Titel:
Load forecast with Neural Networks and Wavelets

Theodor Hugosson

Abstract:
The aim of this thesis is to improve the prediction of the power
consumption in Sweden 24 hours ahead. The technique used is to denoise the
power load with wavelets and then feed the variables needed to describe the
load to a neural network. The choice of variables to feed the neural
network is determined with a test called deltatest. To evaluate the
results, the power load is also predicted using a linear time-series
approach.

The result is that the use of waveletbased techniques does not seem to
improve the quality of the prediction of the power consumption.