ARTICLE IN PRESS

Renewable Energy 33 (2008) 1435–1443

www.elsevier.com/locate/renene

Review

Neuro-fuzzy dynamic model with Kalman filter to forecast irradiance

and temperature for solar energy systems
Maher Chaabenea,b,, Mohsen Ben Ammara,b
a
The High Institute of Technological Studies (ISET), Sfax, Tunisia
b
Unité de commande de machines et énergies renouvelables CMER, ENIS, Tunisia
Received 7 December 2006; accepted 15 October 2007
Available online 26 November 2007

Abstract

This paper introduces a dynamic forecasting of irradiance and ambient temperature. The medium term forecasting (MTF) gives a daily
meteorological behaviour. It consists of a neuro-fuzzy estimator based on meteorological parameters’ behaviours during the days before,
and on time distribution models. As for the short term forecasting (STF), it estimates, for a 5 min time step ahead, the meteorological
parameters evolution. It is ensured by the Auto-Regressive Moving Average (ARMA) model of the MTF associated to a Kalman filter.
STF uses instantaneous measured data, delivered by a data acquisition system, so as to accomplish the forecast. Herein we describe our
method and we present forecasting results. Validation is based on measurements taken at the Energy and Thermal Research Centre
(CRTEn) in the north of Tunisia. Since our work delivers accurate meteorological parameters forecasting, the obtained results can be
easily adapted to forecast any solar conversion system output.
r 2007 Elsevier Ltd. All rights reserved.

Keywords: Meteorological forecasting; Modelling; Neuro-fuzzy; ARMA; Kalman filter

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1436
2. The estimator principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1436
3. The MTF model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1437
3.1. The neuro-fuzzy estimator approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1437
3.2. Meteorological behaviour estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1438
3.3. Irradiance distribution model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1439
3.4. Ambient temperature distribution model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1439
4. Short term forecasting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1439
4.1. Kalman filter approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1439
4.2. Application to meteorological parameter forecasting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1440
5. Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1441
6. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1443
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1443

Corresponding author. Rte Mahdia Km 2, BP88A, ElBustan 3099-Sfax, Tunisia. Tel.: +216 74431425; fax: +216 74431386.
E-mail address: maher.chaabane@isetsf.rnu.tn (M. Chaabene).

0960-1481/$ - see front matter r 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.renene.2007.10.004
 ARTICLE IN PRESS
1436 M. Chaabene, M. Ben Ammar / Renewable Energy 33 (2008) 1435–1443
1. Introduction
Accurate forecasting of irradiance and ambient tem-
perature behaviours during daylight is required in most
solar energy applications, particularly in design methods,
in system characterization and in decision making for
energy management. Different models were introduced in
literature so as to forecast irradiance. Evaluations, with
respect to measured data uncertainty, of the accuracy of
many model approaches where no measurements are
available were proposed [1]. More complex modelling
methods gave empirical models or spatial and temporal
variations [2,3]. Measurements based models were also
proposed using either METEOSAT images [4] or meteor-
ological station measurements [5]. Recent methods for
forecasting daily total irradiance such as adaptive wavelet
network [6] and fuzzy logic [7] were introduced. Since these
researches aimed to offer tools for solar plants sizing, they
estimated yearly, monthly and daily irradiance relying on
long term data which is not always available for all sites.
Other researches used weather forecasts provided via
e-mail by meteorological institutes [8,9] in order to build
solar plants control. Even if forecasted meteorological
parameters are truthful, the method remains unsuitable for
all solar applications as it depends on Internet connection
state.
Although produced results are significant, they present
many insufficiencies. First, most researches were interested
in irradiance modelling without considering ambient
temperature. Second, they did not estimate correctly and
easily in short term the quantity of solar energy collected
which prevented real time plant energy forecast and
management. In addition, since meteorological parameters
are stochastic signals, the introduced methods could not
provide accurate forecasting following an unexpected
weather disturbance. This paper reveals firstly a medium
term forecasting (MTF) of irradiance and ambient
temperature relying on their behaviours during the days
( d − 1) G1 Tmin 1 Tmax 1
DB(d − 1) = (d −2) G2 Tmin 2 Tmax 2
( d − 20 ) G20 Tmin 20 Tmax 20
Meteorological
Neuro-Fuzzy
parameters
estimator
distribution models
Meteorological behaviour estimation
~ ~
G( d )
~ ~
E ( d ) = Tmax ( d )
~ Tmf (d , t )
Tmin ( d )
Fig. 1. Medium and short term forecasting principle.
before. This task is ensured by a neuro-fuzzy estimator.
The obtained MTF is modelled by an Auto-Regressive
Moving Average (ARMA). A Kalman filter provides
short term forecasting (STF) using the ARMA model
and the measured data of meteorological parameters since
sunrise. Once entered to an input/output solar plant model,
the proposed MTF and STF models allow daily and
instantaneous realistic forecasting of the plant energy
supply. The developed models have been tested and
validated by analysing the normalized root mean square
error (NRMSE) and the normalized mean bias error
(NMBE).
2. The estimator principle
The meteorological estimator consists of two types of
dynamic forecasting:
 MTF : it delivers an estimated time distribution for the
day (d) of irradiance (I~mf ðd; tÞ) and ambient temperature
(T~ mf ðd; tÞ).
 STF: it gives, during the same day d for a time step
assumed to be 5 min, the real time forecasting of irra-
diance (I^sf ðdÞkþ1jk ) and ambient temperature (T^ sf ðdÞkþ1jk ,
Fig. 1).
The MTF consists of Gaussian distributions of the
~
cumulated irradiance GðdÞ and the maximum and mini-
mum ambient temperatures ðT~ max ðdÞ; T~ min ðdÞÞ during the
day d, which are represented by the vector EðdÞ.~ ~
EðdÞ is
estimated on the basis of the meteorological behaviour
during the last 20 days DBðd  1Þ, using a neuro-fuzzy
estimator. A Kalman filter provides the STF by carrying
out the MTF from the ARMA model and by consider-
ing the k last measurements of meteorological para-
meters represented by Y ðd; kÞ which are taken during the
same day d.
Meteorological station
..., I ( d , k − 1 ), I ( d , k )
Y( d ,k ) =
...,T ( d , k − 1 ),T ( d , k )
ARMA model
A(q-1), C(q-1)
Medium term forecasting Short term forecasting
~ I mf (d , t ) Iˆsf (d ) k +1 k
X mf (d , t ) = ~ Xˆ sf (d ) k +1 k =
Tˆsf (d ) k +1 k
 ARTICLE IN PRESS
M. Chaabene, M. Ben Ammar / Renewable Energy 33 (2008) 1435–1443 1437

3. The MTF model
Three meteorological parameter amounts are estimated
for the day d: the cumulated irradiance, the minimal
ambient temperature and the maximal one. The estimation
is based on the ANFIS neuro-fuzzy estimator. The
obtained amounts are exploited so as to determine a time
function for each meteorological parameter using two time
distribution models [11,12].
3.1. The neuro-fuzzy estimator approach
The Takagi and Sugeno model uses fuzzy inputs and rules
so as to provide a powerful tool for modelling complex non-
linear problems when combined with a network structure as
in the adaptive network fuzzy inference system or ANFIS
[10]. ANFIS is a class of adaptive multi-layer feed-forward
networks, applied to non-linear forecasting where past
samples are used to forecast the sample ahead. It generates
fuzzy rules from an input–output data set (called training
data). A typical fuzzy rule in a Takagi and Sugeno fuzzy
model has the format: If x is Ai and y is Bi then z ¼ f ðx; yÞ
where Ai and Bi are fuzzy sets in the antecedent; z ¼ f ðx; yÞ
takes usually a polynomial form. Consider such a model that
contains two rules:
Rule 1: If X is A1 and Y is B1 ; then f 1 ¼ p1 x þ q1 yþ r1 ,
Rule 2: If X is A2 and Y is B2 ; then f 2 ¼ p2 x þ q2 yþ r2 ,
where fpi ; qi ; ri g are the consequent parameters that are
determined during the training process. According to
Takagi and Sugeno model, ANFIS uses the least-squares
method to identify these parameters. The main objective of
ANFIS is to optimize the parameters of a given fuzzy
inference system by applying a learning procedure using a
set of input–output data pairs (training data). The
parameter optimization is done in a way such that the
error measure between the desired and the actual output is
minimized [11,13].
Fig. 2(a) illustrates graphically the fuzzy reasoning
mechanism to derive an output f from a given input [x,y].
The firing strengths w1 and w2 are usually obtained as the
product of the membership grades of the premise part, and
the output f is the weighted average of each rule’s output.
ANFIS (Fig. 2(b)) is composed of five layers. Each layer
contains several nodes described by the node function.
Adaptive nodes, denoted by squares, represent the para-
meter sets that are adjustable in these nodes, whereas fixed
nodes, denoted by circles, represent the parameter sets
that are fixed in the system. The output data from the
nodes in the previous layers will be the input in the present
layer. Fig. 2(b) shows the corresponding ANFIS structure
where nodes within the same layer perform functions of the
same type. Note that Oji denotes the output of the ith node
in jth layer. To illustrate the ANFIS procedures, we
consider the above system having two inputs [x,y] and one
output f. The node function in each layer is described
below.
Layer 1: Each node in this layer generates a membership
grade of a linguistic label. The node function of the ith
node is a generalized bell membership function:
O1i ¼ mAi ðxÞ ¼ 1=f1 þ jðx  ci Þ=ai j2bi g,

A1 B1

w1 f1 = p1 x + q1 y + r1
X Y
A2 B2

w2 f2 = p2 x + q2 y + r2
X Y

x y

Layer 1 Layer 2 Layer 3 Layer 4 Layer 5

x y
A1
w1 w1 w1 f1
x
Π N
A2
w1 f1 + w2 f2
Σ f=
w1 + w2
B1
y Π N = w1 f1 + w2 f2
w2 w2 w1 f2
B2
x y

Fig. 2. (a) Takagi and Sugeno model. (b) ANFIS architecture.

 ARTICLE IN PRESS
1438 M. Chaabene, M. Ben Ammar / Renewable Energy 33 (2008) 1435–1443
where x is the input to node i; Ai the linguistic label (small,
large, etc.) associated with this node; and fai ; bi ; ci g the
parameter set that changes the shapes of the membership
function, ANFIS employs gradient descent to fine-tune
them during the training process. Parameters in this layer
are referred to as the premise parameters.
Layer 2: Each node in this layer calculates the firing
strength of a rule via multiplication:
O2i ¼ wi ¼ mAi ðxÞmBi ðyÞ; i ¼ 1; 2
Layer 3: Node i in this layer calculates the ratio of the ith
rule’s firing strength to the total of all firing strengths:
wi
O3i ¼ wi ¼ ; i ¼ 1; 2
ðw1 þ w2 Þ
Layer 4: Node i in this layer computes the contribution
of ith rule towards the overall output using the node
function:
O4i ¼ wi f i ¼ wi ðpi x þ qi y þ ri Þ,
where w̄i is the output of layer 3 and fpi ; qi ; ri g is the
consequent parameter set to be determined during the
training process.
Layer 5: The single node in this layer computes the
overall output as the summation of contribution from each
rule:
P behaviour characterizing the day d, we consider the base
X wi f
O5i ¼ wi f i ¼ Pi .
i i wi
ANFIS in Fig. 2(b) has, as the basic learning rule, the
backpropagation gradient descent algorithm (the same
used in feed-forward neural nets) which calculates recur-
sively the error signals from the output layer backward to
the input nodes [13]. From this architecture, it is seen that
given the values of premise parameters, the overall output
function can be expressed as a linear combination of the
consequent parameters:
f ¼ w1 f 1 þ w2 f 2 ¼ ðw1 xÞp1 þ ðw1 yÞq1 þ ðw1 Þr1
þ ðw2 xÞp2 þ ðw2 yÞq2 þ ðw2 Þr2 .
Thus, ANFIS uses a mixture of least mean squares
estimation (to determine consequent parameters) and
backpropagation (to learn the premise parameters). A step
in the learning process has two passes: In the first pass
training data is brought to the inputs, the premise
parameters are assumed to be fixed and the optimal
consequent parameters are estimated by an iterative least
mean squares procedure. In the second pass the patterns
are propagated again, but this time the consequent
parameters are assumed to be fixed and backpropagation
is used to modify the premise parameters [13].
In time series estimation with ANFIS, two data sets are
required so as to predict one step ahead: ‘‘training’’ and
‘‘test’’ data. Sub-clustering is used to generate ANFIS
structure automatically. Training phase involves an itera-
tive procedure, which seeks to calculate optimum values of
system parameters by minimizing the sum of squared
differences between model predictions and training data
values. Once the training output model f is determined, a
set of testing input–output data is applied to this model in
order to compute the ANFIS prediction output.
3.2. Meteorological behaviour estimation
Time series modelling is fundamentally different from
the conceptual modelling of systems because a time series is
interpreted as an output of an unknown system. Neither
the system characteristics nor the input functions are
known. Consequently time series analysis must content
itself with estimating future output values by means of their
own past values. The principle consists of using a time
series, with a time step p, of n last measurements up to time
t, ½mðt  ðn  1ÞpÞ; . . . ; mðt  pÞ; mðtÞ as training data and
½mðt  ðn  2ÞpÞ; . . . ; mðtÞ; mðt þ 1Þ as testing data, to
estimate during a sliding time interval the p step ahead
future values of the time series ½mðt ~  ðn  2ÞpÞ; . . . ; mðtÞ;
~
~ þ 1Þ. Here mðt þ 1Þ is fixed arbitrarily at any value so
mðt
as to achieve the ANFIS testing process. Thus, the retained
output estimated value p step ahead is mðt ~ þ 1Þ. In our case
we consider, for the training phase, the day as time unit
(t ¼ d  1: ending day), n ¼ 20 (20 days ago), and p ¼ 1
(each day). In order to estimate the meteorological
of measurements DBðd  1Þ as training data and DBðdÞ as
testing data:
2 3
ðd  1Þ G 1 T min1 T max1
6 7
6 ðd  2Þ G 2 T min2 T max2 7
6 7
DBðd  1Þ ¼ 6 6 .. .. .. .. 7
7
6 . . . . 7
4 5
ðd  20Þ G 20 T min20 T max20
¼ ½V 1 V2 V3 V 4 ,
2 3
ðdÞ G T min T max
6 7
6 ðd  1Þ G 1 T min2 T max1 7
6 7
DBðdÞ ¼ 6 .. .. .. .. 7,
6 . . . . 7
4 5
ðd  19Þ G 19 T min19 T max19
where V1 is the day order vector and V2–V4 are,
respectively, the measured vectors during the last 20 last
days of the cumulated solar radiations, the minimal
temperatures and the maximal ones. G; T min ; T max are
supposed as the measured values of climatic parameters
during the day d; they take any values in order to
accomplish the ANFIS testing process.
Considering the neuro-fuzzy estimator approach
(Section 3.1), the ANFIS architecture applied to our case
is given by Fig. 3. In order to increase the prediction
accuracy, 10 partitions are fixed, which select 10 member-
ship functions, for each input. In view of Fig. 3, 20 rules are
delivered by layer 4 to layer 5 in an attempt to determine
the ANFIS output f V i . In Fig. 3, ANFIS is executed
 ARTICLE IN PRESS
M. Chaabene, M. Ben Ammar / Renewable Energy 33 (2008) 1435–1443 1439

Vi1 Vi2 ..Vi20 over the day-light of day d,

Ai1 Z DLðdÞ  
Vi1 wi1 wi1 pt
Π N ~
GðdÞ ¼ I~max ðdÞ sin dt. (2)
Ai10 0 DLðdÞ
Vi1 Vi2 ..Vi20
wi2 This yields
Bi1
Vi2 Π N ~
wi2 Σ fV pGðdÞ
Bi10
i I~max ðdÞ ¼ , (3)
2DLðdÞ
Vi1 Vi2 ..Vi20 with
Ti1
Vi20 wi20 wi20 DLðdÞ ¼ GMTsunset ðdÞ  GMTsunrise ðdÞ, (4)
Π N
Ti10 L þ  cos1 ðtgðdðdÞÞtgðfÞÞ
GMTðdÞ ¼ 12  TeðdÞ þ , (5)
Layer 1 Layer 2 Layer 3 Layer 4 Layer 5 15
Fig. 3. ANFIS architecture applied to meteorological parameter sinð2ðNðdÞ þ 10ÞÞ
estimation. TeðdÞ ¼ 0:123 cosðNðdÞ þ 87Þ  , (6)
6

separately for each vector of measure Vi (as a training data) dðdÞ ¼ 23:45 cosðNðdÞ þ 10Þ, (7)
to calculate its relative model f V i . Consequently, the
NðdÞ ¼ 0:988½DðdÞ þ 30:3ðMðdÞ  1Þ, (8)
estimated value of each climatic parameter is computed
using its specific testing data vector in DBðdÞ and the where GMTðdÞis the sunset and sunrise hours of the day d
ANFIS model f V i . Following the training and the testing with  ¼ 1 for sunset and  ¼ þ1 for sunrise, L the
process, ANFIS delivers the estimated matrix: longitude, f the latitude, TeðdÞ the time equation (in hour),
2 3 dðdÞ the declination (in degree), DðdÞ the day of the month,
ðdÞ G~ T~ min T~ max
6 7 MðdÞ the month number (January ¼ 1) and NðdÞ the
6 ðd  1Þ G~ 1 T~ min2 T~ max1 7 number of the day for the year.
~ 6 7
DBðdÞ ¼ 6 .. .. .. .. 7 ,
6 . . . . 7
4 5 3.4. Ambient temperature distribution model
ðd  19Þ G~ 19 T~ min19 T~ max19
from which the first line vector EðdÞ, ~ representing the The MTF of the ambient temperature (T~ mf ðd; tÞ) for the
climatic parameters estimated behaviour for the day d, is day d is computed using the following sinusoidal distribu-
deduced: tion [12]:
h i T~ max ðdÞ þ T~ min ðdÞ
~
EðdÞ ~
¼ GðdÞ T~ min ðdÞ T~ max ðdÞ , T~ mf ðd; tÞ ¼
2
 
~
where GðdÞ is the estimated amount of cumulated T~ max ðdÞ  T~ min ðdÞ pðt  1Þ
þ sin , ð9Þ
irradiance, T~ min ðdÞ and T~ max ðdÞ are, respectively, the 2 12
estimated minimal and maximum ambient temperatures.
where t is the time in decimal hours since sunrise (t ¼ 0 at
The ANFIS computation was conducted using the
sunrise, Eq. (5)) for day d and T~ min ðdÞand T~ max ðdÞ are
Matlab software. The ANFIS training algorithms, includ-
estimated by the neuro-fuzzy meteorological behaviour
ing the gradient method and the least-squares method, ~
estimation vector EðdÞ.
were embedded in the software of Matlab’s fuzzy inference
toolbox. After training, the ANFIS model with forecasting
4. Short term forecasting
function is applied to testing data to obtain the output
forecasting.
4.1. Kalman filter approach
3.3. Irradiance distribution model The dynamic behaviour of a stochastic system is
described by the discrete state equation (Eq. (10)) and the
The MTF of irradiance for the day d is computed using a discrete output equation (Eq. (11)) [14]:
Gaussian distribution:
  X ðk þ 1Þ ¼ AX ðkÞ þ V 1 ðkÞ, (10)
pt
I~mf ðd; tÞ ¼ I~max ðdÞ sin , (1)
DLðdÞ Y ðkÞ ¼ CX ðkÞ þ V 2 ðkÞ, (11)
where t is the time in decimal hours since sunrise (t ¼ 0 at where k is the discrete step, X the variable vector of the
sunrise), DLðdÞ the day length of day d and I~max ðdÞ the system state, Y the output variable vector, A and C are the
value of I~mf ðd; tÞ at t ¼ DLðdÞ=2. Since the estimated ARMA model coefficients which are computed by Eq. (12),
~
amount of solar energy GðdÞ ~
is obtained by integrating IðdÞ V1 and V2 are independent zero mean stochastic noises for
 ARTICLE IN PRESS
1440 M. Chaabene, M. Ben Ammar / Renewable Energy 33 (2008) 1435–1443

which covariance matrix is given by (Eq. (13)): The kth vector of meteorological parameter measure-
ments delivered by the meteorological station during
Aðq1 ÞY ðtÞ ¼ Cðq1 ÞeðtÞ, (12)
day d.
1 1 2 nA
with Aðq Þ ¼ 1 þ a1 q þ a2 q    þ anA q , Cðq Þ ¼ 1 " #
AI ðq1 Þ 0
1 þ c1 q1 þ c2 q2    þ cnC qnC , eðtÞ is the zero mean  1
Aðq Þ ¼ ; and
independent stochastic sequence of variables (white noise), 0 AT ðq1 Þ
q the time series variable, nA and nC are the orders of " #
ARMA model. 1
C I ðq1 Þ 0
" ! #  Cðq Þ ¼ .
 0 C T ðq1 Þ
V 1 ðkÞ T T
Q S
E ðV 1 ðlÞ; V 2 ðlÞÞ ¼ dkl , (13)
V 2 ðkÞ ST R The ARMA models matrixes of MTF vector X~ mf ðd; tÞ
for day d computed by Eq. (12).
where E is the covariance, Q; R; S are elements of the noise " #
covariance matrix. dkl is the Kronecker delta, which is 1 for K I ðkÞ 0
k ¼ l and 0 otherwise.
 KðkÞ ¼ :
0 K T ðkÞ
T
The filtering principle consists of two computation steps:
he matrix of the Kalman filter gain computed using
the measurement update equations and the time update
Eq. (15).
equations. The first step estimates X ðkÞ using measure-
ments up to k (denoted X^ kjk ) then using measurements up
 The estimation of the vector
2 3
to k1 (denoted X^ kjk1 ) by computing the covariance I^sf ðdÞkjk
E½X ðkÞjY ð0Þ; Y ð1Þ; . . . ; Y ðkÞ. The solution of this filtering X^ sf ðdÞkjk ¼ 4 ^ 5
problem is now given as T sf ðdÞkjk

X^ kjk ¼ AX^ kjk1 þ KðkÞ½Y ðkÞ  C X^ kjk1 , (14) is

" #" #
where K is the Kalman filter gain given by AI ðq1 Þ 0 I^sf ðdÞ
X^ sf ðdÞkjk ¼
KðkÞ ¼ ½ASkjk1 C T þ SðkÞ½CSkjk1 C T þ RðkÞ1 . (15) 0 AT ðq1 Þ T^ sf ðdÞ
kjk1
Accordingly, the relative covariance matrices Skjk and " #8" #
K I ðkÞ 0 < Iðd; kÞ
Skjk1 must also be calculated: þ
T T T
0 K T ðkÞ : Tðd; kÞ
Skjk ¼ ASkjk1 A þ QðkÞ  KðkÞ½ðCSkjk1 C þ RðkÞK ðkÞ, 9
" #" #
(16) C I ðq1 Þ 0 I^sf ðdÞ =
 .
with X^ k0 jk0 1 ¼ X̄ 0 and Sk0 jk0 1 ¼ P0 as initial conditions. 0 C T ðq1 Þ T^ sf ðdÞ ;
kjk1
As for the second step, it estimates X^ kþ1jk (Eq. (17)) by
means of X^ kjk (Eq. (14)) and Aðq1 Þ (Eq. (12)), then
calculates Skþ1jk (Eq. (18)) to be used for the next The ðk þ 1Þth vector X^ sf ðdÞkþ1jk represents the STF
estimation: of meteorological parameters during day d for time
step ðk þ 1Þ. X^ sf ðdÞkþ1jk is subsequently computed using
X^ kþ1jk ¼ AX^ kjk , (17) Eq. (17). This vector can be written as follows:
2 3 " #" #
Skþ1jk ¼ ASkjk AT þ QðkÞ  KðkÞ½ðCSkjk C T þ RðkÞK T ðkÞ. I^sf ðdÞkþ1jk AI ðq1 Þ 0 I^sf ðdÞ
X^ sf ðdÞkþ1jk ¼4 ^ 5¼ .
(18) T sf ðdÞkþ1jk 0 AT ðq1 Þ T^ sf ðdÞ
kjk

4.2. Application to meteorological parameter forecasting (19)

The STF algorithm consists of the following steps:
To forecast irradiance and ambient temperature for
short term using the Kalman filter approach (Section 4.1),  Read the matrix DB(d1) that provides the meteor-
we assume ological measurements during the last 20 days.
" #
I~mf ðd; tÞ  Estimate for day d the meteorological behaviour vector
 X~ mf ðd; tÞ ¼ ~ : ~
EðdÞ using ANFIS.
T mf ðd; tÞ
 Compute the MTF time distribution matrix X~ mf ðd; tÞ.
The MTF vector of meteorological parameter values  Compute the ARMA matrixes Aðq1 Þ and Cðq1 Þ
of the day d computed using distributions (Eqs. (1) relative to X~ mf ðd; tÞ.
and (9)).  Repeat from 1 h past sunrise (to get at least 12
" #
. . . ; Iðd; k  1Þ; Iðd; kÞ measurements)
 Y ðd; kÞ ¼
. . . ; Tðd; k  1Þ; Tðd; kÞ
: J Read the past measurements vector Y ðd; kÞ .
J Compute the matrix of the Kalman filter gain KðkÞ.
 ARTICLE IN PRESS
M. Chaabene, M. Ben Ammar / Renewable Energy 33 (2008) 1435–1443 1441

J Estimate X^ sf ðdÞkjk . Since measurement errors are not significant, they were
J Estimate X^ sf ðdÞkþ1jk . neglected when minimizing the Kalman Filter errors. Thus,
 Until sunset. data were modelled independently of measurement errors.
. Forecasting models have been tested during daylight
5. Results and discussion since July 2006. In order to evaluate models’ performances,
daily estimated and measured data were analysed by
The MTF and STF algorithms were implemented at computing the NRMSE and the NMBE factors [2] for
CRTEn which is equipped by a meteorological station. each forecasting type
Measurements are delivered by means of a data acquisition qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
system connected to two meteorological sensors. The 1
PN 2
N i¼1 ðestimated value  measured valueÞ
station technical characterization is NRMSE% ¼ 1
P N
 100,
N 1 measured value
 Temperature sensor: type PT100 with accuracy error of
0.5%. PN
 Global irradiation sensor: type CIMEL/CEI180 with i¼1 ðestimated value  measured valueÞ
NMBE% ¼ PN  100,
accuracy error of 1%. 1 measured value
 The station overall estimated uncertainties in the
measurements is 1.2%. where N is the number of reading points. The NMBE
 The time step of measured data is instantaneous one provides information on the MTF performance. As for
sample every 5 min. NRMSE, it gives information on the STF performance.

8000

7000
Solar Irradiation [W/m2]

6000

5000

4000

3000
NMBE=0.22%
2000 G:measured
G:estimated
1000
180 200 220 240 260 280 300
Day of the year

45
Tmax-measured
Tmin-measured
40 Tmax-estimated
Tmin-estimated
NMBE Tmax=0.04%

35
Temperature [°C]

30

NMBE Tmin =0.20%

25

20

15

10
180 200 220 240 260 280 300
Day of the year

Fig. 4. Estimated and measured data of the (a) cumulated irradiance and (b) ambient temperature.
 ARTICLE IN PRESS
1442 M. Chaabene, M. Ben Ammar / Renewable Energy 33 (2008) 1435–1443

Figs. 4(a) and (b) give the estimated and measured
amounts for the cumulated irradiance and the minimal
and maximal ambient temperatures during the test period.
It can be seen that estimated and measured values present
good concordance (NMBE ffi 0:2%). The estimated vector
~
EðdÞ is used to hold MTF and STF. For clarity, forecasts
are analysed over three arbitrary chosen days. Estimated
values and measured data of both irradiance and ambient

Measured Medium Term Forecasting Short Term Forecasting

1000
NMBE =-9.11%
800 NRMSE=3.96%
I (W/m2)

600
400
200
0
6 8 10 12 14 16 18
40
NMBE=2.2%
35 NRMSE=0.87%
T (°C)

30
Day: Auguest, 18, 2006
25
20
6 8 10 12 14 16 18
Time (hour)

1200
1000 NMBE=1.8%
NRMSE=7.42%
800
I (W/m2)

600
400
200
0
6 8 10 12 14 16 18
40
NMBE=-8.17%
35
NRMSE=2.28%
T (°C)

30
25
Day: September, 15,2006
20
15
6 8 10 12 14 16 18
Time (hour)

1000
NMBE=5.4%
800
NRMSE=8.22%
I (W/m2)

600
400
200
0
6 8 10 12 14 16 18
35
NMBE=3.97%
30
NRMSE=2.1%
T (°C)

25
20
Day: October, 22, 2006
15
10
6 8 10 12 14 16 18
Time (hour)

Fig. 5. MTF, STF and measurements of irradiance and ambient temperature during 3 days.
 ARTICLE IN PRESS
M. Chaabene, M. Ben Ammar / Renewable Energy 33 (2008) 1435–1443
temperature are gathered in the same diagram (Fig. 5).
Since MTF provides global behaviour of meteorological
parameters, its relative curves present no fluctuation and
are not correlated enough with measured data. Besides,
STF grants instantaneous correlation in case of signals
stability and oscillations. Hence, STF forecasted curves
follow measured ones even when weather is disturbed
(October 22, 2006), which emphasizes the performance of
the used Kalman filter (Eq. (19)). Obtained NMBE and
NRMSE errors do not exceed 10%,which proves the
efficiency of the adopted method when compared to those
calculated for various standard models [2].
6. Conclusion
A new methodology for dynamic forecasting of meteor-
ological parameters is presented. Two kinds of forecasting
are conducted. An MTF, based on a neuro-fuzzy
estimator, offers daily time distribution of meteorological
parameters relying on the meteorological behaviour during
the days before. In addition, an STF forecasts meteor-
ological parameters for a time step of 5 min by means of
the MTF ARMA model, a Kalman Filter and measure-
ments provided by a meteorological station. Forecasts had
been validated at the Energy and Thermal Research Centre
(CRTEn) in the north of Tunisia. According to the NMBE
and the NRMSE computation, the meteorological estima-
tor carries out obvious estimation of meteorological
parameters. Actually, the MTF is being used to carry
studies in CRTEn on medium term planning of the thermal
energy of a 760 m2 collector plant. As well, a predictive
control of a hybrid system composed of a PVP, a battery
and a diesel generator, is under study using STF.
1443
References
[1] Myers DR. Irradiance modelling and measurements for renewable
energy applications: data and model quality. Energy Int J 2005;30(9):
1517–31.
[2] Li DHW, Cheung GHW. Study of models for predicting the diffuse
irradiance on inclined surfaces. Appl Energy 2005;81(2):170–86.
[3] Chaâbane M, Masmoudi M, Medhioub K. Determination of Linke
turbidity factor from irradiance measurement in northern Tunisia.
Renew Energy 2004;29(13):2065–76.
[4] Chaabane M, Ben Djemaa A. Use of HR Meteosat images for the
mapping of global solar irradiation in Tunisia: preliminary results
and comparison with Wefax images. Renew Energy 2002;25(1):
139–51.
[5] de Miguel A, Bilbao J. Test reference year generation from
meteorological and simulated irradiance data. Sol Energy 2005;78(6):
695–703.
[6] Mellit A, Benghanem M, Kalogirou SA. An adaptive wavelet
network model for forecasting daily irradiance. Appl Energy 2006;
83(7):705–22.
[7] Iqdour R, Zeroual A. A rule based fuzzy model for the prediction of
irradiance. Rev Energies Renouvelables 2006;9(2):113–20.
[8] Prudhomme T, Gillet D. Advanced control strategy of a solar
domestic hot water system with a segmented auxiliary heater. Energy
Build 2001;33:463–75.
[9] Pickhardt R. Nonlinear modelling and adaptive predictive control of
a solar power plant. Control Eng Pract 2000;8:937–47.
[10] Riverol C, Di Sanctis C. Improving adaptive network based fuzzy
inference systems (ANFIS): a practical approach. Asian J Inf Technol
2005;4(12):1208–12.
[11] Chaabene M, Annabi M. A dynamic model for predicting solar plant
performance and optimum control. Energy Int J 1997;22(6):567–78.
[12] Chaabene M, Annabi M. Dynamic thermal model for predicting
solar plant adequate energy management. Energy Convers Manage
1998;39(3–4):349–55.
[13] Jang J-SR. ANFIS: adaptive-network-based fuzzy inference systems.
IEEE Trans Syst Man Cybern 1993;23(3):665–85.
[14] Cathin DE. Estimation, control and the discrete Kalman filter.
Berlin: Springer; 1988.