=Paper=
{{Paper
|id=Vol-2180/paper-83
|storemode=property
|title=Supporting Predictive Models Results Interpretation for Comfortable Workplaces
|pdfUrl=https://ceur-ws.org/Vol-2180/paper-83.pdf
|volume=Vol-2180
|authors=Iker Esnaola-Gonzalez,Jesús Bermúdez,Izaskun Fernandez,Aitor Arnaiz
|dblpUrl=https://dblp.org/rec/conf/semweb/Esnaola-Gonzalez18a
}}
==Supporting Predictive Models Results Interpretation for Comfortable Workplaces==
https://ceur-ws.org/Vol-2180/paper-83.pdf
Supporting Predictive Models Results
Interpretation for Comfortable Workplaces
Iker Esnaola-Gonzalez1,2 , Jesús Bermúdez2 , Izaskun Fernández1 , and Aitor
Arnaiz1
1
IK4-TEKNIKER, Iñaki Goenaga 5, 20600 Eibar, Spain
{iker.esnaola, izaskun.fernandez, aitor.arnaiz}@tekniker.es
2
University of the Basque Country (UPV/EHU), Paseo Manuel Lardizabal 1, 20018
Donostia-San Sebastián, Spain
jesus.bermudez@ehu.eus
1 Extended abstract
Approximately 90% of people spend most of their time in buildings, so feel-
ing comfortable while staying indoors is a must. Although many times being an
overlooked factor, research has proven that having an uncomfortable thermal sit-
uation involves many risks including clinical diseases, health impairments, and
reduced human performance and work capacity. Therefore, there is a need to
establish HVAC (Heating, Ventilation and Air Conditioning) control strategies
that ensure comfortable thermal situations in these environments. Most times,
workplaces are complex buildings which cannot be climatized with rather simple
systems like a thermostat-based reactive one. Thus, KDD (Knowledge Discov-
ery in Databases) processes may be applied by data analysts to create predictive
models that identify optimal HVAC control strategies that will ensure thermal
comfort within a workplace. The EEPSA (Energy Efficiency Prediction Seman-
tic Assistant) process assists data analysts through this KDD process, which
can be arduous and very time-consuming when there is a lack of sufficient do-
main knowledge. For that purpose, it takes leverage of the Semantic Technologies
such as the EEPSA ontology3 which aims to capture all the necessary expert
knowledge related to buildings, sensing and actuating devices, and their corre-
sponding observations and actuations. EROSO (thERmal cOmfort SOlution) is
a framework that combines KDD processes and Semantic Technologies for en-
suring thermal comfort in workplaces. Specifically, EROSO supports the KDD’s
Interpretation phase where Semantic Technologies are used to obtain an expla-
nation of predictive model’s temperature predictions with regards to the thermal
comfort regulations they satisfy. Furthermore, this result interpretation supports
facility managers in the task of selecting the optimal HVAC control strategies.
The EEPSA process facilitates the construction of a predictive model to fore-
cast the temperatures for the upcoming hours within a workplace, according to
the different HVAC control strategies4 used as input. The EROSO framework
3
https://w3id.org/eepsa
4
An HVAC control strategy example may be the activation of the HVAC system at
6:00 at 23◦ C until 14:00.
2 Authors Suppressed Due to Excessive Length
begins with the execution of this predictive model (see (1) in Figure 1). That is,
for each HVAC control strategy used as input for the predictive model, a tem-
perature prediction for the upcoming hours is obtained. Once these predictions
are obtained, a Jena script is triggered (see (2) in Figure 1). This script anno-
tates predictions data according to a domain ontology (the EEPSA ontology)
and stores annotated data in an RDF Store. It also executes a set of predefined
SPARQL Construct rules to classify predictions according to the thermal com-
fort regulations they are forecasted to satisfy. Facility managers use the graphic
interface to select the thermal comfort regulation they want to have at their
workplace (see (3) in Figure 1). This triggers the generation and execution of
a SPARQL query against the RDF Store. This way, HVAC control strategies
that are forecasted to satisfy the selected thermal regulation, are shown to fa-
cility managers. Finally, they select and implement the optimal HVAC control
strategy in their workplace’s BMS (Building Management System)5 (see (4) in
Figure 1).
Fig. 1. EROSO framework’s overview.
The EROSO framework has been implemented and tested in the Open Space,
a large office where over 200 people work on a daily basis and is part of the IK4-
TEKNIKER building located in Eibar (Spain). The implementation makes use
of a predictive model generated as part of a previous research work to predict
Open Space’s temperature for the upcoming 24 hours. 20 different HVAC control
strategies are used as inputs of the predictive model, so 20 different temperature
predictions are obtained. This forecasting process is automatically executed daily
at 17:00, so that the facility manager can make a decision on which HVAC control
strategy to implement, in order to ensure next day’s thermal comfort.
EROSO has been used by the IK4-Tekniker building manager and two work-
ers. They have been surveyed and it has been demonstrated that the usability
of the EROSO framework is overall good. Furthermore, the EROSO framework
recommends HVAC control strategies that may ensure a satisfactory thermal
comfort in the Open Space throughout the working day, while another previ-
ous solution implemented in the same physical space, may have periods when
thermal comfort may not be achieved. The EROSO framework is expected to be
implemented in further workplaces with different comfort requirements.
5
A BMS (Building Management System) is the system in charge of setting HVAC
control strategies in buildings.