Welcome to Scribd. Sign in or start your free trial to enjoy unlimited e-books, audiobooks & documents.Find out more
Download
Standard view
Full view
of .
Look up keyword
Like this
2Activity
0 of .
Results for:
No results containing your search query
P. 1
Short term flood forecasting using RBF static neural network modeling a comparative study

Short term flood forecasting using RBF static neural network modeling a comparative study

Ratings:
(0)
|Views: 48|Likes:
Published by ijcsis
Abstract—The artificial neural networks (ANNs) have been applied to various hydrologic problems recently. This research demonstrates static neural approach by applying Radial basis function neural network to rainfall-runoff modeling for the upper area of Wardha River in India. The model is developed by processing online data over time using static modeling. Methodologies and techniques by applying different learning rule and activation function are presented in this paper and a comparison for the short term runoff prediction results between them is also conducted. The prediction results of the Radial basis function neural network with Levenberg Marquardt learning rule and Tanh activation function indicate a satisfactory performance in the three hours ahead of time prediction. The conclusions also indicate that Radial basis function neural network with Levenberg Marquardt learning rule and Tanh activation function is more versatile than other combinations for RBF neural network and can be considered as an alternate and practical tool for predicting short term flood flow.

Keywords-component; Artificial neural network; Forecasting; Rainfall; Runoff;
Abstract—The artificial neural networks (ANNs) have been applied to various hydrologic problems recently. This research demonstrates static neural approach by applying Radial basis function neural network to rainfall-runoff modeling for the upper area of Wardha River in India. The model is developed by processing online data over time using static modeling. Methodologies and techniques by applying different learning rule and activation function are presented in this paper and a comparison for the short term runoff prediction results between them is also conducted. The prediction results of the Radial basis function neural network with Levenberg Marquardt learning rule and Tanh activation function indicate a satisfactory performance in the three hours ahead of time prediction. The conclusions also indicate that Radial basis function neural network with Levenberg Marquardt learning rule and Tanh activation function is more versatile than other combinations for RBF neural network and can be considered as an alternate and practical tool for predicting short term flood flow.

Keywords-component; Artificial neural network; Forecasting; Rainfall; Runoff;

More info:

Published by: ijcsis on Oct 10, 2010
Copyright:Attribution Non-commercial

Availability:

Read on Scribd mobile: iPhone, iPad and Android.
download as PDF, TXT or read online from Scribd
See more
See less

05/17/2012

pdf

text

original

 
(IJCSIS) International Journal of Computer Science and Information Security,Vol. 8, No. 6, 2010
Short term flood forecasting using RBF static neuralnetwork modeling a comparative study
Rahul P. Deshmukh
Indian Institute of Technology, BombayPowai, MumbaiIndiadeshmukh.rahul@iitb.ac.in
A. A. Ghatol
Former Vice-ChancellorDr. Babasaheb Ambedkar Technological University,Lonere, Raigad, Indiavc2005@rediffmail.com
 Abstract 
—The artificial neural networks (ANNs) have beenapplied to various hydrologic problems recently. This researchdemonstrates static neural approach by applying Radial basisfunction neural network to rainfall-runoff modeling for theupper area of Wardha River in India. The model is developedby processing online data over time using static modeling.Methodologies and techniques by applying different learningrule and activation function are presented in this paper and acomparison for the short term runoff prediction resultsbetween them is also conducted. The prediction results of theRadial basis function neural network with LevenbergMarquardt learning rule and Tanh activation function indicatea satisfactory performance in the three hours ahead of timeprediction. The conclusions also indicate that Radial basisfunction neural network with Levenberg Marquardt learningrule and Tanh activation function is more versatile than othercombinations for RBF neural network and can be consideredas an alternate and practical tool for predicting short termflood flow.
 Keywords-component;Artificial neural network;Forecasting; Rainfall; Runoff;
I.I
NTRODUCTION
The main focus ofthis research is development of Artificial Neural Network (ANN) models for short term floodforecasting, determining the characteristics of different neuralnetwork models. Comparisons are made between theperformances of different parameters for Radial basis functionartificial neural network models.The field engineers face the danger of very heavy flowof water through the gates to control the reservoir level byproper operation of gates to achieve the amount of waterflowing over the spillway. Thiscan be limited to maximumallowable flood and control flood downstream restricting riverchannel capacity so as to have safe florid levels in the riverwithin the city limits on the downstream.By keeping the water level in the dam at the optimumlevel inthe monsoon the post monsoon replenishment can beconveniently stored between the full reservoir level and thepermissible maximum water level. Flood estimation is veryessential and plays a vital role in planning for flood regulationand protection measures.The total runoff from catchment area depends uponvarious unknown parameters like Rainfall intensity, Durationof rainfall, Frequency of intense rainfall, Evaporation,Interception, Infiltration, Surface storage, Surface detention,Channel detention, Geological characteristics of drainagebasin, Meteorological characteristics of basin, Geographicalfeatures of basin etc. Thus it is very difficult to predict runoff at the dam due to the nonlinear and unknown parameters.In this context, the power of ANNs arises from thecapability for constructing complicated indicators (non-linearmodels). Among several artificial intelligence methodsartificial neural networks (ANN) holds a vital role and evenASCE Task Committee Reports have accepted ANNs as anefficient forecasting and modeling tool of complex hydrologicsystems[22].Neural networks are widely regarded as a potentiallyeffective approach for handling large amounts of dynamic,non-linear and noisy data, especially in situations where theunderlyingphysical relationships are not fully understood.Neural networks are also particularly well suited to modelingsystems on a real-time basis, and this could greatly benefitoperational flood forecasting systems which aim to predict theflood hydrograph for purposes of flood warning andcontrol[16].A subset of historical rainfall data from the WardhaRiver catchment in India was used to build neural network models for real time prediction. Telematic automatic raingauging stations are deployed at eight identified strategiclocations which transmit the real time rainfall data on hourlybasis. At the dam site the ANN model is developed to predictthe runoff three hours ahead of time.In this paper, we demonstrate the use of Radial basisfunction neural network (RBF) model for real time predictionof runoff at the dam and compare the effectiveness of differentlearning rules and activation function. Radial basis functionneural network is having a feed-forward structure consisting of hidden layer for a given number of locally tuned units whichare fully interconnected to an output layer of linear units.At a time when global climatic change would seem tobe increasing the risk of historically unprecedented changes in
93http://sites.google.com/site/ijcsis/ISSN 1947-5500
 
(IJCSIS) International Journal of Computer Science and Information Security,Vol. 8, No. 6, 2010
river regimes, it would appear to be appropriate thatalternative representations for flood forecasting should beconsidered.II.M
ETHODOLOGY
In this study different parameters like learning rule andactivation function are employed for rainfall-runoff modelingusing Radial basis function neural network model of artificialneural network.Radial basis functions networks have a very strongmathematical foundation rooted in regularization theory forsolving ill-conditioned problems.The mapping function of a radial basis functionnetwork, is built up of Gaussians rather than sigmoids as inMLP networks. Learning in RBF network is carried out in twophases: first for the hidden layer, and then for the output layer.The hidden layer is self-organising; its parameters depend onthe distribution of the inputs, not on the mapping from the inputto the output. The output layer, on the other hand, usessupervised learning (gradient or linear regression) to set itsparameters.
 
Figure 1. The Radial basis function neural network 
In this study we applied different learning rules to theRBF neural network and studied the optimum performancewith different activation function. We applied Momentum,Deltabar Delta, Levenberg Marquardt , Conjugate Gradient,Quick prop learningrule with activation function Tanh, LinearTanh, Sigmoid and Linear Sigmoid.
Performance Measures:
The learning and generalization ability of the estimatedNNmodel is assessed on the basis of important performancemeasures such as MSE (Mean Square Error), NMSE(Normalized Mean Square Error) and r (Correlationcoefficient)
 MSE (Mean Square Error):
The formula for the mean square error is:
200
Pijij ji
dy MSE  NP
… (1)WhereP = number of output PEs,N = number of exemplars in the data set,
ij
 y
= network output for exemplar i at PE j,
ij
= desired output for exemplar i at PE j. 
 NMSE (Normalized Mean Square Error):
The normalized mean squared error is defined bythe following formula:
22000
 NijijPii j
PNMS NMSE  Nd N 
… (2)WhereP = number of output processing elements,N = number of exemplars in the data set,MSE = mean square error,
ij
= desired output for exemplar i at processingelement j.
r (correlation coefficient):
The size of the mean square error (MSE) can be usedto determine how well the network output fits the desiredoutput, but it doesn't necessarily reflect whether the two sets of data move in the same direction. For instance, by simplyscaling the network output, the MSE can be changed withoutchanging the directionality of the data. The correlationcoefficient (r) solves this problem. By definition, thecorrelation coefficient between a network output x and adesired output d is:
__22__
iiiiiii
 xxd N ddxx N
     
… (3)where
_1
1
 N ii
 xx N 
and
_1
1
 N ii
d N 
94http://sites.google.com/site/ijcsis/ISSN 1947-5500
 
(IJCSIS) International Journal of Computer Science and Information Security,Vol. 8, No. 6, 2010
The correlation coefficient is confined to the range [-1, 1].When r = 1 there is a perfect positive linear correlation betweenx and d, that is, they co-vary, which means that they vary bythe same amount.III.S
TUDY
A
REA AND
D
ATA
S
ET
The Upper Wardha catchment area lies directly in the pathof depression movements which originates in the Bay of Bengal. When the low pressure area is formed in the Bay of Bengal and cyclone moves in North West directions, manytimes this catchment receives very heavy intense cyclonicprecipitationfor a day or two. Occurrence of such events havebeen observed in the months of August and September.Rainfall is so intense that immediately flash runoff, causingheavy flood has been very common feature in this catchment.For such flashy type of catchment and wide variety intopography, runoff at dam is still complicated to predict. Theconventional methods also display chaotic result. Thus ANNbased model is built to predict the total runoff from rainfall inUpper Wardha catchment area for controlling water level of thedam.In the initial reaches, near its origin catchment area is hillyand covered with forest. The latter portion of the river liesalmost in plain with wide valleys.The catchment area up to dam site is 4302 sq. km. Atdam site the river has wide fan shaped catchment area whichhas large variation with respect to slope, soil and vegetationcover. 
Figure 2-Location of Upper Wardha dam on Indian map
 
Figure 3-The Wardha river catchment
Data: Rainfall runoff data for this study is taken from theWardha river catchment area which contains a mix of urbanand rural land. The catchments is evenly distributed in eightzones based on the amount of rainfall and geographical survey.The model is developed using historical rainfall runoff data ,provided by Upper Wardha Dam Division Amravati,department of irrigation Govt. of Maharashtra. Network istrained by rainfall information gathered from eight telemetricrain-gauge stations distributed evenly throughout the catchmentarea and runoff at the dam site.The data is received at the central control room online throughthis system on hourly basis. The Upper Wardha dam reservoiroperations are also fully automated. The amount of inflow,amount of discharge is also recorded on hourly basis. From theinflow and discharge data the cumulative inflow is calculated.The following features are identified for the modeling theneural network .
T
ABLE
I-T
HE PARAMETERS USED FOR TRAINING THE NETWORK
MonthRG1RG2RG3RG4RG5RG6RG7RG8CIF
Month The month of rainfallRain1 to Rain8 Eight rain gauging stations.Cum Inflow Cumulative inflow in damSeven years of data on hourly basis from 2001 to 2007 isused. It has been found that major rain fall (90%) occurs in themonth of June to October Mostly all other months are dryhence data from five months. June to October is used to trainthe network IV.R
ESULT
The different structures of neural network areemployed to learn the unknown characterization of the systemfrom the dataset presented to it. The dataset is partitioned intothree categories, namely training, cross validation and test. Theidea behind this is that the estimated NN model should betested against the dataset that was never presented to it before.This is necessary to ensure the generalization. An experiment isperformed at least twenty five times with different randominitializations of theconnection weights in order to improvegeneralization.The data set is divided in to training , testingand cross validation data and the network is trained for allmodels of Radial basis function neural network for 5000epochs.The performance results obtain on parameters byapplying learning rules Momentum, Deltabar Delta, LevenbergMarquardt , Conjugate Gradient, Quick prop with activationfunction Tanh, Linear Tanh, Sigmoid, Linear Sigmoid arelisted in Table IIthrough Table VI.
95http://sites.google.com/site/ijcsis/ISSN 1947-5500

You're Reading a Free Preview

Download
scribd
/*********** DO NOT ALTER ANYTHING BELOW THIS LINE ! ************/ var s_code=s.t();if(s_code)document.write(s_code)//-->