Energy efficiency model-based Digital shadow for Induction motors: Towards the implementation of a Digital Twin

The 4th industrial revolution requires the tracking and optimization of energy consumption using an intelligent energy management system (IEMS). Such a system needs accurate real-time energy information to operate industrial machines, where Induction Motors (IMs) represent 42.2% of the global energy consumption. This paper addresses the problems of data acquisition for Induction motors in the context of Industry 4.0 (I4.0) where precision, real-time constraints, and optimal operations play a major role in decision-making. A new vision of Digital Shadow (DS) is proposed for the energy efficiency (EE) of IMs. A hybrid model consisting of a data-driven model and a physics-based model is developed to represent the machine efficiency information in real-time (RT). The proposed method is developed based on a two-stage procedure. Firstly, the IM EE model is established by considering Stator joule losses, core losses, rotor joule losses, friction, and windage losses, in addition to the stray losses to create an improved model. This was established based on the double cage model where core loss resistance is added. Secondly, the machine’s losses and efficiency are visualized using the 8 electrical circuit parameters (ECP) from the double cage model in addition to the rated speed and test temperature. The parameters estimation in RT incurred complexities and required the use of Adaptive Neuro-Fuzzy Inference System (ANFIS)-based modeling. The proposed model was developed using Fuzzy Logic Toolbox and Neuro-Fuzzy Designer app from MATLAB by training Sugeno systems. 8 ANFIS MATLAB models are trained to estimate each of the 8 ECP from the double cage model by using standard inputs. The training and testing dataset is constituted by the experimental data of the ECPs were calculated using the FSOLVE function to solve the nonlinear system formed by 60 motors from 8 manufacturers’ data such as voltage, number of pole-pairs, rated output power, rated torque, current, starting current, maximum torque and the power factor. Finally, the proposed method is validated experimentally using a 1.5 KW, 400/230 V, 50 Hz squirrel cage induction motor (SCIM) for linear speed control. The EE, Torque, and losses are visualized through the proposed model using MATLAB/SIMULINK. The mean value of the RMSE for the eight estimated ECPs are 7.57e-05, and 5.73e-01 respectively for training and testing while the mean values of the MAE are 2.06e-05, and 3.79e-01 respectively for training and testing. The errors of the EE estimation w.r.t the measured EE at rated condition, are RMSE = 0.205, and MAE = 0.1671. These results show that the proposed model can be implemented in the industry to monitor the machine EE and its losses.