Data-driven energy prediction modeling for both energy efficiency and maintenance in smart manufacturing systems

Journal Article (2022)







Doc link


Download the digital copy of the doc pdf document


The optimization and monitoring of the energy consumption of machinery lead to a sustainable and efficient industry. For this reason and following a digital twin strategy, an online data-driven energy modeling approach with adaptive capabilities has been proposed and described throughout this paper. This approach is useful in developing robust energy management systems that enhance the energy efficiency of industrial machinery. In this way, the dynamic behavior of their energy consumption is modeled without using phenomenological laws. In contrast, traditional methodologies hardly consider such dynamic behavior or use an exhaustive modeling process. The proposed approach includes an adaptive mechanism to consider the natural degradation of machinery. This mechanism is based on a concept drift detector, which detects when the current consumption of the machine is not correctly represented by the model estimation and adapts the model to account for these new behaviors. The concept drift detector has broad applicability in the face of reducing maintenance costs, measuring the impact and evolution of either abnormal behaviors (e.g., failures) or degradation, and identify which elements change. The proposed methodology has been validated in an industrial testbed. An experiment with three emulated concept drifts was carried out in the testbed. As a result, the proposed adaptive approach obtained more than doubled the t rate of the energy prediction/estimation compared to the non-adaptive model and successfully detected these changes in energy consumption.


automation, control theory, optimisation.

Author keywords

Non-intrusive load monitoring, Data-driven model, Subspace identi cation, Energy models, Concept drift, Digital twin, Gaussian mixture models, Energy eciency, Machine fault diagnosis

Scientific reference

M.A. Bermeo, C. Ocampo-Martínez and J. Diaz. Data-driven energy prediction modeling for both energy efficiency and maintenance in smart manufacturing systems. Energy, 238: 121691, 2022.