Energikonsumtionsprediktion med expanding window approach - En undersökning med syfte att skapa dimensioneringsunderlag för framtida elnätsutbyggnader, i samarbete med Kungälv energi.

Abstract

The impact of electric or hybrid vehicles and photovoltaic systems of a households energy consumption is a growing problem for energy grid owners. As the two increase in popularity the variance of energy consumption between two households in the same area is growing. This generates a major issue for grid owners as they are planning grid connections to future residential areas. To be able to cope with the growing changes in energy consumption the grid owners are dependent on predictions to ensure that the grid can handle the loads generated by the new residential area. In this thesis predictions are generated for the mean peak daily energy consumption, also the difference between households with and without electric vehicles.

The predictions are generated with an expanding window approach and aims to conclude if exogenous variables for the number of electric vehicles and/or mean photovoltaic energy generation improve the predictive power of an ARIMA-based model with only temperature as the exogenous variable. This thesis also aims to examine if there are any difference in mean peak daily energy consumption between households that has photovoltaic system, electric vehicle or neither and if there is a significant difference predicting it. This thesis can not prove that the introduction of variables for number of electric vehicles and/or mean photovoltaic energy generation improve the predictions compared to only using temperature as the exogenous variable. The difference between households with and without electric vehicles is shown to have a increasing trend, this was also predicted with expanding window approach and reached reasonable results.