POWER FLOW PREDICTION USING RECURRENT NEURAL NETWORKS WITH L.S.T.M. AND GENETIC ALGORITHM IN AN IEEE 30 BUS SYSTEM

Abstract

Accurate forecasting of active (P) and reactive (Q) power in electrical networks is fundamental to improving operational reliability and planning in modern power systems. This article proposes a forecasting model based on Long Short-Term Memory (LSTM) neural networks to estimate load demand in an IEEE 30-bus system. The model considers the active and reactive power supplied by the generators as input variables, while the power components of the loads are used as target variables for prediction. To improve performance, a Genetic Algorithm (GA) was used to optimize hyperparameters, which reduced the Mean Absolute Error (MAE) and increased the accuracy of the predictions. The results demonstrate that the proposed approach provides stable predictions of load behavior, highlighting its potential for application in smart grids and microgrid management systems.

DOI: https://doi.org/10.56238/sevened2025.036-051