A Novel Approach for Forecasting Average Temperature Using Artificial Neural Networks (Applied to Benghazi City's Weather)

Abstract

Weather forecasting, a dynamic system, is play an important role for humans' life development spatially in agriculture industries. There are so many techniques were investigated and presented by researchers to give a better estimated results for weather forecasting, some of these techniques were used a traditional techniques that used statistical methods and others were used the improved techniques as using Artificial Neural Networks (ANN) approach and soft computing as has been investigated in this paper. All of these techniques have used the same weather-measured stations location to obtain the predicted results. In this work, the one-day period temperature is forecasted from historical data that collected from surrounding neighboring weather-measured stations at different geographical locations instead of the current location (target station) of the weather-measured station, as well as the missed data at the target weather-measured station were obtained using the Radial Basis Function (RBF) and ANN model approach. This research has applied to simulate the five measured-weather stations at different geographical locations in Libya and Greece country. Finally, results of generating, forecasting and finding of missing data in the main weather-measured station were excellent in general and the idea behind of this work has been approved using Matlab programming language.