Terra Daily — May 22, 2026

Research Worth Reading

• Engineering Hybrid Physics-Informed Neural Networks for Next-Generation Electricity Systems: A State-of-the-Art Review — Reviews integration of physics-informed ML (PIML) in electricity systems, embedding physical laws (e.g., power flow) into neural networks to reduce data needs and improve interpretability for grid monitoring, control, and design. Directly relevant for engineers building hybrid models where purely data-driven approaches fall short.
• Resilient Energy-Based Control for DC Data Centers under Grid and Load Disturbances — Proposes a passivity-based control framework for AC-DC converters in data centers that ensures stability beyond nominal operating points by leveraging total energy balance. Relevant for engineers designing resilient power electronics for high-reliability facilities.
• Assessing global drivers of forest transpiration using clustered machine learning models — Applies clustered ML to identify environmental drivers of forest transpiration across species and climates, aiding climate modeling and ecosystem health prediction. Demonstrates a data-driven approach to understanding plant water use at scale.
• Another first for renewables: Wind and solar outgenerate gas in April — Reports that wind and solar combined produced more electricity than natural gas globally in April 2026, a first per Ember data. Highlights the accelerating shift in generation mix and the narrowing gap over five years.
• Equilibrium-Free Contraction Stability Analysis for Grid-Forming Converter-Based Microgrids — Develops a contraction-based stability analysis for renewable-dominated microgrids that avoids needing equilibrium points, using semi-contraction theory in a symmetry-aware projected state space. Enables more realistic stability assessment of inverter-based systems.

Technology & Innovation

• The Mercedes AMG GT Coupe Is An Absolute BEAST! — Details Mercedes’ upcoming AMG GT Coupe featuring battery cell technology derived from Formula One, indicating high-performance applications of motorsport-derived energy storage. Relevant for engineers interested in battery innovation and performance engineering.

Open Source Projects

• NatLabRockies/WAVES — A discrete-event simulation tool for estimating offshore wind farm lifecycle costs (LCOE) in Python. Useful for engineers evaluating wind energy project economics and lifecycle performance.
• Another Route To Rooftop Solar: The Ann Arbor Solution — Describes a not-for-profit utility model in Ann Arbor offering rooftop solar-plus-storage with no upfront costs and a $600 flat annual fee. Removes financial barriers and could be replicated elsewhere.
• Stellantis to Build EVs for Dongfeng in France — Stellantis will manufacture EVs for Chinese automaker Dongfeng in France, reflecting global shifts in EV production localization. Highlights supply chain and manufacturing trends for engineers tracking industry dynamics.

Policy & Regulation

• Geothermal energy gets boost from new coalition of Western governors — A bipartisan group of Mountain West governors aims to unlock 200 GW of geothermal energy, a 50-fold increase, signaling political momentum for baseload clean energy. Relevant for engineers assessing long-duration clean energy baseload options.

Community Finds

• What The Heck Is Going On With OTA Cuts To EV Range In China? — Reports that some Chinese EV makers use OTA updates to reduce vehicle range, raising transparency concerns in battery management. Highlights tensions between rapid innovation and consumer rights in EV software.

Today’s Synthesis

If you’re working on grid-adjacent software, two threads here are worth stitching together. The Equilibrium-Free Contraction Stability Analysis gives you a mathematical framework for stability in inverter-dominated microgrids without assuming equilibrium — exactly the kind of edge case that breaks conventional control loops. Meanwhile, Engineering Hybrid Physics-Informed Neural Networks shows how to embed power flow constraints into ML models so they respect physical laws while learning from limited data. Pair that with a tool like WAVES for lifecycle cost estimation, and you can build a prototype that jointly evaluates: does this converter topology stay stable under disturbance, and does the LCOE pencil out over 20 years? The practical play: start by validating the contraction-based stability claims on a simple inverter test case, then wrap a PIML power flow model around it to predict performance across operating points, and finally feed those profiles into WAVES for cost sensitivity. Each layer maps directly to a skill you likely already have — control theory, ML model architecture, and simulation — and the intersection is a concrete problem space where climate engineering meets deployable software.