AI funding hits Alberta ops: $3M to optimize battery performance and revenue
SCALE AI, Canada's AI cluster, is putting $3 million into an Alberta project that optimizes battery operations-part of a broader $128.5 million commitment announced this week. "A.I. is no longer theoretical. It is practical, it is deployable, and it is especially delivering real impact today," said Yassine Lazraq, investment director at SCALE AI.
The project is led by Calgary-based Arcus Power and targets a core operations challenge: deciding when to charge and discharge based on market signals, grid constraints, and asset health. The objective is straightforward-grow revenue, cut wear, and improve system performance at scale.
What's being built
Arcus Power is deploying AI to optimize hybrid PV-plus-battery assets, starting with Elemental Energy's solar facility in Cypress County near Medicine Hat. "We are moving from a single battery system to many battery systems, and how do you best decide what action to take for each individual battery to ensure that the whole portfolio is operating in the best manner," said Rhonda Jewett, vice-president of growth and operations at Arcus Power.
A real constraint is already on the table: the facility couldn't always discharge back to the grid because of transmission limitations. The optimization stack is designed to factor in those constraints while preserving asset life and meeting market rules.
Why this matters for operations
- Better dispatch decisions: Align charging and discharging with price signals, congestion, and reserve needs in real time.
- Asset longevity: Limit unnecessary cycling and respect degradation curves, which protects warranties and long-term returns.
- Portfolio control: Move from single-site decisions to fleet-level optimization, including bidding and curtailment strategies.
- Fewer manual fire drills: Standardize SOPs for peak events, outages, and market volatility.
How the optimization works (at a glance)
- Inputs: real-time and day-ahead market data, grid constraints, state of charge, temperature, asset health, and warranty limits.
- Decisions: automated bids and dispatch recommendations by asset and portfolio, with guardrails for safety and compliance.
- Feedback: continuous learning from actual performance versus forecast to refine decisions.
Operations playbook: start here
- Data audit: confirm quality of SCADA, telemetry, market feeds, and maintenance logs. Fix gaps before automation.
- Integration plan: define interfaces to EMS/SCADA, market gateways, and alarms. Set clear failover modes.
- Sandbox first: run shadow dispatch for 4-8 weeks. Compare against your baseline strategy and document lift.
- KPIs locked: agree on targets and alerts before go-live. No ambiguous success criteria.
- SOPs and training: update runbooks for dispatch, outages, and incident response. Train operators on override and rollback.
- Human-in-the-loop: require operator approval for key actions until performance is proven and compliant.
- Security and compliance: review cyber controls, role-based access, audit trails, and market rule alignment.
Metrics to watch
- Gross margin per MWh captured versus market benchmarks
- Cycle count and equivalent full cycles per day/week
- Battery health: degradation rate, temperature excursions, warranty limit breaches
- Curtailment due to transmission limits and related lost opportunity
- Dispatch accuracy: forecast vs. actual revenue and reserve delivery
- Availability and forced outage rates
Risks and guardrails
- Over-cycling for short-term gains: enforce degradation and warranty constraints in the objective function.
- Market whiplash: use scenario tests and caps on exposure during volatile hours.
- Transmission constraints: model limits explicitly; trigger alerts when nearing congestion thresholds.
- Compliance: maintain audit logs of decisions, bids, overrides, and operator approvals.
- Operator trust: provide clear explanations for dispatch choices and simple override controls.
As one University of Calgary computer science professor put it, A.I. still needs oversight. "We need to understand that, A.I. is not perfect. A.I. provides information. It's never 100 per cent correct. Which is why we say that you always need a human in the loop. You always need the human expertise," said Richard Zhao.
Alberta's signal to the market
Calgary MP Corey Hogan framed it this way: "When we used to connect the country, we used to think about railways with steam locomotives, bringing us together from east to west. Well, today it's AI and digital systems that write the next great chapter." Since 2018, SCALE AI has invested more than $25 million in Alberta across seven projects, with this $3 million battery initiative extending that momentum.
Resources
Bottom line for ops: treat AI as a decision engine that needs clean data, clear KPIs, and firm guardrails. Start in shadow mode, prove the lift, and keep humans in the loop.
Your membership also unlocks:
