Stability, the tendency of a system to return to a given operating point following a perturbation, is a very important property in most engineered systems. In the context of supervisory control of spacecraft power systems, it is important that the controller select stable sets of operating points at which the system should operate. In this way, the controller can achieve the system control objectives while maintaining stable operation of the power system. In this study, an automated method to evaluate the small-signal stability of spacecraft power systems about given operating points is presented. Generalized power systems are regarded as interconnections of specific components. Each of these components has specific sets of parameters that correspond to its physical parameters and its operating condition, and the components are represented using component-level systems of differential algebraic equations (DAEs). The automated method described herein combines the DAEs for each of the components based on their interconnections, resulting in a system-level system of DAEs. The method locates the system equilibrium point by solving the system-level system of DAEs and performs linearization about the equilibrium point. This results in a linear small-signal ordinary differential equation model of the system about the equilibrium point. The stability of the linearized model is evaluated to determine the small-system stability of the power system about the equilibrium point. The method is demonstrated for several example systems. The results for these systems are presented along with a comparison to time domain simulation results. It is found that the automated approach is capable of determining the local stability properties of the demonstration systems correctly and efficiently.

Electric power distribution systems composed of power electronics converters are susceptible to instabilities under certain conditions. Small-signal impedance approaches to stability analysis are incapable of predicting large-signal stability properties. Herein, a practical and scalable genetic algorithm based procedure for the estimation of regions of asymptotic stability of power electronics systems is proposed. The procedure is demonstrated on six nonlinear models that range from 6 to 75 state variables. The models represent the dynamics of Naval power electronics-based system components and systems.

An Intermittent Megawatt Generator (IMG) has been designed by Innovative Power Solutions (IPS) to meet the needs of future directed energy loads on high-performance aircraft. These loads significantly impact the electrical, mechanical, and thermal performance of the generator, load, and aircraft. If representative simulation models of the generator and other important subsystems can be obtained, the impact on system performance can be analyzed and optimized before the generator is deployed. The objective of this work was to utilize various modeling techniques to obtain accurate electrical, thermal, and mechanical performance models of the IMG, and to apply these models to analyze dynamic response transients to sudden load changes as seen for directed energy loads. Additionally, the models have been used to optimize the IMG control to mitigate voltage transients during these load changes.

Immittance based methods are often used in the stability analysis of power electronics based systems. Because it is difficult and/or time consuming to develop average value models of some components, it is often desirable to extract immittance data from detailed simulations (simulations in which the switching of the power semiconductors is represented). Traditionally, this is accomplished by introducing a perturbation, extracting the fundamental component of the voltage and current waveforms at the perturbation frequency, from which the impedance at that frequency may be extracted using transform techniques. In this work, an alternate approach is suggested, which offers both reduced computational effort as well as increased accuracy.

This work sets forth a nonlinear average value model (NLAM) of a switched reluctance machine. NLAMs are characterized by state variables that are constant in the steady-state facilitating high-speed simulation and control analysis and design. The main difficulties in deriving such a model for the switched reluctance machine arise from nonlinearities due to magnetic saturation, absence of mutual inductance, and a nonsinusoidal self-inductance profile which prevents the use of a rotational transformation in arriving at a rotor position-invariant machine description. Herein, these difficulties are overcome by introducing a variable representation that approximates machine variables by the inner product of a vector of basis functions and a time-varying coefficient vector. A set of nonlinear differential equations is derived which governs the behavior of the coefficient vectors. These equations are rotor position invariant and feature state variables which are constant in the steady-state. The resulting model is experimentally validated.

The problem of estimating regions of asymptotic stability for nonlinear dynamic systems is considered as an optimization problem. Genetic algorithms are then proposed to solve the resulting optimization problems. Three test systems are used to evaluate the performance of the proposed genetic algorithms. The test systems are 6th, 8th, and 17th order nonlinear power electronics systems. The performance of the genetic algorithms are also compared with that of the classical Broyden-Fletcher-Goldfarb-Shanno (BFGS) algorithm and the simplex method of Nelder and Mead. Time domain simulations of the test systems are performed to validate the results of the optimization algorithms. Issues involved with the successful implementation of genetic algorithms to estimate regions of attraction are discussed. It is observed that genetic algorithms outperform the classical optimization algorithms in estimating regions of asymptotic stability.

This work sets forth a new switched reluctance machine model based on a current dependent inverse inductance representation of the flux linkage versus current and position characteristic. This form permits a particularly simple mathematical description. Based on this relationship, a complete state model of the device is derived. The resulting model is easy to implement and computationally efficient. In addition, both manual and genetic algorithm based methods of parameter identification are set forth. The accuracy of the model is tested via comparison to laboratory measurements of the machine's steady state voltage and current waveforms as well as torque-speed characteristics. The proposed model is shown to be quite accurate.

In this paper, methods for transitioning between different types of wound-rotor synchronous machines models are examined. The specific transitions investigated are (1) finite element to circuit-based machine models and (2) circuit-based to average value models of machine/rectifier systems. Good agreement between simulation results obtained using the different model types is achieved. This work sets the stage for a software environment that enables both component and system level analysis of electrical power systems.