Hydrogen as an energy carrier and fuel cell technology is seen as the most promising alternative in the medium term to achieve a sustainable and clean energy system. To achieve this goal various scientific and technological barriers have to be overcome. This project aims to contribute to the advance of this technology developing a device for optimal management and advanced control of Proton Exchange Membrane (PEM) fuel cells for different applications, both transportation and stationary applications. To achieve this goal both scientific and technological advances are necessary.
The first step to ensure success in the stated objective is to achieve adequate models for both process simulation, and for the adjustment and / or inclusion in advanced controllers that manage these processes. Dynamic models are a key step in any application of process control, more so in those cases where the process is multivariable and nonlinear as in the case of PEM fuel cells. In this way, the aim is to investigate nonlinear dynamic models using several techniques and to compare the results establishing the most appropriate methodology for this process. Several methodologies are intended to investigate: the dynamic models based on Partial Least Squares (PLS) models fitted with robust nonlinear multi-objective evolutionary techniques (MOEA). Additionally, to implement a predictive controller, predictors as FLAP (Large Fuzzy Ahead Prediction) will be investigate.
The next step after model development is to properly set specifications for the controllers, which requires a study of different types of applications and the correct evaluation of the requirements and restrictions. The types of applications to study are mainly mobile applications, in particular, transportation applications (cars, trains, ships, etc..) and stationary applications such as residential, distributed generation, etc.
The final scientific step is the development of advanced controllers and optimal management algorithms for all fuel cell subprocess and for energy transfers between the companion elements in each application. The techniques that will be investigated are the model predictive control adapted for each type of model or predictor for the development: Latent Variable Model Predictive Control (LV-MPC), use of FLAP in MPC and the MPC setting techniques based on multi-objective optimization.
Research in multi-objective optimization applied to all stages of the developments (modeling, control settings, process optimization, etc.) will close designer preferences to the final solution
The final technological step is to translate these algorithms to an embedded platform with industry format that can be incorporated into commercial systems.
KEYWORDS: PEM, Fuel Cell Control, Partial Least Squares, Embebbed control systems