Synthetic data provides a mechanism to accelerate the energy transition by enabling better use of AI for energy systems while eliminating concerns around data privacy. To ensure the synthetic data is useful, models need to be trained using real data.
Synthetic data is artificially generated data that mimics real-world data, used for various purposes like training machine learning models, testing software, and enhancing data privacy. Hence, synthetic data can be artificially generated to mimic real data sets, letting companies create a large amount of diverse training data without spending a lot of money and time.
An example of this is with the LF Energy OpenSynth project. This was developed to enable access to AI-generated synthetic data. As an open source community, OpenSynth is able to make use of datasets contributed by multiple utilities to improve its synthetic data generation capabilities.
The project was originated by the Centre for Net Zero – a non-profit arm of the commercial company Octopus Energy – before being contributed to LF Energy. The system was initially trained using data from Octopus Energy.
In a recent move, RTE, the Transmission System Operator for France has now contributed the first set of non-demand data to the project. The set combines a mixture of non-synthetic data (grid topology), all network components, and synthetic data (injection time-series created by reconstructing using open source aggregated data).
Overview of OpenSynth
OpenSynth is an open data community, originated by Centre for Net Zero (CNZ) and sourced under The Linux Foundation (LF Energy). The technology and programme are designed to democratise access to AI-generated synthetic data and accelerate the decarbonisation of global energy systems.
Initially, OpenSynth focused on synthetic demand data, given CNZ’s expertise in using AI to generate extensive and valuable datasets with diverse use cases for unlocking smart energy systems. The community empowered both holders of raw smart meter data around the world to be able to generate and share synthetic data, and for community members to generate, improve and share algorithms.
Integrating grid data
As OpenSynth has grown, it has absorbed other forms of synthetic and real energy-adjacent data from different international markets for potential inclusion. To build sophisticated models of the energy system, data that is accurate at all levels of the grid hierarchy – real or synthetic – is highly valuable. We have therefore decided to expand the scope of the open data community to maximise its value for current and future users.
The first candidate for non-demand data inclusion was D-GITT & RTE7000. D-GITT (Detailed Grid Inner Topology Timeseries) is a centralised, open dataset hub . The first major contribution was RTE7000, which represents the 7000 nodes of the French transmission grid.
RTE7000 combines a mixture of non-synthetic data (grid topology), all network components, and synthetic data (injection time-series created by reconstructing using open-source aggregated data).
With access to transmission level grid topology data for the full French network, together with time-series load data, researchers, software developers, and energy systems modellers are able to initiate large-scale system studies and AI model development via this open data platform.
Going forwards, this paves the way for further AI-generated synthetic datasets for energy research and modelling as well as providing the basis as a synthetic system data to train advanced AI models.
