Home // ADAPTIVE 2019, The Eleventh International Conference on Adaptive and Self-Adaptive Systems and Applications // View article

Modeling of Automotive HVAC Systems Using Long Short-Term Memory Networks

Authors:
Peter Engel
Sebastian Meise
Andreas Rausch
Wilhelm Tegethoff

Keywords: BEV; Applied Machine Learning; HVAC; LSTM; ANN;

Abstract:
Adaptive and fast-calculating HVAC and climate models are gaining increasing importance in the automotive development process. Physically motivated thermal models achieve high quality results, but have a disadvantage in terms of their computing speed due to their complexity. One possible approach for the fast and precise simulation of thermal systems is deep learning with artificial neural networks. This paper aims to determine the extent to which neural LSTM are suitable for modeling the complex dynamic behavior of vehicle air conditioning. For this purpose, a physical reference model of a passenger car air conditioning system including a vehicle cabin is set up in the simulation environment Dymola with the component library TIL Suite. Furthermore, a model structure of a LSTM-based deep neural network to map the dynamic thermal behavior correctly is proposed. For the purpose of training the ANN, the overall system has been broken down into subsystems. The subsystems are individually trained open-loop and then linked to form a closed-loop overall model. For evaluation purposes, models with the same model structure but based on feedforward network (FFN) architectures are implemented, trained and tested.

Pages: 48 to 55

Copyright: Copyright (c) IARIA, 2019

Publication date: May 5, 2019

Published in: conference

ISSN: 2308-4146

ISBN: 978-1-61208-706-1

Location: Venice, Italy

Dates: from May 5, 2019 to May 9, 2019