Chaire Science des Systèmes et - lgi.ecp.frli/materials/booklet.pdf · Chaire Science des Systèmes et Défi Énergétique Fondation Européenne pour les Énergies de Demain - Électricité

ChaireScience des Systèmes et

Défi Énergétique Fondation Européenne pour les Énergies de Demain - Électricité de France (EDF)

www.ssde.fr

The Chair on Systems Science and the Energy Challenge (SSEC) is shared between Ecole Cen-trale Paris (ECP, www.ecp.fr) and Supelec (www.supelec.fr), and is supported by the European Foundation for New Energy of Electricite’ De France (EDF, www.edf.fr ), a leading electrical com-pany in the World.

Ecole Centrale Paris (ECP) and Supelec are two of the top five French engineering schools (“Grandes Ecoles”). They recruit students with excellent analytical skills through a national com-petition, after two years of intensive training in Fundamental Science (maths, physics…). They

deliver engineering degrees at the Master of Science and PhD levels. ECP and Supelec share a strong record in terms of academic research and have strong connection with industry.

The Chair activities have focused on the development, implementation and use of computational models, methods and algorithms for the analysis of the failure behavior of complex energy systems and the related uncertainty. Then, the re-search topics of interest for the Chair cover aspects related to reliability, availability and maintainability (RAM) engineering, risk assessment, safety and security evaluation, vulnerability analysis.

Currently, two main research axes are under development:

Axis 1: Characterization of the aging and failure behavior of power production plant components•Axis 2: Electrical network systems•

Overall, the research is intended to be exploratory and methodological. The Chair aims at complementing and integrating the different approaches to complex system analysis (from abstract network theory to detailed simulation, from analytical to empirical modeling, from probabilistic to non-probabilistic uncertainty analysis) in order to be able to effectively res-pond at the specific needs of the different analyses that can be performed (design, operation, maintenance, protection, etc.) on the different objects of analysis (components, plants, networks, etc.).

The expected outcomes of the research are advanced, proven concepts and methods of analysis. The tangible outcomes are expected to be:

Publications on international, peer-reviewed scientific journals•Presentations at internationally recognized conferences•Softwares for proof of concepts and methods (the practical implementation and/or commercialization of the tools •developed is not within the objective of the research).

As of today, the SSEC Chair team consists of two full time members, a senior professor, director of the Chair (Enrico Zio) and one assistant professor (Yanfu Li), two post-docs (Carlos Ruiz Mora and Valeria Vitelli) and a dynamic group of valuable PhD students.

Chaire « Sciences des Systèmes et

Défis Energétiques (SSDE) » Fondation Européenne pour les Energies de Demain — Électricité de France (EDF)

PRESENTATION The Chair on Systems Science and the Energy Challenge is a research and education initiative shared between ECP and SUPELEC, with the support of EDF. Its activities focus on the development, implementation and use of computational models, methods and algorithms for the analysis of the failure behavior of complex energy systems and the related uncertainties. The rresearch topics of interest relate to complex systems modeling, reliability, availability and maintainability (RAM) engineering, risk assessment, safety and security evaluation, vulnerability analysis, failure diagnostics and prognostics. The research spirit is intended to be exploratory and methodological. The Chair aims at complementing and integrating the different approaches to complex system analysis (from abstract network theory to detailed simulation, from analytical to empirical modeling, from probabilistic to non-probabilistic uncertainty analysis) to effectively respond at the specific scope of the different analyses (design, operation, maintenance, protection, etc.) on the different objects of analysis (components, plants, network systems, critical infrastructures, etc.).

RESEARCH SPIRIT ��Methodological ��Exploratory ��Systemic RESEARCH AREAS ��Modeling ��Simulation ��Optimization

RESEARCH LINES 1. Aging and failure processes in components of energy production plants Component failure prognostics Ronay AK, Jie LIU, Valeria VITELLI Component degradation and maintenance modeling and simulation Yan-Hui LIN, Yan-Fu LI

2. Energy network systems Agent-based modeling Elizaveta KUZNETSOVA, Carlos Ruiz MORA Yan-Fu LI Complexity Science Yi-Ping FANG, Tairan WANG System-of-Systems approach to External Events Risk Assessment Elisa FERRARIO, Chung-Kung LO Optimization under uncertainty Rodrigo MENA, Carlos Ruiz MORA, Yan-Fu LI

Yan-Fu LI Assistant Professor

Enrico ZIO Professor

Chair Director Two post-docs

Eight PhD students

Valeria VITELLI

Carlos Ruiz MORA

Ronay AK Yi-Ping FANG

Elisa FERRARIO Elizaveta

KUZNETSOVA

Jie LIU Chung-Kung LO

Rodrigo MENA

Tairan WANG

The Team Two full-time faculty members

COLLABORATIONS Beihang University, China City University, Hong Kong Denmark Technical University, Denmark Électricité de France (EDF) R&D Federal University of Pernambuco, Brasil Institute of Nuclear Energy Research, Taiwan Lund University, Sweden Politecnico di Milano, Italy Politecnico di Torino, Italy Universidad Federico Santa Maria, Chile University of Stavanger, Norway



RESEARCH LINE(S): Methods of Optimization of Complex Systems Under Uncertainties

RESEARCH TOPICS: �� Modeling and optimization for the allocation and planning of distributed ge-

neration systems under uncertainties. Considering: �� The installation of multiple renewable power generators and storage devi-

ces in the distribution network. �� The uncertain behavior of renewable sources (solar irradiation, wind

speed, electric vehicles), failures and repairs of the generation units and feeders, charging and discharging of storages, variability of loads, market evolution and fluctuation of grid power supply.

�� The complexity given by the physical and functional interrelations between the entities that compose the system.

�� Assessment of overall network performance in terms of reliability, voltage quality, and operational and technical constraints satisfaction.

RESEARCH METHODS: �� CURRENT

�� Multi-State Modeling: components operating/mechanical states and uncertainty sources (generation and failure/repairing transition rates) identification, and stochastic representation by Markov Chains Approach and Probability Distribution Functions.

�� Multi-Objective Optimization Modeling: construction of objective functions, minimization of overall net-work investment and operating costs and maximization of network reliability.

�� FUTURE �� Robust Optimization Approach: two-stage Stochastic Programming subject to stochastic, technical and

network topology constraints. Nested optimal Power Flow analysis. �� Non-dominated Sorting Genetic Algorithm-II (NSGA-II): Pareto Front determination. �� Integration of Monte Carlo Method: multiple scenarios generation for systemic performance evaluation.

RESEARCH RESULTS AND APPLICATIONS: �� OBTAINED

�� Multi-state modeling framework: stochastic representation of the components operating/mechanical states and their physical-functional interrelation.

�� Two-stage stochastic formulation: multi-objective optimization with nested Power Flow Analysis. �� EXPECTED

�� Assessment of the impact of transforming the Mixed-Integer Non-Linear Problem (MINLP) Three phase Power Flow analysis into a Mixed-Integer Linear Problem (MILP), solvable with standard cut-and-brunch methods, to govern the complexity of the system and handle the multiple sources of uncertain-ty.

�� To obtain a tractable Robust Optimization model: system design immune to time dependent operation uncertainties.

�� Methodology for treating the uncertainties in order to optimize the expected performance of the distri-buted generation network based on the allocation and sizing of renewable distributed generators.

�� Realistic industrial settings application.

SPONSORS: Comisión Nacional de Investigación Científica y Tecnológica CONICYT, Gobierno de Chile.

Rodrigo Mena

Carlos Ruiz Mora

Yan-Fu Li



RESEARCH LINE: Agent-based modeling RESEARCH TOPIC: �� Modeling framework for energy management in microgrids with of locally in-

stalled renewable generators and storage facilities. �� Agent-based modeling and simulation of microgrid individual actors, each

with explicit goals and interactions with the environment. �� Reinforcement learning → to exploit the environment → decision of optimal

scenario to maximizing reward → multi-objective goals (costs, renewable energy, etc).

�� Modeling of the interactions among agents → detection of potential system equi-libria → operation point in which all agents are happy and do not want to deviate.

RESEARCH METHODS: �� CURRENT

�� Multi-criteria learning of different objective functions. �� Markov Chain modeling of stochastic wind conditions (real wind speed data

set). Integration of the uncertainties related to mechanical failures. �� FUTURE

�� Extension of duration of planning scenario to increase learning efficiency and rate of goals achievement.

�� Development of case study with multiple intelligent agents with sharing of information, conflicting objectives, different interconnection frame-works etc.

�� Equilibria modeling.

RESEARCH RESULTS: �� OBTAINED

�� Formulation of rewards → multiple objective functions → selection of dis-counted frameworks to estimate rewards from different scenarios.

�� Learning performance in presence of uncertainties, i.e., stochastic wind conditions and random mechanical failures.

�� EXPECTED �� Improvement of multi-criteria learning algorithm → optimal use of infor-

mation under complex, uncertain and changing environments. �� Multiple intelligent agents → modeling of emerging effects. �� Multiple intelligent agents → study of potential equilibria to design incen-

tives for promoting the most efficient ones (maximum social welfare).

COLLABORATIONS: �� ECONOVING International Chair in Eco-Innovation, REEDS International Centre for

Research in Ecological Economics, Eco-Innovation and Tool Development for Sustainability, University of Versailles Saint Quentin-en-Yvelines.

�� “Programme Gare” of ECONOVING Chair and SNCF, with participation of Alstom Grid, GDF Suez, Italce-menti, Saur and Ademe.

�� Institute for Energy and Environment, University of Strathclyde, Glasgow, Scotland.

Elizaveta Kuznetsova

Carlos Ruiz Mora

Yan-Fu Li

Fig. 3

Fig. 1

Fig. 2



RESEARCH LINE(S): Prediction and prognostics

RESEARCH TOPICS: Prediction with uncertainty quantification.

Kernel-based pattern recognition methods for prognostics of nuclear components’ Estimated Time To Failure (ETTF).

Nonlinear regression methods for prognostics; prediction of functional patterns.

RESEARCH METHODS AND APPLICATIONS: �� CURRENT

Feature selection methods and correlation analysis for the pre-treatment of input data.

Multi-objective genetic algorithm approach for the estimation of Neural Network (NN)-based Prediction In-tervals (PIs) (application to wind speed data).

Probabilistic Support Vector Machine for fault prognosis of electrical components.

Local Gaussian Processes (LGP) for multivariate nonlinear trend estimation.

�� FUTURE

Ensemble methods for improving accuracy and robustness of NN-based PIs.

Bayesian Neural Networks and Particle Filtering for prognostics.

Functional prediction and uncertainty quantification models for high dimensional signals.

RESEARCH RESULTS: �� OBTAINED

Data-driven method for determining the optimal values of the parameters of a NN maximizing coverage probability (CP) and minimizing the prediction interval width (PIW) in Pareto optimality sense. Evaluation of the method via a synthetic case study of literature, and a real case study of short-term wind speed predic-tion.

State-of-the-art review of existing probabilistic support vector machines methods for ETTF prediction.

�� EXPECTED

Definition of a general strategy for input data pre-treatment.

Real case studies: wind power forecasting, load pre-diction, etc.

Energy components ETTF prediction.

COLLABORATIONS:

SPONSORS: EDF R&D, Département Simulation et Traitement de l’information pour l’Exploitation des systèmes de Pro-duction (STEP).

China Scholarship Council (CSC).

Ronay AK

Jie LIU

Valeria VITELLI

Fig. 1: Estimated PIs for one hour ahead wind speed prediction (dashed lines), and actual wind speed data (solid lines).


Défis Energétiques (SSDE) » Fondation Européenne pour les Energies de Demain — Électricité de France

RESEARCH LINE: Complex Energy Systems Modelling and Analysis

RESEARCH TOPIC:

Vulnerability and risk analysis of electric power system, e.g. identification of the critical components, cascading failure modeling.

Decision making and modeling uncertainty for the multi-criteria analysis of com-plex energy systems.

RESEARCH METHODS:

�� CURRENT:

Complex network theory; Critical components identification using unsupervised spectral clustering algo-rithm; All-Hazard Approach; Ordinal methods for decision making.

�� FUTURE:

Monte Carlo simulation of cascading failures.

RESEARCH RESULTS:

�� OBTAINED

New method for identifying critical components based on unsupervised clustering algorithm: components on the clusters borders are more critical (Figure 1).

Introduced a qualitative ranking method of energy system (Figure 2).

�� EXPECTED

�� Cascading failure process in a Hierar-chical model of a network obtained by successive clustering. Model analysis to design protections against propaga-tion of cascading failures after their outbreak.

�� Decision making framework for sup-porting the analysis of system vulner-ability.

SPONSORS:

China Scholarship Council

Tairan WANG

Figure 1 Figure 2

Yiping FANG



RESEARCH LINE(S): External Events Risk Assessment (Seismic).

RESEARCH TOPICS: Safety analysis of a critical plant, e.g., a nuclear power plant, exposed to risk from external events, e.g., earthquakes.

System-of-Systems analysis of the interdependent infrastructures that support the critical plant.

Uncertainty analysis and advanced simulation.

RESEARCH METHODS AND APPLICATIONS:

��CURRENT

System analysis: Muir Web and Fault Tree Analysis.

Quantitative evaluation of the model: Monte Carlo simulation.

Uncertainty analysis: study of advanced simulation method (e.g., stochastic collocation).

�� FUTURE

System analysis: Muir Web, Graph Theory and flow diagrams.

Uncertainty analysis:

�� Uncertainty representation frameworks: probabilistic and non-probabilistic (e.g., possibilistic). �� Uncertainty propagation methods: pure probabilistic approach (Monte Carlo simulation) and Hybrid

Monte Carlo simulation (combination of probabilistic and possibilistic approaches, and combination of stochastic collocation and Monte Carlo simulation).

RESEARCH RESULTS:

��OBTAINED

Method to investigate the contribution of the interdependent infrastructure systems to the safety of a critical plant.

��EXPECTED

Methods for optimizing controllable characteristics of the system of systems for increasing the safety of the critical plant, including in the analysis the spatial correlation and the recovery time of all the elements of the system of systems.

Methods for uncertainty treatment in External Events Risk Assessment.

COLLABORATIONS: Politecnico di Milano.

SPONSORS: Taiwan Government Scholarship.

Ecole Centrale Paris Scholarship.

Chung-Kung LO Elisa FERRARIO

RESEARCH LINE: Component degradation and maintenance modeling

RESEARCH TOPICS:

Modelling of competing degradation processes in energy components, by inhomogeneous Markov chains with time-dependent transition rates influenced by physical factors/mechanisms.Developing Bayesian networks to represent the interdependences among the influencing factors/mechanisms, accounting for the associated uncertainties.

� Integrating the Bayesian networks into a stochastic Petri net-based framework for the simulation of com-ponent degradation.

RESEARCH METHODS AND APPLICATIONS: � CURRENT

�� Multi-state physic modelling and simulation of the degra-dation processes by stochastic Petri nets.

�� Numerical methods of solution of the inhomogeneous Markov chains.

� FUTURE �� Study of the uncertainties in the influencing factors/

mechanisms. �� Application to a nuclear component degradation problem.

RESEARCH RESULTS:

OBTAINED�� Multi-state physic model and stochastic petri net-based

simulation framework.�� Comparison of different methods to solve inhomogene-

ous Markov chains.EXPECTED�� Development of methods for handling uncertainties of

different types.�� New simulation algorithms in the presence of uncer-

tainty and dynamics.

COLLABORATIONS: Beihang University. SPONSORS: China Scholarship CouncilEDF R&D, Département Maitrise du Risque Industriel (MRI).



Yan-Hui LIN

Fig 1

Fig 2

Yan-Fu LI

ÉCOLE CENTRALE PARISGrande Voie des Vignes92295 CHÂTENAY-MALABRY CedexTél : 01 41 13 16 06 - Fax : 01 41 13 12 72www.ecp.fr

SUPÉLEC3 rue Joliot Curie Plateau de Moulon

91190 GIF-SUR-YVETTE Tél : 01 69 85 15 20

www.supelec.fr

Documents

Chaire Science des Systèmes et - lgi.ecp.frli/materials/booklet.pdf · Chaire Science des Systèmes et Défi Énergétique Fondation Européenne pour les Énergies de Demain - Électricité