The github repository can be found here: https://github.com/mmfill/electric-motor.git
The blog can be found here: https://medium.com/@matthias.fill/how-to-improve-the-electric-car-250cd92f2793
Short description:
Looking at the data from Paderborn Universität to explain motor temperature and torque of an electric motor by other parameters.
The data can be found here: https://www.kaggle.com/wkirgsn/electric-motor-temperature
Needed packages in Python3:
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
import random
from sklearn.linear_model import LinearRegression
from sklearn.model_selection import train_test_split
from sklearn.metrics import r2_score
files in repository:
- electric-motor.ipynb Jupyter notebook
Motivation:
Personal: I had to to a project for the Udacity Data Science Nanodegree. I chose the dataset of the electric motor, because the topic interests me, but I did not know a lot about. General: Internal parameters like temperature and torque are important features to optimize the efficiency of an electric motor. But they are hard to measure in a driving vehicle. Can we predict both features with "outside" and easy-to-measure parameters like outside temperature, current or voltage? For this data was collected at 2Hz for different runs which lasted 1-6 hours.
Results: All results were achieved with a simple linear regression model. Current i_q is ideal to predict torque. This feature alone explains 99.36% of the variance of torque with a simple linear regression model. This is true when looking at the data of a single run. Motor temperature is more difficult to predict and the given parameters explain about 58% of the variance of motor temperature. This is true when looking at the data of a single run. When combining the data of different runs the explained variance of motor temperature drops to 34% while the explained variance of torque drops to 99.32%.