< Back to previous page

Project

Personalized Products Through Anticipative Methods Based on Historical Information and System Profiles

Human interaction with control systems that make use of substantial energy is typically sub-optimal, as these interactions are normally not able to determine a good compromise between functionality and energy savings. In this regard, a system with an intelligent control can reduce its consumption greatly while maintaining a high degree of comfort. Such an intelligent control system can find the optimal balance between savings and comfort by accurately predicting the expected usage of the system. In that way, the system can make the right decisions in advance. In literature, several methods have been applied to perform usage prediction; however, the most of them are case dependent and in some cases can be considered as black boxes. The objective of this research thesis is to provide a framework to perform usage prediction that is general enough to cope with different types of applications while providing valuable information about the system (data oriented). This framework is a two-step process: (1) profiling; and, (2) forecasting. The profiling step analyses the system’s historical usage data to determine the typical usage profiles. Additionally, in this step, aspects such as seasonal and calendar influences are also investigated to discover the usage profiles. In the forecasting step, the current usage of the system, as well as its usage of the previous days or weeks, is analysed to detect in which type of profile the system is being used. Every usage profile contains valuable information such as usage probabilities and their uncertainty in function of time. Even though human interaction with such systems can be stochastic, repetitive patterns can be discovered and exploited to predict future usage, thus helping the smart control to find the right balance between savings and comfort.

Date:11 Mar 2014 →  31 Dec 2019
Keywords:Personalized products
Disciplines:Business administration and accounting, Management
Project type:PhD project