by/par Claro A. Noda Diaz, fondateur de/founder of Noda Technologies AB
Dans cette collaboration spéciale, l’équipe AL13 est fébrile de pouvoir collaborer l’entreprise Noda Technologies AB, dans la diffusion des résultats de sa récente recherche scientifique portant sur les systèmes d’électrification québécois et les procédés d’électrolyse dans nos alumineries.
Le présent texte a été rédigé et vous est présenté en anglais. Un résumé en français est disponible.
[FR] Résumé : Le système électrique du Québec entre dans une nouvelle phase, portée par l’électrification, les interconnexions et l’émergence rapide de grands consommateurs d’énergie tels que les centres de données liés à l’intelligence artificielle. Dans ce contexte, la contrainte ne réside plus uniquement dans la production d’électricité, mais dans l’utilisation et la coordination efficaces des ressources existantes.
Cet article soutient que la flexibilité industrielle contractable ne relève pas seulement de la volonté opérationnelle, mais constitue une propriété émergente liée à l’observabilité et au contrôle sécurisé des procédés. En améliorant la compréhension en temps réel du procédé d’électrolyse, les alumineries peuvent évoluer vers une exploitation pilotée par l’état et contribuer à l’optimisation du réseau sans compromettre la stabilité du procédé.
Note de l’éditeur – Cet article s’adresse aux producteurs d’aluminium, aux services publics, aux fournisseurs d’équipements, aux décideurs politiques et aux chercheurs. Le vocabulaire technique est rigoureux, mais chaque concept est présenté en s’appuyant sur la logique du système.
[ENG] Abstract: Québec’s electricity system is entering a new phase, driven by electrification, interconnections and the rapid emergence of large-scale energy consumers such as AI datacenters. In this context, the constraint is no longer only electricity generation, but the efficient utilization and coordination of existing resources.
This article argues that contractable industrial flexibility is not simply a matter of operational willingness, but an emergent property of observability and safety-bounded control. By improving real-time understanding of the electrolysis process, smelters can evolve toward state-aware operation — operation guided by measurable internal process state — and contribute to grid optimization without compromising process stability.
Editorial note – The article is written for aluminium producers, utilities, equipment suppliers, policymakers and researchers. The technical vocabulary is kept precise, but each concept is introduced through system logic.
Québec’s energy geography is changing in value
Québec’s power system is structurally distinctive. Large hydroelectric resources are concentrated in the north and northeast, while major demand centres and export interfaces sit farther south. Electricity moves across long transmission corridors toward Montréal, Ontario, New York and New England.
Aluminium smelters occupy a strategically interesting position inside this geography. They are not located only at the edge of the system; many are embedded along major industrial and transmission corridors in Saguenay–Lac-Saint-Jean, Côte-Nord and the St. Lawrence Valley.
That position matters more today than it did in the past. Electrification, export opportunities and new large loads — including AI datacenters — increase pressure on clean, firm electricity. The question is no longer only how much electricity can be generated. It is how effectively the system can use, allocate and coordinate the electricity it already has.
Figure 1 — Québec’s energy geography gives embedded industrial loads strategic value. Hydroelectric generation is concentrated in remote northern regions, while demand centres and export interfaces lie farther south. Aluminium smelters are positioned along major transmission corridors, creating a structural opportunity: large industrial loads are not only consumers, but potentially controllable elements within the system. Conceptual illustration by the author (Noda Technologies AB).
The constraint is not only energy supply — it is system utilization
Power systems are typically built to serve peak demand, not average demand. This creates a structural mismatch: infrastructure must be sized for rare high-load conditions, while generation and transmission assets may be underused during many other periods.
New baseload demand intensifies this challenge. Hyperscale AI datacenters, for example, are often capital-intensive, fast-moving and supported by high willingness to pay for firm electricity. Aluminium smelting is also capital-intensive and long-term, but it operates in a very different margin environment. That difference can create new allocation pressure around clean, reliable power.
Industrial demand flexibility offers another lever. In addition to expanding generation or transmission, effective system capacity can also be increased by improving utilization of existing assets — provided that flexible loads can be trusted.
Figure 2 — Power systems are built for rare peaks, leaving most capacity underutilized. Generation and transmission capacity are sized to meet infrequent peak demand, resulting in structurally low average utilization. Industrial flexibility can increase effective system capacity by shifting demand within this envelope, reducing the need for additional infrastructure. Conceptual illustration by the author (Noda Technologies AB).
Click the image to expand
Flexibility is not one thing
Figure 3 — Flexibility in smelting is governed by time scale, not just magnitude. Flexibility in aluminium smelting spans multiple time horizons, from seasonal and day-ahead planning to shorter operational adjustments. Short perturbations may be absorbed due to thermal and electrochemical inertia, while sustained deviations require tighter control of internal state variables such as heat balance and bath condition. Improved observability expands the range of safe and usable flexibility across these time scales. Conceptual illustration by the author (Noda Technologies AB).
Click the image to expand
Flexibility is often discussed as a single concept, but operationally it spans several time scales. In market-based systems, these may be described as day-ahead, intraday and balancing layers. In Québec, the institutional structure is different, but the physical time scales still matter for planning, winter peaks, intertie flows, operational reserves and curtailment decisions.
Short perturbations may be tolerated; sustained deviations require tight control of internal state.
In aluminium smelting, this distinction is critical. Short perturbations in electrical input may be tolerated because the Hall–Héroult cell exhibits thermal and electrochemical inertia, meaning that heat balance and internal electrochemical conditions evolve over finite time scales rather than responding instantaneously to changes in electrical input. Sustained deviations are different. They interact with bath chemistry, heat balance, ledge stability, current distribution and recovery requirements.
This is why scalable flexibility in smelting should not be framed as generic fast response. In practice, flexibility in smelting is likely to emerge first through seasonal, scheduled and hours-to-sub-hour orchestration, with faster response depending on the level of observability, control and recovery guarantees that can be achieved.
Contractability emerges from observability
A flexible load becomes valuable to the grid only when its response can be measured, bounded, verified and recovered from. In other words, contractable flexibility — flexibility that can be specified, verified and integrated into system-level agreements — is not only a matter of agreements, but an emergent property of observable, safety-bounded industrial processes.
In conventional smelter operation, operators have good control of production, but limited real-time visibility into the full internal state of the cell. The safe limits exist, but they are difficult to observe continuously. That makes dynamic flexibility risky, particularly when deviations are sustained.
Better sensing, state estimation and physics-informed modelling can change the decision boundary by improving observability — the ability to measure or reliably infer internal process variables in real time. In this context, the relevant question is not flexibility for its own sake, but how to define a safe operating envelope: how much power can change, how fast, for how long and with what recovery path.
Figure 4 — Flexibility becomes usable only when internal state is observable. Conventional operation relies on limited visibility of internal cell state, constraining the ability to modulate load safely. Enhanced sensing and state estimation make key variables — such as current distribution, thermal balance and bath condition — observable, enabling definition of safe operating envelopes and transforming potential flexibility into measurable and contractable system value. Conceptual illustration by the author (Noda Technologies AB).
Click the image to expand
Smelters belong in a portfolio, not in a simple analogy
A modern grid does not need every resource to behave the same way. It needs complementary resources that can be coordinated according to their strengths, costs and constraints.
Wind and solar add low-carbon energy, but remain non-dispatchable. Hydro provides firm, controllable energy, but stored water has opportunity value. Batteries are highly controllable, but limited in duration and scale relative to very large industrial loads. AI datacenters introduce a new baseload demand class with high willingness to pay.
AI datacenters are often treated as rigid electricity consumers because uninterrupted compute has very high economic value. However, some forms of flexibility may become possible if workloads can be slowed, deferred or redistributed during periods of grid stress. In that sense, flexibility depends not only on the physical infrastructure, but also on how computation, operations and service commitments are coordinated.
Figure 5 — Grid flexibility emerges from coordination, not substitution of resources. Different grid resources provide flexibility with varying degrees of controllability, duration and uncertainty. Aluminium smelters occupy a unique position: large, electrified loads with limited flexibility under conventional operation, but increasing usable flexibility as internal state becomes observable and controllable. Their value emerges in coordination with other resources rather than substitution. Conceptual illustration by the author (Noda Technologies AB).
Click the image to expand
Aluminium smelters occupy a distinctive position in that portfolio. They are already large, electrified and embedded in the system. Today, their flexibility is limited by uncertainty. With state-aware operation, the same physical scale can become more usable.
Québec’s opportunity is ecosystem leverage
Québec has the ingredients for a distinctive form of industrial flexibility: a hydro backbone, large aluminium loads, interconnections to neighbouring systems and a deep aluminium ecosystem. This suggests that smelters may be better understood not as generic demand response, but as candidates for a more structured pathway toward contractable flexibility.
That pathway requires more than market signals. It requires measurement, modelling, industrial validation, compatible power-supply and rectifier controls, and a shared definition of what “safe flexibility” means at plant and system levels.
Earlier work presented for ICSOBA 2025 explored electricity-market-driven optimization of aluminium smelting operations and showed that economic and system value may be possible when operational constraints are respected. The next step is deeper: connecting grid orchestration with observable cell state and verified safe operating envelopes.
In an era of electrification and new baseload demand, unlocking existing flexibility may become as strategically important as building new capacity. For Québec’s aluminium sector, this is not merely an operational question. It is a chance to help define what modern, cost-effective and decarbonized industrial electricity systems can become.
Technical basis
This article builds on Noda Technologies’ research direction in electricity-market-driven smelting optimization, electrolysis observability, safe operating envelopes and contractable industrial flexibility.
Suggested reference: Claro A. Noda Diaz, “Electricity Market-Driven Optimisation of Aluminium Smelting Operations,” TRAVAUX 54, Proceedings of the 43rd International ICSOBA Conference, 2025.
Author – Claro A. Noda Diaz is founder of Noda Technologies AB, a Sweden-based company developing sensing and inference infrastructure for aluminium electrolysis. Noda’s work focuses on making high-power industrial processes observable, safety-bounded and eventually contractable as flexibility resources.
