Forecasting Hourly Energy Consumption Using LSTM: A Deep Learning Approach for Indian States

Especially in light of growing needs and the pressing need for sustainable resource use, precise energy consumption forecasting forms a foundation of contemporary energy management systems. This research offers a deep learning approach based on Long Short-Term Memory (LSTM) to forecast hourly energy consumption in several Indian states. Using historical usage data, the model is individually trained and assessed for every state to spot trends, patterns, and anomalies. Using preprocessing methods including normalization, missing value imputation, and sequence creation, the model performs well in many areas. Strong predictive accuracy is demonstrated by the top-performing states, including Madhya Pradesh (R² = 0.6573) and Dadra and Nagar Haveli (R² = 0.6615) .The results indicate that real-time energy demand prediction benefits from LSTM models, which also make decision-making aids for grid stability and policy design .