Urban heat island and electrical load estimation using machine learning in metropolitan area of rio de janeiro

This study presents two innovative machine learning-based models: one for daily electrical load forecasting in the State of Rio de Janeiro and another for monthly forecasting for each Light concessionaire substation in the Metropolitan Area of Rio de Janeiro (MARJ). The utilized data include (1) daily electrical load data from the National System Operator (ONS) for the State of Rio de Janeiro spanning four years (2017-2020); (2) monthly electrical load data from 84 Light substations over 11 years (2010-2020); and (3) maximum, minimum, and mean air temperatures. Additionally, remotely sensed land-surface temperature (LST) based on Landsat data from 1984 to 2020 is employed to reconfigure the existing meteorological network in MARJ based on the distribution of Light substations. The regressive machine learning algorithms undertook training using a cross-validation procedure in 500 daily and 500 monthly training-testing experiments. Results for daily-ONS and monthly-Light loads show average correlations (hindcast in parentheses) of the fitted models of 0.85±0.09 (0.83±0.07) and 0.89±0.05 (0.91±0.06), respectively. The model’s Mean Absolute of Error (MAE) values correspond to a percentage error of about 4.03% (daily) and 4.83% (monthly). According to the monthly electrical load behavior revealed, when the temperature changes from 23 to 26℃ at MARJ, it rises roughly from 1.92 x 10⁶ ± 67227.4 kWh to 2.70 x 10⁶ ± 90198.5 kWh. A cluster analysis was conducted to optimize the meteorological station network, considering the locations of the existing 18 meteorological stations, the 84 Light electrical load substations, and the urban heat island cores. This analysis identified seven specific locations for installing new meteorological stations, aiming to improve the spatial resolution of electrical load modeling in MARJ. This insight opens paths for future research to refine the network and further enhance the models’ precision in capturing the intricate relationships between meteorological factors and electrical loads in the region.