Experimental data driven neural-network prediction model of charge/discharge duration in phase-change thermal energy storage

One approach to mitigating intermittency of solar photovoltaic (PV) energy production is converting it to heat and storing it in a phase change material (PCM)-based thermal energy storage (TES) system. Optimizing these integrated systems (e.g., PV, heat pumps, and PCM-TES) for residential heating is needed. Previous research focused primarily on system-level numerical simulations and data-driven approaches for outlet fluid temperature prediction, neglecting key operational metrics such as charging and discharging time, which are critical for effective control, scheduling, coordination with variable PV, and residential energy management. To address this gap, this study provides: (i) a machine-learning (ML) approach to develop a black-box model predicting charging/discharging time based on inlet fluid temperatures and mass flow rates, trained using high-resolution datasets from commercial thermal batteries (SunAmp Thermino 80i); (ii) experimental testing of the same PCM-TES in a laboratory setting to evaluate unit efficiency and inlet/outlet temperature as a function of time; (iii) validation of model performance against laboratory results; and (iv) comparison with standalone base ML model. The hybrid prediction model employs both Multi-Layer Perceptron (MLP) regression and eXtreme Gradient Boosting (XGBoost) regression to predict charging/discharging duration. The model is configured for both sensor-based and sensor-less input configurations. Results show the model accurately captures non-linear thermal dynamics, with RMSE below 1% for both configurations. Experimental testing yielded an average round-trip efficiency of 0.87. The validated model demonstrates absolute error < 1%, confirming practical prediction performance and reduced RMSE % compared to base model. The proposed framework is suitable for integration into simulation software for scheduling PCM-TES charging based on solar energy availability and electrical tariffs in residential applications.

Authors

Institutions

Publication Details

Journal
Journal of Energy Storage
Published
2026-11-16
DOI
https://doi.org/10.1016/j.est.2026.125699
Primary Topic
Phase Change Materials Research
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Experimental data driven neural-network prediction model of charge/discharge duration in phase-change thermal energy storage

Joshua M. Pearce, Shafquat Rana, Amanullah Quamer
2 links
Journal of Energy Storage
Phase Change Materials Research
article

Experimental data driven neural-network prediction model of charge/discharge duration in phase-change thermal energy storage

Joshua M. Pearce, Shafquat Rana, Amanullah Quamer
article en

Abstract

One approach to mitigating intermittency of solar photovoltaic (PV) energy production is converting it to heat and storing it in a phase change material (PCM)-based thermal energy storage (TES) system. Optimizing these integrated systems (e.g., PV, heat pumps, and PCM-TES) for residential heating is needed. Previous research focused primarily on system-level numerical simulations and data-driven approaches for outlet fluid temperature prediction, neglecting key operational metrics such as charging and discharging time, which are critical for effective control, scheduling, coordination with variable PV, and residential energy management. To address this gap, this study provides: (i) a machine-learning (ML) approach to develop a black-box model predicting charging/discharging time based on inlet fluid temperatures and mass flow rates, trained using high-resolution datasets from commercial thermal batteries (SunAmp Thermino 80i); (ii) experimental testing of the same PCM-TES in a laboratory setting to evaluate unit efficiency and inlet/outlet temperature as a function of time; (iii) validation of model performance against laboratory results; and (iv) comparison with standalone base ML model. The hybrid prediction model employs both Multi-Layer Perceptron (MLP) regression and eXtreme Gradient Boosting (XGBoost) regression to predict charging/discharging duration. The model is configured for both sensor-based and sensor-less input configurations. Results show the model accurately captures non-linear thermal dynamics, with RMSE below 1% for both configurations. Experimental testing yielded an average round-trip efficiency of 0.87. The validated model demonstrates absolute error < 1%, confirming practical prediction performance and reduced RMSE % compared to base model. The proposed framework is suitable for integration into simulation software for scheduling PCM-TES charging based on solar energy availability and electrical tariffs in residential applications.

Journal of Energy StorageVol. 183
Lovely Professional University (IN), Western University (CA), Ivey Business School (CA)
Natural Sciences and Engineering Research Council of Canada
Affordable and clean energy, Industry, innovation and infrastructure, Climate action
Phase Change Materials Research
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon