---
updatedAt: 2025-11-11T22:13:06.000Z
agentTools:
  projectIndex: https://forecastdocs.modoenergy.com/llms.txt
---

# New-build, Retrofit, and Retirement Decisions

We attempt to build a realistic evolution of the capacity stack given both net-zero goals but also current commercial (and operational) realities. For example, we do not forecast a system with no gas on it in 2035 as we see that would lead to loss of load - which we see as a red line.

### Economics inform capacity growth and decline

Hurdle rates for NPV and IRR calculations are based on [analysis by Europe Economics](https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/910814/Cost_of_Capital_Update_for_Electricity_Generation_Storage_and_Demand_Side_Response_Technologies.pdf), with adjustments made to account for higher interest rates since this was published.

* We make retirement and retrofit decisions for **unabated CCGTs**. CCGTs that are not retrofitted with CCS are retired in the year of their maximum cumulative cashflows. A plant is retrofitted with CCS if it is technically feasible:
  * close to a planned carbon storage cluster,
  * profitable (NPV is greater than 0),
  * preferable to continued unabated generation (i.e., there is a higher NPV for retrofit).
* We make new build decisions for **abated CCGTs** (ie CCS). New abated CCGT plants are built when IRR exceeds the hurdle rate. We allow a maximum of 3 abated CCGT build decisions in each model iteration (ie across the forecast horizon). We assume that newly built unabated CCGT plants have a capacity of 0.9 GW. For the first 5GW of CCS, we assume a hurdle rate of 0%. For every subsequent GW that is built, we assume an incremental jump in the hurdle rate of 0.3% as the technology matures.
* We make new-build decisions for **gas reciprocating engines**. New plants are built when IRR exceeds the hurdle rate. New-build gas reciprocating engines have capacities of 100MW, and we allow a maximum of 10 new-build decisions for each, in each iteration. Existing OCGTs and gas reciprocating engines are retired in the year of their maximum cumulative cashflows. We don't build any new OCGTs.
* We make **new build decisions** for **batteries**. New batteries are built when IRR exceeds the hurdle rate. We allow a maximum of 10 new-build decisions over the time horizon per iteration, and apply a learning rate to the capacity of batteries we can build per year, starting at 3GW and by 2040 this is 4.8GW. We assume that newly built batteries have a capacity of 0.2 GW.

**For example,** when we say we allow 10 new build decisions for storage in each model iteration, it means we could build 10 new batteries in iteration 1, i.e., 1 per year from 2030 to 2040. In iteration 2, we build another 10 - giving 2 per year from 2030 - 2040. This could push spreads down in the price model - reducing IRR - so in iteration 3 we get 1 per year from 2030 - 2037, ie 7 new systems. Or, 17 new systems in total.

As we loop through multiple iterations of the model, it converges on a suitable build-out: one that responds to system needs without saturating the revenues (causing low IRRs) and avoids an overbuild.

<Image alt="An example of varying technology build-out. This is taken from a development run and is not used within the final forecast" align="center" src="https://files.readme.io/537dfc4f4020b98a51b809f195d407f359684dfe63bc4c46a53c711ea1e4b412-aec1fa2-ccgt_capacity_evolution.png">
  An example of varying technology build-out. This is taken from a development run and is not used within the final forecast
</Image>

All decisions are reversible. This means that any under or overbuild of a generation type can be corrected in the next iteration. This results in an iterative process where capacity build-out converges to an optimal solution.

![](https://files.readme.io/eac3bd0-image.png)