SIGMA computes a real-time fault probability from power-quality signals and flags incipient faults ~12 seconds before onset. It is interpretable, sub-5 ms per window, and grounded in grid physics.
Self-clearing arcing and resistive events precede permanent failures. They show up as brief voltage sags, current spikes, and harmonic bursts lasting only a few cycles — below the magnitudes that trip a conventional relay, so they go unseen.
Incipient faults sit beneath protective settings. Relays designed to trip on large, sustained faults never see them coming.
Renewable intermittency and nonlinear data-center loads make modern microgrids volatile, with fast power-quality swings that mask developing faults.
Traditional schemes act after a fault. Resilient operation needs proactive detection that recognizes the precursor signature in time.
SIGMA sits between deterministic observers and opaque deep nets: a lightweight, interpretable logistic model that merges measurable physical indicators with probabilistic learning.
Short RMS dips from arcing or partial shorts.
Arcing injects nonlinear harmonics; THD rises sharply.
Arc or insulation faults shift the V–I phase angle.
Highlights abrupt sags and recoveries.
Captures the current spikes typical of arc ignition.
A logistic core maps the feature vector to a calibrated fault probability \(p(t)\) every sample.
Coefficients are constrained \(\beta_i \ge 0\) to enforce monotonicity: worse power quality always implies higher fault risk.
Parameters are learned by \(\ell_2\)-regularized maximum likelihood over labeled windows.
A sliding window updates \(p(t)\) on every sample for continuous, online estimation.
The probability stream is classified into operating states that drive alerting.
Probability rises smoothly in anticipation of a fault rather than spiking only after it occurs.
An optional extended Kalman filter models nominal feeder behavior; residuals emphasize deviations and suppress noise.
Residual filtering cut the false-alarm rate by roughly half in testing.
A real-time pipeline from sensing and simulation through inference to alerting — benchmarked at under 5 ms per window.
Arduino + INA226 sensors sample \(V\), \(I\), power factor, and THD at 1 Hz on a 12 V DC microgrid testbed.
A Pandapower model generates labeled fault runs — line-to-line, line-to-ground, overload, and harmonic distortion.
Logistic core with optional EKF preprocessing, on rolling 30 s windows, at under 5 ms latency per window.
Python 3.11, FastAPI, TimescaleDB, and a Streamlit dashboard — output routes into existing monitoring with no new hardware.
On a 12 V DC microgrid testbed and Pandapower digital-twin simulations, leave-one-run-out cross-validation.
lead time before onset
average detection rate
ROC AUC
inference per window
| Fault type | Lead time (s) | Detection (%) | False alarms (/h) |
|---|---|---|---|
| Voltage sag | 12.4 | 95 | 0.09 |
| Overload | 11.8 | 92 | 0.10 |
| Line-to-line / line-to-ground | 13.1 | 94 | 0.08 |
Kalman preprocessing reduced the false-alarm rate by roughly 50%, and probability output rose smoothly in anticipation of each event — prognostic, not reactive.
R. Monemi and S. Monemi · Electrical and Computer Engineering
SIGMA (Smart Infrastructure Grid Monitoring AI) computes a real-time fault probability from power-quality features using a logistic core and sliding windows, with an optional Kalman filter for noise rejection. The approach bridges classical reliability analysis and modern AI-based predictive maintenance.
Ensemble models and adaptive online learning — recursive Bayesian updating — for robustness across operating conditions.
Distributed implementation for multi-node, grid-scale microgrids.
Extension to AI data centers and industrial facilities, where nonlinear loads and the cost of a missed fault are highest.
Engineer at SKM Systems Analysis · M.S. candidate
Built by one man. He designed the physics-informed model, the simulation and sensing pipeline, and the real-time serving layer that connects them, and is lead author of the paper behind it, “SIGMA: A Physics-Informed AI Framework for Predictive Fault Probability Modeling in Microgrids” (R. Monemi and S. Monemi).
At SKM Systems Analysis he validates PTW and PTX power studies. He holds a B.S. in Electrical and Computer Engineering and previously designed electric-distribution infrastructure for SDG&E and for SpaceX propellant generation.
If you work on microgrids, power-system protection, or data-center power and want to talk about SIGMA, reach out directly.
founder@monegrid.com