DYNAMIC DECISION MAKING SYSTEM FOR SMART GRIDS STABILITY USING HYBRID POLICY GRADIENT- REINFORCEMENT LEARNING WITH FUZZY LINEAR PROGRAMMING

Abstract

Smart grid operations need accurate decision-making structures which also adapt to changing circumstances for improved stability assessment. Traditional statistical models together with rule-based systems have difficulties in effectively processing sophisticated and uncertain data from the smart grid because this can result in incorrect stability assessments. The proposed system uses Policy Gradient Reinforcement Learning with Fuzzy linear Programming (PGRL-FLP) model by combining PGRL with FLP for establishing an adaptive smart grid stability classification system. Predictive modeling and pattern recognition systems achieve better discovery of important data relationships through the use of correlation-based feature extraction with smart grid stability datasets. The decision-making system benefits from fuzzy constraints because they allow FLP to manage unpredictable grid situations effectively. Through the combination of PGRL-FLP model, the accuracy improves alongside decreased misclassification errors and better smart grid management enabling better performance of resilient smart grid infrastructure. The proposed model achieves 98.6% accuracy, 97.5% precision, 94.5% recall and 97.4% F1-score.