discharge-normalizer

normalizer.py -- selectively normalizes discharge capacity data

description

normalizer.py is a command line interface (CLI) written in Python 2.7 that reads an input directory containing .csv files of battery cycle data. It then selects data to process and normalizes the discharge capacities of each selected cycle from 0 to 1. This processed data is then output to a nested directory within the original directory, named Normalized_Discharge_Capacities.

Data is selected on the following criteria:

• If the percent difference between the global maximum current (over all cycles) and any current measurement within a cycle is greater than 10%, then only the first cycle and all other cycles with >10% current difference are selected. A significant change in current indicates a change within the cycling parameters (e.g. power consumption), and hence the data should be kept and analyzed during these cycles.

• Otherwise (when no anomalous cycles are detected), then only the first, last, and every 100th trial in between are kept.

usage

On UNIX-like distributions:

Clone/download this git repository and run ./normalizer.py in your terminal once inside the directory.

On Windows:

Clone/download this git repository. Open normalizer.py by double-clicking on the file, and this should automatically spawn a Windows terminal executing this script in python2.7. Alternatively, if Windows does not recognize normalizer.py or if you just hate the Windows terminal like me, you may install the IDLE Python IDE instead and run this module by pressing <F5>. This will allow you to run the script as if you were using a terminal.

future directions

As this is my first professional application of Python, I'm quite proud of it. However, there are many design flaws that can be improved, and I hope to fix them as an exercise to improve my scripting ability in Python.

• Doesn't take full advantage of Python's features. I'm not a big fan of Python thus far. There are a lot of artificial OOP constructs required by the average programmer to memorize, and are completely embedded within the language. However, in exchange, you get a language that makes programming about as easy as building with Legos, as long as you don't need to care about the speed or bloat of your software. I was completely bored out of my mind studying Python, and I only read the first few chapters on data structures and their objects and methods. I'll try to finish my textbook, and apply some of what I've learned to perhaps shorten this script.

• Poor input/exception handling. I didn't do a lot of testing with this, and there are possibly many hidden nuances of user-side implementation I didn't account for. In addition, as errors in Python throw exceptions that completely shut down the program if not caught in a try...except block (rather than having sane exit codes like functions in C), tiny mistakes in the input will break the script and bewilder users. In addition, the input handling could be optimized, I'd like to implement tab-completion and arrow keys if possible.

hydrocodone@t420 ~ % time ./normalizer.py
...
./normalizer.py 29.31s user 0.29s system 78% cpu 37.585 total

• It's slow. While Python inherently is a slow, interpreted language, it shouldn't be this slow. There are a lot of reasons for this. This script creates two temporary files and parses each one before generating the next, and thus a total of three files are parsed for every file generated. Not an intelligent or optimized implementation, but it made the most sense and was easy to implement. Ideally, no temporary files should be generated at all. Reducing the number of for loops used would also improve performance slightly. For 226MB (only 10 batteries) of data, it takes my Thinkpad T420 (Quad-core Intel i5-2520M @3.200GHz) nearly half a minute to process. On the old Windows 7 systems I tested it on, the same 226MB of data took more than a minute. For servers potentially storing gigabytes of unprocessed data, this is unacceptable.

• It's kind of ugly. The lines often exceed 80 characters. Sometimes, this is because of long strings for the CLI, and must be changed manually until I write a script that helps me format them automatically. However, most of the time, there are too many indentations, nested for loops, etc. The readability should be improved.

about

This script was written during my internship at Amprius Corporation, a company specializing in the manufacture of lithium-ion batteries. Hence, it has a rather specific use-case.