| Electric machinery is widely used in our daily lives. There are electric motors in our computers and appliances at home, in tools and robots in manufacturing plants, and in vehicles such as cars, trains and ships. The past few decades have seen major advancements in the field of variable speed drives, consequent to new technologies in power electronics. However, widespread use of such drives pollutes the power grid with undesirable harmonics which compromise the normal operation of sensitive devices such as computers and telecommunication systems. The impact of the variable speed drives on the power grid must therefore be analyzed with simulation tools. Moreover, designing high power drives could also benefit from a simulation tool that would allow prototyping of the drive's controller.; Our goal in this thesis is to develop such a tool, a fully digital real-time simulator dedicated to electric drives. Such a simulator would allow engineers responsible for the design of large drives to prototype the drive's controller and initially test it using a simulated power converter and machine. Such tests would require limited space and equipment, and could be carried out safely without any high power devices.; Our work is based on modeling the drive using the state variable approach. We first describe a method which automatically computes the state equations of any linear electric system, nonlinear components being simulated outside the state-space representation. The method is based on linear graph theory and uses matrix computations extensively. This task is performed efficiently in the Matlab environment. An original technique, used to update the state equations when a switching device changes state, is then described. We then describe a method which yields a unique state-space representation for the entire power stage of a drive. This method allows a simultaneous delay-free solution of all the drive's dynamic equations. An implementation of the trapezoidal integration rule, adapted for time-varying systems, is then described and compared to an integration method recently developed specifically for real-time simulation of stiff systems. Finally, the various methods discussed thus far are implemented in order to yield a real-time simulation of an industrial drive on a parallel computer. Good results are obtained with timesteps of the order of 60 μs, which includes interprocessor communications and inputs and outputs acquisition and conversion. |
