​
Changelog

​
v1.2.0
04-Aug-2025
Model calibration standard published

​
Model Calibration

The table below sets out the calibration standard for all energy system models that will be built at TZ. It applies to both Capacity Expansion (tz-osemosys) and Unit Dispatch (PyPSA). The calibration standard is country- and scenario-agnostic. This means that it applies to all countries and all scenarios. It consists of a list of model results that should be checked against their respective calibration thresholds to consider a model ‘calibrated’. It also specifies the resolution - spatial and temporal - at which these calibration thresholds should be applied. For e.g. Emissions in the base year should be within +/- 10% of actual historical data for the base year. Model calibration is an essential step in building credibility. The goal of calibration is produce models that analysts can use with confidence. A calibrated model will both reproduce historical energy system trends as well as plausible future energy system pathways. A basic calibration covers capacities, generation, emissions, and trade at a relatively coarse resolution (annual for an entire country). As model and data availability improves, calibration will performed at higher resolution (e.g. hourly for each region) and for a wider range of metrics (e.g. electricity prices). Higher standards of calibration will increase the data and modelling requirements for a given electricity system. For e.g., calibrating a model for electricity prices will require improvements to the model representation of ancillary services, capacity markets etc. as well as remuneration data related to each. Below is a step-by-step guide to the model calibration process: 1. Gather historical data:
  • Collect reliable historical data for the calibration period for the relevant metrics.
  • Ensure the data is consistent with the model’s temporal and spatial resolution.
2. Identify calibration parameters (set in the ‘Calibration standard’ table below)
  • Select the model parameters that can be adjusted to improve the fit with historical data.
3. Implement Calibration Strategy
  • Manual Calibration: Adjust parameters, run the model, compare results to historical data, and repeat.
  • [Future work] Automated Calibration: Use optimisation algorithms to find the parameter set that minimises the difference between model results and historical data.
4. Compare model results with historical data:
  • Compare the model outputs with the historical data for the chosen metrics.
  • Calculate error metrics against the calibration threshold for each of the chosen metrics
5. Iterate and Refine:
  • If the model results do not adequately match the historical data, repeat Steps 4 and 5.+
  • Adjust parameters and re-run the model
6. Validation:
  • Validate the calibrated model by comparing its performance for future years and measuring it against the calibration threshold.

​
Calibration Standard

image.png