This paper compares three numerical classification algorithms in terms of their accuracy in performing classification for pre-screening battery cells as part of a pre-acceptance qualification program: support vector machine (SVM), linear discriminant analysis (LDA), and principal component analysis (PCA) pre-processing followed by LDA (PCALDA). The paper augments previous research which examined only one classifier, the simple generalized classifier (SGC). Key findings of the paper are that the SVM, LDA, and PCALDA all outperformed the SGC in terms of overall classifier accuracy in a numerical study. The PCALDA is shown to yield the greatest overall classifier accuracy (97%), while LDA is shown to give the greatest accuracy (98%) in identifying failure-prone cells. Based on these observations, the SVM, LDA and PCALDA methods are potentially promising candidates to perform battery cell pre-screening classification.

This paper describes an experimental and numerical optimization procedure for off-line extraction of parameters for unsymmetrical, single-phase induction machines with capacitor-start operation. In addition to permitting asymmetry between phase winding parameters, the approach requires only average and rms measurements as inputs; electrical phase angle measurements are not required. The method is validated on an experimental motor with transient and steady-state simulations, using fitted parameters obtained using the proposed approach.

Mobile microgrid generator systems can provide power to electrical loads during grid outages and for off-grid applications. These systems are often configured using conventional generator sets, but can also be used with parallel energy storage. The addition of energy storage may provide advantages in terms of power quality and emissions. This paper introduces a procedure to experimentally assess mobile microgrid generator systems operating on natural gas (NG) fuel in conventional and energy storage coupled types, for several power and environmental emissions metrics: thermal efficiency, voltage and frequency stability, harmonic distortion, and air pollution from total hydrocarbons (THC), carbon monoxide (CO), nitrogen oxides (NOx), and carbon dioxide (CO2). The analysis of thermal efficiency and detailed gas composition and engine emissions analyses are described. Also included is a method of synthesizing realistic load profiles for laboratory testing, based on statistical sampling of metered load data. Implementation of the proposed test procedure is experimentally demonstrated on two types of mobile microgrid generator systems of differing engine sizes: a conventional 22 L NG generator set and a hybrid system consisting of an 11 L NG generator set, in parallel with battery energy storage. Significant differences in power quality, fuel usage and emissions metrics were observed between the two systems using the procedure. Measurements using the procedure are also used as inputs to an example economic and environmental cost analysis for a remote microgrid design. These findings suggest the potential usefulness of the procedure for evaluation of competing generator sizes and configurations, which can also provide input to mobile microgrid designs.

This paper introduces an approach for pre-screening manufactured batteries before system deployment, with the goal of reducing higher lifecycle maintenance costs attributed to failure-prone batteries. The method employs a pattern recognition algorithm for classifying parts for acceptance or rejection. The paper describes the classification algorithm and demonstrates its performance on example pre-acceptance test data. Using an example system maintenance concept, the economic and operational impacts of the classification are also demonstrated.

This paper compares several electrical load models for estimating the efficiency of DC vs. AC distribution in micro grids. Candidate models include energy balance, harmonic power flow, and time-domain modeling. Model results are compared with numerical studies and validated with experimental measurements. Based on quantitative and qualitative considerations, the most appropriate load modeling approach for larger-scale DC distribution efficiency studies is proposed.

This paper describes a detailed nonlinear electromagnetic model for predicting current in transformers commonly used to supply electrical power in rural arc welding applications. Current-limiting in these welders during quenched arcs is achieved using a gap-less core design, resulting in high-magnetizing impedance but also a significant degree of magnetic saturation. To accurately predict transformer currents during arc welding, the proposed model includes nonlinear magnetization effects. Parameter identification of the magnetic properties of the transformer core is performed using a population-based search algorithm using open-circuit transformer measurements. The model is validated using measurements on an experimental transformer during live arc weld testing.

While existing pump testing standards assume fixed voltage power supplies, solar pumps are supplied by DC voltage from photovoltaics. In lower-cost implementations of these systems, a lack of voltage or current regulating power electronics results in performance which is affected by the inherent current vs. voltage profile of the photovoltaic modules; this behavior must therefore be emulated during system performance evaluations. This paper describes an experimental method for estimating the efficiency of unregulated solar irrigation pumps and validates the method using an experimental testbed.

We present a new field-extrema hysteresis loss model (FHM) for high-frequency ferrimagnetic materials, along with a parameter identification procedure. The model does not involve solving an ordinary differential equation (ODE) and is asymmetric in that it works well under dc bias conditions. In the proposed model, the loss calculations are based on the extrema values of the fields. The model includes the effects of magnetic saturation as well as frequency effects. The model is comparable in accuracy to the ODE-based Jiles-Atherton model, but retains the convenience and computational efficiency of an empirical model. We demonstrate a procedure to characterize the model parameters using the Jiles-Atherton model. We compare magnetic hysteresis loss calculated by our new model with a full time-domain solution, as well as an empirical model, for a sample high-frequency ferrite. We demonstrate the use of the model, and validate the model, by calculating magnetic loss in an EI core inductor operating as the filter inductor in a buck converter. The model and identification procedure are being endorsed as a useful framework for computing magnetic loss in the context of automated magnetic device design.

Automated, population-based design methods are becoming increasingly popular as design tools. When using these algorithms (e.g., Monte Carlo, genetic algorithms, particle swarm optimization), it is common to require the analysis of 10,000-1,000,000 individual designs. Unique challenges arise when using population-based methods to design magnetic components. Whether the components are to be used in stand-alone applications or in the context of a larger system, several modeling and computational difficulties must be overcome. First, detailed knowledge of intrinsic magnetic material characteristics that will comprise the component must be known. Second, considerable effort is required to obtain an accurate 3D magnetic model that enables rapid calculation of quasi-static magnetizations within the component, as well as accounting for all leakage flux. Third, calculation of hysteresis losses, including frequency effects, in a computationally efficient way is required. This thesis presents several key developments that have been made in the modeling and automated design of ferrimagnetic inductors. A novel procedure was developed to obtain the magnetic characteristics of a ferromagnetic material from commonly available toroidal cores and a new test configuration. A high fidelity equivalent circuit model (HFMEC) was developed that rapidly and accurately predicts the anhysteretic flux vs. current characteristic of an inductor based exclusively on material characteristics and geometry. A static hysteresis loss model (SHM) was then developed that eliminates the need to perform time-domain simulations to calculate magnetic power loss in both zero-bias and dc bias applications. Automated design of a stand-alone inductor, including magnetic core loss, has been performed with the use of the HFMEC, SHM, and a genetic algorithm.

In this paper a procedure is set forth to perform detailed inductor designs using population-based optimization methods. The inductor designs are chosen based on design inputs-i.e., the inductor geometry and core material. The method incorporates recent research in magnetic characterization and advanced inductor modeling. Several case studies are presented for both single and multiobjective optimizations.

Evolutionary design refers to the use of evolutionary computing methods in the design process. Normally, this entails the formulization of the design problem as an optimization problem, which is solved using evolutionary techniques such as a genetic algorithm or particle swarm optimization. This paper provides three examples in the use of this highly effective method to the design of electromagnetic and electromechanical devices.

In this paper, we propose an improved method of characterizing highly permeable magnetic materials. The method is experimentally simple, inexpensive, and more accurate than that in IEEE Standard 393-1991. Our method places special emphasis on determining magnetic characteristics during saturated conditions.

Permanent magnet synchronous machines can be designed to obtain high efficiency and high torque density. Recently, there has been intense interest in the use of genetic algorithms (GAs) to design either part or all of the machine. In this work a highly structured approach to PMSM design which encompasses the machine and the machine control is considered. The number of parameters to be determined is over twice that of most previous work. The procedure set forth is explored for both single and multi-objective optimizations

We present a high-fidelity magnetic equivalent circuit (HFMEC) inductor model that reduces the inaccuracies associated with a traditional MEC approach. The model can accurately predict the flux linkage versus current characteristic in a fraction of the time needed for finite-element analysis. The accuracy, computational efficiency, and simple inputs (consisting of only geometry and material specifications) make the model ideal for automated inductor design.