Terra Daily — June 26, 2026

Research Worth Reading

• Bayesian Changepoint Detection for Smart Sensing of Battery Degradation: Cycle-Level Health Indicators and PyMC Implementation — A Bayesian single‑changepoint model monitors the charge‑time‑to‑discharge‑time ratio to detect accelerated degradation onset in lithium‑ion batteries. The PyMC implementation provides a reusable template for probabilistic health‑monitoring pipelines.
• When the Timetable Breaks: Physics-Anchored Scientific Machine Learning for Cold-Wave-Robust Battery-Electric Bus Operations — WeatherRobust integrates physics constraints with scientific machine learning to model hidden failure modes where cabin heating depletes battery capacity during cold waves. This approach demonstrates how domain‑aware ML can improve reliability of electric transit systems.
• A Bilevel Framework for Data Center-Grid Coordination with DLMPs in Unbalanced Three-Phase Distribution Systems — A bilevel optimization model uses distribution locational marginal prices (DLMPs) to coordinate active‑power demand between data centers and unbalanced three‑phase grids. The framework offers a scalable method for grid‑aware resource aggregation.
• Feasibility-Aware Security-Constrained Unit Commitment via Hybrid Soft Actor-Critic with Quantum-Sampled Features — A three‑layer hybrid framework combines a Bernoulli hybrid soft actor‑critic policy with quantum‑sampled features to solve security‑constrained unit commitment problems at realistic system sizes. This reduces computational expense while preserving feasibility constraints.
• Health feature extraction from battery energy storage system field fault data — Techniques for extracting reliable health indicators from operational field data under variable conditions, enabling early detection of catastrophic faults in grid‑connected battery storage. The work highlights feature engineering for real‑world sensing.
• Huracan: A skillful end-to-end data-driven system for ensemble data assimilation and weather prediction — Huracan implements an ML‑based end‑to‑end pipeline for ensemble data assimilation and weather forecasting, delivering high‑accuracy predictions with substantially lower compute than traditional numerical models. It serves as a reference for operational weather‑prediction systems.

Open Source Projects

• Sunrun, Renew Home, & Tesla Team Up to Deliver 16+ GW of Flexible Home Energy for Data Centers — An aggregation of residential battery storage, smart thermostats, and V2G resources forms a virtual power plant to meet the surging electricity demand of data‑center AI workloads. Engineers can study the orchestration of distributed resources at utility scale.
• Tesla, Sunrun, Renew Home team up on massive 16GW virtual power plant — The partnership creates a large‑scale VPP network targeting 16 GW capacity focused on data‑center hotspots, representing a real‑world deployment of grid‑integration techniques. This case provides insight into the architecture and coordination of heterogeneous distributed energy assets.

Today’s Synthesis

An engineer could combine the Bayesian changepoint model from Bayesian Changepoint Detection for Smart Sensing of Battery Degradation with the bilevel coordination framework described in A Bilevel Framework for Data Center-Grid Coordination with DLMPs in Unbalanced Three-Phase Distribution Systems to create a real‑time health‑aware resource manager for data‑center battery storage. The PyMC template would continuously monitor the charge‑time‑to‑discharge‑time ratio of each battery module, flagging accelerated degradation before it impacts performance. The weather‑aware model from When the Timetable Breaks: Physics-Anchored Scientific Machine Learning for Cold-Wave-Robust Battery-Electric Bus Operations can be repurposed to predict temperature‑induced capacity loss, feeding predictions into the same health‑aware scheduler for additional safety margins. When a degradation signal is detected, the bilevel optimizer can adjust the data center’s active‑power demand using distribution locational marginal prices, shifting load to healthier modules or to the emerging 16 GW virtual power plant built by Tesla, Sunrun, Renew Home team up on massive 16GW virtual power plant . This integrated pipeline would keep data‑center AI workloads powered while extending battery asset life and providing grid‑scale flexibility.