VAV control based on backpropagation neural network

• Main Author: Zhou, JIngwei
• Other Authors: Su Rong
• Format: Theses and Dissertations
• Language: English
• Published: 2018
• Subjects: DRNTU::Engineering::Electrical and electronic engineering
• Online Access: http://hdl.handle.net/10356/76040
• Institution: Nanyang Technological University

The commercial and residential building sectors occupy more than 40% of primary energy consumption in the worldwide and which is continuously increasing yearly due to the increasing population and economic activities. As a tropical country, more than 50% electricity consumption are used by air-conditioning for commercial buildings in Singapore. The GBIC group provided an economical Internet of Things upgrade that implemented the Token Based Scheduling Algorithm to reduce energy consumption in heating, ventilation, and air-conditioning (HVAC) systems in commercial buildings. The IoT prototype is formalized with different hardware, software, their interaction and integration. Different sensor modules are installed to collect environment data include indoor/outdoor temperature, indoor humidity, CO2 level, supply air speed and mass flow rate, the data will be transmitted to zone module for computing minimum cooling energy based on local thermal model. Then the data and computed request are uploaded to MySQL database based on Ethernet communication. The requests are balanced by central scheduler with consideration of several constrains and then tokens are allocated to each zone for next sampling period, which aims at minimizing the total energy consumption of HVAC system. The detailed description of components, software, hierarchy and working processes are presented. Two approaches for VAV control based on backpropagation neural network are introduced which aims at training neural network with labeled historical data and desired output to perform a control decision. First method is trying to predict how different environment parameters will influent the indoor temperature change in next sampling period and do prediction for future temperature change based on real-time environment data. The second method is trying to adjust the VAV system to make the environment parameters approach to desired values based on training error. The detailed explanation about data preparation, data processing, results analysis and challenges faced are provided. In addition, a central sever side GUI and sub-GUI (MATLAB Environment) has completed for administrator to manage owners’ information and enhance monitoring and visualization of environment changes for each zone. And a user-side web GUI (PHP Environment) also has completed for personalized energy management which allows individual owner to control temperature set-point and check the historical environment data.