Yang H.-HHuang K.-CChen W.-TSY-YEN KUO2022-04-252022-04-25202122195491https://www.scopus.com/inward/record.uri?eid=2-s2.0-85123173992&doi=10.23919%2fEUSIPCO54536.2021.9616005&partnerID=40&md5=0a24f7be9d88a5ce488f5723d28823c2https://scholars.lib.ntu.edu.tw/handle/123456789/607311Modeling the dynamics of the induction motor is a crucial problem because induction motors are used widely in several scenarios. However, it is difficult to model the dynamics of the induction motor precisely, because the induction motor system is modeled as the complicated high order non-linear differential equation. To address this problem, we propose a novel residual grid network. The proposed grid connection effectively merges the various levels of feature information. Moreover, previous methods are usually based on complex network architecture with a mass of parameters. It may be infeasible for deploying this application on edge devices in real-world scenarios. Therefore, in the proposed method, we introduce the lightweight strategy with grid connection to reduce the number of parameters. Experimental results show that the proposed network contains fewer parameters but outperforms other existing models and achieves state-of-the-art performance on both simulated and real-world motor data. ? 2021 European Signal Processing Conference. All rights reserved.Grid connectionLightweight modelMotor dynamicsResidual blocksComplex networksDifferential equationsDynamicsNetwork architectureSignal processingElectrical induction motorsGrid connectionsGrid networkHigh-orderHigher-orderInductions motorsMotor systemsResidual blockInduction motorsconference paper10.23919/EUSIPCO54536.2021.96160052-s2.0-85123173992