Research Worth Reading

Forecast and Model Predictive Control of Distributed Energy Resource Aggregators for Net-Demand Balancing — Combines forecasting with Model Predictive Control to coordinate distributed energy resource aggregators (DERAs) for net-demand balancing. Relevant for engineers working on grid integration of renewables, as it addresses supply uncertainty through decentralized control architectures.

Power Grid Topology Control — Explores network-side flexibility via transmission topology switching to manage renewable integration and climate-driven grid stress. Advances in power electronics and communication now make real-time topology control a viable resilience tool for system operators.

SABLE: GPU-Based Power Flow Accelerator for Sparsity-Aware Batched Learning — A GPU-accelerated sparse batched power flow solver designed for differentiable learning frameworks. Enables embedding AC power flow directly into ML pipelines while preserving sparsity — useful for optimization-in-the-loop applications like grid-aware RL or neural operators.

MPC for nonlinear systems: a comparative review of discretization methods — Compares direct multiple shooting, direct collocation, and successive linearization for nonlinear MPC. Practical reference for engineers implementing MPC in energy systems, battery management, or grid control where nonlinear dynamics matter.

Estimating Evolving Functions with Dynamic Gaussian Processes — Introduces Dynamic Gaussian Processes for estimating functions governed by integro-difference equations from discretized linear PDEs. Extends GP regression to evolving spatiotemporal systems — applicable to climate field reconstruction, emissions monitoring, or grid state estimation.

Technology & Innovation

Pioneering grid battery nudges California closer to 24/7 clean energy — The Tumbleweed project in Kern County is the first major U.S. battery installation with 8-hour discharge duration, doubling typical 4-hour lithium-ion systems. A milestone for long-duration storage economics and grid reliability modeling.

Vice President of Software Engineering at Titan Advanced Energy Solutions — Titan is hiring senior software leadership for grid-scale energy storage and battery management systems. Signals growing demand for software-centric approaches to storage optimization, degradation modeling, and fleet orchestration.

Open Source Projects

Waymo Sending Used EV Batteries to Community Clean Energy — Waymo repurposes degraded robotaxi fleet batteries into community energy projects. Real-world case study in second-life battery logistics, health diagnostics, and stationary storage integration at fleet scale.

Mexico Reaches 5 Gigawatts of Distributed Solar Power — Mexico hit 5.165 GW of sub-0.7 MW distributed solar across 600,000+ installations by end of 2025. Deployment data point for distributed PV scaling in emerging markets — relevant for grid modeling, inverter interoperability, and DERMS design.

Today’s Synthesis

The SABLE GPU-accelerated sparse power flow solver makes differentiable AC power flow practical at batch scale — pair it with the DERA MPC paper ’s forecasting + Model Predictive Control formulation for net-demand balancing, and you have the core of an end-to-end trainable controller for distributed energy resource aggregators. The MPC discretization review then gives you a decision framework: direct multiple shooting for stiff battery dynamics, direct collocation for smoother thermal/storage trajectories, or successive linearization for real-time receding-horizon updates on edge hardware. A concrete starting point: implement a batched differentiable MPC layer in JAX/PyTorch using SABLE’s sparse solver as the physics constraint, train the forecaster and policy jointly on historical CAISO + DER telemetry, then benchmark discretization methods against the review’s criteria (constraint satisfaction, solve time, gradient quality). This moves beyond simulation-only RL into deployable model-based control with certified feasibility — exactly the stack storage operators like Titan are hiring for.