Session: 08-14-03: Emerging Technologies in Solar Energy

Paper Number: 94830

94830 - Using Machine Learning Methods Towards Identifying College Campus Load Profiles and Energy Storage Application for Reducing Peak Energy Demand From the Utility Grid 

Efforts to reduce peak energy demand on the utility grid have been a challenge due to unique demand profiles for individual customers such as college campuses, businesses, and homeowners. By identifying patterns in power consumption, customers can better understand ways to more efficiently use the energy provided by the utility grid, therefore reducing costs and carbon emissions. In recent years, machine learning and pattern classification methods are used to improve systems behavior from being manually reactive to being autonomous. Additionally, these autonomous systems are capable of reacting faster. Using methods of machine learning engineers are able to automate the process of predicting outcomes and solving problems before they occur. This work illustrates the application of machine learning in the form of Bayes Estimation, Principle Component Analysis (PCA), and Fishers Linear Discriminant to identify typical power demand profiles for the author’s institution campus buildings. These methods of machine learning are applied to data collected from the campus in 2019 and focuses on identifying trends in power usage for various types of buildings as well as identify optimal times for charging and discharging of an energy storage system (ESS). Application of the algorithms is carried out using MATLAB to better understand the demand profiles of various academic and residential buildings on campus. The data collected for this research accounts for a total of 16 buildings which includes both hourly and daily power consumption readings from their respective meters. The data from campus buildings is analyzed as both an aggregate load and as individual buildings to determine when peak loading occurs on specific buildings and on the campus collectively. Bayes Estimation is used to determine optimal times for charging and discharging of an ESS using training sets from the power consumption data. On the other hand, Principle Component Analysis is used to determine the features from the data that most effectively differentiate between the academic and residential buildings being observed. Fisher’s Linear Discriminant is applied to the dataset for a similar purpose to Bayes Estimation, however the algorithm is used to determine peak vs non-peak  recordings from the hourly power consumption data. Analysis of the results will be used to further understand where and when ESS can be most effective to reduce peak energy demand from the campus on the local utility grid network. The paper presents the process of applying methods of machine learning to the data as well as the results from the mentioned methods.

Presenting Author: Christopher Sweeny Wentworth Institute of Technology

Presenting Author Biography: Christopher Sweeny is a graduate electrical engineering student at Wentworth Institute of Technology in Boston Massachusetts. His degree focus is in power and energy and he has a bachelors of science in mechanical engineering from the same institution. Chris is an active member of the IEEE PES Boston Chapter and a founding member of the Power and Energy Research Laboratory at Wentworth.

Authors:

Christopher Sweeny Wentworth Institute of Technology
Jackson R. Smith Wentworth Institute of Technology
Afsaneh Ghanavati Wentworth Institute of Technology
James R. McCusker Wentworth Institute of Technology