Adaptable short-term forecasting of local-weather parameters

Denis Kotnik (2014) Adaptable short-term forecasting of local-weather parameters. EngD thesis.

Preview

PDF Download (8Mb)

Abstract

The goal of this research is to explore if we could improve the wind speed forecasts, with the regression methods and artificial neural networks. We utilized measurements data, which we obtained from road-weather stations of Direkcija Republike Slovenije za avtoceste and forecast data of INCA system of Agencija Republike Slovenije za okolje. We used the Python programming language for the purpose of data preparation process. In the R programming environment we created the parameter error, which we defined as the difference between the predicted value for time t + ∆t and the measured value in time t+∆t. We predicted the error for subsequent 11 hours with the usage of regression methods and artificial neural networks, then we subtracted it from the INCA-CE predictions and visualised the results. We came to the conclusion that wind speed forecasts for 2014 could be corrected by up to 1 m/s in the early predicted hours with the usage of simple regression methods or neural networks.

Actions (login required)