Assessing AI Safety of Qcells’ Data Center Energy Management System Control Component

Qcells provides clean energy solutions spanning solar technologies, energy storage and software-enabled energy services. Building on its experience supporting complex energy environments, Qcells has expanded its focus to mission-critical facilities, including data centers, where operators must balance safety, reliability, efficiency and increasingly dynamic power demands.

In data center environments, Qcells’ energy management system (EMS) helps automate energy usage oversight and energy bid optimization, drawing on operational data to generate insights that support more informed energy decisions. The EMS is an artificial intelligence (AI)-native, cloud-based optimization platform designed to produce time-bound, explainable energy optimization plans that are reviewed and approved by operators before any execution on physical infrastructure.

Customer challenge

Modern data centers operate at increasing density and complexity, requiring rapid, coordinated decisions across power distribution, cooling and critical electrical infrastructure. In many facilities, legacy energy management tools still rely on manual workflows and one-off integrations, making it difficult to coordinate real-time signals across assets and respond consistently as conditions change. This can contribute to overprovisioning, overcooling and elevated operational risk — especially as operators explore more automated, AI enabled approaches to optimization.

As energy and operational systems become more connected and adaptive, stakeholders increasingly expect independent, third-party substantiation that AI-driven control logic behaves as intended. This includes under abnormal or degraded conditions. In this context, intended behavior is defined by the system’s documented architecture and design walkthroughs, along with the customer’s risk assessment and applicable requirements that govern acceptable risk. Stakeholders also look for practical evidence of human oversight — for example, approval gates before equipment-affecting actions, clear operator override mechanisms, and audit-ready records that show what the system recommended, why it recommended it and what occurred after execution.

To support repeatable, scalable operations, Qcells’ EMS architecture emphasizes standardized digital infrastructure, including:

The Qcells EMS is designed with human-in-the-loop governance as a foundational principle. It continuously reasons over inputs such as weather forecasts, electricity price signals, facility load forecasts and telemetry to generate optimization plans. Recommendations are routed through an operator approval workflow and logged with rationale and outcomes to support auditability.

Qcells sought a third-party evaluation of its EMS AI-driven optimization capability to help demonstrate trustworthiness, transparency and responsible human control in high-density data center environments and to support broader market readiness as AI governance expectations continue to evolve.

UL Solutions approach

Qcells worked with UL Solutions to complete a third-party evaluation of the identified AI-based control component under UL 3115, the Outline of Investigation for Safety of AI-Based Products. UL 3115 provides a structured, evidence-based framework for evaluating AI-powered products across technical, ethical and governance considerations.

Evaluation activities

UL Solutions’ evaluation activities included:

Throughout the evaluation, UL Solutions assessed evidence and demonstrations related to governance and oversight expectations such as operator approval gates, explainable rationale, audit trails and rollback/restore capability.

Results

Qcells successfully completed the UL 3115 evaluation for the identified AI-based control component within the Qcells EMS. This engagement represents the first agentic AI EMS to be assessed for compliance with the UL 3115 framework.

With the evaluation complete, the component was eligible to use the UL Mark in accordance with applicable usage guidelines, signaling that defined evaluation criteria have been assessed through a third-party process.

Customer benefits

The evaluation provided an evidence-based safety case aligned to key UL 3115 expectations for the defined scope, supporting audit-ready artifacts and governance substantiation that can be used in customer and stakeholder discussions. It also supported readiness conversations related to emerging and evolving AI governance expectations, such as the EU AI Act.

The evaluation and certification provide third-party evidence related to governance, transparency and oversight expectations for AI-based operational systems in data center environments, supporting stakeholder confidence as the industry explores more autonomous optimization approaches.