SPUR 2019: Developing Sustainable Energy Storage Using Machine Learning


With the goal of limiting or replacing non-renewable energy sources, extensive research has gone into renewable forms of energy, such as solar and wind power. With this increased research drive, demand has increased, efficiency has increased, costs have dropped, and implementation of these technologies has extended globally. While these sources of renewable energy are in a position to supply world energy demand, one huge problem stands in their way: continuous supply is not always available and storing excess energy generated during peak times is not possible with current battery technologies. As demand for sustainable and “green” energy continues to rise, more efficient and cost-effective batteries are becoming necessary. Typically, battery material development and testing has been a time-consuming endeavor with low odds for success, however new research paths in materials science are beginning to change this. These paths, which the PI is a pioneer in, include utilizing machine learning techniques to predict material performance before synthesis ever occurs. If machine learning techniques are applied to study battery materials, battery efficiencies and cyclability as a function of material combinations can be studied to identify which materials stand to have high success rates and high-performance characteristics when implemented into batteries. There is a long list of materials which need to be assessed for potential applications in batteries. A variety of tasks will need to be undertaken for this research, including gathering, overseeing, and preparing data for use in machine learning algorithms.

Student Role

The student involved with this project will work alongside two NSF funded graduate students on comparing battery materials and material combinations using materials informatics, data mining, and machine learning. The student will work as a point of contact between research groups at the University of Utah and Brigham Young University who are actively gathering data in this area. During this research, the student will organize and edit existing data, help gather additional, new data from literature and online databases, merge existing data sets with newly collected data, and write code in Python to perform data analysis. As data is gathered and sorted, a data-centered approach will be used to find suitable models and develop a method for screening potential battery materials and predicting their performance; specific battery parameters explored will include anode and cathode types, electrolyte materials, and battery cyclability. These calculations can then be used to identify candidates for density functional theory calculations and future synthesis that may follow from this summer research.

Student Learning Outcomes & Benefits

Data science is a rapidly growing field and the student will learn how to employ machine learning techniques to a sustainability focused, materials science based project. The student will learn broad skills in Python coding and data science. Statistical treatment of datasets using SciPy, a Python based statistical tools library, will be used to determine basic statistical descriptions of collected data. The student will also learn how to utilize more advanced machine learning techniques involving Scikit-Learn, one of the most widely used machine learning libraries; these techniques will include principle component analysis and regression, random forest algorithms, neural networks, support vector machines, leave-one-out-validation, and descriptor and algorithm development. The student will gain research team experience, connections across universities, and will become an expert in the field of battery materials and performance assessment, which are the future of global renewable energy.

Taylor Sparks
Assistant Professor

Materials Science & Engineering
College of Engineering
Global Change & Sustainability Center

I will provide significant support during the initial stages of the project, where project planning and management will be critical for success. Here I will guide the student through meaningful research based on an understanding of the global impacts of this project. After the formative stage, my mentoring approach will take on a supportive role, allowing the student to learn techniques on their own terms under the direct guidance of graduate student mentors. This policy enables the student to explore their own interests within the project while also benefiting from the detailed technical expertise of their peers, who will guide the student through best practices in machine learning, data management, and coding techniques. I also offer an open-door mentoring policy: anytime a question or clarification comes up, I am available both in my office and through electronic means, and all questions are welcome and encouraged. I hope to learn as much from the student that joins this project as they learn from being a part of it!