A hybrid model of commercial building cooling load prediction based on the improved NCHHO-FENN algorithm

Abstract

Accurate building cooling load prediction is one of the key conditions for optimal operation, energy efficiency and sustainability studies of large central air-conditioning systems. In this paper, a novel model based on nonlinear chaotic mapping Harris Hawks Optimization (HHO) optimized Fully Elman Neural Network (FENN), called NCHHO-FENN, is proposed for building cooling load prediction. First, the entropy weight method is combined with the grey correlation method to generate evaluation indexes for feature extraction. Then, parameters of the Fully Elman Neural Network prediction model were optimized using the nonlinear chaotic Harris hawk optimization (NCHHO) algorithm, and the hybrid model for cooling load prediction was constructed. Finally, cool loading data from a sizable commercial building in Shaanxi Province was adopted as the research sample for validation. The experimental results show that the Root Mean Square Error (RMSE) and coefficient of determination (R-square) of the NCHHO-FENN prediction model are reduced and improved by 54.82%–6.64%, 49.20%–4.81% and 11.72%–0.46% respectively, compared with the three improved Fully Elman Neural Network models, which explains that harris hawks algorithm can not only be used to optimize the parameters of fully elman neural network but also has superior prediction effect. At the same time, in the case of a few training samples, NCHHO-FENN can still proficiently predict the cooling load in various months, which also shows that it has high generalization. It demonstrates that the proposed method is efficient and reasonable, and can be applied to building cooling load prediction, contributing to theoretical justification for building energy-saving operation and management.