Forecasting Daily Discharge in Dam Using Data Mining Techniques

Abstract

     This research aimed to 1) study the appropriate data mining techniques to predict the amount of water in the dam, and 2) compare the prediction of monthly water in Kiew Lom Dam, Lampang, using data mining techniques. This research used information on the factors affecting the level of water in town. The amount of water flowing into the dam, the amount of water in the dam, emissions and evaporated water.  The Laehi Data were collected daily from the year 2535 - 2559, including 25 years by 9,300. Data were monthly separated to use for forecasting with Forecasting Techniques.

     The results showed that: 1) the data mining techniques appropriate for forecasting the water level in the dam consists of four techniques: Regression (Regression Analysis) method, ANN (Artificial Neural Network: ANN), Maroon Peak (Model Tree: M5P), and Technical Support Vector Machine Co. (SVM); and 2) comparing the results from predictions of water monthly in Kiew Lom dam, Lampang, using, the four techniques data mining techniques of found that. MLB replica trees Five Peel showed the lowest absolute tolerances at 10.56 and with the most appropriate  technique to develop a system for forecasting the water in the dam. Considering the absolute values of a deviation from the technical tolerances ascending order at the following: the model tree M. Light Peak with 10.56. the support vector machine was 10.84, the regression analysis with 11:12, and artificial neural networks with 12.53.