ONE-DAY WORKSHOP: Monitoring, Diagnostics and Control for FC. 4 July 2017 Lucerne (CH) KKL - European Fuel Cell Forum 2017

Transcription

1 ONE-DAY WORKSHOP: Monitoring, Diagnostics and Control for FC Improving fuel cells performance through innovative diagnosis and control 4 July 2017 KKL - European Fuel Cell Forum 2017 Diagnostics and Control for FC Motivations, challenges and main issues Cesare Pianese, Pierpaolo Polverino University of Salerno (UNISA)

2 2

3 Outlines Motivation and objectives; The challenge; Advanced Control and Diagnosis; Algorithm design; On-board implementation issues; Conclusions 3

4 H 2 /FC Systems Integration Reference: 4

5 Motivations & Objectives TOWARDS FULL WORLDWIDE MARKET DEPLOYMENT Reference: US DOE, 2015 Fuel Cell Technologies Market Report (2016) To fulfil FC worldwide deployment the following objectives should be pursued: Higher reliability, durability and availability; Lower production, operation and maintenance costs. DURABILITY INCREASE CURRENT STATUS COSTS REDUCTION 5

6 Objectives breakdown Increase reliability, durability and availability Use innovative materials; Improve stack and system performance; Reduce fault/failure occurrence; Avoid detrimental operations; Maintain functionality under abnormal states. Reduce CAPEX, OPEX and maintenance costs Reduce materials amount and cost; Improve production process efficiency; Reduce quality test costs; Improve overall system efficiency and cost; Implement predictive maintenance. The highlighted objectives can be tackled trough proper diagnostic and control strategies 6

7 The Challenge As the systems become more complex and the performance objectives increase, a better integration of modeling, control and experiments is mandatory. This is a critical step towards system deployment. A multidisciplinary view is required to merge: Control techniques Models methodologies are available. new approaches are required and can be brought from other areas (e.g. planes, cars). Experiments new way of performing characterization and identification via, e.g., statistics (DOE). Testing Software-/Hardware-In-the-Loop (SIL/HIL) 7

8 Advanced Control and Diagnosis FUEL CELL SYSTEM OFFLINE Actuators System Sensors ONLINE Control algorithm actions Controller Data Acquisition Processing raw data For instance: OV measurements VI curves EIS spectra CV diagrams RUL treated data Inferential tool Lifetime estimation Features extraction Features generators degradation features Degradation models Features extraction Features analysis Inferential algorithm LIFETIME ESTIMATION fault DIAGNOSTIC ALGORITHM 8

9 Algorithm design: the V-cycle The ever-increasing complexity of modern technologies drives the development of sophisticated diagnostic and control system architectures based on models and expert knowledge. Model-based algorithms design provides a time and cost-effective approach to development of simple or complex monitoring, diagnostic and control systems. Project Definition The figure illustrates the embedded algorithm V Diagram often used to describe the development cycle. System definition Modelling & Simulation Rapid Prototyping Verification and Validation Targeting Time System Testing Hardware in the Loop Software in the Loop Project Test and Integration 9

10 Algorithms for Design: Control strategies Diagnostic strategies Prognostic strategies Material/shape Optimization Inverse design (not yet exploited for FCs) White/Grey-Box Models: Mean Value Modelling (e.g. ODE) CFD (e.g. PDE: FV, FD, FE, VOF) Molecular & Atomistic Main focus on accuracy On-board uses: Real time monitoring Embedding in control algorithms Embedding in diagnostic algorithms Embedding in prognostic algorithms On-line working conditions optimization Focus on computational speed Design Oriented models can be exploited to supply data/parameters to Application Oriented models Experiments for model validation Grey/Black-Box Models: Regressions Neural Network & Fuzzy Logic Mean Value Modelling (e.g. ODE) 1-D (e.g. PDE) Experiments for models/parameters identification and model validation Hierarchical Approach 10

11 Hierarchical approach Issues: Accuracy, Experiments, Implementability, Computational Speed Solution: Hierarchical Approach and Hybrid Modelling Hierarchical Modelling Physical Input Complexity Computational speed Hybrid Model Models contamination may exploit the feature of each approach to maximize the results Black-box Ident. test # data Output Fast algorithm Number and type of simulated phenomena Fuel cell stack and balance of plant features; Control and diagnostic strategies; etc. Single cell and components (i.e. GDL, GFC, CL and membrane) features and behaviour; etc. System level Cell level Experimental activity easy to perform Detailed physical representation Electrochemical and transport properties; degradation phenomena; etc. Atomic scale phenomena (e.g. adsorption and desorption, activation energies, etc ). Micro-scale level Atomistic level 11

12 On-board measurement issues Conventional Operating variables (e.g., V, I, T, P, etc.) are applied for conventional system monitoring, control and diagnosis. Advanced Improve performance, reliability & maintainability via advanced algorithms based on non-conventional measures (i.e., V-I, EIS, CV). Non conventional measures requires a coordinated control of stack and converter: let the battery/grid compensate for the user load transients and the FC load change; keep optimal FC operations (T, P, etc.) control V & I values to obtain: VI fix one couple of VI values, repeat several times to have a representative part of the VI curve (steady/sweep); EIS fix one couple of VI values perturb the current with periodic signal at several frequencies; CV fix one FC current value and change the voltage dynamically by following a ramp; Careful load control strategy (high level) is requested to manage the diagnostic actions; Design SW to be implemented on-board ensuring great accuracy and performance while maintaining low computations burden. 12

13 Sensors placement, next step Minimal Sensor Set Is the smallest set of sensors that fulfils diagnostic requirements given a model structure Model Diagnostic requirements Sensor placement analysis All possible Minimal Sets of Sensors 11 possible sensors 4 system faults 11 sensor faults Significant system cost reduction PEMFC case Important! Diagnostic scheme is built upon model structure. No inference on system is required 17 minimal solutions 1 example P Polverino et al. (2017) Model-based diagnosis through Structural Analysis and Causal Computation for automotive Polymer Electrolyte Membrane Fuel Cell systems, J. of Power Sources 357:

14 Conclusions Advanced control, diagnostics and prognostics can improve system performance, reliability, durability and availability; help in reducing cost (CAPEX & OPEX). 7+ consortia have tackled the problem since 2010 HEALTH-CODE continues the work done on EIS for PEMFC after D-CODE. DIAMOND has focused on conventional and advanced SOFC systems. INSIGHT implements advanced diagnostic and lifetime techniques for SOFC stack. 14

15 Workshop schedule Welcome Diagnostics and Control for FC motivations, challenges and main issues Description of project DIAMOND Description of project HEALTH-CODE EIS Characterization of O2-fed PEMFC under fault operations Influence of operating and faulty conditions on the EIS spectra of PEMFC-based µ-chp FCS Scaling-up technique for PEMFC EIS, from single cell to stack Operation results for DIAMOND advanced configuration, Standard control versus Advanced control Testing & validation of adv. control & diagnostics for SOFC-not quite business as usual Model-based design of diagnostic tools for conventional and advanced SOFC systems TCO reduction of fuel cell systems thanks to diagnosis and prognosis algorithms Lunch break Health-based control and optimisation of SOFC stack operation State-of-health estimation and prognosis of the remaining useful life in SOFC systems EIS and soft computing techniques for the diagnosis of O2-feed PEMFC Equivalent Circuit Model-based diagnosis of PEMFC via EIS Hardware/software design for on-board fuel cell EIS Advanced control, diagnostics and monitoring for SOFC, the industry perspective Overview of project INSIGHT Discussion among guests, partners, participants 15

16 Thank you all for your attention Lucerne (CH) 16