Browse Prior Art Database

Fuel cell polarization curve estimation using self-organizing neural network

IP.com Disclosure Number: IPCOM000127767D
Publication Date: 2005-Sep-13
Document File: 1 page(s) / 279K

Publishing Venue

The IP.com Prior Art Database

Abstract

The measure of performance most often associated with fuel cells is a polarization curve, which describes the output voltage over a range of current densities for a specific set of operating conditions. Presently implemented and widely accepted method for characterization of the fuel cell stack performance is polarization curve (stack voltage vs. stack current). The proposed solution utilizes self-organizing neural network that is fed data from the dynamic drive, and therefore does not require facilitated test under strict conditions. The self-organizing neural network "unfolds" itself in the Voltage-Current plane and based on the dynamic data provided, it takes shape of the polarization curve. This would enable easier and less time consuming polarization curve acquisition, leading to the easier fuel cell performance degradation monitoring.

This text was extracted from a Microsoft Word document.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 71% of the total text.

Fuel cell polarization curve estimation using self-organizing neural network

The measure of performance most often associated with fuel cells is a polarization curve, which describes the output voltage over a range of current densities for a specific set of operating conditions.

 

Presently implemented and widely accepted method for characterization of the fuel cell stack performance is polarization curve (stack voltage vs. stack current). This polarization curve depends on several parameters (temperatures, pressures, fuel and air flows). The method that is used for polarization curve estimation consists of running a warm fuel cell stack according to the pre determined load profile,  set stack current, wait for certain amount of time (usually minutes) for voltage to stabilize, record the voltage and current, and move to the next load point. This procedure is somewhat difficult to implement on the vehicle. Even in the controlled environment, achieving a fixed load points can be difficult due to the interaction of the other systems (fans may start or stop, cooling pumps may go on or off, etc).

                                        

Traditional Polarization Curve Acquisition Algorithm                                  Proposed Polarization Curve Acquisition Algorithm

The proposed solution utilizes self-organizing neural network that is fed data from the dynamic drive, and therefore does not require facilitated test under strict conditions. The self-organizing neura...