Energy Management System Modeling of DC Data Center with Hybrid Energy Sources Using Neural Network

• Main Author: Althomali, Khalid
• Format: Others
• Published: DigitalCommons@CalPoly 2017
• Subjects: DC Data Centers; Renewable Energy Source; Neural Network; Hybrid Energy; Electrical and Electronics; Power and Energy
• Online Access: https://digitalcommons.calpoly.edu/theses/1701; https://digitalcommons.calpoly.edu/cgi/viewcontent.cgi?article=2909&context=theses

As data centers continue to grow rapidly, engineers will face the greater challenge in finding ways to minimize the cost of powering data centers while improving their reliability. The continuing growth of renewable energy sources such as photovoltaics (PV) system presents an opportunity to reduce the long-term energy cost of data centers and to enhance reliability when used with utility AC power and energy storage. However, the inter-temporal and the intermittency nature of solar energy makes it necessary for the proper coordination and management of these energy sources. This thesis proposes an energy management system in DC data center using a neural network to coordinate AC power, energy storage, and PV system that constitutes a reliable electrical power distribution to the data center. Software modeling of the DC data center was first developed for the proposed system followed by the construction of a lab-scale model to simulate the proposed system. Five scenarios were tested on the hardware model and the results demonstrate the effectiveness and accuracy of the neural network approach. Results further prove the feasibility in utilizing renewable energy source and energy storage in DC data centers. Analysis and performance of the proposed system will be discussed in this thesis, and future improvement for improved energy system reliability will also be presented.