Leveraging AI for Sustainable and Cost-Effective Decentralized Energy in Rural U.S. Regions

Abstract

Sustainable and affordable decentralized power in rural parts of the U.S. is difficult to accomplish because of inadequate infrastructure and transmission costs, and centralized grids are susceptible. This paper suggests an artificial intelligence (AI) hybrid framework combining XGBoost and a neural network to provide precise predictions of renewable generation and operational expenses. More than 13000 electric power plants databases in the United States underwent preprocessing in order to solve the problem of missing data, heterogeneity, and rural-urban disparities. The proposed model (XGBoost+NN) was compared with KNN, Auto encoder, PINN, CNN-LSTM, and BiLSTM-Attention. It out-performed all baselines with MSE of 0.0003, RMSE of 0.0178, R2 of 0.999, and MAPE of 1.36%. The error in predicting costs was only 0.08% and this is both a technical strength and an economic importance. These findings indicate the model’s scalability, interpretability, and possible use to facilitate equitable access to clean and affordable energy in underserved rural populations.