Information Technology Reference
In-Depth Information
an historical dataset of ten years. The FPM uses a sample of fifty days from the
historical dataset (of the month to which belongs the day of the prediction) in
which radiation frosts occurred.
Formally, the minimum temperature is calculated by the following linear re-
gression (LR) equation:
D
0
+
i,
(1)
where
T
p
is the minimum temperature to be predicted,
T
o
represents the mini-
mum temperature and
D
0
the dew point. The parameters
T
o
and
D
0
have to be
acquired the same day of the prediction, two hours after sunset. Finally,
i
is the
LR intercept,
s
T
temperature slope and
s
D
dew point slope.
The values of
s
T
and
i
are calculated from the equations (2) and (3), respectively.
s
T
=
(
T
h
0
−
T
p
=
s
T
∗
T
o
+
s
D
∗
T
h
0
)(
T
m
−
T
m
)
(
T
h
0
−
,
(2)
T
h
0
)
2
s
T
T
h
0
n
,
(3)
where
T
h
0
are historical temperatures registered two hours after sunset,
T
m
min-
imum temperatures that succeed in the night, and
n
is the number of historical
data. Finally,
T
h
0
and
T
m
account for the average data temperatures.
The slope
s
D
is calculated by using the equation (4).
s
D
=
(
D
h
0
−
i
=
T
m
−
D
h
0
)(
R −
R
)
(
D
h
0
−
,
(4)
D
h
0
)
2
where
D
h
0
are historical dew points two hour after sunset and
R
the residuals.
The parameters
D
h
0
and
R
are the average of
D
h
0
and
R
, respectively. Finally,
the residual is calculated with the expression:
R
=
T
m
−
s
T
∗
T
o
+
i
.
4.2 Application Implementation
In order to develop the frost prediction application we implement the Snyder and
de Melo-Abreu [15] FPM, using Java and MySQL. MySQL was used to store
the data from the sensor nodes and the results obtained after running the FPM.
The application was executed using Amazon EC2 instances.
The integration of WSN data with Cloud infrastructures was performed with a
WSN - Cloud integration platform called Sensor Cirrus [17,18,19]. Sensor Cirrus
manages the WSN data using Cloud services and includes the developed frost
prediction application for data processing.
Figure 1 illustrates a scheme of the frost prediction module. The information
collected by WSN sensors in the field is stored in a proper database, as it seen in
process (1). Next, in process (2), the application performs a query to catch the
sample of fifty days. This sample includes all the collected data (temperature,
humidity, solar radiation, wind speed, etc.) by the WSNs. Then, in process (3)
the application retrieves from the sample of fifty days only the FPM input data
(
T
o
,
D
o
, etc.). Finally, in (4) the FPM is executed, resulting in the minimum
temperature that occurs next night through process (5).
