Extreme Scenario Selection in Day-Ahead Power Grid Operational Planning

TL;DR

This work tackles day-ahead grid operational risk arising from forecast errors in load and renewable generation by proposing a novel functional-depth screening approach to identify extreme scenarios before running expensive UC/ED simulations. It treats day-ahead scenarios as full daily curves and applies eight functional-depth notions (ID, MBD, EXD, ERLD, LID, HMD, DQ, RTD) to rank scenarios by extremality across multiple facets, including net load, RS, LS, and VC, with adaptive, task-specific pre-screening. The Texas-7k case demonstrates that depth-based screening can recover a large majority of truly extreme scenarios (e.g., up to ~93% for RS and ~93% for LS with zonal info) while reducing the search space, though VC remains more challenging due to grid topology and seasonality. The study reveals that grid topology and seasonal effects strongly influence extreme events, and suggests directions for improvement such as one-sided depth variants and multifacted ranking to further enhance screening efficiency and accuracy in operational planning.

Abstract

We propose and analyze the application of statistical functional depth metrics for the selection of extreme scenarios in day-ahead grid planning. Our primary motivation is screening of probabilistic scenarios for realized load and renewable generation, in order to identify scenarios most relevant for operational risk mitigation. To handle the high-dimensionality of the scenarios across asset classes and intra-day periods, we employ functional measures of depth to sub-select outlying scenarios that are most likely to be the riskiest for the grid operation. We investigate a range of functional depth measures, as well as a range of operational risks, including load shedding, operational costs, reserves shortfall and variable renewable energy curtailment. The effectiveness of the proposed screening approach is demonstrated through a case study on the realistic Texas-7k grid.

Figures (16)

• Figure 1: Fuel mix in the Texas-7k power grid. Total power capacity is 104.9 GW, with 26.73% from VRE generators.
• Figure 2: Case study workflow to efficiently screen clnSim scenarios necessary for day-ahead power grids operational planning in Vatic.
• Figure 3: Four facets of operational extremality against aggregated hourly net load $f^{\cal N}_i(t)$ on Feb 13, 2018. We highlight the top $m=50$ scenarios in terms of each facet: the more extreme a scenario, the brighter its color.
• Figure 4: Detecting reserves shortfall via 4 different functional depth metrics on Feb 13, 2018. Aggregated daily NL on the $x$-axis vs. daily RS on the $y$-axis. Scenarios are color coded according to the respective depth metric: the brighter the color gradient, the less deep the scenario. Green horizontal (vertical red) line at 30.17 GWh (resp. 39.18 GWh) shows the threshold for the top 50 highest RS (resp. top 50 highest NL).
• Figure 5: Detection accuracy for identifying RS. The bubble swarm plots correspond to different functional depth metrics ${\cal D}$(see Table \ref{['tab:aggShort']}), with each dot representing the accuracy $P^{\cal D}_k$ achieved for the $25$ given simulation dates and the gray horizontal bars denoting the respective mean accuracy $Ave_k(P^{\cal D}_k)$. Symbol size denotes the magnitude of respective RS and the colors represent seasons of the year.