Skip to content

Developing Data-Driven Models for the Thermal Dynamics of Livestock Barns in California Dairy Farms

September 10, 2019
Beri Kohen Behar
Mathematical and Computational Science, Class of 2022

This summer I worked in Powernet, a project focusing on network optimization of distributed energy resources. Since the beginning of winter quarter, I had been working on a sub-project of Powernet which focuses on livestock barns in dairy farms. The goal of this sub-project is to optimize the livestock barn operations by minimizing the energy cost, demand charge, and battery degradation while also keeping the heat stress on livestock under a certain threshold. I worked with PhD student Lily Buechler to use statistical learning methods to develop empirical models of the thermal dynamics of these livestock barns. Throughout the summer, I worked on data gathered from a barn in California, and from National Weather Service forecasts. Most of my work focused on designing programs that would process this data into more organized forms and use this data to develop accurate models for the indoor temperature and relative humidity in the barn. I investigated the suitability of different data-driven models, which will later be implemented in control algorithms aiming to optimize the operation of electrical loads and energy storage.

I’m glad that I had started working with my advisor before summer began. This way I didn’t spend much time trying to refine my research topic, and I had a good sense of what I had to do during the summer. I had already done some data processing and visualizing during the year, which was also helpful. In the beginning of the summer, I went to the dairy farm that we were collecting data from with the research team, which was one of the highlights of my research. Visiting the livestock barn and talking to farm managers gave us insights about certain trends and anomalies in our data that we weren’t able to explain beforehand. Overall, I liked the independence that my advisor provided me with. The problems I was trying to solve were pre-determined, but I was quite free in trying different methods and designing the programs and models the way I was comfortable with. Although I’m not planning to pursue an academic career, I think attending SUPER was a great choice for my freshman summer. I gained a lot of experience in designing solutions for data processing and modeling problems, and I had the chance to meet an advisor and a lab that I’m looking forward to work with during next year as well.