Tesla Speed Optimizer PySpark EV Telemetry Analysis

Tesla Model 3 · Argonne SMART 2.0 · PySpark Analytics

km/h

optimal

0.173

Combined Cost

202.2 Wh/km

Energy @ 70 km/h

243.5 Wh/km

Best Bin Avg

441K rows

Telemetry Points

29%

Autopilot Advantage

View on GitHub See Results Explore Pipeline

441,771

Total Rows

Test Drives

143

CAN-Bus Signals

10 Hz

Sampling Rate

522

Valid Wh/km Samples

Pipeline Steps

pipeline --steps

> 4-stage PySpark data engineering pipeline — ETL → Analytics → Visualization → BI

INPUT

20 CSV files · 441K rows
143 CAN-bus signals · 10 Hz

→

ETL

Parse · Clean · Filter
Segment · Feature-engineer

→

STORAGE

Parquet (Snappy, 8 shards)
+ CSV aggregates

→

OUTPUT

Charts · Power BI
Optimal speed: 70 km/h

⚙️ STEP 01

01_pyspark_etl_pipeline.py

PySpark ETL Pipeline

Loads 441K rows of raw CAN-bus telemetry across 20 CSV files. Parses pipe-separated columns, filters driving state, segments trips, and engineers Wh/km, speed bins, and driving mode features using Spark window functions.

PySpark Spark SQL Parquet

→ processed_telemetry/ (8 Snappy-compressed Parquet shards)

📊 STEP 02

02_visualizations.py

Visualization Engine

Generates 6 dark-theme Matplotlib/Seaborn charts: energy vs speed curve, autopilot comparison panels, combined cost overlay, SOC analysis, traffic impact, and a full dashboard PNG.

Matplotlib Seaborn SciPy

→ 6 PNG charts in visualizations/

📈 STEP 03

03_powerbi_preparation.py

Power BI Preparation

Transforms aggregated results into a star-schema for Power BI: 4 fact tables + 3 dimension tables. Generates DAX measures and a full POWERBI_INSTRUCTIONS.txt with relationship diagrams.

Pandas Power BI DAX

→ powerbi_data/ star-schema CSVs

🎯 STEP 04

04_optimal_speed_calculator.py

Optimal Speed Calculator

Computes the combined cost function C = 0.5×(energy_cost) + 0.5×(time_cost) across speed bins. Includes a Spark 4.0 CTE fix and a pure Pandas fallback for environments without Spark.

PySpark Pandas NumPy

→ optimal_speed.csv — 70 km/h identified as optimal

Results & Analysis

Key findings from 522 validated Wh/km samples across 11 test drives

Energy Efficiency by Speed Wh/km · lower is better

Other bins Optimal zone (60–80 km/h)

Driving Mode Efficiency Avg Wh/km by mode

276.6

Autopilot
Wh/km

Autopilot ACC Only Manual

Combined Cost Function C = 0.5 × (Energy Cost) + 0.5 × (Time Cost) · minimum at 70 km/h

Energy Cost Time Cost Combined Cost Optimal (70 km/h)

⚡

km/h

Optimal Speed

🔋

202.2

Wh/km

At Optimal Speed

🤖

29%

better

Autopilot vs ACC

📉

0.173

score

Min Combined Cost

Optimal Speed — Full Breakdown

Speed	Wh/km	Energy Cost	Time Cost	Combined
70 km/h	202.2	0.202	0.143	0.173 ✓
80 km/h	295.5	0.296	0.125	0.210
60 km/h	369.1	0.369	0.167	0.268
50 km/h	360.8	0.361	0.200	0.280
40 km/h	356.0	0.356	0.250	0.303

visualizations --charts

> 6 dark-theme charts generated by the pipeline — hover to enlarge

01 · Energy vs Speed (U-Curve)

Energy efficiency vs speed — shows U-curve with optimal zone at 60-80 km/h

Wh/km vs speed curve. Optimal zone (±15 km/h) marked. Physics annotations explain rolling resistance at low speed and aerodynamic drag at high speed.

02 · Autopilot vs Manual vs ACC Comparison

4-panel comparison of energy, speed, jerk and variability across driving modes

4-panel grouped bar chart: energy consumption, average speed, jerk, and speed variability for Autopilot / ACC-Only / Manual modes.

03 · Combined Cost Overlay

Normalized energy cost, time cost and combined cost vs speed with global minimum marked

Normalized energy + time cost overlay. Global minimum at 70 km/h clearly visible.

04 · State of Charge Analysis

SOC percentage vs distance with linear regression and range estimate

SOC% vs distance + linear regression for range estimation. Energy consumption histogram with mean/median lines.

05 · Traffic Impact

Energy efficiency and speed by traffic density

Wh/km and avg speed by traffic density: No Traffic → Light → Moderate → Heavy.

06 · Full Dashboard

5-panel summary dashboard: optimal speed, autopilot comparison, traffic, jerk, SOC KPI

Complete 5-panel summary dashboard combining all key findings in a single view.

tech --stack

> Production-grade data engineering tools from ingestion to insight

# distributed-compute

⚡

PySpark 4.0.1

Window functions · Spark SQL · Parquet I/O

🗃️

Apache Parquet

Snappy compression · 8 partition shards

# data-processing

🐼

Pandas

Fallback optimizer · Power BI prep · star-schema

🔢

NumPy · SciPy

Normalization · linear regression · range estimates

# visualization

📊

Matplotlib · Seaborn

6 dark-theme charts · physics annotations · dashboards

📈

Power BI

Star-schema · 4 fact + 3 dim tables · DAX measures

# data-source

🚗

Argonne SMART2.0

2020 Tesla Model 3 · 20 test drives · 441K rows · 10 Hz CAN-bus

🔌

CAN-Bus Signals

143 columns: BMS, DAS, GPS, LIDAR, Radar, Autopilot state

github --stats

> repository

> built-with

> contributor-activity