15 Wildfire Assessment for Transmission System Threats

Faculty Mentor: Ali Bidram (Electrical and Computer Engineering, University of Utah)

Wildfires have affected many parts of the world, and with their intensity and frequency only trending upward, solutions to the problems they create, as well as methods to prevent them, must be explored. These fires are often triggered by or damage power lines, which leads to additional environmental damage as well as thousands of people losing power. The purpose of this project is to explore a solution to wildfire prevention and damage containment by creating a prototype of a machine capable of calculating the probability of wildfires occurring near power lines. The method by which this is accomplished is by using microcontrollers, such as Arduinos, that collect environmental sensory information, which is then fed to a Raspberry Pi with a trained machine learning model capable of calculating the probability of wildfires.

Some of the reasons to consider the importance of a project like this include improved grid safety, reliability, and improved fire response. By providing utilities with a better assessment of wildfire risk, the system could provide better information for preventative power outages and reduce the likelihood of unplanned outages caused by wildfires. The result of which could be strengthening the resiliency of power infrastructure while also preventing the broader environmental damage caused by wildfires.

The first portion of the project included research on the causes behind wildfires and the different methods that are currently deployed to predict or prevent wildfires. What was discovered during this portion of the project is that a significant number of wildfires are caused by human activity, of which power line faults contribute significantly. What was also discovered is that many existing wildfire risk indexes aren’t always predictive or local, which is a problem that this project is intended to tackle directly. During this phase of the project, machine learning models were explored as well. Many different machine learning models exist, and so it was important to discover the advantages and disadvantages that each one of these had to offer. Lastly, during this phase of the project, much of the data collection for model training was done.

The second portion of the project includes building a physical prototype of the machine capable of reading the data from sensors and feeding this data to the Raspberry Pi. Before anything was put together, research into what sensors were needed, and exactly which models were needed, had to be completed. The method by which that was determined was by researching which environmental factors were seen as the most essential to predicting fire risk; in the end, this included acquiring sensors capable of measuring temperature, humidity, air quality, wind speed, wind direction, and rainfall. These sensors were then connected to an Arduino capable of reading the input given by each. After data collection is completed, it is then uploaded to a physically connected Raspberry Pi.

Much of the previously mentioned work has been completed; however, at this moment, a deeper analysis into which machine learning model should be used is still being made. Refinement and improvement of the physical prototype and the code that is being used to read sensor input are also being explored. Next steps beyond that could include field testing as well as expanding datasets and developing more advanced prototypes. Developing a fully functional prototype would help communities and environments from the devastating effects of wildfires, so further work will be explored.

Bibliography

Abatzoglou, J. T., & Williams, A. P. (2016). Impact of anthropogenic climate change on wildfire across western US forests. Proceedings of the National Academy of Sciences, 113(42), 11770– 11775. https://www.pnas.org/doi/full/10.1073/pnas.1607171113

NASA Prediction Of Worldwide Energy Resources (POWER). (n.d.). POWER Data Access Viewer. NASA Langley Research Center. Retrieved July 31, 2025, from https://power.larc.nasa.gov/

Texas A&M Engineering Experiment Station. (n.d.). How power lines cause wildfires. Wildfire Mitigation. Retrieved July 31, 2025, from https://wildfiremitigation.tees.tamus.edu/faqs/how-power-lines-cause-wildfires

Udemy. (n.d.). Machine Learning A-Z™: AI, Python & R + ChatGPT Bonus [Online course]. Retrieved July 31, 2025, from https://www.udemy.com/course/machinelearning/?couponCode=MT300725C

U.S. Forest Service. (n.d.). Fire occurrence statistics – Enterprise Data Warehouse (EDW). Retrieved July 31, 2025, from https://data.fs.usda.gov/geodata/edw/datasets.php?xmlKeyword=firestat

Wildland Fire Communicators and Educators. (n.d.). Power Lines and Wildfires: Understanding the Risks. Western Fire Chiefs Association. Retrieved July 31, 2025, from https://wfca.com/wildfire- articles/power-lines-and-wildfires/