Condition-based vegetation management for electric utilities. Meridian scores every line span by wildfire/outage risk, forecasts when each span will breach its required clearance, and tells a vegetation-management planner where the next trimming dollar buys down the most risk — replacing fixed "trim everything every N years" cycles with a data-driven, budget-aware plan.
Live demo: https://sattvitripathy.github.io/meridian/
All data is synthetic, seeded, and fictional — no real utility data. The modelled utility (Sierra Crest Power & Electric) spans the Sacramento Valley up into the Sierra Nevada foothills, across the CPUC High Fire-Threat District Tier 1 → 3 gradient.
Trees and brush growing into power lines are the leading cause of distribution outages and a major wildfire ignition source. Utilities are obligated (CPUC GO 95, NERC FAC-003) to keep vegetation clear of conductors. Most still trim on fixed cycles — expensive and blind to which spans are actually dangerous now. Meridian moves the planner from cyclical to condition-based trimming.
Every span scores 0–100 by blending five normalized factors (weights are tunable in Scenario compare):
| Factor | Proxy | Why |
|---|---|---|
| Encroachment | current tree-to-conductor clearance vs. required envelope | how close to failure now |
| Growth | species growth rate × time since last trim | how fast the gap closes |
| Fire threat | CPUC HFTD tier (1 / 2 / 3) | turns an outage into a catastrophe |
| Criticality | customers downstream + voltage class | a fault here hurts more |
| Access | terrain slope & crew reachability | response time & cost |
Time-to-violation = clearance headroom ÷ growth rate — converts a score into a deadline ("breaches in ~8 months"), which is how planners actually think.
- Risk dashboard — hybrid map (stylized schematic or real California basemap), every span colored by VRI, linked to a ranked work-priority list and a deep span-detail drawer (clearance gauge, factor breakdown, recommended action).
- Budget optimizer — greedy risk-per-dollar selection: given a budget, pick the set of spans that buys down the most risk / protects the most customers / covers the most Tier-3. The spend-efficiency frontier plots the same budget spent on a fixed longest-since-trim rotation alongside the optimized plan — the gap between the two markers is the case for condition-based trimming.
- Portfolio analytics — 24-month projected-violations curve (no-action vs. funded plan), risk by fire tier, highest-risk circuits, risk by species, clearance-margin distribution.
- Crew dispatch (lite) — turn high-risk spans into work orders, auto-assign to crews, track them across a Backlog → Scheduled → In-progress → Completed board.
- Compliance register — auditable list of clearance violations & imminent breaches with GO 95 / FAC-003 references and remediation deadlines. Export CSV or print.
- Scenario compare — save tuned weight/budget models, apply any saved scenario back onto the live model, and compare two head-to-head.
Every view is deep-linkable (#/optimizer, #/compliance, …) and so is any span
(#/span/SPN-1451) — browser back/forward work as expected.
Plain, dependency-light static PWA — no backend, no build step.
index.html— app shellcss/styles.css— lavender themejs/data.js— seeded synthetic data engine (deterministic; ~570 spans, 21 circuits, 6 substations) plus the VRI scoring functionsjs/app.js— all views, scoring, map rendering, persistencesw.js+manifest.webmanifest— installable, offline-capable PWAgenerate-icons.js— zero-dependency Node PNG icon generatorserver.js— tiny static server for local previewtest/— unit tests for the scoring core (node --test)docs/— Node scripts that generate the Word product guide & interview brief
State (work orders, tuned weights, scenarios) persists in localStorage. The geographic
basemap uses Leaflet + OpenStreetMap/CARTO tiles (the schematic view works fully offline).
node server.js # → http://localhost:5174
# or: python -m http.server 8125To regenerate the app icons: node generate-icons.js.
Run the scoring-model unit tests (no dependencies, Node ≥ 20):
node --testDeploy note: the service worker serves navigations network-first and assets
stale-while-revalidate, so new deploys are picked up on the next visit. Bump the
CACHE version in sw.js when you change the precached asset list.
Built as an exploratory prototype. Basemap © OpenStreetMap contributors © CARTO.