`optimex` Changelog¶

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[Unreleased]¶

Fixed under-counting of installation impacts. One installed unit now delivers its production temporal distribution over its whole lifetime instead of in every operating year, so a process with an n-year operation window no longer amortizes its construction and end-of-life impacts over n times the output it actually delivers. Optimization results change for any process whose operation window spans more than one time step (#61)
var_installation and var_operation both count process units now, so OperationCapacity compares them directly. existing_capacity and the deployment and operation limits are in the same units: values calibrated as an annual capacity need multiplying by the number of operating years (#61)
Postprocessing capacity outputs (get_production_capacity, plot_capacity_balance, plot_utilization_heatmap) now report annual capacity, directly comparable with production. get_installation reports installed units, a lifetime quantity (#61)

Sped up the whole pipeline a lot: the notebooks/methanol_and_iron.ipynb case study goes from 613 s to 44 s end to end (#64)
Changed results: background_inventory.cutoff now defaults to None. The old default truncated background inventories before aggregating them, biasing flow amounts low by up to 30% (#64)
Background databases are now calculated in parallel by default, and elementary flows without a characterization factor are dropped from the optimization model. See background_inventory.retain_flows and restrict_to_characterized_flows (#64)
Building and solving the model is much faster too: on the same case study, model build 103.6 s to 1.7 s, solve wall clock 55.7 s to 0.1 s and PostProcessor.get_dynamic_inventory() 19.8 s to 0.2 s, with identical results (#64)
get_dynamic_inventory() no longer returns rows whose amount is exactly zero (#64)
Flow limits that cannot take effect are now reported: naming a flow that is not in the model raises in ModelInputManager.override() (pointing at retain_flows), and create_model() warns about limits it has to ignore or that can never bind (#64)
Background inventories are cached on disk per (project, database, activity), so a new session reuses them instead of rebuilding every technosphere matrix. Switch it off with background_inventory.use_disk_cache and clear it with lca_processor.clear_lca_caches(include_disk=True) (#64)

Added foreground_db_name argument to LCAConfig
Fixed LCAConfig.foreground_db_name being ignored by LCADataProcessor, which silently fell back to the database named "foreground". A config naming a non-existent database now raises instead of falling back.

Added Vintage-dependent foreground parameters: Model how process characteristics change based on installation year (vintage). Supports two approaches:
Explicit values per vintage via foreground_*_vintages fields
Scaling factors via vintage_improvements field