Season 05 · Episode 02 · Energy · 44 min
AI in Energy: The Forecast That Keeps the Lights On.
Before the algorithmic grid had a name, Dr. Lin Hua was building it. Her first conversation on the show covers renewables, chaos and the strange discipline of predicting a machine the size of a continent.
October 11, 2023·With Dr. Lin Hua
▶44:31§01Chapters
Introduction: Weather, Demand and Everything Between
Why Renewables Broke the Old Forecast
A Continent as a Time Series
The Morning Ritual of Being Checked
Batteries, Ducks and Negative Prices
What Comes After Prediction
§02Show Notes
This is where the show's energy coverage began. In 2023, Dr. Lin Hua was newly arrived at GridLogic from the public sector, and the proposition she describes here still sounded radical: that the grid's future depended less on new generation than on better prediction. Wind and solar had turned supply itself into weather, and the deterministic forecasting tools of the fossil era were, in her phrase, 'confidently wrong at exactly the wrong moments'.
The technical heart of the episode is her explanation of probabilistic forecasting for a lay audience, still the best we have aired. A useful forecast, she argues, is not a number but a shape: a distribution honest about its own ignorance. Her description of the morning ritual, in which yesterday's predictions are scored publicly in front of the control-room staff, explains more about building institutional trust in AI than any governance framework published since.
Listeners who know her season seven appearance will hear the foreshadowing everywhere: the insistence that accuracy is necessary but insufficient, the operators treated as the model's examiners rather than its audience. Two years later, one doubted week would prove every warning in this episode correct.
“Operators don't need my certainty. They need my honest uncertainty, priced in megawatts.”
§03Transcript Extract
NISHANT:
You left a senior government post to join a forecasting startup, which your colleagues apparently considered a demotion. Why did you jump?
DR. HUA:
Because the job changed underneath the title. When generation was coal and gas, supply was a decision and only demand was a guess. Renewables made supply a guess too. Suddenly the grid was two colliding weather systems, and the institution best placed to manage that was not the one with the biggest turbines but the one with the best forecast. I went where the leverage moved.
NISHANT:
You say a forecast should be a shape, not a number. Unpack that for someone who has never thought about their electricity.
DR. HUA:
If I tell you tomorrow's demand will be 41 gigawatts, I have told you almost nothing, because I will be wrong and you don't know by how much. If I tell you there is a five percent chance it exceeds 44, you can decide, today, whether that risk is worth the cost of standing up another plant. Operators don't need my certainty. They need my honest uncertainty, priced in megawatts.
NISHANT:
And when the model earns that trust, what's the next frontier?
DR. HUA:
Prediction is passive. The frontier is orchestration: batteries charging against the forecast, industrial loads shifting to soak up the solar afternoon, prices doing the choreography. The grid stops being a machine we react to and becomes one we conduct. Ask me back in a couple of seasons and I will tell you how that went.
§ Keep listening
More on Energy.
S07 · E06 · 46 min
Powering the Algorithmic Grid
AI is balancing the world's most complex machine — the electricity grid — in real time. What happens when it fails?
S07 · E05 · 55 min
The Death of the Script?
Generative tools are rewriting Hollywood's economics. We talk to the people on both sides of the picket line.
S07 · E04 · 49 min
Pricing the Unpriceable
Inside the algorithms now valuing trillions of dollars of real estate — and the bias problem nobody is talking about.