Accurate power demand forecasting is essential for effective energy resource planning and system operation, ensuring a reliable and efficient supply of electricity to consumers. Traditional forecasting methods often fall short in capturing complex consumption patterns, leading to the growing adoption of machine learning-based approaches. This study aims to develop a robust and precise medium-term energy consumption forecasting model for student housing facilities. The model utilizes wind speed, temperature, and humidity as input variables, while the target output is the corresponding energy consumption. Data for model training (70%), testing (20%), and validation (10%) were collected from real-time energy monitoring devices installed in the building, complemented by meteorological data from the nearest weather station. To enhance predictive accuracy, a renowned metaheuristic optimization technique, the Biogeography-Based Optimization (BBO), was integrated with an adaptive neuro-fuzzy inference system (ANFIS). Additionally, the influence of hyperparameter selection was analyzed using a fuzzy c-means (FCM)-clustering technique. Experimental results demonstrated that the FCM-clustered hybrid ANFIS-BBO model with three clusters yielded the most accurate predictions compared to the standalone ANFIS model. This study emphasizes the importance of hybrid techniques and effective hyperparameter tuning in enhancing predictive modeling performance, particularly for energy forecasting applications.