Integrity-first control of direct air capture systems under carbon certifiability constraints

[Geoengineering News]

https://www.sciencedirect.com/science/article/abs/pii/S0306261926010846

Authors: Ahmet Çakanel

16 July 2026

https://doi.org/10.1016/j.apenergy.2026.128430

Highlights

•Carbon certifiability enforced in real time as an operational DAC constraint.

•Closed-form integrity gate with proven admissibility and robustness bounds.

•Intra-day CI variability, not mean intensity, governs certifiable availability.

•Causal supervisor rivals perfect-foresight constrained scheduling, without foresight.

•Validated on 3730 grid-days (Turkey, Germany, USA) with open code and tests.

Abstract

Direct air capture (DAC) systems are increasingly deployed with the goal of delivering net-negative carbon emissions through permanent CO2 storage or utilization. However, capturing CO2 from the atmosphere does not guarantee that the removal remains certifiable under conservative measurement–reporting–verification (MRV) frameworks. Grid electricity with variable carbon intensity, auxiliary power demand, and MRV uncertainty margins interact during cyclic operation to create transient periods of net-positive emissions that accumulate as a liability over time. This cumulative liability, referred to as carbon debt, must remain bounded relative to physical capture if reported removals are to be defensible under third-party verification standards. This study introduces Carbon-Integrity-Constrained Control (CICC), a supervisory framework that treats carbon certifiability as an operational constraint enforced in real time, rather than as a retrospective accounting adjustment. Admissible operation is characterized by a bounded debt-to-capture ratio (δ≤δmax). We show that the per-cycle carbon debt is an affine function of grid carbon intensity (empirically, R2=0.9999), which yields a closed-form admissibility gate and supports formal guarantees: under the gate, certifiability is provably maintained (δ≤δmax), and a robustness bound quantifies the loss under imperfect carbon-intensity forecasts. The framework is evaluated on real hourly grid data from three contrasting electricity systems—Turkey (EPIaş, 2020–2022), Germany (ENTSO-E/OPSD, 2019–2022), and the United States (EIA Lower-48, 2019–2021)—spanning 3730 grid-days, and benchmarked against a static carbon-intensity filter, a throughput-optimal model predictive control (MPC) strategy, and a constrained MPC that respects the integrity bound with perfect foresight. Fixed-cycle operation violates integrity constraints on 98.2% of Turkish, 77.9% of German, and 99.9% of US days, while throughput-optimal MPC—despite high availability—fails on a comparable share, confirming that integrity must be enforced as an explicit constraint rather than emerging from optimization. Under CICC-IF, admissibility rises to 97.7% (Turkey), 70.7% (Germany), and 94.7% (USA). Crucially, this admissibility is achieved by curtailing operation to certifiable windows: on the carbon-intensive Turkish and low-variability US grids it reflects near-total, and defensible, non-operation (availability 4.8% and 4.3%; mean capture below 1.1 kg/day), whereas on the German grid the controller sustains substantial certifiable operation (30.9% availability). The causal CICC-IF supervisor matches or exceeds the availability of the constrained, perfect-foresight MPC scheduler on every grid, indicating that the simple online rule is competitive with perfect-foresight constrained scheduling despite using no look-ahead. A central finding is that intra-day carbon-intensity variability, not mean intensity, governs certifiable availability: Germany and the USA exhibit nearly identical mean CI, yet CICC-IF sustains 30.9% availability in Germany (whose renewable-driven grid offers clean windows) versus only 4.3% in the smoother US grid. Sensitivity analysis shows availability rising from 6.0% to 77.8% as δmax relaxes from 0.15 to 0.40, and a persistence forecast reproduces the perfect-foresight results almost exactly, confirming that the framework requires no clairvoyant prediction. These findings establish that supervisory control decisions directly shape whether DAC operation remains compatible with conservative certification expectations, and that integrity enforcement may require accepting reduced operational availability—particularly on carbon-intensive or low-variability grids—as the defensible outcome.

Source: ScienceDirect