Research on the generation method of outdoor design parameters for summer air-conditioning based on machine learning and dynamic time warping

Abstract

In order to combat global climate change and promote sustainable development, nations globally are advancing towards carbon neutrality. Building energy-saving is vital for carbon neutrality, crucially involving the alignment of air conditioning design parameters with local climates. Currently, standards around the world generally use a statistical non-guaranteed rate method for selecting air conditioning design parameters, neglecting the correlation between dry-bulb temperature (DBT) and wet-bulb temperature (WBT). This results in imprecise parameters, impacting system design and energy efficiency optimization. This study uses Support Vector Machine Regression (SVR) and Kernel Density Estimation (KDE) to model the complex relationship between outdoor meteorological parameters, improving air conditioning outdoor design parameters by integrating the nuanced relationship between DBT and WBT through the air state non-guaranteed rate. In addition, the same hourly variation coefficients are used across different regions in China, ignoring regional differences. The paper adopts the Dynamic Time Warping (DTW) algorithm to enhance design parameter accuracy by reflecting temporal climatic variations, thus optimizing air conditioning energy efficiency. The results show that this method provides more accurate outdoor design parameters than those derived from Chinese standards and the ASHRAE Handbook, thereby better representing the hourly changes in regional climates. The use of EnergyPlus simulations on a representative office building model clearly demonstrates that the cooling load on the design day is significantly reduced compared to the Chinese standard method, thereby confirming both the practical feasibility and the improved accuracy of the research method.