INTELLIGENT ANALYSIS OF INDOOR CLIMATE FOR INCREASED WELL-BEING

Get Complete Project Material File(s) Now! »

Intelligent analysis of indoor climate for increased well-being

A smart environment may control indoor environment factors to optimize energy consumption as well as productivity and well-being. This often involves control of HVAC in buildings. By control and reduction of the consumption of energy, an optimal indoor climate can be maintained for both environment and cost purposes. For example, by detection of occupancy, the HVAC system and lighting in the building can be optimized (Candanedo & Feldheim, 2016).
To make a smart environment further understand the context of the environment and make automated decisions, intelligent analysis is necessary. Automated support can be provided by analysis of indoor climate data from sensors with algorithms and machine learning techniques like data mining (Cook & Krishnan, 2014).
Drawing from the pervasive computing vision of Weiser (1999) and other research motivated by it, Davies and Clinch (2017) identify a new research area called pervasive data science. It is characterized by “a focus on the collection, analysis (inference) and use of data (actuation) in pursuit of the vision of ubiquitous computing” (Davies & Clinch, 2017, p. 1). They consider smart environments to be one of the most obvious applications of pervasive data science and include algorithms for processing pervasive sensor data as a topic of research. In practice, this could involve smart environments which utilize sensor data to understand the context of the environment and make autonomous decisions regarding its climate.
There are several approaches to control an indoor climate in a smart environment. Merabet, Essaaidi, Benhaddou, Khalil and Chilela (2018) developed a model, which with environment sensor data as well as information about the occupants, could predict their thermal comfort and automatically adjust the environment accordingly. With the same purpose in mind, Peng et al. (2018) used supervised and unsupervised learning to develop a control strategy which responded to occupant behavior and successfully controlled the cooling system in an office.

Problem description

The problem is associated to how the indoor environment can be modeled with relevant data to enable use of machine learning e.g. autonomous decisions, pattern recognition or predictions.
In order to utilize sensor data from the environment for machine learning solutions, it is necessary to model the specific environmental factors. Hence, to model indoor climate in an office environment, suitable factors need to be considered. Humans prefer different types of indoor climate and it may be difficult to create statistical models with subjective data based on perception. Useful connections can be drawn between objective data that is indicated by research to affect the brain capacity e.g. CO2 and temperature (Vehviläinen et al., 2016), as well as humidity. There are many different factors that can be studied and modeled to draw connections, and one can use statistical models to determine which relationships are relevant.
Specifically, this thesis aims to contribute to the knowledge base by modeling conference room environments, using clustering to categorize meeting room quality, and evaluating the clustering performance. This knowledge can be used in a smart office environment to enhance the comfort and productivity. Furthermore, it can provide means to create more energy efficient environments by establishing awareness of the indoor climate.

Related work

The literature shows different ways in which smart environments are used to monitor human activity, analyze human behavior’s impact on indoor climate, and decrease energy consumption. It also shows how environmental data can be analyzed with K-means clustering.
Mozer et al. (1995) conducted research on home automation using machine learning. They saw potential in a system that operates the home and adapts to the behavior of the occupants, both to meet their needs and the energy consumption goals of the home. They explored neural network reinforcement learning and prediction techniques as tools to use adaptive control in a residence that was equipped with multiple sensors and actuators.
To study how human behavior affects indoor air quality, Lin et al. (2017) collected behavior data from sensors and chemical indoor air quality measurements in two smart home environments. They used machine learning algorithms to see what indoor air quality factors were impacted by smart home features. The results showed that there is a strong relationship between human behavior and indoor air quality. This result is useful knowledge to us when we select factors to measure and analyze.

1 Introduction 
2 Background
2.1 SMART ENVIRONMENTS
2.2 INDOOR CLIMATE
2.3 INTELLIGENT ANALYSIS OF INDOOR CLIMATE FOR INCREASED WELL-BEING
2.4 PROBLEM DESCRIPTION .
3 Related work 
4 Aim and Scope
5 Method .
5.1 EXPERIMENT DESIGN
5.2 DATA COLLECTION
5.3 FEATURE ENGINEERING
5.4 CLUSTERING
5.5 EVALUATION .
6 Results 
7 Discussion
7.1 K-MEANS RESULTS
7.2 EVALUATION RESULTS
7.3 CATEGORIZATION OF MEETINGS
7.4 VALIDITY THREATS
8 Conclusions 
References

GET THE COMPLETE PROJECT
Categorizing conference room climate using K-means PAPER WITHIN

Related Posts