PEM fuel cell prognostics under variable load: A data-driven ensemble with new incremental learning

Proton Exchange Membrane Fuel cells (PEMFC) are one of the most promising fuel cell technologies, which qualify for variety of applications as power generation source. The Prognostics & Health Management of fuel cell is an emerging field, which is paving the way for large scale industrial deployment of PEMFC technology. More precisely, prognostics of PEMFC become a major area of focus nowadays that enables predicting the behavior of PEMFC to produce actionable information to extend its life span. This paper contributes the first application on data-driven prognostics of PEMFC stack under variable load for combined heat and power generation (μCHP). In brief, an ensemble structure of Summation Wavelet-Extreme Learning Machine models is proposed with a new incremental learning scheme, to achieve long-term predictions on stack state of health (SOH) and to give confidence for better decisions. The proposed prognostics model is validated on data from PEMFC stack used for a μCHP application under variable load profile for a complete year. A thorough comparison on SOH predictions results clearly shows the significance of proposed prognostics model, which can predict with few learning data for a long-term prognostics horizon around 650 hours with high accuracy and low uncertainty.