Evaluating prediction models for electricity consumption
This paper presents a system for visualizing electricity consumption
data along with the implementation of a pattern recognition approach for peak
prediction. Various classification algorithms and machine learning techniques are
tested and discussed; in particular, Support Vector Machine (SVM), Gaussian
Mixture Model (GMM) and hierarchical classifiers. Most notably, the best
algorithms are able to detect 70% of the peaks occurring within the next 24 hours.
Also, various ways of correlating energy consumption are considered in the present
context. Finally, a few directions for future work are discussed.