CHAPTER 2 REVIEW OF LITERATURE

Similar documents
EE-588 ADVANCED TOPICS IN NEURAL NETWORK

Weather Forecasting using Soft Computing and Statistical Techniques

Prediction of Hourly Solar Radiation in Amman-Jordan by Using Artificial Neural Networks

Chapter-1 Introduction

Weather Forecasting Using ANFIS and ARIMA MODELS. A Case Study for Istanbul

Prediction of Monthly Rainfall of Nainital Region using Artificial Neural Network (ANN) and Support Vector Machine (SVM)

DESIGN AND DEVELOPMENT OF ARTIFICIAL INTELLIGENCE SYSTEM FOR WEATHER FORECASTING USING SOFT COMPUTING TECHNIQUES

A Hybrid ARIMA and Neural Network Model to Forecast Particulate. Matter Concentration in Changsha, China

Rainfall Prediction using Back-Propagation Feed Forward Network

SOIL MOISTURE MODELING USING ARTIFICIAL NEURAL NETWORKS

To Predict Rain Fall in Desert Area of Rajasthan Using Data Mining Techniques

ANN and Statistical Theory Based Forecasting and Analysis of Power System Variables

COMPARISON OF CLEAR-SKY MODELS FOR EVALUATING SOLAR FORECASTING SKILL

Research Article Weather Forecasting Using Sliding Window Algorithm

Rainfall is the most important climate element affecting the livelihood and wellbeing of the

ESTIMATION OF SOLAR RADIATION USING LEAST SQUARE SUPPORT VECTOR MACHINE

Short Term Load Forecasting Using Multi Layer Perceptron

LONG - TERM INDUSTRIAL LOAD FORECASTING AND PLANNING USING NEURAL NETWORKS TECHNIQUE AND FUZZY INFERENCE METHOD ABSTRACT

Importance of Numerical Weather Prediction in Variable Renewable Energy Forecast

Short Term Load Forecasting Based Artificial Neural Network

AN INTERNATIONAL SOLAR IRRADIANCE DATA INGEST SYSTEM FOR FORECASTING SOLAR POWER AND AGRICULTURAL CROP YIELDS

Temperature Prediction based on Artificial Neural Network and its Impact on Rice Production, Case Study: Bangladesh

USE OF FUZZY LOGIC TO INVESTIGATE WEATHER PARAMETER IMPACT ON ELECTRICAL LOAD BASED ON SHORT TERM FORECASTING

CHAPTER 2 DATA AND METHODS. Errors using inadequate data are much less than those using no data at all. Charles Babbage, circa 1850

Estimation of Precipitable Water Vapor Using an Adaptive Neuro-fuzzy Inference System Technique

Short-term wind forecasting using artificial neural networks (ANNs)

Forecasting Drought in Tel River Basin using Feed-forward Recursive Neural Network

Application of Text Mining for Faster Weather Forecasting

Seasonal to decadal climate prediction: filling the gap between weather forecasts and climate projections

SHORT TERM LOAD FORECASTING

Weather and climate outlooks for crop estimates

WMO Aeronautical Meteorology Scientific Conference 2017

Estimation of Pan Evaporation Using Artificial Neural Networks A Case Study

Water information system advances American River basin. Roger Bales, Martha Conklin, Steve Glaser, Bob Rice & collaborators UC: SNRI & CITRIS

Stochastic Hydrology. a) Data Mining for Evolution of Association Rules for Droughts and Floods in India using Climate Inputs

DEVELOPMENT OF A LARGE-SCALE HYDROLOGIC PREDICTION SYSTEM

Projected Change in Climate Under A2 Scenario in Dal Lake Catchment Area of Srinagar City in Jammu and Kashmir

NSF Expeditions in Computing. Understanding Climate Change: A Data Driven Approach. Vipin Kumar University of Minnesota

Seasonal forecasting of climate anomalies for agriculture in Italy: the TEMPIO Project

A Recommendation for Tropical Daily Rainfall Prediction Based on Meteorological Data Series in Indonesia

Thunderstorm Forecasting by using Artificial Neural Network

Real Time wave forecasting using artificial neural network with varying input parameter

Application and verification of ECMWF products 2013

El Niño / Southern Oscillation

Application and verification of the ECMWF products Report 2007

Climate Prediction Center Research Interests/Needs

Prediction of Seasonal Rainfall Data in India using Fuzzy Stochastic Modelling

TOOLS AND DATA NEEDS FOR FORECASTING AND EARLY WARNING

Indian Weather Forecasting using ANFIS and ARIMA based Interval Type-2 Fuzzy Logic Model

MURDOCH RESEARCH REPOSITORY

On Improving the Output of. a Statistical Model

Possible Applications of Deep Neural Networks in Climate and Weather. David M. Hall Assistant Research Professor Dept. Computer Science, CU Boulder

Data and prognosis for renewable energy

Predicting Floods in North Central Province of Sri Lanka using Machine Learning and Data Mining Methods

5B.1 DEVELOPING A REFERENCE CROP EVAPOTRANSPIRATION CLIMATOLOGY FOR THE SOUTHEASTERN UNITED STATES USING THE FAO PENMAN-MONTEITH ESTIMATION TECHNIQUE

EE-588 ADVANCED TOPICS IN NEURAL NETWORK

Postprocessing of Numerical Weather Forecasts Using Online Seq. Using Online Sequential Extreme Learning Machines

International Journal of Computer Science Trends and Technology (IJCST) Volume 3 Issue 3, May-June 2015

Solar radiation analysis and regression coefficients for the Vhembe Region, Limpopo Province, South Africa

Appearance of solar activity signals in Indian Ocean Dipole (IOD) phenomena and monsoon climate pattern over Indonesia

ATM S 111, Global Warming Climate Models

not for commercial-scale installations. Thus, there is a need to study the effects of snow on

SARIMA-ELM Hybrid Model for Forecasting Tourist in Nepal

MODELLING OF TOOL LIFE, TORQUE AND THRUST FORCE IN DRILLING: A NEURO-FUZZY APPROACH

ATMOSPHERIC MODELLING. GEOG/ENST 3331 Lecture 9 Ahrens: Chapter 13; A&B: Chapters 12 and 13

A Support Vector Regression Model for Forecasting Rainfall

Application and verification of ECMWF products: 2010

NEURO-FUZZY SYSTEM BASED ON GENETIC ALGORITHM FOR ISOTHERMAL CVI PROCESS FOR CARBON/CARBON COMPOSITES

A FUZZY NEURAL NETWORK MODEL FOR FORECASTING STOCK PRICE

NON-STATIONARY & NON-LINEAR ANALYSIS, & PREDICTION OF HYDROCLIMATIC VARIABLES OF AFRICA

Ground-based temperature and humidity profiling using microwave radiometer retrievals at Sydney Airport.

A SELF-TUNING KALMAN FILTER FOR AUTONOMOUS SPACECRAFT NAVIGATION

Temporal global solar radiation forecasting using artificial neural network in Tunisian climate

The document was not produced by the CAISO and therefore does not necessarily reflect its views or opinion.

A Feature Based Neural Network Model for Weather Forecasting

Introduction of Seasonal Forecast Guidance. TCC Training Seminar on Seasonal Prediction Products November 2013

Enhancing Weather Information with Probability Forecasts. An Information Statement of the American Meteorological Society

Optimal Artificial Neural Network Modeling of Sedimentation yield and Runoff in high flow season of Indus River at Besham Qila for Terbela Dam

Radial basis function neural networks model to estimate global solar radiation in semi-arid area

ESTIMATION OF HOURLY MEAN AMBIENT TEMPERATURES WITH ARTIFICIAL NEURAL NETWORKS 1. INTRODUCTION

Fuzzy Systems. Introduction

Please, Donald Trump, don't send climate science back to the pre-satellite era

Direct Normal Radiation from Global Radiation for Indian Stations

Regionalization Techniques and Regional Climate Modelling

A. Pelliccioni (*), R. Cotroneo (*), F. Pungì (*) (*)ISPESL-DIPIA, Via Fontana Candida 1, 00040, Monteporzio Catone (RM), Italy.

Climate Change in the Pacific: Scientific Assessment and New Research Volume 1: Regional Overview

A Hybrid Wavelet Analysis and Adaptive. Neuro-Fuzzy Inference System. for Drought Forecasting

Optimum Neural Network Architecture for Precipitation Prediction of Myanmar

Flux Tower Data Quality Analysis in the North American Monsoon Region

Journal of of Computer Applications Research Research and Development and Development (JCARD), ISSN (Print), ISSN

Project Name: Implementation of Drought Early-Warning System over IRAN (DESIR)

Finding Multiple Outliers from Multidimensional Data using Multiple Regression

Construction and Analysis of Climate Networks

Products of the JMA Ensemble Prediction System for One-month Forecast

S e a s o n a l F o r e c a s t i n g f o r t h e E u r o p e a n e n e r g y s e c t o r

Comparative Analysis of ANFIS, ARIMA and Polynomial Curve Fitting for Weather Forecasting

Comparative Study of ANFIS and ARIMA Model for Weather Forecasting in Dhaka

Regional climate projections for NSW

Deep Learning Architecture for Univariate Time Series Forecasting

Transcription:

CHAPTER 2 REVIEW OF LITERATURE 2.1 HISTORICAL BACKGROUND OF WEATHER FORECASTING The livelihood of over 60 per cent of the world's population depends upon the monsoons, of which the Asian summer monsoon is the largest. Accurate predictions of the monsoons, at least a season in advance, are therefore crucial for the monsoon regions. Furthermore, the Asian summer monsoon is a key component of the earth's climate system, having important tele-connections with global weather and climate (Walter Maner, 1997). Following the Great Indian Drought of 1877, H.F. Blanford, who had established the India Meteorological Department in 1875, issued the first seasonal forecast of Indian monsoon rainfall in 1884. Later, in the early part of the 20th century, Sir Gilbert Walker initiated extensive studies of global teleconnections which led him to the discovery of Southern Oscillation. Walker introduced, for the first time, the concept of correlation for long-range forecasting of the Asian summer monsoon and his findings are relevant even today. More than 100 years later, forecasts of the Asian summer monsoon is still being made using statistical regression, often with remarkable success. General circulation models are also capable of capturing some of the features of the Asian summer monsoon and may be able to give improved short-term forecasts. Generally, there are two methods that are used in weather forecasting one is empirical approach and other is dynamical approach (Lorenz, 1969). The empirical approach is based upon the occurrence of analogues and is often referred to by meteorologists as analogue such as changes in barometric pressure, current weather conditions, sky condition to determine the future conditions (Ozelkan, 1996). This approach normally is useful for predicting local-scale weather if recorded case are very large in number. Dynamical approach is based upon equation and forward simulations of the atmosphere and is often referred to as computer modeling which involves pattern 44

recognition skills, knowledge of model performance and knowledge of model biases (Lorenz, 1969). This approach is normally useful for modeling large-scale weather phenomena and may not predict short-term weather efficiently. Most of the weather forecasting systems are combined techniques of empirical approach and dynamical approach. However, not much attention has been paid to the use of soft computing in weather forecasting. Weather forecast systems are among the most complex equation systems that computer has to solve. A great quantity of data, coming from satellites, ground stations and sensors located around our planet send daily information that must be used to foresee the weather situation in next hours and days all around the world. Weather reports give forecast for next 24, 48 and 72 hours for wide areas (Pasero, 2004). Weather forecasts provide critical information about future weather. There are various techniques involved in weather forecasting, from relatively simple observation of the sky to highly complex computerized mathematical models (M. Tektas, 2010). 2.2 APPROACHES FOR WEATHER FORECASTING Numerical weather prediction is the prediction of weather phenomena by the numerical solution of the equations governing the motion and changes of condition of the atmosphere. Numerical weather prediction techniques, in addition to being applied to short-range weather prediction, are used in such research studies as air-pollutant transport and the effects of greenhouse gases on global climate change. The first operational numerical weather prediction model consisted of only one layer and therefore it could model only the temporal variation of the mean vertical structure of the atmosphere. Computers now permit the development of multilevel (usually about 10 20) models that could resolve the vertical variation of the wind, temperature and moisture. These multilevel models predict the fundamental meteorological variables for large scales of motion Lunagariya et al. (2009) made an effort to verify the weather forecast from NCMRWF. Analysis was carried out weekly, seasonal as well as yearly basis using various numerical verification techniques like ratio score, usability analysis and 45

correlation approach during 2006-07 and 2008-09. The forecasts were found within usability range for some parameters but for other parameter improvement is still possible. The complexities in the relationship between rainfall and sea surface temperature (SST) during the winter monsoon (November-January) has been observed by Goutami Chattopadhyay et al. (2008). Evaluation is done statistically using scatter plot matrices and autocorrelation functions. Linear as well as polynomial trend equations were obtained and it was observed that the coefficient of determination for the linear trend was very low and it remained low even when polynomial trend of degree six was used. An exponential regression equation and an artificial neural network with extensive variable selection were generated to forecast the average winter monsoon rainfall of a given year using the rainfall amounts and the sea surface temperature anomalies in the winter monsoon months of the previous year as predictors. The artificial neural network was generated in the form of a multi-layer perceptron with sigmoid non-linearity and geneticalgorithm based variable selection. Both of the predictive models were judged statistically using the Wilmot s index, percentage error of prediction and prediction yields. The statistical assessment revealed the potential of artificial neural network over exponential regression. Dawid (1984) explain in his paper that the purpose of statistical inference is to make sequential probability forecast for future observation rather than to express information about parameters. Therefore, there is a need of an approach which is better than statistical inference method. However, Glahn et. al. (1972) prove that Model Output Statistics (MOS) technique is an objective weather forecasting technique which consists of determining a statistical relationship between a predict and variable forecast by a numerical model at some projection time. It is the determination of the weather related statistics of a numerical model. Glahn has applied this technique, together with screening regression to the predication of surface wind, probability of precipitation, maximum temperature, cloud amount and conditional probability of frozen precipitation. The result is compare by the national weather system over teletype and facsimile. It was concluded that MOS is useful technique in objective weather forecasting. Therefore, in the proposed research statistical regression as multidimensional response surface tool is applied to forecast local monsoonal precipitation. 46

Allen and Vermon (1951) defined objective forecasting system as one which can produce one and only one forecast from a specific set of data. It does not depend for its accuracy upon the forecasting experience or the subjective judgment of the meteorologist using it. Subjective judgment is, of course, used in the development of the system. From all above review, it is clear that there is a need of much better approach which can handle weather parameters more intelligently rather that a crisp theory. The implementation of ANN, an important Soft Computing methodology in weather forecasting has started by Hu (1964). Özelkan and Duckstein (1996) compared the performance of regression analysis and fuzzy logic in studying the relationship between monthly atmospheric circulation patterns and precipitation. Cook and Wolfe (1991) developed a neural network to predict the average air temperatures. Fuzzy logic can also be of great use in the atmospheric data analysis and prediction. Being capable of dealing with linguistic variables, this methodology can be utilized in analyzing atmospheric variables. Liu and Chandrasekar (2000) developed a fuzzy logic and neurofuzzy system for classification of a hydrometeor type based on polarimetric radar measurements where fuzzy logic was used to infer a hydrometeor type and the neural network-learning algorithm was used for automatic adjustment of the parameters of the fuzzy sets in the fuzzy logic system according to the prior knowledge (Mehmet Tektas, 2010). Neural networks and fuzzy inference systems have been widely used in several intelligent multimedia applications. Artificial Neural Network (ANN) learns from scratch by adjusting the interconnections between layers. Fuzzy Inference System (FIS) is a popular computing framework based on the concept of fuzzy set theory, fuzzy if-then rules, and fuzzy reasoning. Integrating ANN and FIS have attracted the growing interest of researchers due to the growing need of adaptive intelligent systems to meet the real world requirements (Abraham, 2001). Gholam abbas et al. (2009) revealed in his study that soft computing techniques are promising and efficient. The root mean square error by using Fuzzy inference system model was obtained 52 mm. Further he stated that unlike conventional artificial intelligence techniques the guiding principal of soft computing is to exploit tolerance for imprecision, uncertainty, robustness, partial truth to achieve tractability and better rapport 47

with reality. Liong et al. (2008) explained in his paper that Neural networks offer a number of advantages, including requiring less formal statistical training, ability to implicitly detect complex nonlinear relationships between dependent and independent variables, ability to detect all possible interactions between predictor variables and the availability of multiple training algorithms. Disadvantages include its black box nature, greater computational burden, proneness to over-fitting and the empirical nature of model development. Bae et al. (2007) apply monthly weather forecasting information in improvement of monthly dam inflow forecasts. The ANFIS (Adaptive Neuro-Fuzzy Inference System) is used to predict the optimal dam inflow, since it has the advantage of tuning the fuzzy inference system with a learning algorithm. A subtractive clustering algorithm is adopted to enhance the performance of the ANFIS model, which has a disadvantage in that the number of control rules increases rapidly as the number of fuzzy variables increases. The use of multi variable Adaptive Neuro Fuzzy Inference System (ANFIS) in predicting daily rainfall using several surface weather parameters as predictors is investigated by Edwin (2008). He found out that relative humidity is the best predictor with a stable performance regardless of training data size and low RMSE amount especially in comparison to those from other predictors. Other predictors show no consistent performances with different training data size. Performances of ANFIS reach a slightly above 0.6 in correlation values for daily rainfall data without any filtering for up to 100 data in a time series. The performance of ANFIS is sensitive to the magnitude and scale differences among predictors, thus suggesting introducing a transforming and scaling factor or functions. The development of neural network model of Back Propagation type is focused in Kassem (2009) paper. He estimates the hourly total and diffuse solar radiation on horizontal surface and result is compared during testing phase statistically by three Methods; Root Mean Square Error (RMSE), Mean Bias Error (MBE), and Coefficient Of Determination (R), giving a value of 4.284 W/m, -0.60 W/m and 0.9309, respectively when estimating hourly total solar radiation. In contrast, the RMSE, MBE and R values of estimating hourly diffuse solar radiation were 2.450 W/m, -3.449 W/m and 0.7802, 48

respectively. These results indicate that this methodology can be used to estimate the hourly total and diffuse solar radiation on horizontal surface. The fuzzy sets approach is realistic and flexible. It may offer a better approach to information transfer than does the classification of climate into discrete sets. (Bratney, 1985). Imran Maqsood et. al. (2002) investigated the development of a reliable and efficient neuro-computing technique to forecast the peak weather in Vancouver, British Columbia, Canada. For developing the models, they used one year's data comprising of daily maximum temperature, wind-speed and visibility. They explained in their paper how neural network models could be formulated using different learning methods and then investigated whether they can provide the required level of performance, which are sufficiently good and robust to provide a reliable model for practical peak weather forecasting. Experiment results demonstrate that neuro-forecast models shown a very good prediction performance and the approach is effective and reliable. Imran Maqsood et. al. (2005) has applied soft computing models to hourly weather analysis in southern Saskatchewan, Canada. They conclude that accurate weather forecasts are necessary for planning our day-to-day activities. However, dynamic behavior of weather makes the forecasting a formidable challenge. They presented a soft computing model based on a radial basis function network (RBFN) for 24-h weather forecasting of southern Saskatchewan, Canada. The model was trained and tested using hourly weather data of temperature, wind speed and relative humidity in 2001. The performance of the RBFN was compared with those of Multi-Layered Perceptron (MLP) network, Elman Recurrent Neural Network (ERNN) and Hopfield Model (HFM) to examine their applicability for weather analysis. Reliabilities of the models were then evaluated by a number of statistical measures. The results indicate that the RBFN produces the most accurate forecasts compared to the MLP, ERNN and HFM. A comparative study of medium-weather-dependent load forecasting using enhanced artificial/fuzzy neural network and statistical techniques has done by M. M. Elkateb et. al. (1998). Monthly peak load demand of Jeddah area for the past nine years was used for investigation. The first seven years data was used for training while the prediction was carried out for the following two years. First, Minitab statistical software 49

package was used for peak load prediction using Autoregressive Integrated Moving Average (ARIMA) technique, and an average error value of 11.7% is achieved. Next, an Artificial Neural Network (ANN) was utilized and several suggestions are implemented to build an adaptive form of ANN. Direct ANN implementation shown poor performance. Also, Fuzzy Neural Network (FNN) was also examined but showed comparatively better performance. The modeling of the trend of peak load demand is incorporated by introducing time index feature and that clearly enhanced the performance of both ANN (6.8% error) and FNN (4.7% error). Neural Networks are good at recognizing patterns; they are not good at explaining how they reach their decisions. Neural Networks can only come into play if the problem is expressed by a sufficient amount of observed examples. These observations are used to train the black box. On the one hand no prior knowledge about the problem needs to be given and it is not straight forward to extract comprehensible rules from the neural network's structure (Fuller, 1995). Fuzzy logic systems, which can work with imprecise information, are good at explaining their decisions but they cannot automatically acquire the rules they use to make those decisions. A fuzzy system demands linguistic rules instead of learning examples as prior knowledge. Furthermore the input and output variables have to be described linguistically. If the knowledge is incomplete, wrong or contradictory, then the fuzzy system must be tuned (Fuller, 1995). Since there is not any formal approach for rule construction, the tuning of membership function is performed in a heuristic way. This is very time consuming and error-prone. Hybridization of systems combining Fuzzy logic, Neural Networks are proving their effectiveness in a wide area of real world problems. Every intelligent system has particular ability to learn that make them suited for particular problems and not for others (Fuller, 1995). Moreover, it is clear that the performance of ANFIS can predict weather condition accurately. Gholam Abbas (2009), Maqsood et al. (2004), Paras et al. (2008), Singh et al. (2006), Abraham et al. (2004), uses temperature, relative humidity and vapour pressure 50

as the input of weather forecasting to train and test the difference and contributions to the outcome of the prediction. 2.3 GAPS IN EXISTING RESEARCH After a comprehensive study made on the existing literature, a lot of limitations/gaps have been found in the area of weather forecasting: Majority of work reported for weather forecasting problems has been done using various statistical methods like Curve Fitting, Regression Analysis, ARIMA model etc. which have their own limitations. Hence a more attention is required towards a new approach for weather forecasting. It may conclude from the literature survey that a lit bit attention has been paid in South Western monsoonal seasonal precipitation prediction. Hence a more research is required towards this. Most of the works reported on weather forecast has paid focus on objective weather forecasting system, not so much attention has been given for the subjective weather forecasting which produces more accurate results. Most of work with the weather forecasting using soft computing is done. There is limited work towards hybridization of Neural Network and Fuzzy System in weather forecasting. Hence more emphasis is required towards it. 2.4 CONCLUSION OF LITERATURE REVIEW From the survey of literature, it is concluded that soft computing techniques especially Neural Network and Fuzzy System has become interesting preference for researchers to solve weather forecasting problems. Development of hybridization of Neural Network and Fuzzy System are still the major issues related to weather forecasting which include forecasting of Local Monsoonal Precipitation (LMP). Therefore, in the present work, weather forecasting system for LMP problem with very good performance measures including very less Root Mean Square Error (RMSE) have been considered. An attempt has been made to develop hybrid algorithm that is based on combination of powers of two algorithms Neural Network and Fuzzy System for weather forecasting of Local Monsoonal Precipitation has been done. 51

2.5 OBJECTIVES OF PROPOSED RESEARCH WORK To Identify and characterize different Agrometeorological parameters that can be used as input variable for Neuro Fuzzy Technique using SPSS. To propose a new approach using hybridization of Fuzzy Inference System and Neural Networks on weather data for Agrometeorological weather forecasting process, optimizing agricultural production using ANFIS editor in MatLab. Assessment of the feasibility of proposed Soft Computing approach for Agrometeorological weather forecast and compare its performance analysis on the basis of Mean Squared Error (MSE) and Root Mean Squared Error (RMSE) with existing Best Statistical Model using various statistical tools in MatLab. 2.6 PROPOSED APPROACH In the proposed research soft computing approach is applied to monsoonal weather analysis in Hisar, Haryana, India. A new approach is proposed that use concept of fuzzy logic in meteorology system. One method to understand such type of inconsistent problem is by using a technique that learns the pattern using the previous data like an adaptive neural network system until it reaches the required level of training error. Such technique is possible by using a hybridization of a neural network and Fuzzy Inference System. This is Adaptive Neuro Fuzzy Inference System called ANFIS. ANFIS is integration of Artificial Neural Networks and fuzzy logic methods. It has the inherent quality to capture the benefits of both these methods in a single framework. ANFIS eliminates the basic problem in fuzzy system design (defining the membership function parameters and design of fuzzy if then rules) by effectively using the learning capability of ANN for automatic fuzzy rule generation and parameter optimization. (Bacanli et al., 2009). These limitations have been a big reason behind the formulation of an intelligent hybrid systems that overcomes the limitations of neural networks and fuzzy systems. Fuzzy systems required to have an automatic adaption procedure which is comparable to neural networks. Hybridizing both approaches should include advantages and exclude disadvantages of both the techniques. 52