Intelligent data systems for building energy workflow: Data pipelines, LSTM efficiency prediction and more

Abstract

The data collection process for thermal energy storage (TES) system is largely still and restricted to data collection only. This leaves a gap to study the transient state physical process of charge and discharge as it proceeds. In addition, these devices are restricted and cannot perform on spot model fitting, prediction and other data curation techniques. This paper demonstrates the application of intelligent data layer with neural networks for evaluating and predicting end to end performance of heat pump integrated stratified thermal energy storage (TES) system. The data modelling – acquisition, curation, and transformation is done in situ (dynamically). The key objectives are:•A method to demonstrate the application of data-layer framework to visualize in real-time energy efficiency of TES. To fit the second law of thermodynamics-based exergy model. This will help engineers to intuitively understand the energy efficiency of their devices using novel data pipeline.•To demonstrate the use-case of hyper-tuned DL framework of LSTM to predict the energy efficiency in the process loop.Predicted results show tuned correlation with the parametrized experimental data, even during the load phase, where substantial amount of math (convection/mixing) is present for the network to learn (train/test).