Economic Assumptions

Economic Assumptions#

This page documents how GreenBubble translates technology investment data into annual capital charges, how the discount rate is applied, and how brownfield initial conditions are parameterised.

—

Technology-data source#

Investment costs, fixed O&M rates, efficiencies, and technical lifetimes come from the technology-data repository (branch pypsa-eur_AA), which is a fork of the PyPSA-Eur technology-data catalogue.

The repository provides separate CSV files for each 5-year planning horizon: 2020, 2025, 2030, 2035, 2040, 2045, 2050. All cost projections are in constant real EUR of a fixed base year (i.e. they represent technology learning curves, not nominal price inflation — see Real costs and currency).

The CSV files are downloaded automatically by the retrieve_tech_data rule (all years at once) and stored in data/technology-data/outputs/. A git blob SHA is cached alongside each file so that Snakemake detects upstream changes without a manual --forcerun.

Project-specific overrides (compressor sizing, GLS-specific equipment) are defined as ("technology", "parameter") tuples in scripts/technology_inputs.py and merged into the cost table via helpers.merge_into_costs() before the annuity is computed.

—

Capital cost annualisation#

The annual capital charge for a technology with investment cost I (EUR/MW), fixed O&M rate f (% of investment per year), and technical lifetime L (years) is:

\[\text{capital\_cost} = I \cdot \left[ \text{annuity}(r, L) + \frac{f}{100} \right]\]

where the annuity factor is:

\[\text{annuity}(r, L) = \frac{r}{1 - (1+r)^{-L}}\]

r is the project discount rate (see Discount rate).

This is computed in helpers.read_costs() and stored in the fixed column of the cost DataFrame. PyPSA then multiplies fixed by the optimised capacity (p_nom_opt) to obtain the annual capital expenditure in n.statistics.capex().

Amortization period

By default the annuity denominator equals the technical lifetime L. Setting amortization_period in config.yaml to a positive number (e.g. 15) overrides L for all new expandable capacity:

amortization_period: 15   # recover new investments over 15 years

A shorter amortization period → higher annual charge → harder to invest. null (default) restores the technical-lifetime behaviour.

For existing capacity the effective period is always amortization_period (warn if it exceeds the asset’s remaining technical lifetime — reinvestment may be implied; see Brownfield initial conditions).

—

Discount rate#

discount_rate in config.yaml is a real rate — it excludes inflation. All cost data are expressed in constant real EUR of a fixed base year, so comparing costs across planning years (2020 vs 2030) is valid without any price-level adjustment.

Note

Do not use a nominal rate (which includes expected inflation). Mixing real costs with a nominal discount rate would systematically over-penalise future costs.

Typical real discount rates for energy projects range from 5 % (public finance) to 10 % (private equity). The default is 7 %.

—

year_investment#

year_investment selects which year’s cost CSV is used for new expandable capacity. For example, year_investment: 2030 loads data/technology-data/outputs/costs_2030.csv.

This is independent of the energy/weather year (En_price_year) used for electricity prices and capacity factors.

Available values: 2020, 2025, 2030, 2035, 2040, 2045, 2050.

—

Brownfield initial conditions#

When initial capacity > 0 in n_config.yaml, an existing (EXI_) component is added to the network. Three parameters control its annual capital charge:

construction_year

The year the asset was built. Used to look up the investment cost at the actual build year: I(construction_year). Technology costs change between years (learning curves), so an older plant typically cost more than a plant built today.

If not set (null), defaults to year_investment - 10 (i.e. 10 years before the current planning year), capped at 2020 (the earliest available cost data).

remaining_investment_fraction

The fraction of I(construction_year) that is financially still outstanding. This is independent of the technical remaining lifetime: a fully paid-off asset has remaining_investment_fraction = 0 even if it still has many years of useful life ahead.

0 (default) = sunk cost; the EXI_ component carries no annual CAPEX. 1 = the full original investment is still outstanding.

Annual charge formula

\[\text{EXI\_capital\_cost} = \text{rif} \times I(\text{construction\_year}) \times \text{annuity}(r,\, \text{amortization\_period})\]

where rif = remaining_investment_fraction and the effective amortization period comes from amortization_period in config.yaml (or the technical lifetime if null).

Note

If the asset’s remaining technical lifetime (\(\text{construction\_year} + L - \text{year\_investment}\)) is shorter than the amortization period, a UserWarning is issued: re-investment within the planning horizon may be implied.

Example

Existing electrolysis unit, commissioned in 2020, 60 % of original investment still outstanding, no new capacity to be built on top:

# config/n_config.yaml
electrolysis:
  initial capacity: 5       # MW_el
  expansion: false
  construction_year: 2020
  remaining_investment_fraction: 0.6

With year_investment: 2030, amortization_period: null, discount_rate: 0.07, and a 25-year technical lifetime:

  • remaining_lifetime = 2020 + 25 − 2030 = 15 years

  • I(2020) is read from costs_2020.csv

  • annuity(15, 0.07) ≈ 0.110 (Python call: annuity(n, r))

  • Annual charge = 0.6 × I(2020) × 0.110 EUR/MW/year

Relationship to amortization_period and remaining lifetime

remaining_lifetime

amortization_period

Outcome

> amortization_period

set

Recovery accelerated; asset financially clear before technical end-of-life.

< amortization_period

set

Warning issued; implies re-investment before full recovery.

= amortization_period

null (default)

Standard case; annuity uses remaining technical life.

—

Real costs and currency#

All monetary values in GreenBubble are expressed in real EUR of a fixed base year (the base year is inherited from the technology-data repository). The cost trajectories from 2020 to 2050 represent technology learning (e.g. falling solar costs) — not changes in the general price level.

Consequence: there is no need to inflate or deflate investment costs between construction_year and year_investment. I(2022) and I(2030) are already in the same real EUR, so comparing or dividing them is financially consistent.

USD-denominated technologies are converted at the USD_to_EUR exchange rate set in config.yaml.