Wildfire Risk De-energization (PSPS) and Public Safety Automation Workflow

Manual PSPS decisions involve cross-departmental meetings to reconcile disparate weather, fire risk, and asset health data. A custom workflow ingests these hyper-local signals (wind, humidity, fuel moisture indices) in real-time, applies utility-specific risk models, and generates ranked circuit de-energization recommendations. This compresses a multi-day, consensus-driven process into a continuous, data-driven operation, enabling faster response to rapidly changing fire conditions.

Blunt, circuit-level shutoffs maximize customer outages and invite regulatory scrutiny. An automated workflow enables segment-level granularity, de-energizing only the highest-risk spans while keeping adjacent customers online where safe. By integrating with OMS/ADMS and a digital twin, it simulates the customer impact (SAIDI/SAIFI) of each proposed action, allowing operators to balance wildfire mitigation with service reliability, directly reducing potential fines and customer compensation costs.

Manual notification processes are error-prone and slow, damaging public trust. The workflow orchestrates personalized, multi-channel (SMS, email, IVR) notifications based on the precise de-energization plan, pulling data from CIS and outage management systems. It simultaneously generates and routes regulatory reports (CPUC, PUC) and alerts to first responders, ensuring a consistent, auditable communication trail that mitigates reputational and legal risk.

The workflow doesn't just recommend—it executes. With approved sequences, it sends secure switching orders to remote-controlled devices via SCADA/DMS integration, performing controlled de-energization. Crucially, it then plans and orchestrates the restoration sequence, prioritizing critical infrastructure and validating line safety before re-energization. All actions are gated by operator approval, with full observability into device status and a rollback path for any command.

Post-event regulatory investigations demand exhaustive evidence. The workflow automatically captures the complete decision chain: ingested risk data, model outputs, recommended actions, operator approvals, executed switching orders, and communication logs. This immutable, linked audit trail is packaged for regulators, turning a reactive, document-scrambling exercise into a proactive compliance asset that demonstrates due diligence and operational rigor.

The ultimate ROI is in catastrophic risk avoidance. By enabling faster, more precise shutoffs, the workflow directly reduces the probability of a utility-caused ignition. This can be modeled to quantify reductions in expected wildfire liability—often measured in billions. Furthermore, a documented, automated risk-mitigation program can positively influence utility insurance premiums and reinsurance terms, creating direct financial upside beyond operational savings.

This agent ingests real-time feeds from weather APIs (wind, humidity), fire risk indices (e.g., NFDRS), and asset health telemetry (e.g., transformer temperature, line sag). It runs predictive models to score circuit segments, generating a ranked recommendation list for de-energization. The architecture ties into SCADA and OMS for live grid topology, ensuring recommendations are operationally feasible and avoid creating new reliability risks.

Upon final approval, this component triggers multi-channel notifications. It personalizes messages based on customer segment (medical baseline, C&I) and preferred channel (SMS, email, IVR), pulling data from the Customer Information System (CIS). It also automates mandatory filings with public safety commissions (e.g., CPUC), generating audit-ready reports of notifications sent, a critical control for regulatory compliance and liability protection.

The core execution layer. It takes the approved switching sequence, validates it against the SCADA/OMS digital twin for safety, and then issues commands to remote-controlled switches and reclosers via DMS/ADMS APIs. The workflow includes pre- and post-switching checks (e.g., verifying load transfer, confirming de-energization via smart meter 'last gasp' signals) and logs every action with millisecond timestamps for the definitive event record.

Manages the return-to-service process. After the all-clear from weather/fire agencies, this agent coordinates the 'patrol, isolate, restore' sequence. It generates optimized patrol routes for field crews using geospatial data, sequences the re-energization of circuits to avoid inrush currents, and integrates with Mobile Workforce Management (MWM) systems to dispatch and track crew progress, ensuring a safe, phased restoration.

A critical governance component. It monitors the entire workflow for anomalies—like a failed switch operation, a critical facility within a shutdown zone, or a conflicting operational constraint—and routes them to the appropriate human role (System Operator, Public Safety Manager) in a prioritized queue via ServiceNow or a custom dashboard. This ensures the system escalates only what's necessary, preserving automation speed while maintaining ultimate human oversight.

This operational intelligence component aggregates data from the entire event: risk predictions vs. actual fire proximity, customer notification effectiveness, switching performance, and outage duration. It processes this data to generate ROI reports (e.g., wildfire risk mitigated vs. customer minutes interrupted) and automatically retrains the risk assessment models, creating a closed-loop system that improves decision accuracy with each PSPS event.

The workflow ingests and fuses real-time data streams—including hyper-local weather (wind speed/direction, humidity, temperature), fire risk indices (NFDRS, ERC), real-time fault indicators, and asset health telemetry—to calculate a dynamic, circuit-segment-level risk score. This triggers a de-energization recommendation only when pre-defined safety thresholds are breached, replacing manual, spreadsheet-based analysis and reducing decision latency from hours to minutes.

Upon a trigger, the system automatically executes a multi-channel notification cascade via SMS, email, and IVR, personalized by customer segment (critical care, medical baseline, general). It generates and logs required regulatory reports (CPUC, PUC) in the background, ensuring audit-ready compliance and reducing the manual coordination burden on public affairs and customer service teams during high-pressure events.

The architecture integrates with the Distribution Management System (DMS) and SCADA to generate and validate a safe switching sequence. With operator approval, it executes remote-controlled switching orders to isolate targeted circuit segments, managing the sequence to avoid creating new instability. This controlled automation minimizes human error in high-stress scenarios and ensures the shutoff is precisely scoped to the assessed risk area.

Post-event, the system monitors ground truth (weather stations, patrol reports) and asset inspections to determine when re-energization is safe. It then orchestrates a phased restoration sequence, prioritizing critical infrastructure and avoiding inrush current issues, by sending validated switching orders back to the DMS. This moves restoration from a manual, crew-led process to a system-coordinated one, accelerating the return of service.

Critical decision points—initial trigger confirmation, switching execution, and final restoration—are designed as mandatory human approval gates within the workflow (e.g., in a platform like LangGraph). The system provides the supervising engineer with all supporting risk data, simulated outcomes, and a clear recommendation, preserving operational control and accountability while leveraging automation for speed and data synthesis.

Every PSPS event generates a rich dataset for improvement. The workflow automatically packages switching logs, customer contact outcomes, weather correlations, and patrol findings. This data feeds back into the risk model for retraining and into operational dashboards for after-action reviews, creating a closed-loop system that incrementally improves prediction accuracy and operational effectiveness with each event.