Artificial Intelligence and Electrical & Electronics Engineering: AIEEE Open Access
Scientific Machine Learning for Power Grids
Abstract
M. Temple-Raston
Power grid operators need real-time situational awareness under conditions where historical training data does not reliably predict what is coming next. With renewable variability and demand shifts there is simply no time to retrain models. Scientific machine learning (sml) is the analytical solution to the science-of-counting (in units) that produces a set of scientific (thermodynamic) measurements for any time-series, available immediately, with no training, as a cloud- based “multi-meter”. The measurements are presented for power consumption data (New York State, industrial grid in MWhs, monthly, 2010-2026). The measurements are then used as unbiased evidence to investigate hypotheses in non- equilibrium dynamics for power consumption data in NYS. Balance and control strategies are understood and developed with access to scientific time-series measurements.