Welcome to meteosynth’s documentation!

meteosynth generates synthetic environmental time-series (solar irradiance, air temperature, wind speed, …) from historical meteorological data using Markov-chain models driven by kernel density estimation. Its primary target is Monte-Carlo simulation of energy-project output and requirements, where many statistically-plausible weather years are needed rather than a single deterministic profile.

Getting started

• Quickstart
  • Installation
  • The 30-second version
  • End-to-end: a synthetic daily ensemble
  • Choosing a training subset: month or rolling window
  • Fetching real data from PVGIS
  • Running the test suite
  • Next steps
• MetDataProcessor Workflow
  • High-Level Workflow Overview
  • 1. Data Ingestion and Validation
  • 2. Seasonal Slicing Strategies (The Slicing Routes)
  • 3. Pivoting to Daily Wide Format
  • 4. Transition Pair Construction and Simulator Fitting
  • 5. Synthetic Generation
• Theory
  • The problem
  • First-order Markov assumption
  • From samples to a joint density (2D KDE)
  • Conditioning and sampling
  • Singularity handling
  • An initial distribution, then transitions
  • Crossing midnight
  • Choosing the training subset
  • Model variants
  • Assumptions and limitations
• Examples
  • 1. Exploratory analysis
  • 2. Synthetic generation
  • 3. Rolling day windows
  • 4. A full synthetic year
  • 5. Leap years and the end of February
  • 6. A Monte-Carlo sketch for energy output
  • Interactive use

API reference

• API Reference
  • Simulators
  • Daily Series Generators
  • Data Processing
  • Plotting Utilities

Indices and tables

• Index

• Module Index

• Search Page