Abstract: The article presents methods for synthesizing a stator flux estimator for an induction machine using artificial neural networks (ANN). The requirements for control characteristics imposed on designers of electric car drives necessitate the use of control algorithms that allow for shaping the response of the electromagnetic torque generated in the drive machine.The applied control methods FOC (Field Oriented Control), DTC (Direct Torque Control), and
DTC-SVM (Space Vector Modulated DTC) use information about the position of the machine’s magnetic fluxes. One way to estimate these fluxes is to build a mathematical model based on artificial intelligence methods. Recently published proposals for new methods of creating a base of input signals for such an approximator prompted the authors to compare their effectiveness in relation to a specific control system. Using the DTC-SVM system as an example, algorithms using delay lines, expert filters [2], and a method for orthogonalizing ANN input signals [4] were compared.
1. INTRODUCTION
The currently growing interest in electric or hybrid-powered cars creates new areas for the mass use of variable-speed drives. Requirements regarding dynamic properties while maintaining high reliability, compactness, and low cost pose new challenges for designers and engineers. At the same time, control algorithms for various types of motors are being improved. The synthesis of modern control systems for converter drives often requires information about the position of the machine’s magnetic flux vector. In a standard induction machine, this quantity is not measured and requires the implementation of an algorithm for its reconstruction. Examples of mass-produced induction traction drives for cars include, among others,
REVA, AC Propulsion (tzero), and Tesla Motors.
State variable reconstruction algorithms enabling the implementation of drive systems with direct torque control (DTC) or field-oriented control (FOC) are a constant subject of research. A number of proposed solutions use algorithms from the field of artificial intelligence. (…)
The article contains 25,998 characters.
Source: Logistyka Magazine
Neural magnetic flux estimator of a drive machine for an electric car – PDF file for download