Four grid operators already run the asset library LF Energy just admitted at Sandbox

Sandbox is where a project's paperwork gets built. LF Energy defines it as the entry point for work that warrants experimentation and still has to establish governance, security practices and a broader contributor community ^[4]. So the tier describes how far the collaboration has come. AssetLife came in at Sandbox with four operators running it in production, one of them a gas transmission network ^[6]. It is also the only one of the four new projects that LF Energy lists with production users at all ^[21].

What AssetLife actually does is unglamorous and useful. It reads infrastructure records and applies survival analysis, stochastic modeling and renewal theory to compare the expected cost of preventive replacement against repair and running equipment to failure ^[5]. A carbon shadow price can be folded into the same calculation, so societal cost sits next to maintenance spend ^[7]. According to LF Energy, utilities usually do this in internal spreadsheets or proprietary planning tools ^[8]. If your team maintains a Weibull fit and a replacement-cost model in a workbook, this is the adopt-or-keep-building decision, and it now has a named alternative with a Python API.

About 10 minutes is 600 seconds; divide by 0.2 seconds and the graph network returned its topology optimization roughly 3,000 times faster than the mixed-integer linear programming baseline ^[20]. It also found a smaller capacity improvement, and LF Energy describes the work as an R&D demonstration and not a control-room deployment ^[14]. The experiment ran on French transmission data with more than 7,000 buses ^[12]. For that speedup to mean anything on your system you would need a network of comparable size and heterogeneity, a topology problem you are willing to solve approximately, and an operator who will accept a worse capacity answer in exchange for the latency. EnerGNN is built with JAX and Flax and models the varied components and connections of a real transmission network instead of collapsing the grid into interchangeable nodes and edges ^[11].

CityLearn is a bench. Its value is that rule-based control, model-predictive control and reinforcement learning get scored on the same datasets and the same performance measures across a district of batteries, heat pumps, electric vehicles and solar ^[9]. Contributors are the University of Texas at Austin's Intelligent Environments Lab, TU Eindhoven, Politecnico di Torino and Concordia University ^[10]. Smart HEMS Benchmark aims at the same trick for home energy-management systems, covering equipment sizing, daily dispatch and virtual-power-plant support, with its open source launch planned for late 2026 ^[15].

The stated model is that utilities contribute software they already run, vendors build commercial products on top, and LF Energy carries the governance and security work that makes the arrangement acceptable to legal departments and regulators ^[19]. That model started with an unscheduled visit: in the mid-2010s Shuli Goodman drove to the Linux Foundation's San Francisco office without a meeting to argue that utilities needed the shared development already normal in cloud and telecom, and her conversations with RTE and others led to the July 2018 launch ^[16]. Goodman died of cancer on January 3rd, 2023 ^[17]. Alex Thornton, previously CTO of Cleartrace and before that at Locus Energy and Sunrun, took over as executive director in December 2023 ^[18].