacities for all the other three methods especially for the RFNN
method. Additionally, if the SOC of the average method reaches its
lowest limit 10% and highest limit 90%, the discharge and charge
functions will be disabled by the control algorithm of the BESS,
and the smooth control of the BESS fails. Thus, the power quality
of the connected grid is degraded.
To further compare the smoothing performance of various
smoothing methods, the standard deviation of the smoothed power
is defined as follows:
h
(Pwind(ti) Pout(ti)) /h
2
(27)
i=1
where h is the total number of sampling of 24h. The standard
deviations of smoothed power at various smoothing methods are
shown in Table 2 where the RFNN method still has the smallest
value. In addition, the grid active power fluctuation limit per
minute is set to 10% of the rated power. The satisfaction of 10%
grid power fluctuation limit using the proposed RFNN is shown in
Fig. 6, where 1min average window is shown in Fig. 6a and 10
min average window is shown in Fig. 6b. Therefore, the impact to
the grid power quality of the RFNN power smoothing control is
also promised. Although the FOLPF has very similar standard
deviation with a much simpler implementation, the time constant
should be tuned off-line for different wind data in order to achieve
favourable wind power smoothing control. On the other hand, the
online training ability of the RFNN can ensure good wind power
smoothing control for different wind data.
4Experimental set-up and experimentation
4.1 Experimental set-up
The block diagram of the digital signal processor (DSP)-based
BESS including two TMS320F28335 control boards DSP1 and
DSP2 and a Personal Computer (PC)-based wind power generator,
IET Renew. Power Gener., 2017, Vol. 11 Iss. 2, pp. 398-407
The Institution of Engineering and Technology 2016
is shown in Fig. 7. In this paper, only the active power of the wind
power, smoothed power and power supplied by the BESS
considered. The difference of the wind power Pwind and smoothed
power Pout is supplied by the BESS Pinv. Therefore, in Fig. 7, the q-
axis current control is responsible for the active power control of
the BESS by using the active power current command iqo, which is
the difference between the output of the RFNN iqg and the active
power component of the wind power generator iqg. The d-axis
current control is responsible for the reactive power control by
using the reactive power current command ido and is set to be zero.
Furthermore, the installed locations of the two power meters are
also demonstrated in Fig. 7. The values of the two DC bus
capacitors and the three filter inductors of the inverter are 1300F
and 1.6mH, respectively. In addition, the photographs of the PC-
based wind power generator and the DSP-based BESS are shown
in Fig. 8.
For the 2kW PC-based wind power generator, a current-
controlled PWM inverter is adopted. Moreover, modelling of the
wind power generator, field-orientation and synchronisation
algorithm are realised via Simulink. In addition, three actual wind
power curves with 600s are built in the wind power generator
which are the variations of the wind speed 7101210m/s, 912
86m/s and 810528m/s.
Both the bidirectional interleaved DC/DC converter and 3-arm
3-level inverter are implemented using an intelligent power
modules (SK30MLI066) manufactured by Semikron Co. with a
switching frequency of 10kHz. The capacity of the inverter is 5
kW and the DC bus voltage is 430V. Moreover, the battery is
consisted of 30 leadacid batteries with 12V/12Ah. In addition, to
show the effectiveness of the wind smoothing control with small
number of neurons, 6 and 9 neurons are adopted at the membership
and rule layers in the RFNN by empirical rules to achieve fast
dynamic responses of the smoothing control and to reduce the
execution time simultaneously.
4.2 Experimentation
The experimental results using the proposed RFNN smoothing
method at the wind speed variation condition 7101210m/s are
shown in Fig. 9 where the wind power and the smoothing power
are shown in Fig. 9a; the measured three-phase currents at the wind
power generator are shown in Fig. 9b; the measured three-phase
currents at the grid side are shown in Fig. 9c; and the measured
three-phase output currents of the inverter are shown in Fig. 9d.
Moreover, the experimental results using the proposed RFNN
smoothing method at the wind speed variation condition 91286
m/s are shown in Fig. 10 where the wind power and the smoothing
power are shown in Fig. 10a; the measured three-phase currents at
the wind power generator are shown in Fig. 10b; the measured
three-phase currents at the grid side are shown in Fig. 10c; and the
measured three-phase output currents of the inverter are shown in
Fig. 10d. Furthermore, the experimental results using the proposed
RFNN smoothing method at the wind speed variation condition 8
10528m/s are shown in Fig. 11 where the wind power and the
smoothing power are shown in Fig. 11a; the measured three-phase
currents at the wind power generator are shown in Fig. 11b; the
measured three-phase currents at the grid side are shown in
Fig. 11c; and the measured three-phase output currents of the
inverter are shown in Fig. 11d. From the experimental results of
various wind variation conditions, the effectiveness of the proposed
intelligent wind power smoothing control is obvious. The
phenomena of time delay are very limited at all test conditions. In
addition, from the measured three-phase currents at the wind power
generator, the measured three-phase currents at the grid side and
the measured three-phase output currents of the inverter, the
difference between the three-phase currents at the wind power
generator and the three-phase currents at the grid side is
compensated by the three-phase output currents of the inverter at
all test conditions.
In this paper, the sampling intervals of the interrupt service
routines (ISRs) of the control algorithms for the bidirectional
interleaved DC-to-DC converter by DSP1 and the three-level NPC
403
Fig. 5 Comparison of smoothing methods
(a) Average method, (b) Moving average method, (c) Low-pass filter method, (d) RFNN method, (e) Enlarged responses of average method, (f) Enlarged responses of moving
average method, (g) Enlarged responses of low-pass filter method, (h) Enlarged responses of RFNN method, (i) SOC of average method, (j) SOC of moving average method, (k)
SOC of low-pass filter method and (l) SOC of RFNN method

inverter by DSP2 as shown in Fig. 7 are both 0.1ms. Moreover,
considering the relatively slow variation of the wind power, the
sampling interval of the ISR of various smoothing control
algorithms by DSP2 is 40ms. In addition, the execution or
computing time of the C program in the TMS320F28335 32bit
floating-point DSP can be obtained by the clock tool of Texas
Instruments Code Composer Studio v5 program editing interface.
The total operation cycles and execution time of the FOLPF and
RFNN are compared in Table 3. As a result, the total execution
time of the RFNN is much <40ms. Therefore, there is no extra

404 IET Renew. Power Gener., 2017, Vol. 11 Iss. 2, pp. 398-407
The Institution of Engineering and Technology 2016
Fig. 6 Fluctuation percentage of the injected power
(a) 1min average window, (b) 10min average window

Fig. 7 DSP architecture of BESS

hardware cost requirement for the implementation of the RFNN

method, except the extra coding time for the RFNN method.
Table 1Required energy capacity and power capacity of
battery at various smoothing methods 5Conclusions
Average Moving average FOLPF RFNN An intelligent wind power smoothing control using RFNN has
energy capacity, kWh 2.64 0.576 0.539 0.43 been successfully developed in this paper. By using the excellent
power capacity, kW 1.125 1.0 0.915 0.877 approximation and online training abilities of the RFNN, a
smoothed power curve with very limited time delay has been
obtained. Moreover, the difference of the actual wind power and
Table 2Standard deviation of smoothed power at various smoothed power is supplied by the BESS. Comparing with the
smoothing methods other smoothing methods, a minimum energy capacity of the
Average Moving average FOLPF RFNN battery has been achieved using the RFNN power smoothing
kW 0.468 0.339 0.291 0.277
control. From the experimental results of various wind variation
conditions, the effectiveness of the proposed intelligent wind

IET Renew. Power Gener., 2017, Vol. 11 Iss. 2, pp. 398-407 405
The Institution of Engineering and Technology 2016
Fig. 8 Photographs of experimental set-up
(a) PC-based wind power generator, (b) DSP-based two stages BESS

Fig. 9 Experimental results using RFNN smoothing method at wind speed Fig. 10 Experimental results using RFNN smoothing method at wind
7101210m/s speed 91286m/s
(a) Wind power and smoothing power, (b) Measured three-phase currents at wind (a) Wind power and smoothing power, (b) Measured three-phase currents at wind
power generator, (c) Measured three-phase currents at grid side, (d) Measured three- power generator, (c) Measured three-phase currents at grid side, (d) Measured three-
phase output currents of inverter phase output currents of inverter

power smoothing control has been clearly shown. In addition, the 6Acknowledgment
fluctuation of injected power to the grid per minute is kept below
The author acknowledge the financial support of the Institute of
10% of the rated power to maintain the grid power quality at all
Nuclear Energy Research of Taiwan, R.O.C. through its grant
test conditions. The major contributions of this paper are: (i) the
NL1040300.
development of a BESS which is consisted of a bidirectional
interleaved DC/DC converter and a 3-arm 3-level inverter; (ii) the
development of an intelligent wind power smoothing control using
RFNN; and (iii) the minimum energy capacity of the battery and a
small fluctuation of the grid active power being achieved
simultaneously by using the RFNN wind power smoothing control.

406 IET Renew. Power Gener., 2017, Vol. 11 Iss. 2, pp. 398-407
The Institution of Engineering and Technology 2016
 [3] Liserre, M., Sauter, T., Hung, J.Y.: Future energy systems: integrating
renewable energy sources into the smart power grid through industrial
electronics, IEEE Ind. Electron. Mag., 2010, 4, (1), pp. 1837
[4] Iyasere, E., Salah, M., Dawson, D., et al.: Optimum seeking-based non-linear
controller to maximise energy capture in a variable speed wind turbine, IET
Control Theory Appl., 2012, 6, (4), pp. 526532
[5] Jannati, M., Hosseinian, S.H., Vahidi, B., et al.: Mitigation of windfarm
power fluctuation by adaptive linear neuron-based power tracking method
with flexible learning rate, IET Renew. Power Gener., 2014, 8, (6), pp. 659
669
[6] Ise, T., Kita, M., Taguchi, A.: A hybrid energy storage with a SMES and
secondary battery, IEEE Trans. Appl. Supercond., 2005, 15, (2), pp. 1915
1918
[7] Nomura, S., Ohata, Y., Hagita, T., et al.: Wind farms linked by SMES
systems, IEEE Trans. Appl. Supercond., 2005, 15, (2), pp. 19511954
[8] Kinjo, T., Senjyu, T., Urasaki, N., et al.: Output levelling of renewable
energy by electric double-layer capacitor applied for energy storage system,
IEEE Trans. Energy Convers., 2006, 21, (1), pp. 221227
[9] Li, X., Hu, C., Liu, C., et al.: Modeling and control of aggregated super-
capacitor energy storage system for wind power generation. IEEE Ind.
Electron. Conf., Orlando, FL, USA, 1013 November 2008, pp. 33703375
[10] Han, X., Chen, F., Cui, X., et al.: A power smoothing control strategy and
optimized allocation of battery capacity based on hybrid storage energy
technology, Energies, 2012, 5, (5), pp. 15931612
[11] Jiang, Q., Wang, H.: Two-time-scale coordination control for a battery
energy storage system to mitigate wind power fluctuations, IEEE Trans.
Energy Convers., 2013, 28, (1), pp. 5261
[12] Jiang, Q., Gong, Y., Wang, H.: A battery energy storage system dual-layer
control strategy for mitigating wind farm fluctuations, IEEE Trans. Power
Syst., 2013, 28, (3), pp. 32633273
[13] Wang, W., Mao, C., Lu, J., et al.: An energy storage system sizing method
for wind power integration, Energies, 2013, 6, (7), pp. 33923404
[14] Aouss, G., Pu, L.: Activereactive optimal power flow in distribution
networks with embedded generation and battery storage, IEEE Trans. Power
Syst., 2012, 27, (4), pp. 20262035
[15] Aouss, G., Pu, L.: Flexible optimal operation of battery storage systems for
energy supply networks, IEEE Trans. Power Syst., 2013, 28, (3), pp. 2788
2797
[16] Li, X.: Fuzzy adaptive Kalman filter for wind power output smoothing with
battery energy storage system, IET Renew. Power Gener., 2012, 6, (5), pp.
340347
Fig. 11 Experimental results using RFNN smoothing method at wind [17] Lin, F.J., Wai, R.J.: Hybrid control using recurrent fuzzy neural network for
linear induction motor servo drive, IEEE Trans. Fuzzy Syst., 2011, 9, (1), pp.
speed 810528m/s 102115
(a) Wind power and smoothing power, (b) Measured three-phase currents at wind [18] Lin, F.J., Tan, K.H., Fang, D.Y., et al.: Intelligent controlled three-phase
power generator, (c) Measured three-phase currents at grid side, (d) Measured three- squirrel-cage induction generator system using wavelet fuzzy neural network
phase output currents of inverter for wind power, IET Renew. Power Gener., 2013, 7, (5), pp. 552564
[19] Fox, B., Flynn, D., Bryans, L., et al.: Wind power integration: connection
and system operation aspects, IET Power Energy Serv., 2007, 50, pp. 167
Table 3Total operation cycles and execution time at 169
various smoothing methods [20] Serkan, D., Amin, H., Alireza, K.: Comparative analysis of bidirectional
three-level DCDC converter for automotive applications, IEEE Trans. Ind.
Control algorithms FOLPF RFNN Electron., 2015, 62, (5), pp. 33053315
total operation cycles 75 9525 [21] Hiroyuki, K., James, S., Seana, M., et al.: Design and performance
total execution time 0.0005ms 0.0635ms evaluation of a 200C interleaved boost converter, IEEE Trans. Power
Electron., 2013, 28, (4), pp. 16911699
[22] Sheikh, M.R.I., Muyeen, S.M., Takahashi, R., et al.: Minimization of
fluctuations of output power and terminal voltage of wind generator by using
STATCOM/SMES. Proc. 2009 IEEE Bucharest Power Tech Conf.,
7References Bucharest, pp. 16
[1] Shafiullah, G.M., Oo, A.M.T., Ali, A.B.M.S., et al.: Potential challenges of [23] Coleman, M., Lee, C.K., Zhu, C., et al.: State-of-charge determination from
integrating large-scale wind energy into the power grid a review, Renew. EMF voltage estimation using impedance, terminal voltage, and current for
Sustain. Energy Rev., 2013, 20, pp. 306321 leadacid and lithium-ion batteries, IEEE Trans. Ind. Electron., 2007, 54, (5),
[2] Jannati, M., Hosseinian, S.H., Vahidi, B., et al.: A survey on energy storage pp. 25502557
resources configurations in order to propose an optimum configuration for
smoothing fluctuations of future large wind power plants, Renew. Sustain.
Energy Rev., 2014, 29, pp. 158172

IET Renew. Power Gener., 2017, Vol. 11 Iss. 2, pp. 398-407 407
The Institution of Engineering and Technology 2016