A DC-area church contacted OAITI about performing a statistical analysis of their energy usage data. The church knew there were valuable insights contained within the data, but needed some assistance in creating a workflow and process that transforms the energy data into an appropriate format, and produces results that are interesting, interpretible, and actionable.
The church provided three separate data tables, of varying formats and structures. The three tables were:
- Gas Usage
- Electricity Usage
- Temperature Data
The levels of aggregation also varied between datasets, with electricity usage being measured at an hourly level, and gas usage at a daily level.
We followed a standard data analysis procedure, defining key terms, restructuring the data sources so that gas, temperature, and electricity data could be seamlessly merged, performing an exploratory analysis to produce interpretible visuals, and doing some basic modeling routines in order to assess various factors that impact energy usage
The church took our findings and used them to produce recommendations for reducing energy usage. One key finding that was noted was a weekly seasonality in electricity usage, spiking at expected times (such as Sunday morning service attendance) but also during events that some members of the church were unaware of. This provided insight into when the church might expect usage to spike and decline, both at a weekly and a daily level. Recommendations were provided to reduce costs by cutting down on gas usage during the summer months. The costs saved would be redirected to more beneficial programs run by the church.
Reducing our impact on the environment is a key component of creating a sustainable future. As data scientists, we have a carbon footprint that results from our reliance on technology to perform our work. In particular, as Deep Learning and other intensive GPU-based tasks continue to rise in popularity, the impact on the environment will grow worse over time. If we can offset this impact by helping to raise awareness about energy usage, both at OAITI and in other organizations, we can help to make more informed decisions that have a practical impact on the planet that we inhabit.
The slides are provided below:
Watch a video presentation of this analysis
Download the source code
About the Author
Nicole Wright studied Chemical Engineering at Iowa State University before becoming a Research Associate at Proliant Biologicals. She continued her education with a certificate of Data Analysis from Colorado State University Online and follows her passion for data science as a volunteer with OAITI.