Steven R. Shaw

Dynamic models and model validation for PEM fuel cells using electrical circuits

This paper presents the development of dynamic models for proton exchange membrane (PEM) fuel cells using electrical circuits. The models have been implemented in MATLAB/SIMULINK and PSPICE environments. Both the double-layer charging effect and the thermodynamic characteristic inside the fuel cell are included in the models. The model responses obtained at steady-state and transient conditions are validated by experimental data measured from an Avista Labs SR-12 500-W PEM fuel-cell stack. The models could be used in PEM fuel-cell control related studies.

Transient event detection in spectral envelope estimates for nonintrusive load monitoring

This paper describes the theoretical foundation and prototype implementation of a power system transient event detector for use in a nonintrusive load monitor (NILM). The NILM determines the operating schedule of the major electrical loads in a building from measurements made at the electric utility service entry. The transient event detector extends the applicability of the NILM to challenging commercial and industrial sites. A spectral preprocessor for use in the transient event detector is introduced first. Then, the transient event detection algorithm is developed. The performance of the algorithm is illustrated with results from a prototype event detector. >

Nonintrusive Load Monitoring and Diagnostics in Power Systems

This paper describes a transient event classification scheme, system identification techniques, and implementation for use in nonintrusive load monitoring. Together, these techniques form a system that can determine the operating schedule and find parameters of physical models of loads that are connected to an AC or DC power distribution system. The monitoring system requires only off-the-shelf hardware and recognizes individual transients by disaggregating the signal from a minimal number of sensors that are installed at a central location in the distribution system. Implementation details and field tests for AC and DC systems are presented.

A dynamic PEM fuel cell model

This paper proposes a relatively simple, physically motivated, dynamic electrical terminal model of a proton exchange membrane (PEM) fuel cell. The dynamic model is obtained by extending a static current voltage description to include temperature dependence, and by dynamically modeling the temperature of the membrane. Model performance is validated using experimental data collected from a 500-W commercial PEM stack.

Identification of induction motor parameters from transient stator current measurements

This paper describes three methods for estimating the lumped model parameters of an induction motor using startup transient data. A three-phase balanced induction motor is assumed. Measurements of the stator currents and voltages are required for the identification procedure, but no measurements from the motor shaft are needed. The first method presented applies simple models with limited temporal domains of validity and obtains parameter estimates by extrapolating the model error bias to zero. This method does not minimize any specific error criterion and is presented as a means of finding a good initial guess for a conventional iterative maximum-likelihood or least-squares estimator. The second method presented minimizes equation errors in the induction motor model in the least-square sense using a Levenburg-Marquardt iteration. The third identification method is a continuation of the Levenburg-Marquardt method, motivated by observed properties of some pathological loss functions. The third method minimizes errors in the observations in the least-squared sense and is, therefore, a maximum-likelihood estimator under appropriate conditions of normality. The performance of the identification schemes is demonstrated with both simulated and measured data, and parameters obtained using the methods are compared with parameters obtained from standard tests.

Estimation of variable-speed-drive power consumption from harmonic content

Nonintrusive load monitoring can be used to identify the operating schedule of individual loads strictly from measurements of an aggregate power signal. Unfortunately, certain classes of loads present a continuously varying power demand. The power demand of these loads can be difficult to separate from an aggregate measurement. Variable-speed drives (VSDs) are industrially important variable-demand loads that are difficult to track non-intrusively. This paper proposes a VSD power estimation method based on observed correlations between fundamental and higher harmonic spectral content in current. The technique can be generalized to any load with signature correlations in harmonic content, including many power electronic and electromechanical loads. The approach presented here expands the applicability and field reliability of nonintrusive load monitoring.

Transient event detection for nonintrusive load monitoring and demand side management using voltage distortion

This paper describes a simple system that can be used for autonomous demand-side management in a load site such as a home or commercial facility. The system identifies the operation of individual loads using transient patterns observed in the voltage waveform measured at an electric service outlet. The theoretical foundation of the measurement process is introduced, and a preprocessor that computes short-time estimates of the spectral content of the voltage waveform is described. The paper presents several example measurements demonstrating the ability of the system to obtain estimates of the spectral content of the voltage waveform.

A Kalman-Filter Spectral Envelope Preprocessor

This paper presents a Kalman-filter approach for computing spectral envelopes of current waveforms for nonintrusive load monitoring on the electric utility. Spectral envelopes represent time-varying frequency content and phase of the current relative to the voltage. Thus, the techniques presented in this paper may be applicable to a variety of lock-in measurement and signal processing techniques. The lock-in and computational performance of the proposed method favorably compares to previous efforts. The performance is demonstrated with data from the field.

Comparison and Identification of Static Electrical Terminal Fuel Cell Models

Six steady-state fuel cell electrical terminal models are compared using experimental data from an Avista Laboratories SR-12 500 W proton exchange membrane (PEM) fuel cell. The paper begins by reviewing the physical effects in a fuel cell. The proposed electrical terminal models are introduced in terms of these physical effects, parameterized for identification and compared using the measured data.

Detection and Diagnosis of HVAC Faults via Electrical Load Monitoring

Detection and diagnosis of faults (FDD) in HVAC equipment have typically relied on measurements of variables available to a control system, including temperatures, flows, pressures, and actuator control signals. Electrical power at the level of a fan, pump, or chiller has been generally ignored because power meters are rarely installed at individual loads. This paper presents two techniques for using electrical power data for detecting and diagnosing a number of faults in air-handling units. The results from the two techniques are compared and the situation for which each is applicable is assessed. One technique relies on gray-box correlations of electrical power with such exogenous variables as airflow or motor speed. This technique has been implemented with short-term average electrical power measured by dedicated submeters. With somewhat reduced resolution, it has also been implemented with a high-speed, centralized power meter that provides component-specific power information via analysis of the step changes in power that occur when a given device turns on or off. This technique was developed to detect and diagnose a limited number of air handler faults and is shown to work well with data taken from a test building. A detailed evaluation of the method is presented in the companion paper, which documents the results of a series of semiblind tests. The second technique relies on physical models of the electromechanical dynamics that occur immediately after a motor is turned on. This technique has been demonstrated with submetered data for a pump and for a fan. Tests showed that several faults could be successfully detected from motor startup data alone. While the method relies solely on generally stable and accurate voltage and current sensors, thereby avoiding problems with flow and temperature sensors used in other fault detection methods, it requires electrical data taken directly at the motor, downstream of variable-speed drives, where current sensors would not be installed for control or load-monitoring purposes.