Black box modelling of HVAC system : improving the performances of neural networks

TL;DR

This paper presents a method to enhance neural network models of HVAC systems by using sensitivity and spectral analysis to optimize architecture and improve prediction accuracy, especially in dry coil conditions.

Contribution

It introduces a novel approach combining sensitivity analysis and spectral analysis to optimize neural network models for HVAC system prediction tasks.

Findings

Optimized neural network architecture for HVAC modeling

Improved prediction accuracy in dry coil conditions

Effective relevance assessment of input variables

Abstract

This paper deals with neural networks modelling of HVAC systems. In order to increase the neural networks performances, a method based on sensitivity analysis is applied. The same technique is also used to compute the relevance of each input. To avoid the prediction errors in dry coil conditions, a metamodel for each capacity is derived from the neural networks. The regression coefficients of the polynomial forms are identified through the use of spectral analysis. These methods based on sensitivity and spectral analysis lead to an optimized neural network model, as regard to its architecture and predictions.