A Comprehensive Framework for Long-Term Resiliency Investment Planning under Extreme Weather Uncertainty for Electric Utilities

Emma Benjaminson

cs.AIApr 2, 2026

A Comprehensive Framework for Long-Term Resiliency Investment Planning under Extreme Weather Uncertainty for Electric Utilities

Authors: Emma Benjaminson

Organizations: Boston Consulting Group’s AI Science Institute, Pittsburgh, PA, USA

Abstract

Electric utilities must make massive capital investments in the coming years to respond to explosive growth in demand, aging assets and rising threats from extreme weather. Utilities today already have rigorous frameworks for capital planning, and there are opportunities to extend this capability to solve multi-objective optimization problems in the face of uncertainty. This work presents a four-part framework that 1) incorporates extreme weather as a source of uncertainty, 2) leverages a digital twin of the grid, 3) uses Monte Carlo simulation to capture variability and 4) applies a multi-objective optimization method for finding the optimal investment portfolio. We use this framework to investigate whether grid-aware optimization methods outperform model-free approaches. We find that, in fact, given the computational complexity of model-based metaheuristic optimization methods, the simpler net present value ranking method was able to find more optimal portfolios with only limited knowledge of the grid.

Figures & tables

Fig. 1: This work compares a model-free and a model-based approach to solving a two-stage capital prioritization problem for electric utilities. Model-free approach: The potential investments are ranked by NPV, which is calculated using a baseline simulation from the digital twin to estimate operational costs and benefits, and then scheduled on a FIFO basis. Model-based approach: The NSGA-II algorithm [ 15 ] finds the Pareto frontier of optimized portfolios over two objectives, using the digital twin as an evaluation function. An ILP method is used to optimize the schedule of implementing the investments. Digital twin of the grid: The digital twin computes power flows over the grid layout given known generation, load and extreme weather events. Multiple simulations can be run to obtain summary statistics and confidence intervals for KPIs across different categories including resilience, growth and affordability.

Fig. 2: Test case 1 results: Comparison of optimized portfolio and schedule for model-free vs model-based approaches. NPV + FIFO: Lines 7 and 5 are buried, addressing the two most critical lines on the left-most feeder branch. NSGA-II + ILP: Gantt chart ordering prioritizes projects with the largest marginal benefit. Portfolio includes both burying some overhead lines (not line 7) and upgrading others.

Metric	NPV + FIFO		NSGA-II + ILP
	Median $\pm$ IQR	95% Confidence Interval	Median $\pm$ IQR	95% Confidence Interval
Cost of total unserved energy [$]	$1\,310$ $\pm$ $1\,320$	[ $1\,130$ , $1\,700$ ]	$1\,450$ $\pm$ $1\,410$	[ $1\,290$ , $1\,890$ ]
SAIDI-MED [min]	$496$ $\pm$ $512$	[ $456$ , $646$ ]	$569$ $\pm$ $452$	[ $518$ , $717$ ]
SAIFI-MED	$1.08$ $\pm$ $1.06$	[ $1.01$ , $1.39$ ]	$1.38$ $\pm$ $0.865$	[ $1.24$ , $1.67$ ]
CAIDI-MED [min]	$459$ $\pm$ $172$	[ $408$ , $470$ ]	$421$ $\pm$ $115$	[ $377$ , $439$ ]

TABLE I: Results for Test Case 1 (Varying with Extreme Weather)

Metric	NPV + FIFO	NSGA-II + ILP
Total investment portfolio cost [$]	3720	$17\,900$
Total number of investments	$2$	$7$
Total days to complete work	$2$	$5$
Cost of total resistive losses [$]	$994$	494
Growth in plant value [$]	$-531$	15300
Average capacity headroom [%]	$97.2$	97.4

TABLE II: Results for Test Case 1 (Fixed Across All Scenarios)

Metric	NPV + FIFO		NSGA-II + ILP
	Median $\pm$ IQR	95% Confidence Interval	Median $\pm$ IQR	95% Confidence Interval
Cost of total unserved energy [$M]	$2.10$ $\pm$ $2.01$	[ $2.22$ , $4.06$ ]	$2.12$ $\pm$ $1.98$	[ $2.21$ , $4.05$ ]
SAIDI-MED [hr]	$24.9$ $\pm$ $25.9$	[ $24.9$ , $34.9$ ]	$24.6$ $\pm$ $23.7$	[ $24.6$ , $34.6$ ]
SAIFI-MED	$0.934$ $\pm$ $0.613$	[ $0.879$ , $1.12$ ]	$0.939$ $\pm$ $0.613$	[ $0.875$ , $1.11$ ]
CAIDI-MED [hr]	$27.4$ $\pm$ $12.9$	[ $26.3$ , $30.6$ ]	$26.9$ $\pm$ $12.4$	[ $26.0$ , $30.4$ ]

TABLE III: Results for Test Case 2 (Varying with Extreme Weather)

Metric	NPV + FIFO	NSGA-II + ILP
Total investment portfolio cost [$]	450 000	$467$ 500
Total number of investments	$2$	$4$
Total days to complete work	$25$	$29$
Cost of total resistive losses [$]	196 000	196 000
Growth in plant value [$]	$-187$ 000	-142 000
Average capacity headroom [%]	95.9	95.9

TABLE IV: Results for Test Case 2 (Fixed Across All Scenarios)

Fig. 3: Test case 2 results: The NPV + FIFO portfolio reduced the median unserved energy in Feeder 5 subnet as compared to the NSGA-II + ILP portfolio, and vice versa for Feeder 3.

Category	Metric	Definition
Resilience	Cost of total unserved energy	The total cost of energy not delivered to meet load demand due to outages caused by extreme weather.
	SAIDI-MED	Average total duration of power interruptions per customer per year due to extreme weather events.
	SAIFI-MED	Average total number of outages experienced per customer per year due to extreme weather events.
	CAIDI-MED	Average outage duration experienced per customer per year due to extreme weather events.
Growth	Growth in distribution plant value	The change in the total value of distribution assets over the planning period, accounting for depreciation (modeled as a straight line).
Reliability	Average capacity headroom of lines	Average available margin before reaching thermal limits on lines.

TABLE V: Performance metrics used to evaluate optimization approaches

Parameter	Value
Total budget	$20,000
Total time	1 year
Total technicians	14
Simulation start/end	2016-01-01 – 2016-12-31
Possible investments	9 underground lines upgraded to higher-rated conductors; 4 overhead lines buried.
Outage date for unserved energy calculation	2016-07-01

TABLE VI: Parameters for Test Case 1

Parameter	Value
Total budget	$500,000
Total time	90 days
Total technicians	16
Simulation start/end	2016-01-01 – 2016-03-30
Possible investments	9 underground lines upgraded to higher-rated conductors; 3 overhead lines buried.
Outage date for unserved energy calculation	2016-07-01

TABLE VII: Parameters for Test Case 2

Fig. 4: Test case 2 results: NPV + FIFO: Lines 8 and 39 are buried, addressing the two most critical overhead lines on radial branches. NSGA-II + ILP: The optimized portfolio did not include line 39 and instead buried line 19 which was not critical as it was located on a loop branch.

Power Systems

Power systems research studies the planning, operation, control, and protection of electrical grids and their integration with energy resources.

See all 231 papers

S3TS: Stochastic Scenario-Structured Tree Search for Advanced Planning Under Uncertainty
Jun 1, 2026Fabio Pavirani, Bert Claessens, Pierre Pinson +1Stochastic Optimization Monte Carlo Tree Search
IDLab Ghent university – imec, Technologiepark Zwijnaarde 126, 9052 Gent, Belgium · Beebop.ai, Belgium · Imperial College London, London, UK
Long-term Power Grid Planning via Answer Set Programming
May 19, 2026Antonio Ielo, Francesco Doria, Sandra Castellanos-Paez +3Long-Horizon Planning Power Systems
University of Calabria, Italy · University of Huddersfield, UK
Decision-Focused Scenario Generation and Selection for Efficient and Robust Grid Dispatch
Jul 7, 2026Yangze Zhou, Yihong Zhou, Thomas Morstyn +1Distributionally Robust Optimization Decision-Focused Learning

Jun 1, 2026J/K move · Enter open · S save

S3TS: Stochastic Scenario-Structured Tree Search for Advanced Planning Under Uncertainty

Fabio Pavirani, Bert Claessens, Pierre Pinson, Chris Develder

IDLab Ghent university – imec, Technologiepark Zwijnaarde 126, 9052 Gent, Belgium · Beebop.ai, Belgium · Imperial College London, London, UK

Stochastic Optimization Monte Carlo Tree Search Decision-Making under Uncertainty

Open paper AskPDF arXiv

Effective scheduling in the energy sector is essential to ensure the reliable operation of electrical grids and their connected assets by, for instance, optimizing the dispatch of generation units and storage systems. An effective planning strategy must (a) accommodate advanced and potentially non-linear system models -- exploiting the increasing data availability of modern grids, and (b) explicitly handle uncertainties arising, for instance, from the integration of renewable energy sources. While existing approaches can address either non-linearity (e.g., Monte Carlo Tree Search) or uncertainty (e.g., stochastic mathematical optimization), there is a lack of planning techniques capable of addressing both challenges simultaneously. To bridge this gap, we propose a Stochastic Scenario-Structured Tree Search (S3TS) algorithm that explicitly represents uncertainty through scenario trees while enabling the integration of advanced non-linear models. We evaluate S3TS on a simulated demand response signal publication problem, largely mimicking the imbalance settlement mechanism in Belgium. The results demonstrate near-optimal performance in linear, analytically tractable settings, with costs within 14% of the mathematically optimal solution conditioned to the scenario trees. In highly non-linear scenarios, S3TS significantly outperforms baseline methods, achieving cost reductions of up to 51% and 5.4% compared to a myopic algorithm and deterministic MCTS, respectively.

Abstract

Power Systems

Explore similar work

S3TS: Stochastic Scenario-Structured Tree Search for Advanced Planning Under Uncertainty

Long-term Power Grid Planning via Answer Set Programming

Decision-Focused Scenario Generation and Selection for Efficient and Robust Grid Dispatch