Mapping the Residential Electric Future of Salta: Comparison between Stochastic Models and Artificial Neural Networks for Annual Hourly Residential Electrical Consumption Simulation

Abstract

The simulation of annual hourly electrical consumption in homes in the city of Salta can be carried out through a dynamic stochastic model or the use of artificial neural networks (ANNs). The stochastic model considers the natural variability of factors such as climate and economic activity, using probabilistic techniques to forecast electrical demand. On the other hand, ANNs employ machine learning algorithms to analyze complex patterns in historical data and generate predictions. The choice between both models depends on data availability, processing capacity, and simulation objectives. While the stochastic model provides a probabilistic understanding of electrical demand, ANNs can capture non-linear relationships and subtler patterns, leading to more accurate predictions in certain circumstances. In Salta, where climate and economic activity are highly variable, combining both approaches could provide a more comprehensive simulation of annual electrical consumption. Although the stochastic model shows statistically superior results, ANNs are more computationally efficient, suggesting that the choice depends on specific contextual considerations and available resources.