Autology - the automated ontology generator

Starting point / motivation

The greatest challenge for future energy systems is to coordinate the available energy from fluctuating renewables with the energy demand. On the demand side, the building sector plays a major role: the building stock in the EU is currently still energy-intensive and largely inefficient. It is responsible for 40% of final energy consumption and 36% of CO2 emissions. Smart energy services such as predictive control or fault detection and diagnosis (FDD) could significantly reduce the energy consumption of heating and cooling systems (up to 30%) and at the same time improve the quality of the indoor climate.

Contents and goals

Despite considerable development progress in recent years (e.g. in the field of AI - artificial intelligence), the application of these innovative energy services still falls far short of the potential. One reason is the enormous manual preparation effort: the semantic description (ontology) of data points is of central importance for the scaling of energy services. It contains the functionalities of entities (devices, actuators, measuring points, etc.) and their relationship to each other and to the higher-level system (e.g. building). This is where the Autology project comes in through the use of artificial intelligence.

Methods

The overarching project goal is the automated extraction and generation of metadata for the creation of ontologies from the building automation system using innovative, AI-based approaches. These so-called ontology bootstrapping methods have hardly been investigated in the field of building automation, but the application would bring decisive advantages for operational automation and the scaling of innovative energy services for new and existing buildings and thus goes far beyond the state of the art.

Expected results

As a result of the project, software solutions are to be created that 

1. extract metadata from an existing BACnet in a highly or fully automated manner, enter this into a new or existing ontology, and link it to any existing metadata,
2. that, with a very high level of accuracy and hit rate, classification of measured value time series along several dimensions (sensor type, measured variable, unit of measure, etc.) and
3. recognizing the physical structures and hierarchies behind sensors with very high accuracy and accuracy using measured value time series.

Project management

DiLT Analytics GmbH

Project or cooperation partners

Graz University of Technology, Institute of Software Technology

DiLT Analytics - Theresa Kohl
Rosenberggürtel 22
A-8010 Graz
Tel: +43 (664) 454 05 26
E-Mail: office@dilt.at
E-Mail: theresa.kohl@dilt.at
Website: www.dilt.at