Authors:
Tuomo Alasalmi
;
Jaakko Suutala
and
Juha Röning
Affiliation:
Oulu University, Finland
Keyword(s):
Single-point Sensing, Pattern Recognition, Machine Learning, Energy Efficiency, Context-awareness, Smart Home.
Related
Ontology
Subjects/Areas/Topics:
Energy and Economy
;
Energy Monitoring
;
Energy Profiling and Measurement
;
Energy-Aware Systems and Technologies
Abstract:
External single-point appliance load monitoring gives detailed information about appliance electricity use
without expensive or intrusive installation. This is vital for a wide distribution of practical solutions. Current
research has focused on improving the load disaggregation algorithms, whereas consumers would benefit most
from a good feedback system, even if the energy usage estimates are not perfect. A good feedback system can
motivate consumers to save energy from 10% to 15%. In an ongoing project on energy efficient living at the
University of Oulu, we have developed a real-time application using a non-intrusive appliance load monitoring
algorithm. The algorithm is based on thresholding, kNN-classifier, and on-and-off event matching. Accuracy
of the developed system is in line with other similar work and provides a real-time operation. In a test setting,
events were detected with 96.1% accuracy and the total energy estimate differed from the actual consumption
by 11.3%. With
such a solution, consumers can easily see the energy used by different appliances and can
make energy saving decisions because they can see the effects of their actions immediately. This kind of
technologies will play a key role if ever increasing energy saving targets set by international contracts are to
be met.
(More)