Context:

Path2GFM is an extension of LF energy GridFM project, integrating various open source packages to enable power system synthetic data generation.

GridFM aims at developping a foundational model specialised for power systems. Initially, pretraining of the GridFM focused on reconstructing iid steady state load flow calculation results. However, current software tooling in GridFM and initial pretraining task have two main limitations:

• First, pretraining was conducted using public data with an ‘academic’ load flow calculation tool. To move closer to an operational use case, it is of practical and strategic importance to rely on a load flow tool that is “industrial-grade” on top of being open-source, free to use and scalable on a computation cluster.
• Second, power transfers on Hydro-Québec’s transmission network are primarily limited by dynamic constraints, such as voltage collapse, which are not detected by load flow calculations, unlike thermal constraints. Therefore, for an operational use case, it is necessary to predict the results of dynamic simulations.

These two technological barriers require the integration of PowSyBl library and Dynawo into GridFM.

A third reason for Path2GFM is the exploration of highly scalable solutions for network simulation and the possibility to compare simulation results between tools like PSS/E, PowSybl, and Dynawo.

Objectives:

Path2GFM has 3 main goals:

• The integration of PowSyBl’s load flow calculation module (Open load flow) into GridFM.
• The integration of Dynawo (dynamic simulator) into GridFM.
• The comparison between Dynawo vs PSS/E on Hydro-Québec‘s simplified HQ29 network.

Project partners:

The project now gathers three industrial partners: Artelys, HydroQuebec and RTE.