1 ANN-Based Optimal Energy Control of Induction Motor in Pumping Applications Osama S. Ebrahim 1 , Ali S. Algendy 2 , Mohamed A. Badr 3 , and Praveen K. Jain 4 Queen’s university 1,4 , Canada, Ministry of Irrigation and Water Resources 2 , Egypt, Ain-Shams University 3 , Egypt osama.bayoumy@queensu.ca 1 , mabadr05@yahoo.com 3 , praveen.jain@queensu.ca 4 Abstract- This paper investigates the opportunity for energy saving in a 3-phase induction motor (IM) driving pump load and proposes an improved loss model control (LMC). Compared with other power loss reduction algorithms for IM, the presented one has the advantages of fast response, high accuracy, and simplicity of implementation. The performance of LMC depends mainly on the accuracy of modeling the motor drive and losses. In this paper, a detailed loss-model for the IM drive has been developed. The model considers inverter voltage harmonics and magnetic saturation effects using closed-form equations. On that basis, an ANN controller is synthesized and learned offline to determine the optimal flux level that achieves maximum drive efficiency. Simulation and experimental studies are performed on 5.5 kW test motor using proposed control scheme. The test results are provided and compared with the fixed flux operation to validate the effectiveness. Index Terms----- Artificial Neural Network, efficiency optimization, induction motor drive, loss model control, PWM harmonic loss, and magnetic saturation. I. INTRODUCTION Rapid increases in energy prices and environmental pollution are recently having significant impacts on the community. It, therefore, becomes imperative that major attention be paid to improve system’s efficiency. The fluid pumping is perhaps the most common process in industrial plants and household applications. Using variable-speed electric motor drive (VSD) to control fluid flow instead of throttling valves (or bypassing) saves a substantial amount of energy. Besides, the VSD prevents fluid hammering phenomenon that causes severe mechanical stress in the pipe system, or even damage, by offering smooth start/stop pumping [1]. Induction motors (IMs) have the advantages of high reliability, ruggedness, and low cost of the machine manufacturing. On the other hand, advances in power switching devices and digital signal processors have significantly matured the voltage-source inverters (VSIs) with associated pulse width modulation (PWM) techniques. As a result, PWM-VSI fed IM has been well established as the foremost structure for the ac VSD systems. Power losses in the IM drive are greatly dependent on control strategies. Fast torque response is not a crucial requirement in the pumping applications. Therefore, IM drives in such plants are usually based on the scalar (V/f) control method as illustrated schematically in Fig. 1a [2]. In this method, the ratio of the stator voltage to frequency and hence machine flux are maintained constant as long as the speed is below rated. Although the method does satisfy application requirements, the constancy of the flux deteriorates motor efficiency, in particular at low speeds with partial load. ΙΜ PWM VSI Speed control ω r ref ω r + − ω e Vs ω s * (a) (b) ΙΜ PWM VSI Speed control ANN loss-model controller ω r ref ω r + − ω e Vs ω s * × R s Is + ψ s opt Fig. 1: IM scalar control. (a) Classical V/f control. (b) Proposed optimal energy control. In an effort to improve IM efficiency, various flux control methods have been developed. These methods can be broadly classified into two topologies; search control and loss-model based control (LMC). The basic principle of the search controller is to measure the input power and then iteratively search for the flux level (or its equivalent variables) until the minimum input power is detected for a given torque and speed [3]-[5]. Major drawbacks of the search controller are the slow convergence rate and flux/torque ripples. The LMC computes losses by using a machine model and selects an optimum flux level that minimizes the losses [6]–[17]. LMC approach is fast and does not produce torque ripples. However, the accuracy depends mainly on the correct modeling of the motor drive and the losses. For instance in [6]-[8], closed-form equations 2009 IEEE Electrical Power & Energy Conference 978-1-4244-4509-7/09/$25.00 ©2009 IEEE