Background
Modern electrical grids tend to have higher penetration of intermittent energy sources. The grid infrastructure is also more susceptible to physical damage due to increased extreme weather events. Grid operators have to run multiple power flows for different scenarios in real time to continuously monitor the system states and make necessary adjustments. Computational complexity of existing power flow algorithms increases with the grid size. A trade-off between memory usage and execution time has to be made in existing algorithms that degrades the precision and speed.
Invention Description
Researchers at Arizona State University have developed a multi-faceted algorithm combining machine learning, topological clustering, and GPU-accelerated parallel processing to optimize power flow calculations in electrical grids. It reduces computational and memory overhead by clustering operational scenarios, maintaining constant-sized Jacobian matrices, and leveraging a sparse difference admittance matrix. A deep learning model predicts initial voltage angles to accelerate convergence, enabling real-time, high-precision grid analysis despite the complexities introduced by renewable energy integration and grid vulnerabilities.
Potential Applications:
- Electric power grid analysis software
- Enhanced decision support systems
- Smart grid and microgrid optimization platforms
Benefits and Advantages:
- High-speed – increased calculation speed through GPU parallel processing
- Efficient – handling of multiple operational scenarios through clustering
- Accurate – enhanced precision without compromising computational speed