Forecasting Fuel Supply Inventory in Remote Areas using Deep Learning Approaches

Electricity is an essential commodity, yet power disruptions remain a critical challenge in remote regions. This research addresses the need for preemptive management of fuel supply for power plants in remote areas to prevent blackouts. The study evaluates the efficacy of three deep learning approaches: Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), and a combined CNN-LSTM approach. The results indicate that the CNN-LSTM method outperforms the others, demonstrating the lowest Root Mean Square Error (RMSE) and the highest $\mathrm{R}^{2}$ value. This finding suggests that the CNN-LSTM model is the most effective for forecasting and mitigating power outages in these regions. Implementing this method could enhance the reliability of electricity supply, thereby ensuring continuous and stable power services for communities in underserved areas.