Development of Risk-Integrated Load Demand Estimation and Prediction Models for Sustainable Rural Microgrid Planning

Despite increased microgrid adoption in Nigeria’s rural electrification, projects often fail due to power imbalances from inaccurate load demand prediction. Current models use Urban or Industrial data, ignoring rural consumption patterns, and lack risk assessment frameworks. This study developed load demand models with risk management for Gwam’s rural microgrid in Niger State, Nigeria. Exploratory factor analysis narrowed 14 input variables to 8 significant factors, ranked by fuzzy analytical hierarchy process, where the top five: previous hour load, temperature, humidity, hours of day, and holidays were selected and used in developing the prediction models. Four models were developed: ANFIS, ANFIS-PSO, ANFIS-GA, and ensemble ANFIS. The ensemble model outperformed others (MAPE: 14.69%, approximately 6% superior to conventional ANFIS). A risk management model categorized predictions as low (<10%), medium (10 – 20%), and high (>20%) risk. The ensemble model had 58.02% low-risk and 20.24% high-risk predictions, vs. 44.19% and 31.71% for ANFIS. This risk-aware framework supports operators in predicting failures, optimizing resources, and decision-making under uncertainty for sustainable rural electrification.