I'm attending the FERC Technical Conference today and the observations of Jim Robb, NERC CEO, articulated the very concerning challenges we face today with regard to the difficulty of accurately forecasting both load and supply in the face of a changing grid mix, i.e. supply (IBR), and load (AI load growth/Data Centers and electrification).
Transmission and Generation needs to support growing loads and overall Resource Adequacy that addresses reliability within the constraint of affordability.
Jim's comment with regard to Natural Gas fuel supply MUST be addressed if we want to ensure reliability during Winter months. This is especially true for New England and Massachusetts in particular.
There are no simple answers to these challenges.
NOTE: Many of the topics discussed in today's FERC Technical meeting will be discussed during this Power Session on November 3. So if you want to know how AI is affecting grid operations and markets, you will want to join in this Energy Central Power Session.
One interesting item I learned today is the work underway with Emerald AI and their advanced flexibility controlling energy demand with data centers. I had no idea this capability existed today. I was very impressed by the testimony of Dr. Varun Sivaram, Emerald AI CEO. He's articulate and knowledgeable - he even referred to SCED dispatch signals during his testimony - wow! I like his enthusiasm, but the situation has some very challenging real-world issues to address.
It's is very interesting, indeed, but I have many questions about how this Emerald AI technology works in practice in both normative and non-normative states. My comments and concerns with the Emerald AI proposed solution are based on my own experiences working across the energy industry since 1990 (both Gas and Electric grids) including with a ISO/RTO Grid Operator for 14 years with a Demand Response Market Program. My thoughts go in two directions:
How do you ensure that your own AI based "flexibility management program" performance can be accurately modeled in system planning and monitored for measured operating performance? How do you model an Emerald AI data center as a resource in a system planning study? What market products would be offered in wholesale markets (capacity, energy, reserves, regulation)?
The Emerald AI solution is a 100% software solution, based on the testimony of Dr. Varun Sivaram. How do you ensure that your Emerald AI software is cyber-secure and resilient to an active cyber attack coordinated by other AI software? What is your fail safe solution? [Software vulnerabilities are a well known cyber risk - all software is subject to this risk]
I think it would help to see Emerald AI's response to the new NAESB Cybersecurity Procurement Contract Addendum. This would give me the insights I'm looking for with regard to Emerald AI's cybersecurity commitments that critical infrastructure operators, i.e. grid operators need for assurance.
I'm wondering how the Emerald AI assets will participate in wholesale markets. Will you have different market offerings based on what energy load is being managed in these data centers, i.e. HVAC, Water pumps, in addition to the actual computing resources electricity usage, used in AI Data centers. Will the resources controlled by Emerald AI software commit to pay for performance programs?
In summary, the Emerald AI solution is interesting, but there are a lot of practical questions that remain. In 2009, ISO New England described the challenges faced by grid operators in this presentation, how does Emerald AI fit in this "Smart Grid picture" and address the issues and concerns cited in this presentation: https://github.com/rjb4standards/Presentations/raw/refs/heads/master/2009-0428-SmartGrid-Roadmap-Summit%5B1%5D.ppt
I'm not trying to rain on the Emerald AI parade, but I have a lot of ground truth experience working with grid operators about these very real risks and challenges. These matters will need to be addressed; transparency, resilience, risk management, reliability and affordability are key factors.