Impact of Climate Change on Cities: Analyzing Building Energy Use in Future Climate Scenarios Through a Hybrid Computational Approach

Under global warming, cities need accurate predictions of building energy use to ensure energy supply and guide urban planning and policy decisions. Traditional data-driven methods struggle to project future scenarios due to the lack of training data reflecting anticipated climate conditions. This study presents a novel approach by integrating the Degree Day Model (DDM) with a Knowledge-Based Neural Network (KBNN). The DDM quantifies heating and cooling demands based on temperature deviations, providing domain-specific insights that enhance the KBNN's ability to extrapolate. We trained our KBNN using monthly energy consumption data from over a thousand New York City buildings, along with current and projected meteorological and building characteristics. The hottest and coldest months each year were reserved for testing, while the remaining data were used for training and validation. Benchmark models, including Multiple Linear Regression (MLR), Support Vector Machines (SVM), Random Forest (RF), and Artificial Neural Networks (ANN), were also evaluated. Our KBNN achieved superior accuracy, with a Mean Absolute Percentage Error (MAPE) of 12.66% and a coefficient of determination (R²) of 0.935, outperforming all benchmarks. Our model predicts that a 2°F increase in outdoor temperature could raise NYC building energy demand by 3.7% during the hottest summer months, while a 4°F increase could lead to an average rise of 7.6%, with some buildings experiencing up to 17.5%. This variability indicates that buildings respond differently to temperature changes. The methodology can identify inefficient buildings needing retrofitting, helping cities adapt to climate-driven energy demands.