Power Quality Monitoring and Forecasting Using LSTM and ANFIS Methods for Isolated Power System Networks

Abstract

In the present environment, the concept of Power Quality (PQ) is of high interest for any power system researcher. The majority of loads at the consumer end are designed using power electronic components and also integration of renewable energy resources, which introduces PQ issues in terms of voltage and current. Hence, appropriate monitoring and forecasting methods of PQ parameters are essential for maintaining grid reliability and stability. In this paper, we introduce a data-driven method for forecasting PQ disturbances using LSTM networks and ANFIS models in a power system environment. A dataset generated from MATLAB/Simulink simulations of an isolated distribution system with nonlinear loads is used for model training and evaluation. The models are evaluated using key performance metrics, including Mean Absolute Error (MAE), Mean Squared Error (MSE), and Coefficient of Determination (R²). Additionally, a comparative benchmarking analysis with conventional machine learning models, including Multilayer Perceptron (MLP), Random Forest (RF), and XGBoost (XGB), is presented. Experimental results demonstrate that LSTM achieves superior predictive accuracy (R² > 0.998), while ANFIS offers competitive performance and interpretability. These findings underscore the potential of AI-based techniques for proactive PQ monitoring and forecasting in power system environments.