Comprehensive Forecasting of Smart Grid Stability Using Advanced Deep Learning and Machine Learning Models

This study comprehensively examines the ability to forecast the stability of smart grids using sophisticated deep learning and machine learning models. We investigate different approaches, such as Bidirectional Gated Recurrent Unit (BiGRU), Multi-Layer Perceptron with Extreme Learning Machine (MLP-ELM), and innovative fusion models that combine Artificial Neural Networks (ANN) with Support Vector Machines (SVM). The performance of these models was evaluated by comparing parameters such as accuracy, precision, recall, and F-measure. The dataset is an improved version of the Electrical Grid Stability Simulated Dataset from the Karlsruher Institut für Technologie. It consists of 60,000 observations and includes twelve primary prediction variables. Before training and testing the models, we conducted thorough exploratory data analysis (EDA) and preprocessing. The solution we present utilizes a Sequential model that incorporates multiple optimizers and activation functions. This approach achieves a maximum accuracy of 99.48% when using the Adamax optimizer. The results indicate the substantial capacity of deep learning methods to improve the dependability and effectiveness of smart grid stability forecasts. This work offers a novel deep learning methodology for predicting the stability of smart grids, attaining a prediction accuracy of 99.48 % and exhibiting enhanced performance relative to leading methodologies. Comprehensive testing across several grid settings demonstrates the effectiveness of the proposed tactics, rendering it very beneficial for real-time stability forecasts.