Research Worth Reading

  • Building Power Grid Models from Open Data: A Complete Pipeline from OpenStreetMap to Optimal Power Flow — A five-stage pipeline that constructs complete, OPF-solvable transmission network models entirely from OpenStreetMap data. Solves the critical barrier of restricted grid data access in the US, enabling reproducible power systems research without proprietary datasets. If you’ve worked with geospatial data pipelines or network graph construction, this is directly in your wheelhouse.

  • Thinking fast and slow – decision intelligence for power systems — A decision intelligence framework for power systems spanning milliseconds to long-term planning, built to handle uncertainty from intermittent renewables and distributed energy resources. The multi-timescale optimization problem here is substantial — relevant if you have a background in control systems, reinforcement learning, or stochastic optimization.

  • Interpretable Neural Networks to Predict Momentum Fluxes of Orographic Gravity Waves — Trains interpretable neural networks on ERA5 reanalysis data to improve parameterization of orographic gravity wave momentum fluxes in Earth system models, which are too computationally expensive to resolve explicitly. The interpretability focus means you can inspect what the model learned rather than treating it as a black box — a meaningful step for ML-in-climate work.

  • Collective and nonlinear structure of wind power correlations — Analyzes 5 years of data from an 80-turbine wind farm, revealing universal collective and nonlinear correlations that cause excess persistency and intermittency in aggregated farm output. Critical for anyone building energy models or grid integration tools — wind farm-scale power behavior is not simply the sum of individual turbines.

Technology & Innovation

  • AI Data Centres Need Big Batteries But Lithium Isn’t Fit-For-Purpose — Argues that AI data centers’ bursty, high-ramp power profiles make lithium-ion suboptimal for grid-scale storage, and advocates for alternative chemistries. If you’re evaluating storage solutions for data center grid integration, this is a useful framing of the mismatch between current battery tech and actual load profiles.

  • Better Flight Planning Can Cut Fuel & Contrail Warming — Optimized flight routing and altitude selection can reduce both fuel burn and contrail formation — a major non-CO2 warming contributor — without waiting for new fuels or aircraft. A near-term, software-driven lever for aviation emissions reduction. The routing optimization problem here is a concrete application for anyone with operations research or ML-for-optimization skills.

  • Europe’s quest for green steel — Covers Europe’s push toward hydrogen-based direct reduction and electric arc furnace technologies for steel decarbonization. Highlights the engineering and scale challenges in tackling one of the most carbon-intensive industrial sectors — relevant for mechanical and process engineers eyeing heavy industry decarbonization.

  • LANDKING Makes Electric Trucks Look Inevitable at Auto China — Reports a visible surge in electric trucks on Guangzhou roads, with LANDKING showcasing new models at Auto China 2026. Signals accelerating commercial EV adoption in China’s heavy-duty transport sector, with implications for global freight decarbonization timelines and supply chain engineering.

Open Source Projects

Today’s Synthesis

The pipeline for constructing transmission models from OpenStreetMap data Building Power Grid Models from Open Data solves a concrete problem for anyone evaluating where large-scale storage fits into grids. Combine that with the observation that AI data centers’ bursty load profiles make lithium-ion storage suboptimal AI Data Centres Need Big Batteries But Lithium Isn’t Fit-For-Purpose , and you have a clear modeling task: plug actual load profiles into OPF-solvable grid models, test alternative storage chemistries under realistic conditions, and see where the bottlenecks are. The wind farm correlation research Collective and nonlinear structure of wind power correlations adds another layer — aggregated renewable output is not simply additive, so storage sizing models that treat wind as a sum of independent turbines will be wrong. Building grid models that account for these nonlinearities before optimizing storage placement is a tractable project for someone with geospatial data pipeline experience and an interest in power systems.