Power Load Data

This dataset contains three regional subsets comprising the electricity load and corresponding weather conditions collected from Australia, Panama, and Austria. Each subset includes time-synchronized records of power consumption along with key meteorological variables such as temperature, humidity, wind speed, and other locally available weather indicators. The data were gathered from regional grid operators and national meteorological sources to reflect the actual operating and environmental conditions of each region.

Experiments were conducted in this paper using two publicly available datasets: ISO-NE and Southern China. The ISO-NE dataset, sourced from New England, USA, comprises 103,776 hourly records spanning from March 1, 2003, to December 31, 2014. In contrast, the Southern China dataset covers the period from January 1, 2012, to January 10, 2015, with a 15-minute resolution, resulting in 106,176 samples. Both datasets incorporate a range of variables, including temperature, rainfall, humidity, power load, and day-type classifications.

ISO New England power load data

Existing AC/DC power flow computations necessi- tate sequential convergence-oriented trial-and-error under vari- ous DC control modes, rising computational burden. This paper thus proposes a physics-guided multi-agent graph learning (PG- MAGL) method towards real-time power flow analysis with DC control mode adaptation. The tailored graph structure with built- in DC control modes and state variables is firstly advanced to ensure topology adaptability. Then, MAGL is proposed to enable adaptive jump over DC control modes.