---
title: 'pyFECONs: Fusion Costing Framework'
url: https://www.emergentmind.com/topics/pyfecons
type: topic
---

# pyFECONs: Fusion Costing Framework

Searching arXiv for the specified pyFECONs papers to ground the article in the cited literature.
pyFECONs is an open-source Python framework for fusion power-plant costing that emerged from the spreadsheet-based Fusion Economics code (FECONs) and was refactored to align with the IAEA-GEN-IV EMWG-EPRI code-of-accounts lineage while preserving explicit physics-to-economics linkages. In its documented form, the framework maps subsystem estimates into standardized accounts, computes levelized cost of electricity, and maintains an auditable trail from physics inputs through geometry, subsystem sizing, and cost roll-up. The 2024–2025 Clean Air Task Force International Working Group extensions retain the ARPA-E-aligned chart of accounts and baseline workflow, but add modular capabilities for architecture-specific cost drivers, probabilistic costing, safety-informed costing, and finance/value analysis [2601.21724] [2602.19389].

## 1. Origin, lineage, and scope

pyFECONs consolidates fusion power-plant costing work performed in support of ARPA-E from 2017 through 2024. Early efforts applied ARIES-style cost-scaling relations to generate Nth-of-a-kind estimates and were calibrated through a pilot study with Bechtel and Decysive Systems. Subsequent work expanded the methodology to treat indirect costs explicitly and to evaluate cost-reduction pathways for non-fusion-island systems through design-for-cost practices, modularization, centralized manufacturing, and learning. In 2023 the capability was refactored into a standards-aligned, auditable Python implementation, replacing key ARIES-derived scaling relations with bottom-up subsystem models for dominant fusion cost drivers coupled to physics-informed power balances and engineering-constrained radial builds [2601.21724].

The framework inherits the IAEA/GIF/EPRI lineage of cost accounts, recast for fusion by ARPA-E. At its core, costs are divided into direct construction, indirect and owner’s costs, and supplementary and financial costs. The CATF IWG work extends that deterministic backbone rather than replacing it. The documented implementation states that CATF IWG created a branch in which the COA and baseline workflow remain untouched, while optional modules “hook” into the standard outputs [2602.19389].

Within this lineage, pyFECONs is both a costing framework and an audit structure. Every cost line in the final report is mechanically linked to the physics input, intermediate quantities such as $V_{\rm blanket}$ and $P_{\rm driver}$, the COA subaccount, and the cost basis. Template-driven LaTeX generation guarantees that tables in the PDF are bit-for-bit traceable to the Python-computed state [2602.19389].

## 2. Computational architecture and auditable workflow

The documented illustrative file and module layout centers on `pyfecons/` with `plant.py`, `power_balance.py`, `radial_build.py`, `bop_sizing.py`, `cost_accounts.py`, `lcoe.py`, subsystem modules such as `magnets.py`, `lasers.py`, `power_supplies.py`, and `materials.py`, utility modules such as `constants.py` and `geometry.py`, and `examples/` and `tests/` directories. The top-level orchestration class is `FusionPlant`, which accepts a dictionary of global inputs and calls, in sequence, `PowerBalance`, `RadialBuild`, `BOPSizing`, subsystem cost modules, `CostAccount`, and `LCOE`. It exposes a single `.run()` method that returns a structured report with account breakdown, power summary, and LCOE [2601.21724].

The workflow is organized as a physics-to-economics pipeline. In the base presentation, the sequence is: load inputs; compute power balance; perform radial build; size containment building and site structures; size balance-of-plant and generate plant layout; assemble overnight construction cost; roll up indirect, owner, supplementary, and financial accounts; annualize costs and compute LCOE; and emit LaTeX tables or JSON/CSV output. In the CATF presentation, the same logic is condensed into six auditable steps: plasma power balance; radial build; dominant driver-sizing; balance-of-plant equipment and buildings; overnight capital-cost roll-up into the COA; and annualized costs to baseline LCOE [2601.21724] [2602.19389].

The framework’s core classes reflect this sequencing. `PowerBalance` implements the steady-state energy partition and computes $P_{\rm th}$, $P_{Egross}$, $Q_{\rm eng}$, and $P_{Enet}$. `RadialBuild` applies engineering-constrained thicknesses and minimum shielding requirements and outputs the volumes and areas needed by downstream cost models. `BOPSizing` is anchored to NETL 2019 cost and performance baselines and scales turbines, generators, pumps, condensers, heat rejection, and buildings around the heat-island geometry. `CostAccount` maps all component costs into the IAEA/GIF/EMWG/EPRI “10-level” chart of accounts, while `LCOE` forms accounts 70–90 and computes levelized cost of electricity in $\$/\mathrm{MWh}$ and $\mathrm{c}/\mathrm{kWh}$ [2601.21724].

## 3. Cost-accounting standard and subsystem costing

The top-level “10-level” accounts in the documented implementation are 10 Pre-construction costs, 20 Capitalized direct costs, 30 Capitalized indirect service costs, 40 Capitalized owner’s costs, 50 Capitalized supplementary costs, and 60 Capitalized financial costs. Within Account 20, the direct-cost subaccounts include 21 Structures & improvements, 22 Heat-island plant equipment, 23 Turbine plant equipment, 24 Electric plant equipment, 25 Miscellaneous plant equipment, 26 Heat rejection, 27 Special materials, 28 Digital twin / simulators, and 29 Contingency on direct capital [2601.21724].

| Account | Name | Scope |
|---|---|---|
| 10 | Pre-construction costs | Pre-construction |
| 20 | Capitalized direct costs | Direct plant capital |
| 30 | Capitalized indirect service costs | Engineering, PM, commissioning |
| 40 | Capitalized owner’s costs | Owner’s capital and staff items |
| 50 | Capitalized supplementary costs | Transportation, spares, decommissioning |
| 60 | Capitalized financial costs | Escalation, interest, fees |

Account 22 contains the main fusion-plant equipment structure. Its documented subaccounts include 22.1 Reactor/heat-island components, 22.2 Main & secondary coolant systems, 22.3 Auxiliary cooling, 22.4 Radioactive waste treatment, 22.5 Fuel processing & handling, and 22.7 Instrumentation & control. Within 22.1, the listed components include 22.1.1 First wall & blanket, 22.1.2 Shield, 22.1.3 Coils or Driver, 22.1.6 Vacuum system, and 22.1.7 Power supplies [2601.21724].

The implementation is data-driven. The Python `CostAccount` class reads a data-driven chart of accounts, every subsystem module returns a cost object that is slotted into the correct subaccount, and roll-up uses simple sums for Accounts 10 and 20 together with bottom-up formulas for Accounts 30–60. For example, the material-volume model for first wall, blanket, shielding, and vacuum vessel components is

$$
C_{\rm mat}=\frac{V\;\rho\;c_{\rm raw}\;m}{10^6},
$$

where $V$ is volume, $\rho$ is density, $c_{\rm raw}$ is raw-material cost, and $m$ is a manufacturing multiplier. Power supplies in Account 22.1.7 scale from a reference ITER supply cost and can include a learning-curve credit,

$$
C_{\rm ps}
= C_{\rm ref}
\Bigl(\frac{P_{\rm ps}}{P_{\rm ref}}\Bigr)^{\beta_{\rm ps}}
\times L_N,
\qquad
L_N = N^{\log_2(1 - \mathrm{LR})},
$$

with $\mathrm{LR}\approx0.10\text{–}0.15$ [2601.21724].

The balance-of-plant costing is likewise standardized. `BOPSizing` scales the turbine island to $P_{\rm th,e}$ and $\eta_{\rm th}$, generators and switchyard to $P_{Egross}$, heat rejection to condenser heat load, and buildings and site facilities by $\$/\mathrm{m}^2$ factors. The generic balance-of-plant form is

$$
C_{\rm BOP,i}=k_i\,P_{\rm th}^{\alpha_i},
$$

with parameters calibrated per subsystem from NETL 2019 cost and performance baselines [2601.21724].

## 4. Physics-informed workflow and architecture-defining drivers

A defining feature of pyFECONs is that costing is not independent of plant physics. `PowerBalance` computes alpha and neutron channels from fusion power, then thermal power, gross electric output, recirculating loads, and net electric output. `RadialBuild` defines the geometry of the plasma region, first wall, blanket, shield, coil envelope, and bioshield, enforces surface heat-loading and neutron-shielding constraints, and produces the areas and volumes used by material and equipment modules. This ensures that changes in physics propagate into building size, material volumes, coil size and cost, and balance-of-plant capacity rather than being inserted as independent cost edits [2601.21724].

The CATF reorganization deepens this linkage through three architecture-defining cost-driver tracks for Magnetic Fusion Energy, Inertial Fusion Energy, and Magneto-Inertial Fusion Energy. The generic placeholder in Account 22.1.3 is treated as a controlled swap-point and replaced by a full cost-account development for the dominant driver in each class. In each architecture, the only change to the COA is that the generic placeholder “Coils or Lasers or Pulsed Power” is replaced by the appropriate driver account with full bottom-up detail; balance-of-plant accounts remain common [2602.19389].

| Fusion family | Account 22.1.3 replacement | Driver basis |
|---|---|---|
| MFE | Magnets | Magnetic-field requirements |
| IFE | Lasers/Driver | Pulse energy, repetition rate, wall-plug efficiency $\eta$ |
| MIFE | Pulsed Power (Driver) | Installed \$ per J, module replication, lifetime/derating |

For MFE, Account 22.1.3 “Magnets” is a bottom-up costing of TF/PF/CS coils driven by stored energy, current density, and structural fraction. The documented bill of materials includes superconductors, copper stabilizer, steel, and insulation, together with manufacturing steps such as winding, impregnation, joints, and cryostat integration, plus explicit labor factors. The base superconducting magnet roll-up is

$$
C_{\rm mag}
= m_{\rm mfr}\bigl(C_{SC}+C_{\rm Cu}+C_{\rm SS}+C_{\rm ins}\bigr)+C_{\rm struct}.
$$

For IFE, Account 22.1.3 “Lasers/Driver” is sized from pulse energy, repetition rate, and wall-plug efficiency and is decomposed into gain media, pumping systems, optics, controls, and thermal management. For MIFE, Account 22.1.3 “Pulsed Power (Driver)” uses an installed \$ per J cost basis for capacitor banks, switches, and chargers; module replication logic; and lifetime/derating logic, while Account 22.1.7 captures support infrastructure such as charging supplies and pulse-forming networks [2601.21724] [2602.19389].

## 5. Probabilistic, safety-informed, and finance extensions

The CATF IWG extensions convert the deterministic costing backbone into a modular environment for uncertainty, safety, and finance analysis. The probabilistic layer applies three multiplicative uncertainty factors to selected COA elements,

$$
C = C_0\,U_{\rm mat}\,U_{\rm TRL}\,U_{\rm LR},
$$

where $C_0$ is the deterministic baseline from pyFECONs. The implementation uses log-normal samples for each factor and Monte Carlo propagation to generate cost distributions at both account and plant level. Materials price uncertainty is constructed from geometry-derived mass and uncertain unit cost, TRL-based maturity uncertainty uses a log-normal factor whose $\sigma$ decreases as TRL increases, and learning-curve uncertainty adopts the standard experience-curve form

$$
C(N)=C_1\,N^b,
\qquad
\mathrm{LR}=1-2^b.
$$

Sampling may be independent or use user-specified correlations, yielding plant-level cost distributions, account-level uncertainty attribution, and AACE-aligned “cone of uncertainty” bands [2602.19389].

The safety-informed layer is a modular “safety basis” that maps fusion-relevant hazards into standardized cost accounts. The implemented hazard taxonomy includes plasma disruptions, rapid structural oxidation, direct radiation exposure, cooling accidents, corrosion and activated material transport, cryogenic hazards, tritium and dust release, supplementary heating beam hazards, vacuum-system radiological streams, radwaste handling and storage, fuel-handling hazards, licensing-pathway implications, and insurance/risk posture implications. Screening metrics and proxies include Relative Consequence Indices, first-wall coating cost proxies, bioshield thickness via Dose Transmission Factor, tritium inventory versus site-boundary distance as a land-cost proxy, and detritiation and radwaste storage proxies. The resulting costs map into ACC 20, ACC 21, ACC 22.1, ACC 22, ACC 52, ACC 53, ACC 54, and ACC 62 [2602.19389].

The finance and value module complements LCOE with investment and planning measures computed from the same COA-mapped outputs. The documented formulas include NPV,

$$
\mathrm{NPV}(r)=-I_0 + \sum_{t=1}^T\frac{CF_t}{(1+r)^t},
\qquad
CF_t=R_t-C_t,
$$

IRR, MIRR, WACC,

$$
\mathrm{WACC}=\frac{E}{E+D}R_e + \frac{D}{E+D}R_d(1-\tau),
$$

TLCC, revenue requirements, EAC, payback period, benefit-cost ratios, savings-investment ratios, IRP/DSM tests, surplus metrics, residual value, follow-on value, and adjusted present value. The base pyFECONs LCOE implementation annualizes O&M, fuel, and capital charge and reports levelized cost in $\$/\mathrm{MWh}$ and $\mathrm{c}/\mathrm{kWh}$ [2601.21724] [2602.19389].

## 6. Validation, reported scenarios, and analytical significance

The documented validation record includes a 2017 EPC pilot with Bechtel and Decysive Systems in which ARIES-derived scaling outputs were calibrated against Bechtel’s internal EPC models. In that benchmark, balance-of-plant costs agreed within $\sim 10\%$ in key BOP categories once escalated to a 2017 dollar basis. A 2019 Lucid Catalyst deep dive separated non-fusion-island costs and reported that BOP plus indirect costs often dominate $(> 60\% \text{ of TCC})$ in early concepts. The same line of work projected up to $15\%$ reductions in non-core direct costs from modularization and centralized manufacturing pathways [2601.21724].

A worked-example comparison against MARS/ARIES reported the following relative breakdowns: direct costs 71% for pyFECONs versus 79% for MARS/ARIES; structures/site 12% versus 13%; reactor equipment 20% versus 29%; turbine plant 8% versus 8.7%; indirect services 1% versus 0.6%; owner’s costs 8% versus 6.7%; supplementary 10% versus 7.7%; financial 8% versus 6.5%; and total capital cost of \$2 173 M versus \$9 100 M for the differently scaled worked example. The documented interpretation is that, although absolute numbers differ, the relative breakdown by account is within a few percentage points of the published ARIES/MARS reference [2601.21724].

The CATF notebook examples extend these benchmarks into pilot-plant and NOAK scenario analysis. A notional MFE pilot plant with high TRL and no learning yields LCOE $\approx 180\ \$/\mathrm{MWh}$ with a P50 band $\pm 40\%$. A stylized NOAK scenario with cumulative capacity doubled ten times, 10% learning rate, and material cost variance $\pm 20\%$ yields median LCOE $\approx 90\ \$/\mathrm{MWh}$ with a P90–P10 band $[75,110]\ \$/\mathrm{MWh}$. Safety adders such as bioshield and detritiation shift capital by $+5$–$10\%$ depending on hazard inputs, and insurance can be multiplied by a scenario factor $\phi(\mathcal{R})\in[0.8,1.5]$. Under $\mathrm{WACC}=8\%$ and $T=30\ \mathrm{y}$, the simple capital recovery factor is reported as $0.088$, with revenue requirement $R=C+I_0\cdot0.088$, and IRR $\approx 9\%$ for a contracted PPA at $95\ \$/\mathrm{MWh}$ [2602.19389].

Taken together, these documented features define pyFECONs as a standards-aligned, provenance-complete environment for cross-architecture comparability, drill-down fidelity, uncertainty quantification, safety integration, and finance-coupled interpretation. The framework guides analysis from fusion physics assumptions to cost accounts, LCOE, risk bands, safety adders, and investment metrics while preserving the underlying audit trail [2601.21724] [2602.19389].

Source: https://www.emergentmind.com/topics/pyfecons