Siemens Energy Accelerates Power Grid Asset Simulation 10,000x Using NVIDIA Modulus

Power Grid Asset Simulation: How Siemens Energy and NVIDIA Are Revolutionizing Energy Management

Summary

Siemens Energy has partnered with NVIDIA to develop AI surrogate models for power grid asset simulation, achieving a 10,000x acceleration. This breakthrough enables real-time thermal insights, enhancing grid reliability and reducing downtime. The collaboration focuses on transformer bushings and gas-insulated switchgears, critical components in modern power grids.

The Challenge of Modern Power Grids

The world’s energy system is becoming increasingly complex and distributed due to the rise of renewable energy sources, decentralization of energy resources, and decarbonization of heavy industries. Energy producers face the challenge of optimizing operational efficiency and costs within hybrid power plants that generate both renewable and carbon-based electricity. Grid operators have less time to dispatch energy resources optimally, making efficient asset management crucial.

AI Surrogates for Transformer Bushings

Siemens Energy is developing an AI surrogate model for transformer bushings to predict hotspot temperatures in real-time under varying conditions. The neural network, trained using NVIDIA Modulus, considers heat generation, propagation, and loss to act as a virtual sensor. This approach saves costs on monitoring systems and allows grid assets to adjust based on temperature profiles, enhancing efficiency.

Training Strategy for Transformer Bushings

The neural network for bushings was trained with input load and ambient conditions to predict temperature profiles. By optimizing load management based on temperature rather than fixed rates, grid assets can be operated more dynamically and efficiently.

AI Surrogates for Gas-Insulated Switchgears (GIS)

Siemens Energy is also creating AI-based surrogate models for GIS to predict thermal behavior under different operational conditions. These models help operators maintain safe temperature levels in switchgear components to prevent overheating.

Training Strategy for GIS

To predict temperature dynamics accurately in GIS components, an optimized graph neural network (GNN) in NVIDIA Modulus was chosen. The model was trained on multi-GPU setups using message-passing algorithms to predict temperature dynamics for 10 hours under unseen conditions, enabling better assessment of short-time overload scenarios.

The Impact of AI Surrogates

The AI surrogate models developed by Siemens Energy and NVIDIA offer several benefits:

• Enhanced Grid Reliability: Real-time thermal insights help prevent overheating and reduce downtime.
• Improved Operational Efficiency: Dynamic load management based on temperature profiles optimizes energy supply chain and demand predictions.
• Cost Savings: Virtual sensors reduce the need for physical monitoring systems.

Table: Key Benefits of AI Surrogates in Power Grid Asset Simulation

Table: Comparison of Traditional vs. AI-Based Power Grid Asset Simulation

Table: Key Components of AI Surrogates in Power Grid Asset Simulation

Conclusion

The collaboration between Siemens Energy and NVIDIA marks a significant advancement in power grid asset simulation. By leveraging AI surrogate models, energy producers can optimize operational efficiency, enhance grid reliability, and reduce costs. This breakthrough underscores the potential of AI and high-performance computing in transforming energy management.