Alfred University and Ranial Systems Get $65,000 to Build an Edge-AI Controller for C&I Microgrids

Industrial & C&I Microgrids Microgrids Research
Marin Holt
August 19, 2026
0
14
8 minutes read

NENY grant: $65,000
Round total (4 partnerships): $315,000
NYSERDA lab grant (2024): $466,853
Donated GE Vernova software: $2.8 million

Alfred University and Ranial Systems Inc. won a $65,000 grant from New Energy New York in July 2026. The money funds a combined battery and microgrid controller. Four New York partnerships split $315,000 in the program’s latest Technical Assistance round. Work started the week of June 29 and ends in November. The target is a System-on-Chip controller. It merges energy management and battery management on one chip, for grid-scale storage, commercial and industrial microgrids, and OEM integration.

What the Money Buys

$65,000 is small for a combined hardware and software program. Five months of work covers AI model training, sensor integration, and hardware-in-the-loop testing. The grant does not pay for the test bench. That came earlier. A $466,853 NYSERDA grant established the Advanced Power Grid Lab at Alfred in 2024. GE Vernova donated $2.8 million in software licenses. Ranial brings CognitEMS, a patented edge-AI platform for battery and energy management. The public money buys integration work on top of hardware that already exists.

Chandan Kumar, assistant professor of computer science, leads the project. Xingwu Wang, professor of electrical engineering, is co-principal investigator. One graduate student in electrical engineering works on the team.

Why It Matters

Microgrid controllers are where the margin sits. Batteries, inverters and switchgear are increasingly commodity hardware. Dispatch logic separates a microgrid that saves money from one that does not. Most C&I microgrid controllers run rule-based schedules. Commissioning engineers tune them once and rarely revisit them. Kumar’s group is testing unsupervised learning, which adapts without labeled training data. That matters for C&I sites. Every load profile differs, and nobody holds years of labeled fault data to hand a model.

New York is also buying domestic supply. NENY pairs Binghamton University with the New York Battery and Energy Storage Technology Consortium. Its job is to build a battery supply chain inside the state. A $65,000 award is a cheap option on a controller that would otherwise come from abroad.

Path to Commercialization

The project ends in November with a tested controller design, not a product. Ranial owns CognitEMS and would carry it to market. The stated targets are grid-scale battery storage, C&I microgrids, and OEM integration. Each needs different certification. A controller that dispatches a grid-scale battery answers to interconnection and protection rules. A behind-the-meter microgrid controller does not face those rules. Neither Alfred nor Ranial has named a pilot host, a utility partner, or a ship date.

Hardware-in-the-loop testing runs at the GE Vernova Advanced Power Grid Lab and the Edibon Power Systems Simulation Lab. That is a real gate. It is also a simulated one. Controllers that pass in the loop still meet surprises on live feeders.

Critical Perspective

A $65,000 grant and a November deadline set the honest scale of this work. Five months buys a tested design, not a certified product, and the project’s own timeline says so. The harder question is what happens after November. NENY’s Technical Assistance Program hands out small awards to move a technology one step forward. Nothing in the award commits Ranial to a pilot, and no utility or site host is named.

Unsupervised learning is the interesting claim and the hardest one to verify. A controller that adapts without labeled data also adapts without a ground truth to check it against. Grid operators and insurers ask how a dispatch decision was reached. A model that shifts its own behavior over a battery’s life makes that answer harder, not easier. The team has not said how it will log or bound those decisions.

Hardware-in-the-loop testing at the GE Vernova and Edibon labs proves the controller survives a simulation the team wrote. Real C&I sites bring unbalanced phases, nuisance trips and meter data that arrives late or not at all. Alfred’s own framing, that the work bridges academic research and industrial application, is the claim to hold the team to in November.