Digital twins are a crucial component of the digital transformation puzzle. The goal of digital twin technology is to create an accurate virtual/digital replica of objects, processes and sometimes even the entire business eco-systems to improve the efficiency and sturdiness of corporate processes while generating economic value through real-time analytics-based prediction and prevention.
With the recent advances in information and communication technologies, the digital twinning concept is attracting the attention of both academia and industry across sectors. In the energy sector, these virtual models have gained significant importance due to the increasing complexity of systems, high penetration of renewables into microgrids and the rise of prosumers in the energy ecosystem.
The use of digital twinning technology can help in creating high-fidelity forecasting models for enabling predictive analysis of the massive sensor data, learning prosumer behavioural patterns and for real-time monitoring of the assets in a complex power system networks.
This blog introduces DIPS (Distributed and Intelligent Power System) digital twins as one such technological solution that is a digital representation of the Monash microgrid edge (IoT) power system having a synchrophasor network of smart devices (such as PMUs). The virtual model mirrors the production environment for this power system into a synthetic/virtual world.
A simulated copy of the edge, sensors, services and compute environments have been created for development and testing purposes. The environment mimics how connections and interactions between different system components and services will occur in the real world, without the need for their physical counterparts.
The environment stimulates data engineering workflow (data collection, processing, storage & archival of the actual DIPS data platform) using simulated sensor data and provides an integrated decision-making support by training machine learning (ML) algorithms on the simulated and historic power data to produce high-fidelity ML models. The machine learning models help generate precise forecasts and valuable insight which can be applied back to the real-time grid data thereby optimising the dynamics of change in the energy eco-system before the production starts.
The virtual and production environments share the same DevOps methodologies and CI/CD automation to implement consistent and fast deployment and test the end-to-end integration of the data engineering & analysis workflow.
This end-to-end virtual environment will help monitor the steady state of the high-streaming devices, the data pipelines but it helps identify any abnormalities before the system goes into production thereby ensuring a highly reliable energy supply with the least amount of service disruption and downtime.
Please stay tuned with our energy blogs to receive updates on the DIPS platform, as an open-source initiative to share our advances!
AUTHORS
Dr Ayesha Sadiq, Research Software Specialist, Monash eResearch Centre
Dr Steve Quenette, Deputy Director, Monash eResearch Centre
Mitchell Hargreaves, Junior Deep Learning Engineer, Data Futures Institute
Ai-Lin Soo, Senior Project Officer, Monash eResearch Centre