HYDROPOWER & FLOOD YEAR PREDICTION USING GREY MODELS

Abstract

Forecasting plays an important role in decision making activity in the field of water resources such as hydropower and flood estimation. There have been many forecasting techniques developed and applied to the foregoing areas, with the objective of minimizing the error in the forecast values. Development and subsequent application of the forecasting techniques have undergone the natural evolution process. One of the latest technique developed for forecasting is Grey Models. GM (1,1) type of grey model is the most widely used in the literature, pronounced as "Grey Model First Order One Variable". This model is a time series forecasting model and in order to reduce possible errors and to improve the accuracy, one can establish a GM(1,1) model using the error sequence. The modified model is named as the remnant GM(l,1) model. Here, forecast accuracy typically measured using the Mean Absolute Percent Error or MAPE. Statistically MAPE is defined as the average of percentage errors. In this dissertation, GM (1,1) and Remnant GM (1,1) are applied to the discharge data for 24 years (1981-2004) on -sarju river, Khutani Small Hydro Project. In order to minimize the time lag between actual and predicted data of the model GM (1,1), it is proposed a combination of grey model with three-point moving average (GM+3PMA) and exponential smoothing technique (GM+ES). The models which have been obtained are compared each other to determine which model has the highest level of accuracy. It will be used as a precise prediction model to predict the future discharge data and discharge prediction which are obtained will be used as hydropower output prediction at current site. Another applications of the grey model are to determine flood year and wet year forecasting. Grey prediction model developed using threshold value that represents the sequence of data based on the classification of flood and extreme flood on the data. The maximum flood data for 58 years are used to flood year prediction and for wet year prediction using the annual rainfall data for 27 years at the Gola River at Kathgodam, District Nainital, Uttarakhand State, India. In this dissertation performances of the various modified grey models in time series prediction have been compared. It is seen that the performance of the grey predictors can be further improved by taking into account the error residuals (Remnant GM(1,1)).The model accuracy examination results show that Remnant GM(l ,1) model is able to make accurate predictions for forecasting.