Three autoregressive-neural network hybrid models for energy efficiency estimation of induction motors. Issue 1 (7th January 2019)

• Record Type: Journal Article
• Authors: Sertsöz, Mine; Fidan, Mehmet; Kurban, Mehmet
• Is Part Of: Compel. Volume 38:Issue 1(2019)
• Page Start: 431
• Page End: 451
• Publication Date: 2019-01-07
• Subjects: Energy efficiency -- Estimation -- Autoregressive model -- Back-propagation -- Induction motors; Electrical engineering -- Data Processing -- Periodicals; Electrical engineering -- Mathematics -- Periodicals; Electrical engineering -- Periodicals; Electronics -- Data Processing -- Periodicals; Electronics -- Mathematics -- Periodicals; 621.3
• DOI: 10.1108/COMPEL-02-2018-0093
• Languages: English
• ISSNs: 0332-1649

Abstract

Abstract : Purpose: Improvements on the energy efficiency of the induction motors bear on not only these motors but also on the whole industry as a result of preference of these types of motors. In recent projects, energy efficiency of the induction motors is approaching to 90 per cent. The first necessary condition of the efficiency improvements is an accurate estimation of energy efficiency. This study aims to estimate the energy efficiency of induction motors by using three innovative estimation methods. Design/methodology/approach: Data for 307 motors were taken from three different companies and their torque, power, power factor and speed data were used. Three hybrid models were created by estimating the error of three autoregressive (AR)-based efficiency estimation models with the back-propagation artificial neural network (ANN) structure. In these proposed hybrid models, the AR models were supported with artificial neural networks to obtain a minimum estimation error. These three hybrid models were called as AR1-ANN, AR4-ANN and residual-ANN. Findings: Without hybridization of AR models by back-propagation ANNs, the best estimation result was obtained by residual model. On the other hand, for the proposed hybrid models, the best estimation was obtained by AR1-ANN, followed by AR4-ANN and finally the residual-ANN according to ME values. Practical implications: Proposed AR-ANN hybrid models relieve of longtime experiments for the energy efficiency measurement of induction motors. Furthermore, these AR-ANN models give more accurate results than the available methods in the literature. Engineering value of this research is three different issues in finding energy efficiency. The first one is minimizing of the test cost, the second one is no requirement the test equipment and the third one is not interrupting the motor. Every company that needs motors can use these estimation methods due to the advantages. Originality/value: Novel three AR-ANN hybrid models for energy efficiency estimation were studied. These novel methods give better response than the other methods which were used for estimation of induction motors in the literature.