Paper 28- Fuzzy Inference System for Weather Prediction

advertisement
Fuzzy Inference System for Weather Prediction
Riyaz P A
Sandeep Kartha
Computer Science and Engineering
Mar Athanasius College of Engineering
Kothamangalam, Kochi, India
riyazaahil@gmail.com
Computer Science and Engineering
Mar Athanasius College of Engineering
Kothamangalam, Kochi, India
sandu999@gmail.com
Gurusamy P
Surekha Mariam Varghese
Computer Science and Engineering
Mar Athanasius College of Engineering
Kothamangalam, Kochi, India
gurumprp@gmail.com
Computer Science and Engineering
Mar Athanasius College of Engineering
Kothamangalam, Kochi, India
surekha@mace.ac.in
Abstract—Climate change is one of the most serious
environmental threats faced by mankind nowadays. It affects the
livelihood of people mainly through its impact on agricultural
productivity, including its direct impact on food production. Also
it significantly reduces the growth rate of a faster developing
country like India, where agriculture is the primary occupation of
many. Among them, majorities are illiterate and deprived of
computer literacy. In this context, if we are able to develop a
system, which automatically predicts the rainfall for the benefit of
farmers, will be very useful. The weather data, being inherently
fuzzy in nature, requires highly complex processing based on
human observations, satellite photography, or radar followed by
computer simulations. This paper describes the idea of developing
a weather prediction tool for Kerala using fuzzy inference rule
based system.
Keywords—Fuzzy Logic; Rainfall; Inference system;Kerala;
Weather; prediction
I.
INTRODUCTION
A weather forecast is a scientific estimate of the future state
of the atmosphere where state of Weather forecast is based on a
number of weather parameters. Number of parameters and their
significance vary from place to place and between seasons [1].
In Kerala, the weather parameters with significant variation for
weather forecasting are cloud cover, wind, temperature,
humidity, pressure and precipitation.
Weather prediction is very complex as it depends on various
factors. In weather forecasting, a Meteorologist must know
about the existing weather condition over a large area before he
can make a reliable forecast. The accuracy of his forecast
depends largely upon his knowledge of the prevailing weather
conditions over a very wide area. The forecast decision is based
on various forecasting tools. The basic tool of a weather
forecaster is the Weather MAP. The weather map depicts the
distribution patterns of atmospheric pressure, wind, temperature
and humidity at the different levels of the atmosphere. There are
two types of the basic weather map namely, the surface map and
the upper-air maps. There are five standard levels of the upperair maps that are constructed twice daily at twelve-hourly
interval. The surface maps are made four times daily at sixhourly intervals.
II.
FUZZY LOGIC
Fuzzy Logic is a very powerful method to model complex
and imprecise systems and is very useful for knowledge-based
systems and linguistically communicated systems. Fuzzy
models are very effective for representing both objective and
subjective knowledge such as definitional, causal, statistical,
and heuristic knowledge. Fuzzy system is very suitable for
representing the weather data, which is inherently fuzzy in
nature. The weather prediction statements are represented in the
form of if premises then conclusion rule forms, and output is
taken by using fuzzy approximate reasoning methods.
Considering all these conditions, fuzzy set theory helps to
manage complexity and uncertainties and gives a user- friendly
visualization of the system construction and working model.
Even though, there are many weather forecasting systems
available, it is important to have weather prediction tool for the
benefit of Kerala farmers.
A general fuzzy system has basically four components:
fuzzification, fuzzy rule base, fuzzy output engine, and
defuzzification (Fig. 1). Fuzzification converts each piece of
input data to degrees of membership by a look-up in one or more
several membership functions. The key idea in fuzzy logic, in
fact, is the allowance of partial belongings of any object to
different subsets of the universal set instead of belonging to a
single set completely. Partial belonging to a set can be described
numerically by a membership function, which takes on values
between 0 and 1 inclusive. Intuition, inference, rank ordering,
angular fuzzy sets, neural networks, genetic algorithms, and
inductive reasoning can be among many ways to assign
membership values or functions to fuzzy variables. This
intuitive approach is used rather commonly because it is simple
and derived from the innate intelligence and understanding of
human beings. Fuzzy membership functions may take on many
forms, but in practical applications simple linear functions, such
as triangular ones, are preferable. We used triangular
membership function, trapezoidal membership function and pi
membership function for our system.
Fig. 1. Schematic Representation of Fuzzy Inference System
The fuzzy rule base contains rules that include all possible
fuzzy relations between inputs and outputs. These rules are
expressed in the IF–THEN format. In the fuzzy approach, there
are no mathematical equations and model parameters. All the
uncertainties, nonlinear relationships, or model complications
are included in the descriptive fuzzy inference procedure in the
form of IF–THEN statements. There are basically two types of
rule systems, namely, Mamdani and Sugeno [3]. Depending
upon a problem under consideration, a user can choose the
appropriate rule system. According to the Sugeno rule system,
the consequent part of the fuzzy rule is expressed as a
mathematical function of the input variable and such a system is
more appropriate for neurofuzzy systems [4]. In the Mamdani
rule system, however, the consequent part of the fuzzy rule is
also expressed as verbally. Mamdani model is used for this
system.
Defuzzification converts the resulting fuzzy outputs from the
fuzzy inference engine to a number. There are many
defuzzification methods, such as COG, bisector of area (BOA),
mean of maxima (MOM), left-most maximum (LM), and rightmost maximum (RM), etc. [3]; [6]. The MOM, LM, and RM
methods disregard the shape of the fuzzy set and hence they are
employed in particular problems [3]. The BOA method picks the
abscissa of the vertical line that divides the area of the combined
output fuzzy subset in two equal halves. In the COG method the
crisp output value is the abscissa under the center of gravity of
the combined output fuzzy subset. The COG method is the most
commonly used defuzzification method [3].
III.
RELATED WORK
Weather is naturally ambiguous and uncertain in nature and
historically weather data has been used as an indicator to analyze
the impact of climate on human activity and vice versa. Weather
event predictions are performed on weather data generally for
farmers, flood, drought, traffic control and cyclone updates, etc.
to help all stakeholders avoid losses and gain financial benefits.
Case-based reasoning has been employed to forecast airport
weather. A rough set based fuzzy inference system has been
used to perform weather prediction.
Association rules have been extracted to discover the
abnormal weather phenomena as well as to make weather
prediction. Fuzzy Logic approach has been adopted to predict
weather. Weather prediction has been used for traffic controlling
applications. Fuzzy expert systems have been employed to
forecast winds and predict weather. Forecasts for the present
case can be based on similar past cases using the analog method.
Many researchers have applied fuzzy logic for more accurate
prediction of weather conditions such as lightning
[18], temperature [19], solar radiation [19], cloud ceiling
heights and horizontal visibility [18], flood [20], thunderstorm
[21], drought [24], traffic control [25] and cyclone [22] etc.
Many of them are focused on specific areas such as lightning
prediction for South Korea [18], cloud visibility at the airports
of Canada [19], temperature prediction for Bombay city [18]
used fuzzy logic based rule based reasoning based on five
parameters relative humidity, total cloud cover, wind direction,
temperature and surface pressure to predict rainfall events for
two Egyptian meteorological stations. These researches prove
that fuzzy logic is desirable to merge the experiences of
forecasters and theoretical studies with efficiency and the
accuracy of the computer systems. [19]. All these systems have
used fuzzy system as a viable alternative to the classical
complex; costly and time-consuming weather prediction
techniques. Other soft computing techniques are also used for
rainfall prediction. Guhathakurta has used ANN for monsoon
rainfall prediction in different districts of Kerala [23].
IV.
METHODOLOGY
This system is based on modeling, analysis, synthesis and
simulation of various weather factors. Main component of this
weather prediction is fuzzy model implemented in the
MATLAB 2009b software. First the input variables are defined,
then the model calculates the value from these input variables
based on the defined rules and returns output variable called
Rainfall. This variable is useful for farmers. In order to keep the
model as simple as possible only three attributes were used for
each variable – low, medium and high except for seasons. These
rules record the basic relationships that have been determined
from the analysis of the past data and from the values of input
variables in the key past situations.
We were looking for a simple algorithm or method that could
be used by farmers who are not skilled enough to use
complicated weather prediction system. Many of the decision
support methods were not reliable or simple enough. That led us
to do this research where a simple model with several input
variables. After the rules have been defined the surface viewer
can be used to visualize the dependency between input variables
and the output variable.
Fig. 2. Fuzzy Inference System (FIS) Editor with Input and Output Variables of Weather Prediction System
A. Data Collection
The weather details of Kochi from January 2012 to
December 2012 are collected. This collection includes the
temperature, humidity, wind speed, total cloud and monsoon
details. The data is collected from India Meteorological
Department [14].
2. The input variable humidity has three linguistic variables
low, medium and high(Fig 4). The minimum is taken as 0 % and
maximum is taken as 100 %. Low linguistic variable using
TrapMF and other variables using TriMF.
B. Data Analysis
The collected data is analyzed to find the range of low,
medium and high values of temperature, humidity, wind speed
and total cloud.
C. Fuzzification of Input/Output parameters
There are different kinds of membership functions are
available in Fuzzy Inference System. Using linguistic variables
represents the variables of FIS. Three linguistic variables
namely “low”, “medium” and “high” are used in this system.
There are five input variables and one output variable for our
system. Input variables are Temperature, Humidity, Wind
speed, Total cloud and Seasons. Output variable is Rainfall and
shown in Fig.2.
Fig. 4. Membership Function Plots of Input Variable Humidity.
3. The input variable wind speed has three linguistic
variables low, medium and high. The minimum numerical value
is taken as 0 kmph and maximum numeric value is taken as 50
kmph. All linguistic variables are using TriMF.
1. The input variable temperature has three linguistic
variables low, medium and high (Fig 3). The minimum
numeric value is taken as -10◦C and maximum numeric value
is taken as 50◦C. All linguistic variables are using TriMF.
Fig. 5. Membership Function Plots of Input Variable Wind Speed.
4. The input variable total cloud has three linguistic variables
low, medium and high. The minimum numerical value is taken
as 0 octa and maximum numeric value is taken as 15 octa. All
linguistic variables using TriMF.
Fig. 3. Membership Function Plots of Input Variable Temperature.
E. Inferencing
This prediction system is using the mamdani inference
model. Mamdani type inference, expects the output
membership functions to be fuzzy sets. After the aggregation
process, there is a fuzzy set for each output variable that needs
defuzzification. A common and useful defuzzification
technique is Centre of gravity (COG) method.
A sample fuzzy rule developed is
Fig. 6. Membership Function Plots of Input Variable Total Cloud.
5. The input variable seasons have four variables winter,
summer, southwest monsoon and northeast monsoon. All
variables are using PiMF. There are 4 seasons in a year in
Kerala.
E.g.: IF ((Temperature is LOW) AND (Humidity is HIGH)
AND (Wind Speed is HIGH) AND (Total Cloud is HIGH) AND
(Seasons is Southwest monsoon)
THEN (Rainfall is HIGH).
V.
IMPLEMENTATION AND RESULT
This system is implemented in MATLAB 2009b. The
specific and generic rules are inserted via rules editor. The input
to the system is automatically taken, that’s why the system
named as automated. The system date is converting into the
seasons input.
The surface view of interaction between two input variables
with the output variable rainfall is shown below.
Fig. 7. Membership Function Plots of Input Variable Seasons.
6. The output variable rainfall has four linguistic variables
no rain, low, medium and high. All variables are using TriMF.
Fig. 9. Interaction of Total Cloud and Humidity.
Fig. 8. Membership Function Plots of Output Variable Rainfall.
D. Rule Generation
Analyzing the collected data from the Internet, books and
newspaper and we develop the rules for this fuzzy inference
system. Insert the rules into FIS via rules editor in MATLAB.
Different research papers are analyzed for the weather
prediction system. There are two types of rules - generic and
specific rules. In this system, a variety of rules are using for
more accurate output.
Fig. 10. Interaction of Seasons and Total Cloud.
Fig. 11. Interaction of Seasons and Humidity.
Fig. 14. Performance of our system.The Y-axis represents the amount of
rainfall in mm and X-axis represents the days in june 2014.
some days small variations are occurred. Overall, the
performance of our system is good.
VII. FUTURE WORK
Now a days smartphones have different kinds of sensors like
temperature sensor, humidity sensor etc. Make this system
better by leveraging the smartphones.
Weather prediction is influence by lot of factors. The height
of place is one of the important factors. Adding more input
factors to the system will provide more accurate results.
Fig. 12. Interaction of Temperature and Total Cloud.
VIII. CONCLUSION
Fuzzy inference system will be useful for farmers in Kerala
to know about the rainfall. Rainfall has a dramatic effect on
agriculture. Different plants need varying amount of rainfall to
survive.
The amount of rainfall is dependent on number of factors.
The fuzzy inference system incorporates the different aspects
and ambiguities in these factors for accurate prediction of
rainfall. It is observed that the incorporation of fuzzy improved
the accuracy of the system to a great extent.
REFERENCES
Fig. 13. Interaction of Wind speed and Total Cloud.
VI.
[1]
PERFORMANCE EVALUATION
The rainfall data of Kochi is collected from IMD website of
June 6 to June 12 in 2014. The performance our system with the
actual data is shown as a Line graph in Fig.14.
We compare our result with the actual result from the IMD
Thiruvanthapuram results.Fig.14 illustrates the evaluation. In
[2]
[3]
[4]
Riyaz P.A., Sandeep Kartha, Gurusamy P., Surekha Mariam Varghese.
“Automated Fuzzy Logic Weather Prediction System”, proceedings of
National Conference on New trends in Electronics, Computing and
Communication(NTECC 2014), April, 2014.
Agboola A.H., Gabriel A.J., Aliyu E.O., Alese B.K. “Development of a
Fuzzy Logic Based Rainfall Prediction Model”, International Journal of
Engineering and Technology Volume 3 No. 4, April, 2013.
Gokmen Tayfur and Vijay P Singh, “ANN and Fuzzy Logic Models for
Simulating Event-Based Rainfall-Runoff”, Journal of Hydraulic
Engineering, ASCE, Vol. 132, No. 12, December 1, 2006.
Sen, Z. “Fuzzy modelling in engineering.” Class notes, Civil Engineering
Faculty, Istanbul Technical Univ., Istanbul, Turkey in Turkish, 1999
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
Chang, L. C., Chang, F. J. & Tsai, Y. H.,Fuzzy exemplar-based inference
system for flood forecasting. Water Resour. Res. 41, W02005,
doi:10.1029/2004WR003037, 2005
Benmiloud T, “Improved adaptive neuro-fuzzy inference system”. Neural
Comput Appl 21:575–582, 2012
L. A. Zadeh, “Fuzzy sets”, Information and Control, vol. 8, no. 3, (1965),
pp. 338–353
Matlab: v7.9.0 (R2009b), 2009. Documentation, the MathWorks, Inc.
Mamdani E.H. and Assilian S. ,"An experiment in linguistic synthesis
with a fuzzy logic controller" International Journal of Man-Machine
Studies, Vol. 7, No. 1, pp. 1-13., 1975
Mitra, A.K., Nath, Sankar, Sharma, A.K.,“Fog forecasting using rulebasedfuzzy inference system.” Journal of the Indian Society of Remote
Sensing, 36, 243–253., 2008
Arnaud, P., Fine, J.A., Lavabre, J.,. “An hourly rainfall generation model
applicable to all types of climate.” Atmospheric Research 85, 230–242.,
2007
Edvin and Yudha, “Application of Multivariate ANFIS for Daily Rainfall
Prediction: Influences Of Training Data Size”: Makara, Sains, Volume
12, No. 1, April 2008: 7-14 7.
B.K.Hansen,Weatherpredictionusingcasebasedreasoningandfuzzysettheory, Master of Computer Science Thesis,
Technical University of Nova Scotia, Canada, 2000.
India Meteorological Department’s web-site :www.imd.gov.in
Wong, K.W., Wong P.M., Gedeon T.D. and Fung C.C., “Rainfall
prediction model using soft computing technique”: Soft Comput. Fusion
Foundat. Methodol. Appli.7:434-438, 2003
T.J. Ross, Fuzzy Logic with Engineering Applications, 2nd edition, Wiley
Black- well, 2004.
Kuk, Bongjae, Hongil Kim, Jongsung Ha, Hyokeun Lee, Gyuwon Lee, A
Fuzzy Logic Method for Lightning Prediction Using Thermodynamic and
Kinematic Parameters from Radio Sounding Observations in South
Korea, Weather and Forecasting, Volume 27, 205–217, Feb 2012.
[18] Dipi A. Patel and R.A. Christian, Proposed Framework of Temperature
Prediction Using Clustering Based Fuzzy Logic, Global Journal of Engg.
& Appl. Sciences, 3 (1), pp:1-3 , 2013
[19] Bjarne Hansen, A Fuzzy Logic–Based Analog Forecasting System for
Ceiling and Visibility, pp:1319- 1330, weather and forecasting, Volume
22, December 2007.
[20] Shu, C., Ouarda, T.B.M.J., 2008. Regional flood frequency analysis at
ungauged sites using the adaptive neuro-fuzzy inference system, Journal
of Hydrology, Volume 349, pp: 31–43, January 2008.
[21] Pin-Fang Lin, Pao-Liang Chang, Ben Jong-Dao Jou, James W.
Wilson, Rita D. Roberts. (2012) Objective Prediction of Warm Season
Afternoon Thunderstorms in Northern Taiwan Using a Fuzzy Logic
Approach. Weather and Forecasting 27:5, 1178-1197, October 2012
[22] D. Sarewitz, A.R. Pielke, R. Byerly,Prediction: Science, Decision Making
and the Future of Nature,Island Press (2000)
[23] Guhathakurta, P., Long-range monsoon rainfall prediction of 2005 for the
districts and sub-division Kerala with artificial neural network,Current
Science 90, 773–779, 2006
[24] W.H. Quinn, D.O. Zopf, K.S. Short, R.T.W. Kuo-Yang, Historical trends
and statistics of the southern oscillation, El Nino, and Indonesian droughts
Fishery Bulletin, 76(3) , pp. 663-677, 1978.
[25] B. Krause, C. von Altrock, M. Pozybill,Fuzzy logic data analysis of
environmental data for traffic control, Proceedings of the Sixth IEEE
International Conference on Fuzzy Systems, vol. 2 , pp. 835–838, 1997.
[26] Somia A. Asklamy, Khaled Elhelow, I.K Youssef, M.Abd El-Wahab,
Rainfall events prediction using rule-based fuzzy inference system,
Atmospheric Research , pp: 228–236, Elsevier, 2011.
[27] Malik Shahzad Kaleem Awan, Mian Muhammad Awais, “Predicting
weather events using fuzzy rule based system”,Applied Soft Computing
11 (2011) 56–63, Elsevier
[28] Jantzen, J. “Design of fuzzy controllers.” Technical Rep. No.98-E864,
Dept. of Automation, Technical Univ. of Denmark,Denmark, 1999.
Download