In [1]:
import IPython.core.display as di
# This line will hide code by default when the notebook is exported as HTML
di.display_html('<script>jQuery(function() {if (jQuery("body.notebook_app").length == 0) { jQuery(".input_area").toggle(); jQuery(".prompt").toggle();}});</script>', raw=True)
# This line will add a button to toggle visibility of code blocks, for use with the HTML export version
di.display_html('''<button onclick="jQuery('.input_area').toggle(); jQuery('.prompt').toggle();">Show/Hide code</button>''', raw=True)
In [2]:
#Allow the created content to be interactivelly ploted inline
%matplotlib inline
#Establish width and height for all plots in the report
#pylab.rcParams['figure.figsize'] = (18, 6) #width, height
In [3]:
#Import needed libraries
import os
from os.path import join, getsize
import pandas as pd
from cycler import cycler
import matplotlib.pyplot as plt
from IPython.display import display
import numpy as np
import collections
import matplotlib as mpl
inline_rc = dict(mpl.rcParams)
#the next cell enables plotting tables without borders
In [4]:
%%html
<style>
table,td,tr,th {border:none!important}
</style>

Summary report of the CO$_2$MPAS WLTP to NEDC CO$_2$ emission simulation model - real vehicles (v.1.2.2)

Visit the CO2MPAS home page

In [5]:
#Specify the output folder and file containing the CO2MPAS summary output file.
folder = r'D:\co2mpas-version-trials\20160413\out_reals'
file = '20160415_094043-summaryAN.xlsx'
infile = join(folder, file)
df=pd.read_excel(infile, 'summaryAN', header=[0, 1, 2], index_col=[0], skiprows=[3])
In [6]:
#Gather and name the basic variables used in the report according to their name in the CO2MPAS output file
NEDC = df['nedc']['prediction']['co2_emission value']
NEDCt = df['nedc']['target']['co2_emission value']
UDC = df['nedc']['prediction']['co2_emission UDC']
UDCt = df['nedc']['target']['co2_emission UDC']
EUDC = df['nedc']['prediction']['co2_emission EUDC']
EUDCt = df['nedc']['target']['co2_emission EUDC']
#Obtain the case number and vehicle model from the input file
df['vehicle'] = df.index
cases = df['vehicle'].str.split('_').str[-1].astype('int')
model = df['vehicle'].str.split('_').str[0]
#Create a dataframe with this data
valuesDF = pd.DataFrame({'NEDC': NEDC,'NEDCt':NEDCt, 'dNEDC':NEDC-NEDCt,'UDC': UDC,'UDCt':UDCt, 'dUDC':UDC-UDCt,'EUDC': EUDC,'EUDCt':EUDCt, 'dEUDC':EUDC-EUDCt,'Case':cases,'Model':model})   
#calculate percentages
valuesDF['NEDC error [%]'] = pd.Series((valuesDF.dNEDC/valuesDF.NEDCt*100), index=valuesDF.index)
valuesDF['UDC error [%]'] = pd.Series((valuesDF.dUDC/valuesDF.UDCt*100), index=valuesDF.index)
valuesDF['EUDC error [%]'] = pd.Series((valuesDF.dEUDC/valuesDF.EUDCt*100), index=valuesDF.index)
valuesDF = valuesDF.dropna()

Section 1. Performance of the model. All vehicles and test cases.

Error statistics for CO$_2$ emission per driving cycle

Error statistics for NEDC, UDC, and EUDC CO$_2$ emission [gCO$_2$ km$^{-1}$]

In [7]:
#Create a dataframe with the NECD, UDC, EUDC error statistics
errorsDF = pd.DataFrame(index=['Averages','StdError','Median','StdDev','Variance','Kurtosis','Skweness','Range','Minimum','Maximum','Sum','Count','Confidence level (95%)'], columns=['NEDC [gCO$_2$ km$^{-1}$]','UDC [gCO$_2$ km$^{-1}$]', 'EUDC [gCO$_2$ km$^{-1}$]'])
errorsDF.loc['Averages'] = pd.Series({'NEDC [gCO$_2$ km$^{-1}$]':round(valuesDF.dNEDC.mean(),2), 'UDC [gCO$_2$ km$^{-1}$]':round(valuesDF.dUDC.mean(),2), 'EUDC [gCO$_2$ km$^{-1}$]':round(valuesDF.dEUDC.mean(),2)})
errorsDF.loc['StdError'] = pd.Series({'NEDC [gCO$_2$ km$^{-1}$]':round(valuesDF.dNEDC.sem(),2), 'UDC [gCO$_2$ km$^{-1}$]':round(valuesDF.dUDC.sem(),2), 'EUDC [gCO$_2$ km$^{-1}$]':round(valuesDF.dEUDC.sem(),2)})
errorsDF.loc['Median'] = pd.Series({'NEDC [gCO$_2$ km$^{-1}$]':round(valuesDF.dNEDC.median(),2), 'UDC [gCO$_2$ km$^{-1}$]':round(valuesDF.dUDC.median(),2), 'EUDC [gCO$_2$ km$^{-1}$]':round(valuesDF.dEUDC.median(),2)})
#errorsDF.loc['Mode'] = pd.Series({'NEDC [gCO$_2$ km$^{-1}$]':round(valuesDF.dNEDC.mode().iloc[0],2), 'UDC [gCO$_2$ km$^{-1}$]':round(valuesDF.dUDC.mode().iloc[0],2), 'EUDC [gCO$_2$ km$^{-1}$]':round(valuesDF.dEUDC.mode().iloc[0],2)})
errorsDF.loc['StdDev'] = pd.Series({'NEDC [gCO$_2$ km$^{-1}$]':round(valuesDF.dNEDC.std(),2), 'UDC [gCO$_2$ km$^{-1}$]':round(valuesDF.dUDC.std(),2), 'EUDC [gCO$_2$ km$^{-1}$]':round(valuesDF.dEUDC.std(),2)})
errorsDF.loc['Variance'] = pd.Series({'NEDC [gCO$_2$ km$^{-1}$]':round(valuesDF.dNEDC.var(),2), 'UDC [gCO$_2$ km$^{-1}$]':round(valuesDF.dUDC.var(),2), 'EUDC [gCO$_2$ km$^{-1}$]':round(valuesDF.dEUDC.var(),2)})
errorsDF.loc['Kurtosis'] = pd.Series({'NEDC [gCO$_2$ km$^{-1}$]':round(valuesDF.dNEDC.kurtosis(),2), 'UDC [gCO$_2$ km$^{-1}$]':round(valuesDF.dUDC.kurtosis(),2), 'EUDC [gCO$_2$ km$^{-1}$]':round(valuesDF.dEUDC.kurtosis(),2)})
errorsDF.loc['Skweness'] = pd.Series({'NEDC [gCO$_2$ km$^{-1}$]':round(valuesDF.dNEDC.skew(),2), 'UDC [gCO$_2$ km$^{-1}$]':round(valuesDF.dUDC.skew(),2), 'EUDC [gCO$_2$ km$^{-1}$]':round(valuesDF.dEUDC.skew(),2)})
errorsDF.loc['Range'] = pd.Series({'NEDC [gCO$_2$ km$^{-1}$]':round((valuesDF.dNEDC.max()-valuesDF.dNEDC.min()),2), 'UDC [gCO$_2$ km$^{-1}$]':round((valuesDF.dUDC.max()-valuesDF.dUDC.min()),2), 'EUDC [gCO$_2$ km$^{-1}$]':round((valuesDF.dEUDC.max()-valuesDF.dEUDC.min()),2)})
errorsDF.loc['Minimum'] = pd.Series({'NEDC [gCO$_2$ km$^{-1}$]':round(valuesDF.dNEDC.min(),2), 'UDC [gCO$_2$ km$^{-1}$]':round(valuesDF.dUDC.min(),2), 'EUDC [gCO$_2$ km$^{-1}$]':round(valuesDF.dEUDC.min(),2)})
errorsDF.loc['Maximum'] = pd.Series({'NEDC [gCO$_2$ km$^{-1}$]':round(valuesDF.dNEDC.max(),2), 'UDC [gCO$_2$ km$^{-1}$]':round(valuesDF.dUDC.max(),2), 'EUDC [gCO$_2$ km$^{-1}$]':round(valuesDF.dEUDC.max(),2)})
errorsDF.loc['Sum'] = pd.Series({'NEDC [gCO$_2$ km$^{-1}$]':round(valuesDF.dNEDC.sum(),2), 'UDC [gCO$_2$ km$^{-1}$]':round(valuesDF.dUDC.sum(),2), 'EUDC [gCO$_2$ km$^{-1}$]':round(valuesDF.dEUDC.sum(),2)})
errorsDF.loc['Count'] = pd.Series({'NEDC [gCO$_2$ km$^{-1}$]':round(valuesDF.dNEDC.count(),2), 'UDC [gCO$_2$ km$^{-1}$]':round(valuesDF.dUDC.count(),2), 'EUDC [gCO$_2$ km$^{-1}$]':round(valuesDF.dEUDC.count(),2)})
errorsDF.loc['Confidence level (95%)'] = pd.Series({'NEDC [gCO$_2$ km$^{-1}$]':2*round(valuesDF.dNEDC.sem(),2), 'UDC [gCO$_2$ km$^{-1}$]':2*round(valuesDF.dUDC.sem(),2), 'EUDC [gCO$_2$ km$^{-1}$]':2*round(valuesDF.dEUDC.sem(),2)})
errorsDF
Out[7]:
NEDC [gCO$_2$ km$^{-1}$] UDC [gCO$_2$ km$^{-1}$] EUDC [gCO$_2$ km$^{-1}$]
Averages 0.17 -1.63 1.2
StdError 0.99 2.09 0.68
Median 1.38 -1.34 1.08
StdDev 4.66 9.78 3.18
Variance 21.68 95.71 10.13
Kurtosis 0.04 0.32 -0.64
Skweness 0.09 -0.15 -0.05
Range 18.97 41.14 11.19
Minimum -7.92 -21.12 -4.6
Maximum 11.05 20.02 6.59
Sum 3.8 -35.86 26.35
Count 22 22 22
Confidence level (95%) 1.98 4.18 1.36

Error statistics for NEDC, UDC, and EUDC CO$_2$ emission [%]

In [13]:
#Create a dataframe with the NECD, UDC, EUDC error statistics
errorsDF = pd.DataFrame(index=['Averages','StdDev','Minimum','Maximum','Cases in ±4%','Cases in ±2.5%'], columns=['NEDC [%]','UDC [%]', 'EUDC [%]'])
errorsDF.loc['Averages'] = pd.Series({'NEDC [%]':round(valuesDF['NEDC error [%]'].mean(),2), 'UDC [%]':round(valuesDF['UDC error [%]'].mean(),2), 'EUDC [%]':round(valuesDF.dEUDC.mean(),2)})
errorsDF.loc['StdDev'] = pd.Series({'NEDC [%]':round(valuesDF['NEDC error [%]'].std(),2), 'UDC [%]':round(valuesDF['UDC error [%]'].std(),2), 'EUDC [%]':round(valuesDF['EUDC error [%]'].std(),2)})
errorsDF.loc['Minimum'] = pd.Series({'NEDC [%]':round(valuesDF['NEDC error [%]'].min(),2), 'UDC [%]':round(valuesDF['UDC error [%]'].min(),2), 'EUDC [%]':round(valuesDF['EUDC error [%]'].min(),2)})
errorsDF.loc['Maximum'] = pd.Series({'NEDC [%]':round(valuesDF['NEDC error [%]'].max(),2), 'UDC [%]':round(valuesDF['UDC error [%]'].max(),2), 'EUDC [%]':round(valuesDF['EUDC error [%]'].max(),2)})
errorsDF.loc['Cases in ±4%'] = pd.Series({'NEDC [%]':round(sum(valuesDF['NEDC error [%]']<abs(4))/len(valuesDF['NEDC error [%]'])*100,1), 'UDC [%]':round(sum(valuesDF['NEDC error [%]']<abs(4))/len(valuesDF['UDC error [%]'])*100,1), 'EUDC [%]':round(sum(valuesDF['NEDC error [%]']<abs(4))/len(valuesDF['EUDC error [%]'])*100,1)})
errorsDF.loc['Cases in ±2.5%'] = pd.Series({'NEDC [%]':round(sum(valuesDF['NEDC error [%]']<abs(2.5))/len(valuesDF['NEDC error [%]'])*100,1), 'UDC [%]':round(sum(valuesDF['UDC error [%]']<abs(2.5))/len(valuesDF['UDC error [%]'])*100,1), 'EUDC [%]':round(sum(valuesDF['EUDC error [%]']<abs(2.5))/len(valuesDF['EUDC error [%]'])*100,1)})
errorsDF
Out[13]:
NEDC [%] UDC [%] EUDC [%]
Averages 0.15 -0.68 1.2
StdDev 3.16 4.75 2.77
Minimum -5.77 -10.01 -3.89
Maximum 6.4 8.41 6.22
Cases in ±4% 90.9 90.9 90.9
Cases in ±2.5% 81.8 72.7 72.7

Distribution of the NEDC, UDC and EUDC errors [%]

In [9]:
mydict = ([('NEDC', 0), ('UDC', 1), ('EUDC', 2)])
mydict = collections.OrderedDict(mydict)
for cycle in mydict:
    if cycle == 'NEDC':
        boxcolor = 'green'
    elif cycle == 'UDC':
        boxcolor = 'blue'
    else:
        boxcolor = 'red'
    # Create a figure instance
    fig = plt.figure(1, figsize=(14, 7))
    # Create an axes instance
    ax = fig.add_subplot(111)
    hist = valuesDF[cycle+' error [%]'].hist(bins=np.arange(min(valuesDF['NEDC error [%]']), max(valuesDF['NEDC error [%]']) + 0.5, 0.5), color=boxcolor)
    hist.set_xlabel(cycle+" error [%]",fontsize=14)
    hist.set_ylabel("frequency",fontsize=14)
    hist.set_ylim(0,5)
    plt.title(cycle+' CO$_2$ emission error distribution', fontsize=20)
    plt.ylabel("frequency",fontsize=18)
    plt.tick_params(axis='x', which='major', labelsize=16)
    plt.tick_params(axis='y', which='major', labelsize=16)
    ax.get_xaxis().tick_bottom()
    ax.get_yaxis().tick_left()
    ax.set_xlim(-10, 10)
    plt.show()

Comparative emission error per driving cycle (%)

In [10]:
#Boxplots of the error distribution
toboxplot = [valuesDF['NEDC error [%]'],valuesDF['UDC error [%]'],valuesDF['EUDC error [%]']]
# Create a figure instance
fig = plt.figure(1, figsize=(14, 7))
# Create an axes instance
ax = fig.add_subplot(111)
# Create the boxplot with fill color
bp = ax.boxplot(toboxplot, sym='', patch_artist=True, whis=10000, showmeans=True, meanprops=(dict(marker='o',markerfacecolor='yellow')))
for box in bp['boxes']:
    # change outline color
    box.set( color='black', linewidth=1)
    # change fill color
    box.set( facecolor = '#b78adf' )
    ## Custom x-axis labels
ax.set_xticklabels(['NEDC', 'UDC', 'EUDC'],fontsize=20)
## Remove top axes and right axes ticks
ax.get_xaxis().tick_bottom()
ax.get_yaxis().tick_left()
#Set y axis title
plt.title('CO$_2$ emission error by driving cycle', fontsize=20)
plt.ylabel("error [%]",fontsize=18)
plt.tick_params(axis='y', which='major', labelsize=16)
ax.set_ylim(-12, 12)
plt.setp(bp['medians'], color = 'purple', linewidth = 2)
plt.show()
print('The purple box represents the 1st and 3rd quartile.\nThe dark purple line is the median.\nThe yellow dot is the mean.\nthe whiskers show the min and max values.')

The purple box represents the 1st and 3rd quartile.
The dark purple line is the median.
The yellow dot is the mean.
the whiskers show the min and max values.

Section 2. Performance of the model. Statistics per vehicle model and case test.

NEDC, UDC, and EUDC CO$_2$ emission error [%] per vehicle model

In [14]:
Carlist = list(sorted(valuesDF['Model'].unique()))
Cidlist = list(range(len(Carlist)))
valuesDF.cid = valuesDF['Model'].replace(Carlist, Cidlist, regex = True)
valuesDF['cod'] = valuesDF.cid
#Create a table with the error statistics for each car model
CarDF = pd.DataFrame(columns=['NEDC error [%]','UDC error [%]', 'EUDC error [%]'])
for x in Carlist:
    Car = valuesDF[valuesDF['Model'] == x]
    CarDF.loc[x] = pd.Series({'NEDC error [%]':round(Car['NEDC error [%]'].mean(),2), 'UDC error [%]':round(Car['UDC error [%]'].mean(),2), 'EUDC error [%]':round(Car['EUDC error [%]'].mean(),2)})
CarDF.columns.name='Error'
display(CarDF)
mydict = ([('NEDC', 0), ('UDC', 1), ('EUDC', 2)])
mydict = collections.OrderedDict(mydict)
for cycle in mydict:
    if cycle == 'NEDC':
        boxcolor = 'green'
    elif cycle == 'UDC':
        boxcolor = 'blue'
    else:
        boxcolor = 'red'
    #plot the emission error per model, and cycle
    fig = plt.figure(1, figsize=(14, 7))
    labels = ['{0}'.format(i) for i in valuesDF['Model']]
    plt.title(cycle+" error [%] per real vehicle",fontsize=20)
    plot = fig.add_subplot(111)
    plot.tick_params(axis='x', which='major', labelsize=14)
    plot.tick_params(axis='y', which='major', labelsize=14)
    plot.set_xlim(0, 23)
    plot.set_ylim(-15,15)
    plot.get_xaxis().tick_bottom()
    plot.get_yaxis().tick_left()
    plt.scatter(valuesDF['Case'],valuesDF[cycle+' error [%]'], color=boxcolor, marker = 'o')
    for label, x, y in zip(labels, valuesDF['Case'], valuesDF[cycle+' error [%]']):
        plt.annotate(label, xy = (x, y), textcoords ='offset points', ha = 'left', va = 'top', size = 12)
    plot.set_xlabel("Vehicle #",fontsize=20)
    plot.set_ylabel(cycle+" error [%]",fontsize=20)
    line1 = plot.axhline(y=-2.5, color='grey', linestyle='-.', label='± 2.5%')
    line2 = plot.axhline(y=2.5, color='grey', linestyle='-.')
    line3 = plot.axhline(y=-4, color='black', linestyle='--', label='± 4.0%')
    line4 = plot.axhline(y=4, color='black', linestyle='--')
    plt.legend(handles=[line1, line3], loc = 3)
    plt.show()
Error NEDC error [%] UDC error [%] EUDC error [%]
AT01 0.81 -3.34 5.00
AT02 3.16 3.42 2.93
AT03 -2.57 -5.72 1.26
AT04 -4.33 -10.01 -0.38
AT05 1.40 2.76 0.13
AT06 6.40 8.41 4.32
AT07 -1.28 -2.72 0.38
MAN01 -5.09 -7.43 -2.73
MAN02 -0.77 -0.66 -0.84
MAN03 3.70 0.35 6.22
MAN04 -2.95 -2.01 -3.66
MAN05 1.22 2.68 -0.10
MAN06 1.53 -1.53 4.57
MAN07 -1.55 -2.06 -1.09
MAN08 1.81 -0.27 3.44
MAN09 0.17 -1.83 2.02
MAN10 1.63 2.27 1.06
MAN11 -5.77 -9.64 -2.03
MAN12 4.13 6.13 2.44
MAN13 -2.74 -1.10 -3.89
MAN14 2.50 4.82 0.72
MAN15 1.97 2.48 1.58