Forecasting of a hydropower plant energy production

This thesis presents several machine learning and a few statistical methods applied for time series forecasting. The time series used in this work as a case study is a result of the energy production of two small hydropower plants. The amount of produced energy is measured each hour during almost two years. The hydropower plants are located in Hyen, Western Norway. One of them is located in mountains with rocky ground so it means that the effect of rain should be immediate and short. The other one is located further in the valley with the pond nearby which means that the stream has water almost during all year and the effect of the rain is not that immediate. Forecasting of energy production for such small power plants is an essential tool for ensuring ongoing power supply for user demand, planning of reserve power supply and transaction between power plants. Applying several techniques to the datasets we have tried to determine the best method to build a one-step-ahead prediction model. Also we have been aiming to estimate the impact of different attributes on the forecasting process.

Перевод пока недоступен