GridPing — Backend

Julia services that simulate smart-meter power flow on local hardware, write readings to PostgreSQL (AWS RDS), and expose the same data to the dashboard via a read-only Lambda API.

Most domestic smart meters have very limited computing power and thus lack the ability to perform detailed power grid analysis. The project team are working together, under the guidance of Dr. Eduardo Cotilla-Sanchez, to develop a GPU-accelerated smart meter alongside grid algorithms to allow for real-time analysis of large power grid cases. This repo is the cloud + simulation backend in that pipeline.

Related repos & site

Piece	Link
Smart meter dashboard (React)	rojaslesly/smartMeter
Landing & pipeline showcase	nikoloda/GridPing-Showcase · gridping.vercel.app
Backend (this repo)	Simulation, RDS writes, Lambda read API

What it does

• Meter poller — loads MATPOWER cases, runs AC power flow (PowerModels + Ipopt), computes local/global PQ and island count, inserts into records / globalRecords
• Case tooling — generate perturbed hourly cases; audit generated and event cases for convergence
• DB helpers — peek or clear SQLite (local) or Postgres (AWS)
• Lambda — SQL queries for the dashboard (latest_bus, bus_24h, latest_global, last_outage); does not run power flow

Simulation runs on your machine (or edge device). RDS stores rows. EC2 is only an SSH jump host for dev writes. Lambda reads RDS for the React app.

Tech stack

Layer	Technology
Simulation	Julia, PowerModels, Ipopt, JuMP
AWS DB	RDS (PostgreSQL), LibPQ
Local DB	SQLite (optional)
API	Node.js Lambda (pg)
Cases	MATPOWER .m under cases/transmission/

Getting started

1. Julia environment

Install Julia and project deps (PowerModels, Ipopt, LibPQ, etc.) as used in scripts/ and src/.

2. Environment file

Contact nikoloda@oregonstate.edu for credentials. Place .env at the repo root (gitignored):

PGHOST=localhost
PGPORT=5433
PGUSER=______________
PGDATABASE=______________
PGPASSWORD=______________

Or a single PG_CONN=... string. connect_pg() in src/DB_AWS_PostgreSQL.jl loads unset vars from .env.

3. AWS tunnel (writes from your laptop)

RDS is not public. Open a tunnel through the bastion EC2 (region us-west-2). Get bastion public IP and RDS endpoint from the AWS console (EC2 → Instances; RDS → Databases). If SSH times out, start the EC2 instance and allow SSH (22) from your IP on its security group.

Leave this terminal open:

ssh -i /path/to/bastion-key.pem -L 5433:<rds-endpoint>:5432 ec2-user@<bastion-ip> -N

Verify:

julia db_manual/peek_postgres.jl

Lambda in production uses DB_HOST, DB_USER, DB_PASSWORD, DB_NAME and talks to RDS directly—no tunnel.

4. Run the meter poller

julia scripts/meter_reading_simulator.jl

Defaults: bus 6, 24 hourly steps, 45 s between readings, Postgres on (USE_AWS_POSTGRES). Toggle SQLite or timing flags at the top of the script.

Optional long-running variant: julia scripts/meter_reading_simulator_forever.jl.

Distribution poller (8500-Node, bus 293471): julia scripts/distribution_meter_reading_simulator.jl.

Distribution debug (single Master-unbal.dss solve, no loop): julia scripts/distribution_meter_reading_simulator_once.jl.

5. Other scripts

julia scripts/check_generated_case_solvability.jl
julia scripts/matpower_case_generator.jl    # generate transmission cases
julia db_manual/clear_postgres_records.jl     # destructive — truncates AWS tables

notebooks/ — early experiments and prototypes (e.g. Postgres init, population); kept as artifacts, not required for the poller flow.

Lambda API (frontend contract)

Implemented in lambda/index.mjs. Matches table layout from DB_AWS_PostgreSQL.jl. Query param query:

Query	Parameters	Returns
latest_bus	bus_id, target_time	Closest meter record for bus
bus_24h	bus_id, target_time	Bus records in prior 24 h
latest_global	target_time	Closest global grid record
last_outage	bus_id, target_time	Latest outage row for bus

target_time: YYYY-MM-DD HH:mm:ss. Response shape: { query, target_time, count, rows[] }.

bus_id values: transmission poller uses 6; distribution meter SX_293471A uses 293471 (distribution_meter_reading_simulator.jl).

Frontend PQ mapping: see frontend/gridData.js (drop-in for smartMeter src/utils/gridData.js).

Both pollers write the same field: power_quality = Δv (p.u.) = meter voltage − ideal voltage at that meter. Non-converged: -100, status = 0.

Ideal case	bus_id
case2383wp.m	6
Master-unbal.dss at min diurnal loadmult	293471

0–100% dial is computed in frontend/gridData.js

Project structure

cases/transmission/ # MATPOWER .m — base, generated_cases/, event_cases/
db_manual/          # peek / clear SQLite or Postgres
lambda/             # Read API for dashboard
notebooks/          # Experimentation / idea testing (historical artifacts)
scripts/            # Poller, case generator, solvability audit
src/                # DB_AWS_PostgreSQL.jl, DB_SQLite.jl

OpenDSS (.dss) cases: use dss-extensions/electricdss-tst for IEEE/EPRI examples and other test circuits (filtered OpenDSS sample scripts).

Contact

Daniel Nikolov — nikoloda@oregonstate.edu