Technical Summary: Resilient Grid Hardening against Multiple Hazards

1. Problem Statement

The increasing frequency and intensity of extreme weather events, driven by climate change, pose severe threats to power system resilience. Infrastructure damage from hazards such as high winds, earthquakes, and vegetation-related incidents leads to prolonged outages and significant economic losses. A critical challenge in grid planning is that traditional hardening strategies often focus on single hazards or single mitigation measures. For instance, undergrounding (UG) lines protects against wind and vegetation but may increase vulnerability to seismic activity or incur high repair costs if an earthquake occurs. Conversely, vegetation management (VM) is effective for overhead lines but does not address seismic risks.

Furthermore, existing optimization models often rely on static, two-stage stochastic programming where initial hardening decisions are fixed for the entire planning horizon. This rigidity fails to account for evolving climate conditions and uncertainties, potentially leading to suboptimal and costly outcomes. The paper addresses the need for a decision-making framework that:

1. Simultaneously considers multiple hazards (specifically high winds and earthquakes).
2. Integrates complementary hardening strategies (long-term undergrounding and short-term vegetation management).
3. Allows for adaptive decision-making to revise initial plans based on realized future scenarios.

2. Methodology

The authors propose an Adaptive Two-Stage Stochastic Optimization (ATS) framework designed for radial distribution systems with distributed energy resources.

2.1. Scenario Tree and Uncertainty Modeling

The problem is modeled using a scenario tree where nodes represent possible realizations of uncertainties (wind events, earthquakes, vegetation failures) over a planning horizon $T$.

• Uncertainty Parameters: The model incorporates random variables for outage duration ($\xi$), the number of events ($N$), and failure probabilities ($\varrho$) for each hazard type.
• Tree Pruning: To ensure computational tractability as the horizon extends, the authors employ a backward reduction method to prune the scenario tree, retaining a representative set of scenarios that approximate the original probability distribution.

2.2. Adaptive Two-Stage Structure

Unlike standard two-stage models where first-stage decisions are fixed, the ATS framework introduces a revision mechanism:

• First-Stage Decisions: Initial decisions regarding which lines to underground ($\alpha_{ln}$) are made at the start.
• Revision Point: At a designated time $t$, the model allows for a single revision of these initial decisions based on observed uncertainty and updated forecasts.
• Second-Stage Decisions: Vegetation management levels ($\beta_{ln}$) are determined annually (short-term) and remain scenario-dependent throughout the horizon.

2.3. Objective Function and Constraints

The objective is to minimize the total expected cost over a 30-year horizon, comprising:

• Investment Costs: Capital costs for undergrounding ($IC^{UG}$) and annual vegetation management ($IC^{VM}$).
• Operational Costs: Outage costs (load shedding) and repair costs associated with wind, earthquake, and vegetation failures.
• Discounting: Future costs are discounted using a factor $\gamma$ derived from inflation and discount rates.

Key Constraints:

• Budget Limits: System-wide caps on undergrounding ($B^{UG}$) and vegetation management ($B^{VM}$) budgets.
• Irreversibility: Once a line is undergrounded, the decision cannot be reversed.
• Mutual Exclusivity: Vegetation management is only applicable to overhead lines; undergrounded lines do not require VM.
• Revision Logic: Constraints ensure that each line undergoes at most one revision decision, and decisions remain consistent across scenarios until the revision point.

2.4. Load Shedding Calculation

To quantify outage costs, the model solves a separate offline optimization problem for each potential line outage. This uses the DistFlow model to calculate the daily load shedding cost ($C^{LS}$) based on power flow constraints, nodal voltage limits, and microturbine ramping capabilities.

3. Key Contributions

The paper makes three primary contributions:

1. Integrated Multi-Hazard Model: A comprehensive optimization model that jointly considers high winds and earthquakes. It explicitly accounts for the trade-offs where undergrounding mitigates wind/vegetation risks but may expose infrastructure to seismic risks, and vice versa.
2. Adaptive Decision Framework: The introduction of an adaptive two-stage stochastic approach that allows for a single revision of long-term undergrounding decisions. This balances the flexibility of multi-stage models with the practicality of two-stage approaches, accommodating evolving risk conditions without frequent, costly reversals.
3. Synergistic Strategy Optimization: The integration of long-term (UG) and short-term (VM) strategies. The model demonstrates how these strategies interact; for example, undergrounding a segment eliminates the need for VM on that segment, creating cost synergies.

4. Simulation Results

The framework was tested on a modified IEEE 22-bus distribution system using MATLAB and Gurobi.

4.1. Optimal Hardening Decisions

• Budget Sensitivity: As the undergrounding budget increases, the model selects more lines, prioritizing those with high load shedding impact and high wind vulnerability but low seismic risk.
• Cost Reduction: Increasing the undergrounding budget significantly reduces total system costs (objective value), as it lowers outage and repair costs.
• Strategy Interaction: As more lines are undergrounded, the required vegetation management budget decreases, confirming the synergistic cost savings.

4.2. Adaptive vs. Static (ATS vs. TS)

• Performance: The Adaptive Two-Stage (ATS) approach consistently outperforms the standard Two-Stage (TS) model.
• Magnitude of Gain: The relative cost savings ($G_{rel}$) of ATS over TS range from 11.85% to 17.83%, depending on the budget and scenario tree size.
• Conditions for Value: The advantage of ATS is most pronounced under higher budgets and larger scenario trees (higher uncertainty). In shorter planning horizons or low-uncertainty scenarios, the value of adaptation diminishes.
• Multiple Outages: In scenarios involving simultaneous outages of multiple lines, the flexibility of ATS provides even greater relative benefits (up to 19.43% improvement), highlighting its robustness in high-impact failure scenarios.

4.3. Multi-Hazard and Multi-Strategy Benefits

• Multi-Hazard Planning: Considering both wind and earthquake risks leads to different, more cost-effective line selections compared to single-hazard models. For instance, lines with high seismic vulnerability are avoided for undergrounding even if they have high wind risks, preventing elevated lifecycle costs. This approach yielded cost savings of 53–59% compared to single-hazard planning.
• Multi-Strategy Planning: Jointly optimizing UG and VM resulted in total savings of approximately $9 million (3.97–5.52% relative to UG-only strategies) by mitigating risks that a single strategy could not address.

4.4. Scalability and Sensitivity

• Scenario Tree Size: Increasing the number of scenarios improves solution quality but increases computation time exponentially. The authors note a point of diminishing returns where marginal cost reductions do not justify the massive increase in computation time.
• Pruning Methods: Backward reduction and forward selection pruning methods yielded nearly identical optimal strategies, validating the robustness of the pruning approach.
• System Size: The model remains computationally tractable when scaled to the IEEE 69-bus system, though computation time increases with system size and line heterogeneity.

5. Significance and Claims

The paper claims that its proposed framework offers a robust and cost-effective approach to long-term grid resilience planning. By moving beyond static, single-hazard models, the framework enables utilities to:

• Make informed investment decisions that account for the complex interplay of conflicting hazards (e.g., wind vs. earthquake).
• Optimize capital allocation by leveraging the synergies between different hardening strategies (undergrounding and vegetation management).
• Enhance adaptability by allowing decision-makers to revise initial plans once, thereby reducing the risk of suboptimal investments in the face of evolving climate uncertainties.

The authors conclude that while the adaptive approach incurs a higher computational burden, the substantial financial benefits and improved resilience justify the effort, particularly for long-term infrastructure planning where flexibility is crucial. They also note that the framework is topology-agnostic regarding the hardening decision, though the load shedding calculation assumes a radial structure (with extensions possible for meshed networks). Future work is suggested to explore the interplay of risk attitudes and asset vulnerabilities in greater depth.