Energy Consumption Dynamical Models for Smart Factories Based on Subspace Identification Methods

Abstract

Given the need of implementing methodologies in industry for the reduction of the energy consumption costs, it is required to create modelling methodologies that, together with the use of new technologies, will allow identifying energy consumption models based on input-output data. These models will later be used to design a suitable model-based control strategy. In this paper, a subspace identification algorithm based on the RQ decomposition approach has been reported, which is both implemented and validated on a test-bench that emulates the energy consumption of an industrial machine within a manufacturing process. Subsequently, the resultant model fitting when using the proposed modelling methodology has been compared with different identification routines included into the MATLAB System Identification Toolbox™, showing, in general, better results for the proposed methodology in this paper, with up to almost 80% of fitting in some cases.

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

“…Previous works have already proposed a methodology for identifying energy models via subspace identification (Sub-ID) algorithms [20]. This methodology represents the machine as a multiple-input and multiple-output (MIMO) system; the inputs U ∈ U are activation/deactivation signals for each device, and the outputs S ∈ R l are the total instantaneous power consumption of the machine for each line.…”

“…This matrix performs as a regressive matrix that solves a least-squares problem, given a linear combination of subspace matrices that describe the data. The state-space matrices are computed through the decomposition of the subspace matrices, using reliable, widely known and available numerical algorithms [20]. The result is a discrete-time linear time-invariant (LTI) state-space representation,…”

“…The contribution of this paper is an online adaptive energy modeling approach in a NILM scheme for manufacturing machines, addressing the variability of real manufacturing systems and their impact on operational performance. This methodology is based on a system identification (SI) algorithm to build mathematical models based on data [20]. These models predict the dynamic behavior of energy consumption of industrial machinery, with high precision and are also fully applicable in the control theory.…”

Adaptive predictive control for peripheral equipment management to enhance energy efficiency in smart manufacturing systems

“…A previous work proposed a methodology for identifying energy models via subspace identification (Sub-ID) algorithms [27]. This methodology is oriented to the identification of multiple-input and multiple-output (MIMO) systems, given that a machine may have several devices to consider.…”