Paper: arXiv 2312.07733

Authors: Mike Ludkovski, Saad Mouti, Glen Swindle

Abstract

We consider the construction of renewable portfolios targeting specified carbon-free (CFE) hourly performance scores. We work in a probabilistic framework that uses a collection of simulation scenarios and imposes probability constraints on achieving the desired CFE score. In our approach there is a fixed set of available CFE generators and a given load customer who seeks to minimize annual procurement costs. We illustrate results using a realistic dataset of jointly calibrated solar and wind assets, and compare different approaches to handling multiple loads.

Complexity vs Empirical Score

  • Math Complexity: 7.5/10
  • Empirical Rigor: 8.0/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper employs advanced mathematical concepts like probabilistic constraints, quantile approximations, and SLSQP optimization, indicating high complexity. It is empirically rigorous, featuring realistic data calibration, scenario-based backtesting, and multi-load case studies that simulate real-world procurement structuring.

Research Flowchart

  flowchart TD
  A["Research Goal: Minimize annual cost of 24/7 CFE procurement"]
  B["Key Inputs: Available CFE generators; Hourly load & renewable data"]
  C["Methodology: Probabilistic framework with CFE score constraints"]
  D["Computational Process: Stochastic optimization across multiple scenarios"]
  E["Outcomes: Optimal renewable portfolio; Cost comparison across strategies"]
  A --> B
  B --> C
  C --> D
  D --> E

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper provides a mathematical optimization framework with probabilistic constraints for structuring carbon‑free electricity portfolios and presents two realistic case studies using fully coupled renewable scenarios.

Source sentence
Our contributions are two-fold: (1) we provide a mathemat-ical framework of optimization with probabilistic constraints for CFE structuring, highlighting the role of the two primary parameters of CFE target and guarantee level; (2) we present two realistic case studies where the underlying scenarios are fully probabilistically coupled and calibrated to a realistic renewable assets.

Assumes 4

  • A fixed set of available carbon‑free electricity (CFE) generators is given. model
    Source sentence
    In our approach there is a fixed set of available CFE generators and a given load customer who seeks to minimize annual procurement costs.
  • A collection of simulation scenarios represents potential fluctuations of loads and generators, making the realized CFE score a random variable. statistical
    Source sentence
    These scenarios represent potential fluctuations that render the realized CFE score a random variable.
  • The CFE target must hold for a given large fraction (guarantee level α) of the scenarios. statistical
    Source sentence
    In line with this, we interpret the CFE target probabilistically, so that it must hold for a given (large) fraction of the scenarios.
  • The piecewise‑linear function ¯R(w) can be approximated as Gaussian via the Law of Large Numbers. statistical
    Source sentence
    Arguing by the Law of Large Numbers that ¯R(w) is approximately Gaussian, one can approximate its 1 −α quantile by ˆµ(w) + Φ−1(1 −α)ˆσ(w), where ˆµ(w) and ˆσ(w) are the estimates of µ(w) = E  ¯R(w)  and σ(w) = q Var � ¯R(w)  , and Φ is the standard Gaussian CDF.

Data

  • ARPA‑E PERFORM Dataset other unknown
    ERCOT region renewable assets (2 solar, 2 wind, hydro)
    Source sentence
    The underlying data and asset names refer to the ERCOT region and the ARPA-E PERFORM Dataset [13].
  • Realistic dataset of jointly calibrated solar and wind assets synthetic unknown
    Source sentence
    We illustrate results using a realistic dataset of jointly calibrated solar and wind assets, and compare different approaches to handling multiple loads.

Costs

Not applicable The paper reports no trading or execution result.

Finds 4

  • The resulting procurement cost per MW of load is 74.268. performance
    Source sentence
    The resulting procurement cost per MW of Load is P i w∗ i ci¯gi/Avet(Lt) = 74.268.
  • The constructed least‑cost portfolio on average procures 125.3% of load, indicating significant over‑generation. performance
    Source sentence
    This is significantly higher than the costs in Table II due to overgeneration: the constructed least-cost portfolio π(·) on average procures 125.3% of Load.
  • Stronger CFE targets (higher pC) increase cost exponentially, while a higher guarantee level α adds less than 0.5% cost. performance
    Source sentence
    Stronger CFE targets, i.e. higher pC are more costly, and so is the higher level of guarantee α. We observe that the guarantee level plays a mild role (less than 0.5% additional cost) in portfolio cost, while costs increase exponentially as CFE target pC →1.
  • The maximum achievable average CFE target in the given setting is 99.34%, beyond which load cannot be fully covered even with all CFE sources. theoretical
    Source sentence
    In the given setting, the maximum achievable average CFE target is 99.34%, so that in most scenarios, even if procuring all available CFE sources, there are hours where load is not fully covered.

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 17 related papers). Coverage is partial, so this is neither replication nor refutation.

Code

No repository by the authors found in the paper.

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.