MULTI-AGENT DECISION SUPPORT SYSTEM FOR RESILIENT DISTRIBUTION GRID

INTRODUCTION

Wildfires and extreme wind events increasingly threaten electric distribution systems and can elevate ignition risk from energized conductors, motivating utilities to implement Public Safety Power Shutoffs (PSPS) (Panteli & Mancarella, 2015). Although PSPS can reduce ignition probability, wide area de-energization may also interrupt communications and essential services needed for evacuation and disaster response (Panteli & Mancarella, 2015; Vugrin et al., 2010). This proposed research develops an explainable, heuristic multi agent decision framework that coordinates controlled feeder sectionalizing and de-energization in high risk zones while maintaining electrical service to designated disaster relief shelters through hardened, renewable powered microgrids (e.g., solar plus storage with optional firm backup) (IEEE Standard Association, 2018). This research integrates multi source information, such as grid telemetry, weather/fire intelligence, into auditable decision rules that support reliable, safety constrained operations under rapidly changing disaster conditions (Vugrin et al., 2010).

IMPLICATION OF STUDY

This study uses a simulation-based dataset that combines (i) distribution grid operational representations and (ii) hazard intelligence. Grid inputs are modeled using feeder and transformer level aggregates consistent with SCADA/AMI derived summaries and line sensor concepts, including switch/breaker status, feeder power flows, voltage indicators, and fault alarms (Zimmerman et al., 2011). Hazard inputs include publicly available weather forecasts (wind speed/gusts), red flag warnings, and fire alert/perimeter estimates from emergency reporting sources (Finney, 2004). These inputs are fused to create circuit level hazard states and ignition risk proxy scores that drive multi agent actions (Vugrin et al., 2010).

Processing and analysis proceed in three stages. First, the research generates scenarios spanning normal operations and wildfire conditions, including wind driven fault likelihood, line damage and section outages, and communication degradation (Vugrin et al., 2010). Second, a rule based multi agent controller executes a safety-first hierarchy: feeder agents isolate the faulted sections, inhibit reclosing during high-risk periods, sectionalize and de-energize segments exceeding hazard thresholds, and attempt conservative rerouting through tie switches subject to thermal, voltage, and radiality constraints (Zimmerman et al., 2011). Shelter microgrid agents island during upstream risk or outages and dispatch local resources to serve tiered critical loads using reserve aware rules, such as minimum state of charge margins and priority-based shedding (IEEE Standard Association, 2018). Third, outcomes are compared with baselines, such as threshold only PSPS and non-coordinated restoration heuristics. The analysis shows reduced risk weighted energized exposure during high hazard intervals, improved continuity of service for shelters/critical loads, and safer restoration with a controlled number of switching operations (Panteli & Mancarella, 2015; Vugrin et al., 2010).

CONCLUSIONS AND FUTURE STUDY

The framework can support utility and municipal resilience planning by linking feeder level switching policies with the siting and operation of shelter microgrids (Vugrin et al., 2010; Panteli & Mancarella, 2015). Because the approach depends primarily on feeder/transformer aggregates and transparent rules, it reduces data and privacy burdens relative to household level monitoring and improves auditability for emergency decision making, including under uncertainty and degraded communications (Vugrin et al., 2010).

This proposed research argues that heuristic multi agent coordination can provide a practical middle ground between manual emergency procedures and complex learning based controllers. This approach can enforce safety constraints by design, incorporate weather and fire intelligence, and prioritize disaster relief zones while limiting unnecessary outages (Panteli & Mancarella, 2015; Finney, 2004). Future work will refine ignition risk proxies, incorporate additional interdependent infrastructure, such as telecom and water system, constraints, and test scalability across multiple feeders and jurisdictions (Finney, 2004; Zimmerman et al., 2011; IEEE Standard Association, 2018).