Information Technology Reference
In-Depth Information
The SOE heating control system aims at integrating the user as an essential part
of the heating control process. At this point, we want to address questions concern-
ing various aspects of human computer interaction. This includes the usability and
acceptance of the developed system with regard to different user groups and/or envi-
ronments. Users are able to access the system using mobile devices and control the
heating process in a fine-grained manner (refer to Fig.
10.5
).
In case that the manually created and automatically optimized heating schedules
differ from each other, the SOE agent will provide recommendations for possible
adaptations. These suggestions are shown as notifications, whereas the user can either
accept the recommended adaptations or reject the automatically generated heating
schedule to manually conduct changes. This is of utter importance, because the
number of falsely classified appliance states is not negligible. False estimates imply
heating during absence or cooling during presence, and are, therefore, undesired.
In our future work, we intend to reduce the number of false estimates and in con-
sequence to improve the appliance classification by using acceptance/rejection as
reward/punishment signal for reinforcement learning strategies. In order to demon-
strate the system outside of our showroom, we use a common notebook to simulate
the smart home and an iPad to show the SOE application.
10.7 Conclusion and Future Work
In this work, we reviewed recent advances in energy disaggregation and adopted
established appliance identification strategies to infer occupancy states for smart
heating control and scheduling. Our proposed approach to appliances state identifi-
cation considers the changes in power consumption as characteristic to classify the
individual devices. In our evaluation, we have shown that the Naive Bayes classifier
is able to achieve relatively high accuracy on the appliance state identification task,
even in unknown environments or households. Furthermore, we explained how to use
the information about identified appliances to infer occupancy states in residential
homes. We exemplified the idea of occupancy-based heating schedules and discussed
the problem of falsely identified appliance states.
The main advantage of our proposed framework for heating control and schedul-
ing is its simplicity in that we refrain from implementing new infrastructure in res-
idential homes, but use given information from available electricity smart meters.
This approach will eventually lead to higher acceptance rates among residents and
provides alternative avenues for novel heating control strategies.
In addition, we demonstrated SOE, a smart heating control system, which inte-
grates the discussed energy disaggregation algorithms to infer appliances states that
indicate presence. Our implementation of the SOE provides insights into practicality
and usability, which are valuable for the intended deployment in real estates.
Since our appliance state identification strategy can replace sensing infrastructure
that is used to identify occupancy states in residential homes, it would also be interest-
ing to compare the energy savings provided by our approach with the performance of
Search WWH ::
Custom Search