Dario Marra

Real-Time Estimation of Engine NOx Emissions via Recurrent Neural Networks

Abstract The paper focuses on the experimental identification and validation of recurrent neural networks (RNN) for virtual sensing of NO emissions in internal combustion engines (ICE). Suited training procedures and experimental tests are proposed to improve RNN precision and generalization in predicting engine NO emissions in transient operation. The reference Spark Ignition (SI) engine was tested by means of an integrated system of hardware and software tools for engine test automation and control prototyping. A fast response analyzer was used to measure NO emissions at the exhaust valve. The accuracy of the developed RNN model is assessed by comparing simulated and experimental trajectories of NO emissions for a wide range of operating scenarios, with an estimation error lower than 2 % throughout the test transients. The results evidence that RNN-based virtual NO sensor offers significant opportunities for improving the performance of SCR after-treatment devices.

Vehicle electrification: A further variable toward integrated intelligent energy systems

Electrification of transportation creates the premises for a strong interaction with the electric grid and the energy system management. While the need of fast and diffuse recharging clashes with the present infrastructure and topology of the grid, the possibility of modulating car recharge and, mainly, of managing the capacity of batteries in both electric vehicle and plug-in hybrid vehicles within the V2G framework offers opportunities, creating a bridge between mobility, electric grid and energy systems. The paper offers a large view on these new trends, with emphasis on the Demand-Response (DR) management systems and on the possible impact on vehicle and powertrain control.

Models for Diagnostic Applications

The correct operation of an SOFC system is ensured by combining optimal design and effective control and diagnostic strategies, to guarantee system efficiency and prevent excessive degradation or undesired faulty states. In this way, system lifetime can increase and market requirements be fulfilled, with a consequent growth in SOFC systems production and market deployment. The aim of a diagnostic algorithm is to detect and isolate undesired events (i.e., faulty states) within the entire system (i.e., stack and ancillaries). During faulty operation, the inference on the system state can feed suitable control strategies in order to drive the system toward a safer operating condition, ensuring in such a way a continuous operation to the final user. The current chapter gives an overview on the development of a suitable diagnostic algorithm, based on a model-based approach. The main features are illustrated and discussed, with focus on the dominant issues to be addressed for their optimal design. The background on model-based diagnosis is summarized along with the basic concepts of diagnostics. Details on the theory behind are available in the main references reported throughout the chapter. Several applications dedicated to an SOFC system are presented to exhibit the diagnostic algorithm capability of suitably detecting and isolating different kinds of faults.

Models for Control Applications

Real-world deployment of SOFC systems entails developing suitable control strategies, which particularly have to guarantee meeting electrical load demand, while limiting as much as possible thermal stresses for ceramic components. In this way, undesirable excessive degradation can be prevented and, in turn, longer lifetime can be achieved. Therefore, the main targets are to control the operating load and manage air and fuel inlet flows so as not to induce severe thermal gradients across fuel cell length, as well as to reduce temperature derivative during both cold-start and shutdown phases. Of course, such control goals are to be pursued taking into account the final application of the SOFC system, depending on which load demand fluctuations considerably vary (e.g., compared to stationary generation, transportation applications exhibit more fluctuating load demand). Therefore, depending on how much articulated is the designed SOFC system, which can particularly include hybridizing components (e.g., batteries and fly wheels) to enable limited power rate operation of the SOFC stack, different control levels must be developed to ensure desired control targets be appropriately met. The current chapter initially focuses on the analysis of the physical relationship between main control and controlled variables, depending on which the multilevel control structure can be appropriately defined. Then, specific analyses are presented and discussed to demonstrate the great potential offered by the model-based approach, to ensure appropriate control strategies be developed for on-field energy-efficient and safe operation of SOFC systems.

A One-Dimensional Modelling Approach for Planar Cylindrical Solid Oxide Fuel Cell

In this paper a 1-D steady-state model of a planar cylindrical Solid Oxide Fuel Cell (SOFC) is described. The SOFC 1-D model developed has been applied for both co-flow and counter-flow configurations. The computational domain selected is a symmetrical single cell slice with an angle of twenty degrees (i.e. one eighteen of the entire cell). The cell has been divided into computational units in the radial direction, for each of them energy, mass and electrochemical conservation equations have been solved. The cell is considered non-adiabatic with heat conduction inside the solid material and convective-radiative heat transfer mechanism between the outer section and the surrounding gases. Moreover, at the cell outlet the residual fuel mixes with the surrounding gases and is completely burnt (afterburning).The 1-D model has been verified making use of literature data generated from 3-D model of a planar cylindrical SOFC. The results obtained confirmed the good performance of the model developed and its applicability in a computational framework for the development of either control or diagnosis algorithm.Copyright