Part 2: Forecasting the Grid: Where Strategy Meets Orchestration
By Velvet Voelz and Oleg Popovsky
This series, From Grid Disjointedness to Grid Orchestration, explores the renewable energy industry’s evolution—from unresponsive assets and fragmented infrastructure to AI-augmented strategies for resilient, responsive, and affordable power that produces stable and attractive returns.
In Part 1 of this series, Oleg Popovsky and I framed how distributed energy resources (DERs) could evolve from static installs into orchestrated, intelligent platforms.
In Part 2: Forecasting the Grid: Where Strategy Meets Orchestration, we dive deeper by turning to the engine that makes orchestration possible: forecasting. Forecasts no longer just inform—they actively shape markets, dispatch, and investment.
Energy market forecasting has always been challenging, but today, it’s increasingly complex. In the past, forecasting required balancing uncertain weather data, shifting fuel prices and imperfect load models. Today’s energy landscape is more dynamic and interconnected. Forecasts shape everything from grid stability to energy arbitrage, from AI-driven dispatch to market design itself.
And as climate change drives extreme weather events and new actors—from storage to distributed resources to hyperscale data centers—join the grid, forecasting is no longer a background exercise. It’s the heartbeat of a reliable, profitable, and affordable power system.
So, what does it take to forecast in an era where tail events are growing bigger, markets are more interconnected, and demand itself is shifting in real time?
The Expanding Use of Forecasts
Forecasting operates across multiple horizons. Long-term models guide transmission and investment planning. Medium-term outlooks inform budgets, risk management, and seasonal strategies. Short-term forecasts drive trading, dispatch, and arbitrage.
And the applications are expanding. DERs and demand response programs now move markets. GenAI load shifting between markets necessitates multi-ISO, real-time iterative models — basically 3D chess instead of 2D grid planning and scheduling. Price signals don’t just shape generation anymore—they reshape demand location, creating feedback loops that forecasters must anticipate or risk herd-driven oscillations across markets.
Goodbye Single Point Predictions, Hello Scenarios
Forecasting must embrace uncertainty—stochastic models that map probabilities, informing risk frameworks, and enabling revenue strategies that flex with reality instead of being pinned to one prediction.
Market Signals: From Peak Demand to Net Demand
Legacy systems were designed around peak demand. But renewable and DER-heavy grids tell a different story: what matters now is net demand and flexibility pricing. Forecasting must evolve to capture not just supply and demand, but co-optimize across energy, services, and constraints. The new signals are dynamic, and the models must be too. For decades, we thought about price signals shaping dispatch of generation. Now, hyperscale data centers are acting like super-loads that can rapidly shift across geographies, and in some cases, turning demand itself into a dispatchable resource.
Evolving Inputs, Smarter Models
The foundation of forecasting is data—but the data we have relied on is changing.
Historical weather patterns no longer align with climate-change related weather volatility. Load and generation behaviors of the past no longer predict how resources and consumers will behave tomorrow. Transmission studies age quickly in fast-changing markets. Nodal forecasts are increasingly important alongside the integration of physical power flow and AI/ML models. Reliability drivers are impacting capacity forecasts in ways we haven't seen before. Data centers introduce demand that can be unpredictable and highly location-sensitive. And renewables paired with storage are reshaping price dynamics, softening the spikes that once signaled stress.
These shifts don’t make forecasting impossible, but they do require models that are more adaptive, flexible, and capable of handling a wider range of outcomes.
AI’s Role in Forecasting 2.0
Forecasting is no longer just about physics—it’s about intelligence. The intelligence isn’t just in the forecast—it’s in how it’s trained, queried, weighted, and applied.
AI is enabling models that spot important anomalies instead of just trends, incorporate policy and behavioral drivers, and adapt continuously to real-time data. Better inputs as well as prompts lead to better forecasts.
Case Study: Forecasting is the new PPA
The investment community isn’t just looking for clean megawatts anymore — they’re looking for predictable, insurable, and orchestrated returns. Forecasting precision has become the dividing line between assets that attract capital and those that remain stranded.
Recent examples prove the point:
Brattle’s VPP Study in California showed that over 100,000 home batteries delivered 535 MW of consistent, additive output during a single peak event. That’s planning-grade reliability from distributed assets — a class once dismissed as “too small to matter.” With reliable forecasting informing charging and dispatch, VPPs become bankable grid resources.
For investors, this changes the calculus. The question is no longer just “what’s the levelized cost of energy?” but “how dynamic is the asset’s value under stress, and can forecasting prove it?” It also changes the magnitude of the additional grid investment needed when optimizing for the for underutilized assets already on the grid.
This is why forecasting isn’t just a technical challenge — it’s a strategic one. Better models don’t just protect the grid, they unlock new contract structures, arbitrage strategies, and capital flows. And in a market where tax incentives are tapering, that precision may define who scales and who stalls.
Why This Matters—More Than Ever
As tax incentives wane, revenue strategies must lean on precision—not subsidies. The shift from selling electrons to selling flexibility is well underway. Arbitrage opportunities are expanding from financial markets into operations themselves.
But here’s the critical point: this new landscape is too complex for legacy forecasting methods.
AI’s role is unique because it can integrate multi-modal inputs—weather, price curves, grid congestion, behavioral patterns, even policy changes—into adaptive, agentic forecasts that continuously learn and improve.
AI doesn’t just predict outcomes; it anticipates responses across thousands of actors on the grid. That means identifying anomalies before they cascade, and opportunities before they disappear.
For investors, AI-enabled probabilistic forecasting translates into insurable revenue streams. For market operators, it reduces volatility and systemic costs. For asset owners, it unlocks flexibility premiums and next-gen contract structures.
That’s why forecasting can’t just be about predicting tomorrow’s prices—it must serve as the AI-driven blueprint for new business models and market structures that ensure reliability, resilience, and affordability. In the energy transition, forecasting isn’t just a tool—it’s the strategy.
This is Part 2 in our ongoing series exploring the evolution of renewable energy infrastructure in the age of intelligent systems.
If you missed Part 1, you can find it here: What If DERs Operated Like Data Centers?
Next up: Part 3, where we’ll explore how the grid has evolved from bi-directional flows to complex, multi-actor dynamics—like 3D chess in motion.
Things to think about:
Where are today’s forecasting models falling short?
What risks or opportunities do you see as DERs, storage, and data centers reshape the grid?
Velvet Voelz, is an energy asset management expert who advances forecasting, revenue optimization and operational strategies in evolving power markets.
Oleg Popovsky is a founder and strategic advisor to cleantech companies focused on energy intelligence, DER monetization, and infrastructure innovation.
This article was originally published on Linkedin.