I was pretty excited by the project idea the other day with Todd, and have been having great conversations with local solar/battery grid decarbonization friends. My project plan has shifted from a battery price map to something more useful: an interactive CAISO SP15 ideal charge time engine, revenue calculator, and curtailment dissipation reduction system.
I'll start with mock data and just focus on the charge time optimiser and the graphs, then expand from there into real data, ML, user-impacted price and curtailment, and decarbonization calculator.
Here is a short PRD of what I am planning:
Purpose
Build a concise demo that proves how grid-scale batteries turn “duck-curve” oversupply into revenue and avoided emissions.
The tool doubles as a practice project and a realistic and useful conversation starter.
Goals
- Predict next-day duck-curve depth (MW and hour).
- Visualise optimal battery charge window, avoided curtailment (MWh / kg CO₂).
- Display net cost / revenue vs. naïve average-price charging.
- Learn about curtailment, refresh D3, and have fun!
Scope
In
- Hourly LMP (mock first, real data later) for SP15.
- Greedy optimiser (upgradeable to MILP).
- Three synced D3 graphs (forecast, curtailment, cost).
Out
- Real-time updates or multi-node expansion.
- Ancillary-services co-optimisation.
- UI for battery degradation cost (future phase).
Deliverables
- Graph A – Forecast – LMP line, battery charge line, curtailment band.
- Graph B – Curtailment Dissipation – energy lost with / without battery.
- Graph C – Net Cost – cost / revenue vs. average-price baseline.
Data Sources
- CAISO OASIS DAM & RTM LMP (XML).
- CAISO net-demand & fuel-mix for carbon factor.
- Mock CSVs to unblock front-end work and as fallback.
Architecture
Hourly CAISO data (or mock CSV) flows through a Next.js API route.
An optimiser computes the charge schedule, and the results feed three D3 graphs rendered in the client.
Phases & Timeline
- Day 5 – Mock data, greedy charge time optimiser, three D3 graphs at
/simulator/curtailment. - Day 10 – CAISO fetch API + RTK Query wiring.
- Day 15 – LightGBM year-long ML regression preprocess, swap in forecasted duck-curve depth.
- Expansions – Upgrade to MILP optimiser, implement variable grid price that increases in response to high charge speed, show carbon-factor overlays.