How Open Source Communities are Enhancing Insights

How Open Source Communities are Enhancing Insights

‍

A few years ago, catching a glimpse of a Tesla on the road was a rare and exciting sight. But today, electric vehicles are becoming more commonplace, and it's estimated that by the end of 2023, there will be 40 million electric vehicles on the world's roads, which is only 3% of the total vehicle fleet. As we move towards electrifying the remaining 97% of vehicles, there will be an increased demand for battery research and development. The industry is working towards improving battery performance, reducing the cost per kilowatt-hour, and utilizing greener chemistries.

To meet the growth timeline of the battery industry, data-driven approaches have become more critical in battery research and development. Experiments and modeling can generate large amounts of battery data, which can then be used to drive the discovery and optimization of new battery materials and architectures. With traditional battery testing, it could take years to reach thousands of cycles. But with modern machine learning approaches, it's possible to shorten the time needed to predict lithium-ion battery lifetimes to just a few weeks of cycling, after just 100 or fewer cycles. This allows for faster insight that can shape your battery program goals while reducing the overall timeline.

Implementing data-driven approaches within your current battery workflow is now easier than ever, thanks to the unsung heroes contributing to open-source battery software. Open-source software is becoming increasingly popular because it's software that is available for free, and the openly available source code can be modified by users. Instead of requiring each person to reinvent the wheel, they can utilize open-source code as a foundation to build upon. After someone has built their new code on top of the original code foundation, they can share their contribution with the community for others to utilize as well. The highly collaborative environment fostered by open-source software results in the accelerated development of innovative software. Additionally, open-source software makes cutting-edge tools and knowledge accessible to all, independent of budgetary restrictions.

Open-source software for battery research and development typically includes data management and analysis, modeling, open-access battery data, and controlling software. Jason Koeller previously published an article overviewing several open-source software contributions to the battery community, and we will expand upon the list here to reflect recent updates. Software can be generally broken down into a few categories: Data Management and Analysis, Modeling, Controlling, and Open-Access Datasets.

Battery testing can generate large volumes of data that can oftentimes be messy and difficult to work with. Various testing protocols, cycler types, and test modes lead to varying data formats. To perform an apples-to-apples comparison of data, you need to clean, label, and harmonize your datasets. Unfortunately, this can be as tedious as cleaning your home. Data management tools such as AmpLabs, Battery Archive, Cellpy, and BEEP automate this process for you. AmpLabs has made battery data management even easier by providing both a full graphical user interface and the ability to accept any type of battery data, independent of the cycler used - so long as your data is in formats. Once the data is in a clean format, you can perform analysis on it to generate insights. The aforementioned programs allow you to perform some level of analysis, and other programs such as impedance.py, a package built for analyzing electrochemical impedance spectroscopy (EIS) data, enable you to take a deeper look at your data.

Battery modeling and simulation are powerful tools for accelerating the R&D process and reducing costs. If you're looking to dive into this area, you can find some helpful blog posts and peer-reviewed publications on the subject. However, modeling battery systems is no easy task - it involves mathematically intensive, complex models that require coding skills. Fortunately, there are open-source packages available that can help you get started. PyBaMM is a popular choice for battery simulation, offering a framework for differential equations, a library of battery models and parameters, and other tools for analysis. With a combination of single-particle and Doyle-Fuller-Newman models, you can predict voltage, temperature, and concentration as experimental protocols are varied. LIONSIMBA: Li-Ion Simulation Battery is another option that uses a MATLAB framework, although it has mostly been merged into PyBaMM. If you're interested in interphases, MPET allows you to simulate porous electrodes with porous electrode theory. Similar to the PyBAMM/LIONSIMBA merger, discussions of a native integration between AmpLabs & PyBaMM have occurred - we suspect the trend for open-source battery packages to consolidate and collaborate will continue, and are excited to see how this will shape the domain. 

As a battery data scientist or modeler, you may find yourself in need of battery data without actually cycling cells. Remember that a model is only as good as the data it uses, so it's crucial to work with high-quality data. Furthermore, the performance of a model can be impacted by the number of data points fed into it, making it essential to have a significant amount of high-quality battery data. Thankfully, AmpLabs and Battery Archive are two data management platforms that offer publicly available repositories of battery data. These repositories are not only helpful for modelers and data scientists, but also provide a great way to publish your battery data alongside your peer-reviewed publications. There are also several individually published datasets available, which can be explored further in a blog post by Abolfazl Shahrooei. 

If you are someone who enjoys working with hardware and none of the aforementioned software packages have sparked your interest, control software could be what you're looking for. BattGenie has developed and recently open sourced Pymacnet, a Python package that allows you to communicate and control Maccor testers at the channel level. Pymacnet provides real-time data logging, monitoring, and alerting capabilities, automated test management, and enables the use of next-generation closed-loop charging methods, all within a community-supported environment that is user-friendly.

At AmpLabs, we're working to create a comprehensive battery software suite that includes data management and analysis tools, modeling tools, and open-source datasets all in one application. We believe that by breaking down barriers to data and knowledge sharing, we can help accelerate battery development and support the transition to green energy. While there are other battery-targeted software options available, we focus on open-source options. If you know of any other open-source battery packages that you think might be relevant to the community, please feel free to comment below.

‍