Load Forecasting in Electrical Grids: Analysis of Methods and their Trends

Main objective of this study is to analyze the progression of load forecasting methodologies for electrical grids, with a focus on identifying trends in performance metrics such as Mean Absolute Percentage Error (MAPE) over time. This analysis evaluates various forecasting approaches, including statistical methods, artificial intelligence, fuzzy logic, ensemble methods, and hybrid systems, to un-derstand their evolution and current state. To achieve the stated goals, the systematic review of scien-tific studies and articles that have the necessary metrics was conducted. From them, it was determined which models were used and what forecasting errors corresponded to them. Also, the publications re-viewed within this study were distributed over time to take into account the dynamics of changes in the results. The most important results are the obtained graphs of the dynamics of forecast of error changes for different models by years, as well as the possible ranges of variation of this error. The results show that, although increasingly complex models are being developed, their accuracy gain remains inconsistent in different application contexts, provided that a single-type architecture is used. Hybrid models demonstrate a significant increase in accuracy, and, therefore, superiority over a single-type architecture. The significance of the obtained results is in the clear illustration of the development of the accuracy of forecasting models. They allow us to determine the optimal vector of evolution of subsequent studies, namely, what type of model should be used to forecast the grid load. This study proves the prospects of using hybrid methods in the area under consideration as well.