Physics-Informed Neural Networks for Heat Pump Load Prediction

Viorica Rozina Chifu, Tudor Cioara (), Cristina Bianca Pop, Ionut Anghel and Andrei Pelle
Additional contact information
Viorica Rozina Chifu: Computer Science Department, Technical University of Cluj-Napoca, Memorandumului 28, 400114 Cluj-Napoca, Romania
Tudor Cioara: Computer Science Department, Technical University of Cluj-Napoca, Memorandumului 28, 400114 Cluj-Napoca, Romania
Cristina Bianca Pop: Computer Science Department, Technical University of Cluj-Napoca, Memorandumului 28, 400114 Cluj-Napoca, Romania
Ionut Anghel: Computer Science Department, Technical University of Cluj-Napoca, Memorandumului 28, 400114 Cluj-Napoca, Romania
Andrei Pelle: Computer Science Department, Technical University of Cluj-Napoca, Memorandumului 28, 400114 Cluj-Napoca, Romania

Energies, 2024, vol. 18, issue 1, 1-20

Abstract: Heat pumps are promising solutions for managing the increasing heating demand of residential houses, reducing the environmental impact when used with renewable energy. Accurate heat load predictions allow the heat pump to operate at the most efficient settings, maintaining comfortable temperatures while reducing excess energy use and lowering operating costs. Data-driven prediction solutions may have difficulty capturing the dynamics and nonlinearities of the thermodynamics involved. The physics-informed models combine the monitored observed data with theoretical knowledge of heat pumps and directly integrate physical constraints, allowing for better generalization and reducing the dependence on large volumes of data. However, they require detailed knowledge of the system topology and refrigerant parameters, which increases the model complexity. Therefore, in this paper, we propose a physics-informed neural network for predicting the heat load of heat pumps that integrates thermodynamics directly into the loss function of the neural network. We model the heat load as a function of the input variables, including the inlet temperature, outlet temperature, and water flow rate. We integrate the function during model training to reduce the model complexity. Our approach increases the accuracy of the predictions compared with data-driven models and generates prediction results that are consistent with the actual physical behavior of the heat pump. The results show superior prediction accuracy, with a 7.49% reduction in the RMSE and a 6.49% decrease in the MAPE, while the R 2 value shows an increase of 0.02%.

Keywords: heat pump; heat load; physics-informed neural networks; thermodynamics; prediction (search for similar items in EconPapers)
JEL-codes: Q Q0 Q4 Q40 Q41 Q42 Q43 Q47 Q48 Q49 (search for similar items in EconPapers)
Date: 2024
References: View references in EconPapers View complete reference list from CitEc
Citations:

Downloads: (external link)
https://www.mdpi.com/1996-1073/18/1/8/pdf (application/pdf)
https://www.mdpi.com/1996-1073/18/1/8/ (text/html)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:gam:jeners:v:18:y:2024:i:1:p:8-:d:1551222

Access Statistics for this article

Energies is currently edited by Ms. Agatha Cao

More articles in Energies from MDPI
Bibliographic data for series maintained by MDPI Indexing Manager ().