Energy and Resource Usage-Aware Building Via Cognitive Internet of Things Agents
Authors
Abstract
There are several ways to improve the economic sustainability of buildings, suchas through embedded systems, or automation, through sensors and actuators. In both
cases, human supervision is omnipresent. On the other hand, the emerging Internet of
Things (IoT) is the inception of a new era where the devices (computers, cell phones,
embedded machines, sensors and actuators etc.) are connected to each other through the
internet, and it can be a practical platform for further automation of energy management.
In this work we show how cognitive software agents, enabled with machine learning
techniques, can support intelligent behaviour for the management of a building’s
infrastructure. Indeed, cognitive models can reduce human efforts in tasks like managing
energy consumption, energy efficiency analysis for potential of energy saving, energy-
aware networking and power management, freeing humans’ attention for more critical
tasks or more abstract level monitoring of the building.
Intelligent management of buildings requires the discovery of energy or resource
consumption patterns; these must be gleaned from the data generated by the large group
of sensors in the building. Such pattern identification and characterisation is a
challenging task, both because of the very-high dimensionality of data (coming from many
sensors) and of the real-time character of the input data stream. Manual processing is
patently unpractical, but a Machine Learning approach seems appropriate. In this work
we use the Regulated Activation Networks (RANs) cognitive model to discover and
characterise such patterns thus enabling the development of Cognitive IoT-based energy
management solutions.