How mannuelf/predict-ev-growth-with-ml works
This repository is a research prototype for EV infrastructure planning, with two principal workflows: forecasting Norwegian EV registrations with LSTMs, and optimizing charging-station placement with PPO on a synthetic demand grid. The forecast workflow loads and preprocesses SSB data, trains or tunes models, and produces recursive forecasts and visualizations. The placement workflow trains an agent in a coverage-and-budget environment and compares its results with random placement. README statements and sampled source support the main connections; unsampled modules are included where documented or indexed, but their internal wiring is not inferred.
Main entry points: train.py Model Training · forecast.py Forecast Runner · train_agent.py Agent Training · plot_stations.py Station Visualizations
Components 17
EV Sales Forecasting
- Baseline LSTM
src/models/baseline_lstm.py - Model Training
src/models/train.py - Hyperparameter Study
src/models/hyperparameter_tuning.py - Recursive Forecast
src/models/recursive_forecast.py - Recursive Tuning
src/models/recursive_tuning.py - Forecast Runner
src/models/forecast.py - Model Checkpoints
checkpoints
Station Optimization
- Placement Environment
src/rl_agent/environment.py - PPO Agent
src/rl_agent/ppo_agent.py - Agent Training
src/rl_agent/train_agent.py
Results & Analysis
- Sales Visualizations
src/visualization/plot_sales.py - Station Visualizations
src/visualization/plot_stations.py - Generated Figures
figures
Data Handling
- Data Processing
src/data_processing.py - EV Data Module
src/models/baseline_lstm.py
Other components
- Researcher
- SSB Workbook
data/07849_20251114-073957.xlsx
Connections 25
- Researcher → Model Training: trains model
README.md - SSB Workbook → Model Training: supplies data
README.md - Model Training → EV Data Module: uses preprocessing
README.md - Model Training → Baseline LSTM: trains
README.md - Model Training → Model Checkpoints: saves weights
README.md - Researcher → Hyperparameter Study: runs study
README.md - Hyperparameter Study → Baseline LSTM: tunes
src/models/hyperparameter_tuning.py - Researcher → Forecast Runner: requests forecast
src/models/forecast.py - Forecast Runner → Baseline LSTM: loads model
src/models/forecast.py - Forecast Runner → Recursive Forecast: generates forecast
src/models/forecast.py - Forecast Runner → Sales Visualizations: renders forecast
src/models/forecast.py - Recursive Forecast → Baseline LSTM: uses model
src/models/recursive_forecast.py - Recursive Tuning → Recursive Forecast: tunes forecast
src/models/recursive_tuning.py - Researcher → Sales Visualizations: generates charts
README.md - Sales Visualizations → EV Data Module: loads and prepares
src/visualization/plot_sales.py - Sales Visualizations → Generated Figures: writes charts
README.md - Researcher → Agent Training: trains agent
README.md - Agent Training → Placement Environment: creates environment
src/rl_agent/train_agent.py - Agent Training → PPO Agent: trains policy
src/rl_agent/train_agent.py - Agent Training → Model Checkpoints: saves agent
README.md - Researcher → Station Visualizations: compares placement
README.md - Station Visualizations → Placement Environment: evaluates coverage
src/visualization/plot_stations.py - Station Visualizations → PPO Agent: runs agent
src/visualization/plot_stations.py - Station Visualizations → Generated Figures: writes charts
README.md - Data Processing → EV Data Module