Carbon-Aware Optimal Power Flow with Data-Driven Carbon Emission Tracing
Zhentong Shao, Nanpeng Yu
TL;DR
The paper tackles the challenge of attributing carbon emissions to nodal loads in power networks by introducing a data-driven, affine carbon tracing framework that yields linear expressions for nodal emissions. By learning generator-to-load distribution factors $\alpha_{n,g}$ via constrained regression and deriving a closed-form LMCE $\mu_n$, the authors embed carbon accounting directly into a DC-OPF as linear constraints, enabling efficient, real-time optimization. Case studies on IEEE test systems demonstrate high fidelity of the learned factors, accurate LMCE rates, and emission reductions with minimal total cost increase when applying carbon-constrained dispatch. The approach offers practical impact for carbon-aware grid operations and demand-side management, with potential extension to multi-period and stochastic settings.
Abstract
Quantifying locational carbon emissions in power grids is crucial for implementing effective carbon reduction strategies for customers relying on electricity. This paper presents a carbon-aware optimal power flow (OPF) framework that incorporates data-driven carbon tracing, enabling rapid estimation of nodal carbon emissions from electric loads. By developing generator-to-load carbon emission distribution factors through data-driven technique, the analytical formulas for both average and marginal carbon emissions can be derived and integrated seamlessly into DC OPF models as linear constraints. The proposed carbon-aware OPF model enables market operators to optimize energy dispatch while reducing greenhouse gas emissions. Simulations on IEEE test systems confirm the accuracy and computational efficiency of the proposed approach, highlighting its applicability for real-time carbon-aware system operations.