All events are in Central time unless specified.

“A Framework to Predict Variability Characteristics in Building Load Profiles”

Architectural Engineering PhD Defense by Sam Moayedi

Date: Time: 3:00 pm–4:00 pm
Peter Kiewit Institute Room: 250
Contact: Durham School, (402) 554-4482, durhamschool@unl.edu
Under the supervision of Dr. Moe Alahmad

To manage building-energy consumption, owners use advancements in monitoring and sensing devices to collect a wide range of energy trends and consumption at different time scales (from sub-minutes, hours, days, weeks, months, or years). Utility-installed Advanced Metering Infrastructure (AMI) measures the energy consumption at lower resolutions (15, 30, or 60 minutes). Generally, the time resolution load profile is often limited by cost, storage, and bandwidth. In contrast, the time resolution of modeled data is limited by the intractability of predicting high-frequency and pseudo-random variations. Because many fields of study in electrical systems require data at varying high-resolution time frames, this project proposes a novel, data-driven approach to predict characteristics of the missing high-resolution information in a low-resolution signal, applicable to both measured and modeled building load profiles through machine learning regression algorithms. These statistical metrics can be used to constrain the potential distribution of the high-resolution signal. Therefore, the proposed framework contributes to the research community by presenting a method that overcomes smart meter constraints via generating the statistics of a missing high-resolution in a given low-resolution building load profile that benefits various applications such as realistic estimation of the peak, load profile bounds, and accurate average load estimations.

Download this event to my calendar