Search

In this work, a data-driven identification method, based on polynomial nonlinear autoregressive models with exogenous inputs (NARX) and the Volterra series, is proposed to describe the dynamic and nonlinear voltage and current characteristics of polymer electrolyte membrane fuel cells (PEMFCs). The structure selection and parameter estimation of the NARX model is performed on broad-band voltage/current data. By transforming the time-domain NARX model into a Volterra series representation using the harmonic probing algorithm, a frequency-domain description of the linear and nonlinear dynamics is obtained. With the Volterra kernels corresponding to different operating conditions, information from existing diagnostic tools in the frequency domain such as electrochemical impedance spectroscopy (EIS) and total harmonic distortion analysis (THDA) are effectively combined. Additionally, the time-domain NARX model can be utilized for fault detection by evaluating the difference between measured and simulated output. To increase the fault detectability, an optimization problem is introduced which maximizes this output residual to obtain proper excitation frequencies. As a possible extension it is shown, that by optimizing the periodic signal shape itself that the fault detectability is further increased. [ABSTRACT FROM AUTHOR]

This paper presents an innovative power hardware-in-the-loop (PHiL) development platform for railway pantograph testing. A novel real-time-capable finite-element catenary model is proposed in a train-fixed moving-coordinate formulation combined with an efficient absorbing boundary layer to accurately depict railway catenary dynamics in the region around the pantograph contact point. The complex catenary dynamics is accurately and efficiently modeled, including nonlinear effects like dropper slackening, and a model-predictive impedance controller realizes the task of accurately emulating the virtual catenary dynamics on a real-world pantograph test rig. • High-fidelity real-time-capable nonlinear finite element catenary model • Absorbing boundary control on a moving Euler-Bernoulli beam under axial pretension • Model-predictive impedance control for catenary emulation on a test rig • Validation of the proposed methods on a full-scale pantograph test rig [ABSTRACT FROM AUTHOR]

Abstract In this work, an online algorithm to estimate the electrochemical impedance spectrum of polymer electrolyte membrane fuel cells (PEMFCs) is presented. The electrochemical impedance, constituting the linear dynamic relation between current and voltage, is described by a discrete time filter model. The unknown model parameters are estimated from streaming time domain data using a recursive algorithm based on the generalized total least squares (GTLS) technique. This allows for flexible noise assumptions, leading to unbiased parameter estimates in the presence of noisy current and voltage measurements. Therefore, the robustness of the proposed method is significantly increased. Broad-band current excitation signals are utilized as to sufficiently stimulate the fuel cell for the online parameter estimation. The model order is selected using Akaike's information criterion (AIC), leading to a parsimonious model. Equivalent circuits are automatically generated using Foster's method of analog circuit synthesis. The applicability and benefits of the proposed algorithm regarding fuel cell diagnostics are demonstrated using measurement data of a small scale single cell (5 cm2 active area) subjected to a broad-band excitation. Highlights • Electrochemical impedance spectrum is described by a discrete time domain model. • An online estimation algorithm is derived based on generalized total least squares. • Equivalent circuits are synthesized from the discrete time domain model. • Validation is done using experimental measurements from a 5 cm2 single cell. • Time resolved impedance information during transient operations are obtained. [ABSTRACT FROM AUTHOR]

Pantograph current collectors, especially for high-speed trains, need to ensure safe contact with the catenary under stringent requirements on the dynamic contact force. A novel, high-dynamic pantograph test rig with accurate virtual catenary emulation is presented that allows for efficient, realistic, and reproducible testing. The complex dynamics of the pantograph/catenary interaction is modeled by a real-time-capable distributed-parameter description in moving coordinates. The proposed test rig controller incorporates model-predictive impedance control to match the desired catenary dynamics. Additionally, it keeps the exchange of the conserved quantities energy and momentum between the real pantograph and the virtual catenary consistent to increase physical trustworthiness of the results, even in high-dynamic test scenarios. The proposed methods are experimentally validated on the full-scale pantograph test rig. [ABSTRACT FROM AUTHOR]