Universal autonomous adaptive control system for HVAC-systems
Buildings consume approximately forty percent of total energy in modern countries. Most of the building energy usage is consumed in HVAC devices.
Therefore, energy efficient HVAC systems have a great potential in conserving energy and diminishing the climate impact. Multiple energy conserving strategies already exists, however, implementing these strategies are typically time consuming and requires a thorough designing. However, the main function of buildings is to provide a comfortable and adequate environment for its users. Therefore, one should conserve no energy by worsening indoor environment considerably.
A growing sensor number in buildings and the internet of things increases the available data from buildings, which computers can use to learn controlling and optimize the operation of building automation system (BAS). In this project a universal autonomous self –adaptive BAS control algorithms is built by using machine learning and existing data to teach the model.
The model will ease implementing energy efficient HVAC control systems, therefore decreasing the climate impact of built environment. In addition, the model will aim to increase indoor conditions at the most of the time. The model is tested in an existing building after developing the model to demonstrate the functioning of the model.
User comfort and energy conserving may be conflicting objectives. In order to find a balance between energy consumption and human comfort one must understand human behavior in buildings. Utilization highly affects to indoor environment, user comfort and energy consumption, thus, information from user behavior and utilization rates are essential to optimize HVAC operation. However, obtaining reliable information from user behavior or utilization is expensive and has possibly privacy concerns.
Thus, the research starts by developing a method to reliably asses occupation number in buildings by analyzing data provided from their existing BAS. This will ease collecting data cost effectively without a need for investing in devices designed to gather user information. Secondly, a dynamic model is developed to predict occupant number in buildings and its spaces, which the model uses as an input amongst the other for the final model. Lastly, the universal autonomous self –adaptive control algorithm is developed and tested in existing buildings. In order to improve the generalizability of the research, over seventy buildings with different usage will be studied. Buildings are mainly public and private offices, museums and schools.