Forecasting Residential Building Heating Load with an Innovative Gaussian Process Regression Method

Abstract

Effectively controlling the heating load (HL) in residential buildings is a vital component of energy conservation and sustainability. This abstract presents a new methodology for predicting HL by incorporating Gaussian Process Regression (GPR) and harnessing the power of two groundbreaking optimization techniques: the Population-based Vortex Search Algorithm (PVS) and the Flow Direction Algorithms (FDA). GPR stands out as a robust machine learning algorithm renowned for its capacity to grasp intricate data relationships. Combining these mentioned optimizers with the GPR model results in a hybrid strategy that harnesses the unique advantages of each element. PVS and FDA are utilized to optimize the GPR’s parameters, thereby elevating its predictive precision. The amalgamation of GPR, PVS, and FDA surpasses conventional techniques and even standalone GPR models regarding predictive precision and convergence velocity. This methodology offers a pragmatic and efficient approach to enhancing the forecast of HL in residential buildings, consequently aiding in better energy management and mitigating environmental impact. The hybrid GPPV model distinguishes itself with its exceptional accuracy when compared to alternative proposed models. Boasting a low RMSE of 1.013 and a R 2 value of 0.990 , GPPV attains the highest performance level. Furthermore, this research paves the way for the exploration of employing nature-inspired optimization techniques alongside neural networks to address a wide array of intricate challenges. The combined influence of GPR and these inventive optimizers highlights the capacity of hybrid models to tackle practical, real-world issues.

Keywords

• heating load
• gaussian process regression
• population-based vortex search algorithm
• flow direction algorithms