Terra Daily — May 18, 2026

Research Worth Reading

• Bridging the climate to energy data gap: simulated annealing for representative climate year selection — Proposes a simulated annealing approach to select representative climate years from hundreds of simulations, addressing the tractability gap for power system models. Directly useful for energy system planners running resource adequacy studies where compute time is limited.
• Enabling Intelligent Bidirectional Charging: A Real-World Communication Interface Between Electric Vehicles, Charging Infrastructure, and a Control Optimizer — Field-validated bidirectional EV charging system in Dresden with a working interface between EVs, chargers, and a control optimizer. Shows what V2G deployment actually looks like in an urban setting.
• Optimizing Chilled Water Systems with Cooling Towers via Virtual Power Metrics and Extremum-Seeking Control — Applies extremum-seeking control to cooling tower fans using a virtual power metric, achieving ~15% energy savings across climate zones in simulation. Chilled water plants are a major industrial load, and the control approach transfers to other thermal systems.
• Njord: A Probabilistic Graph Neural Network for Ensemble Ocean Forecasting — Introduces a probabilistic ocean forecasting model combining deep latent variable frameworks with graph neural networks for uncertainty quantification. Addresses the limitation of deterministic ML ocean models and is relevant to anyone building data-driven forecast systems.
• Fairness-Guaranteed Online Power Allocation Policies for EV Fast Charging Stations — Proposes online power allocation policies with fairness guarantees for EV fast chargers when total port rating exceeds grid or infrastructure capacity. Directly applicable to engineers designing charging networks under real grid constraints.
• SwAIther-Precip: Lead-Time-Aware Bias Correction Enables Kilometer-Scale Downscaling of Global AI Precipitation Forecasts over Switzerland — Demonstrates a lead-time-aware bias correction method that downscales global AI precipitation forecasts (0.25°) to kilometer-scale over complex terrain. Bridges the resolution gap for local hydrological and energy forecasting use cases.

Technology & Innovation

• Solar to overtake coal on Texas grid for the first time ever this year — Solar is projected to generate more electricity than coal in the ERCOT grid for the first time, driven by rapid solar buildout and no new coal plants. A clear signal of how generation mix shifts as build rates change.
• Iceland drills 4.7 km down into volcano to tap clean energy — Iceland Deep Drilling Project accessed supercritical geothermal fluids at extreme depth, achieving significantly higher power output per well than conventional geothermal. A concrete case study in advanced geothermal engineering.

Policy & Regulation

• North Carolina groups fight regulator’s order to cancel solar for 2026 — Clean energy advocates are challenging a regulator’s order halting Duke Energy’s 2026 solar farm investments. The decision adds to the headwinds from tariffs and federal policy changes affecting state-level solar development.
• Ohio utilities report subpar grid reliability as they seek a lower bar — Four of Ohio’s six regulated utilities failed reliability standards in 2025, the 10th consecutive year of at least one failure. Utilities are simultaneously requesting lower reliability benchmarks, raising questions about grid maintenance and investment adequacy.

Today’s Synthesis

As solar overtakes coal on the Texas grid , midday overgeneration and evening ramps become more pronounced. Buildings with chilled water systems—major industrial loads—can shift thermal storage by modulating cooling tower fan operation. The extremum-seeking control paper shows ~15% savings via real-time fan optimization using a virtual power metric, and the approach transfers to other thermal systems. The control approach has been tested across climate zones in simulation, making it applicable to different geographies. Pairing this with bidirectional EV charging lets fleets discharge during peak demand and charge when solar is abundant, while fairness-aware power allocation prevents grid overloads at stations. The Dresden system provides a working interface between EVs, chargers, and a control optimizer, which can be extended to coordinate with building thermal systems. As Ohio utilities report subpar grid reliability , optimizing both thermal and EV loads becomes more critical. Utilities seeking lower reliability benchmarks will face pressure to demonstrate grid flexibility, which building-level optimization can provide. Using the simulated annealing approach to select representative climate years ensures the control strategy performs well across a range of weather scenarios, not just a single year. An engineer could prototype a system that runs extremum-seeking control on cooling towers and coordinates EV bidirectional flow using the same control optimizer. This directly applies ML/control skills to building-scale flexibility in a high-solar grid.