Hydrate Alarm!

Inspiration

We wanted to create a project that incorporates both hardware and software components. That inspired us to look into creating a real-time alarm and dashboard for lease managers' oil wells to immediately provide audio and visual feedback on whether a hydrate has formed. We also wanted to provide data insights into oil well volume, pressure, and temperature trends that can result in hydrate formation, and help save money and a hassle.

What it does

Our tool is a combination of a data analysis tool and a real-time hydrate detection tool. We have a dashboard providing real-time updates from our physical alarm, which detects real-time pressure and temperature. Using these, it relies on physics equations and historical trends to alert the user on whether there is a potential hydrate forming in the well. We also performed data analysis in Python and Excel to identify hydrate formation in the historical data provided to us by EOG Resources.

How we built it

We built an Arduino circuit with pressure and temperature sensors as our alarm, triggering a buzzer and message to be displayed in the case pressure is HIGH and temperature is LOW, conditions for hydrate formation. This alarm will be installed directly into the oil well and alerts the lease manager via our real-time dashboard created in Streamlit, displaying the data from the Arduino. In Excel and Python, we analyzed the given csv data files for volume fill rate and volume outflow rate to determine a linear regression best fit line, calculated the standard deviation of the residuals, and identified potential outliers in the data as hydrates (which were more than 2-3 standard deviations away from the best-fit line) and plotted these on a scatterplot.

Challenges we ran into

We ran into challenges incorporating the hardware and the software before realizing that Arduino data can be streamed directly on Streamlit! We also ran into some errors cleaning up the dataset before creating our ML model.

Accomplishments that we're proud of

Incorporating ML, hardware/circuits, and web development / real-time data streaming all into one app!

What we learned

We learned how to deal with real-life datasets which can have missing data, how to draw trendlines and detect outliers using statistics and ML, how to build and program Arduino circuits, and how to stream Arduino data onto a real-time dashboard.

What's next for Hydrate Alarm!

Analyzing more data related to pressure and temperature to create more accurate predictions for hydrate formation! Enhancing our dashboard to provide better predictions for hydrate formation.