Mo Jamshidi, Ph.D., DEgr. (h.c.)

Transcription

1 System of Systems Engineering: Modeling, Simulation and Control with Applications to Renewable Energy 1 Mo Jamshidi, Ph.D., DEgr. (h.c.) Lutcher Brown Endowed Chaired Professor Member, Univ. Texas System s Chancellor s Council Department of Electrical and Computer Engineering and Director Autonomous Control Engineering - ACE Laboratory The University of Texas, San Antonio, Texas) moj@wacong.org, mo.jamshidi@utsa.edu Innovative Approaches & Researches for System of Systems Engineering Technion, Haifa, Israel July 17, In memory of Dr. Zeev Bonen for his lasting contributions to system security and adpatation.

2 OUTLINE 1. Introduction to System of systems SoS (Cyber-Physical Systems -CPS) 2. The need to go from SoS to SoS Engineering 3. A review of energy sources : Hydro, coal, fossil fuels, nuclear, photovoltaic, wind and geothermal 4. Modeling and Simulation of SoS 5. SoSE Research at the University of Texas, San Antonio and Open Cloud Paradigm 6. Trans-Atlantic Research and Education in SoS An EU-US effort 7. Conclusions and movie

3 Preliminary Comments o Internet has connected people of the world since ~ 1995 o System of Systems (or Cyber-physical systems) is a generalization of connectivity of systems or systems and people or a crossing of cyberspace and physical space! o SoS or CPS is getting into new application cases in a very persisting pace from IT to defense, energy, space, environment, healthcare, services, earth studies, etc.

4 A Definition of SoS One out of many definitions A SoS is an integration of a finite number of constituent systems which are independent and operatable, and which are networked together for a period of time to achieve a certain higher goal. e.g.: lower cost, higher robustness, reliability, etc. Applications: Environment, Energy, Defense, Automation, etc. (Jamshidi, 2005, 2009)

5 SoS(E) SoS Engineering Operational independence of component systems Managerial independence of component systems Geographical distribution Emergent behaviour Evolutionary development processes (Maier, 1998) Directed Acknowledged Collaborative Virtual 5 Dahmann & Baldwin, 2008

6 What is a system of systems Freeways Transportation SoS: Roads +GPS+ ONSTAR Retail businesses Unanticipated benefits of SoS extension beyond MP3 player (Blogs, PODCAST) or Internet purchases Others: ipad iphone ipod Aircraft

7 System of Systems SoS: A meta system consisting of multiple autonomous embedded complex systems that can be diverse in: Technology Context Operation Geography Conceptual frame An airplane is not SoS, an airport is a SoS. A robot is not a SoS, but a robotic colony (a swarm) is a SoS Significant challenges: What is the appropriate mix of independent systems? The operation of a SoS occurs in an uncertain environment Interoperability

8 Application Domains of System of Systems. 1. Homeland Security from borders and ports to natural disasters 2. Planet Earth - GEOSS 3. Military future combat missions 4. Space robot colonies, formation flying objects 5. National and Homeland Security 6. Energy, fossil fuels to renewable 7. Environment 8. Healthcare 9. Transportation 10. Etc.

9 FROM SoS TO SoS ENGINEERING?

10 System Engineering vs. SoS (CPS) Engineering System Engineering is a discipline (at least 5 decades old) BUT SoS (CPS) Engineering (at the present time) is only an opportunity

11 Scholar Hits SoS (CPS) vs. SoSE (CPSE) SoS SoSE Year Chuck Keating, 2006, ODU

12 Challenges SOS SoSE Modeling Control Theory Cyber-Security SOSE Simulation Soft Computing Emergence Architecture ETHICS Net Centricity

13 EXAMPLES of System of Systems

14 Consider The Following (1) A Deployed Force (and Network) Partitioned Due to Terrain Features (2) Perform Analysis Which Yield the Initial Set of Links (3) Local Connectivity is Achieved, but the Forces (and Network) Are Still Partitioned (4) Technical Enablers: Algorithms, Power Profiles Courtesy Monica Stapleton, US Army (5) To Meet Operational Need, Consider Introducing an Airborne Comm Asset ( UAV Flying Overhead or a Satellite Link, Depending on Coverage Constraints)

15 Consider A Change In Operations (1) Introduce a UAV Flying Overhead (2) Perform Analysis to Determine Connectivity of Platforms Within Communications Range (3) Networks Are Merged to Provide Full Network Connectivity (4) Technical Enablers: UAV With Comm Relay Package, Maintenance Infrastructure (Refueling, Launch/Retrieval Mechanisms)

16 Air Transportation Airports An airport can be broken down in to smaller systems. Baggage handling system Air Control Tower Plane w/crew Freight Reservation system Customs Security Airlines

17 Crude Oil Production Wellhead Production Bulk Terminal Storage Fuel Distribution U.S. Vehicle Fleet Refineries Gasoline Distribution Platform Production Barge Tanker Truck Retail Outlets Fuel End Use Crude Oil Logistics Crude Oil Refining Pipeline The primary objective of each system is described in the context of the existing transportation SoS, which moves crude oil from its source to the final processed fuel used by consumers, (Duffy, et al., 2008, Jamshidi, 2009)

18 Energy Smart Homes Home Area Network (HAN) and Wide Area Network (WAN)

19 Smart-Grid Energy SoS Smart Grids

20 Sample of Solar Power Density (W/m 2 ) worldwide in December Courtesy of NASA

21

22

23

24

25 Photovoltaic (PV) Progress a) PV module. b) PV Array c) PV System Various PV systems (US DOE)

26 WIND POWER Contributions from Jose Zayas - SNL

27

28

29

30 MODELING of SoS

31 SoS BIG DATA ANALYTICS I n p u t s.. SoS O u T p u t s Model

32 SoS BIG DATA ANALYTICS Modeling of System of Systems via Data Analytics Case for Big Data in SoS SoS BIG Data PCA Pre-processing (Transformation) Information Data Mining (NN) Knowledge Validation (Post-processing) GA or GP Expert Inference (FL) Refined Knowledge MODEL

33 Objective Advances in sensor technology, the Internet, wireless communication, and inexpensive memory have all contributed to an explosion of Big Data. Recall: SoS are integrated independently operating systems to achieve a higher goal than the sum of the parts.

34 Objectives, Cont d Today s SoS are also contributing to the existence of unmanageable Big Data. Recent efforts have developed a promising approach, called Data Analytics, including: K-means Clustering, Fuzzy Clustering/Fuzzy Inference System Model Generation (FCM,FIS) Neural Network Model Training Principal Component Analysis (PCA) Regression Analysis n... Can use CI and statistical tools of the Data Analytics as they apply to any type of data, here we use photovoltaic (or wind) energy forecasting problem for a micro-grid SoS.

35 Micro-Grid SoS This micro-grid represents a facility scale Cyber- Physical System (CPS) or a SoS consisting of a building with : A solar photovoltaic system A load demand in the form of overall energy consumption A reconfigurable control and acquisition system System Model Internet Ceilometer Ethernet Hub Sky imager CompactRio Sensors for: q Irradiance q Temperature q Rela ve Humidity q Wind Speed q Wind direc on Solar Forecas ng Module Focus of Cyber-Physical System Research Bidirec onal Inverter 1 Solar Genera ng Plant Combiner 1 Combiner 2 Ba ery Satellite (Internet download) Combiner n Merging of technologies to provide dependable renewable power Load Manjili, et al., 2012

36 Source of Data Location Three data sets acquired for the area surrounding Golden, CO were combined for to produce the training data for this exercise. Source 1: Solar Radiation Research Laboratory (SRRL)/National Renewable Energy Laboratory (NREL) Uses over 70 instruments to measure solar conditions and environmental parameters. Also includes 180 images of the sky that can be used to determine current cloud conditions directly Source 2: SOLPOS Data, Measurement and Instrumentation Data Center (MIDC) solpos.html Contains information on solar position and available solar energy Source 3: Automated Surface Observing System (ASOS), the Iowa Environmental Mesonet (IEM) Contains weather parameters in the set SSRL Sky Image

37 Final Data Set PV Data Analysis Goals Consists of data points for half the month of October 2012 Contains approximately 250 variables for each data point Data points containing invalid variable values removed prior to analysis Objective of Analysis Three key parameters representing total received solar energy were selected as the output of the analysis: Global Horizontal Irradiance (GHI) Direct Horizontal Irradiance (DHI) Direct Normal Irradiance (DNI) The goal of the models generated was to predict values for these three parameters 60 minutes in advance. 37

38 PV Data Analysis Setup Due to execution time and memory limitations, the initial set of variables used as the input data set was 13 parameters including Time of Day Cloud Levels Humidity Temperature Wind Speed Current values of GHI, DHI, and DNI Data set further narrowed down to exclude points with very low values of Irradinace - GHI, DHI, and DNI Output Parameters for a Clear Day

39 PV Data Analysis Tools Non-Parametric Model Generation Tools Two tools used to generate models for this SoS: GENFIS3: Fuzzy Inference System Generation Based on Fuzzy C-Means Clustering NFTOOL: Back-Propagation Trained Feed Forward Neural Network GENFIS3 (MATLAB Fuzzy Logic Toolbox) All input and output variables clustered using Fuzzy C-Means Clustering to generate the fuzzy membership functions Rules then generated to best match the training data set Membership functions and rules can be viewed using the MATLAB FIS GUI tools, such as ruleview

40 PV Data Analysis Tools Non-Parametric Model Generation Tools (cont.) GENFIS3 (MATLAB Fuzzy Logic Toolbox) (cont.) Membership functions and rules can be viewed using the MATLAB FIS GUI tools, such as ruleview This sample shows the output produced with five input variables and three output variables. Four fuzzy membership functions produced per variable. Sample Rule View Output

41 PV Data Analysis Tools Sample Output Predictions, 13 Input Parameters

42 PV Data Analysis Tools Non-Parametric Model Generation Tools (cont.) NFTOOL (MATLAB Neural Network Toolbox) By default, uses the Levenberg-Marquardt back propagation method to select weights and biases to minimize the mean squared error performance of a the feed forward network For this exercise, the feed forward network contains one hidden layer consisting of ten neurons Initial Feed Forward Network Architecture

43 PV Data Analysis Tools Non-Parametric Model Generation Tools (cont.) NFTOOL (MATLAB Neural Network Toolbox) (cont.) Performance Training Graphs for 13 Input Parameters Training Performance Curve Network Regression Performance

44 PV Data Analysis Tools Non-Parametric Model Generation Tools (cont.) NFTOOL (MATLAB Neural Network Toolbox) (cont.) Performance Training Graphs for 13 Input Parameters Sample Output Predictions, 13 Input Parameters

45 PV Data Analysis Tools Additional Pre-Processing Efforts To improve the performance of the generated models, two different paths were investigated Nonlinear Derived Parameter Generation Larger Training Data Sets Nonlinear Parameter Generation Objective: To calculate nonlinear parameters from the original set of input parameters Calculated parameters include: x(t) 2 sin(x(t)) cos(x(t)) slope(x(t-60):x(t)) mean(x(t-60):x(t)) stdev(x(t-60):x(t)) slope(x(t-1):x(t))

46 PV Data Analysis Tools Additional Pre-Processing Efforts Nonlinear Parameter Generation (cont.) This graph reflects the performance improvement after incorporating derived nonlinear components into the training data set. Note that the calculation of nonlinear components excludes data points for which the derived calculations are invalid (missing data points, etc.) Regression Performance Improvement with Nonlinear Input Data Set

47 PV Data Analysis Tools Additional Pre-Processing Efforts Larger Training Data Sets Objective: To improve trained model performance by including a larger number of input parameters First, the performance of the initial 13 parameters was compared to a 32 parameter data set. As expected, performance increased accordingly Next, the training of models using the entire input set (approximately 250 variables) was attempted, but neither model generation tool would execute with this number of variables. Principal Component Analysis (PCA) was then used to reduce the dimension of the full initial data set PCA was also conducted on the data set that had been expanded to include all calculated nonlinear parameters

48 PV Data Analysis Tools Additional Pre-Processing Efforts Principal Component Analysis (cont.) PCA used to reduce the dimension of the data while minimizing the information lost Graph Showing Information Contained In Principal Components Data Recovery Using Only 50 out of ~250 Principal Components

49 Results Discussion of Results The best performing model was generated using NFTOOL with an input data set generated as follows: 1. Start with all initial ~250 variables 2. Calculate nonlinear components for each variable, expanding the data set dimension to ~ Perform PCA to drop the dimension of the data to 150 The best performing GENFIS3 generated model used the same input data set, but with a post-pca dimension of 50

50 Results Discussion of Results (cont.) Best Model Performance Best Linear Regression Performance 50 Best Model GHI Prediction Error

51 Results Discussion of Results (cont.) Best Model Performance (cont.) Best Model DHI Prediction Error Best Model DNI Prediction Error 51

52 Discussion of Results (cont.) Table contains results for each training configuration tested Note comparison between configurations should be qualitative Results 52

53 Discussion of Results (cont.) A sub-optimal predictor was constructed in order to show its performance relative to that of the nonparametric models. This predictor was based on Results 53 Performance of Best Non-Parametric to Mean Time Bin Sub-Optimal Predictor

54 Discussion of Results (cont.) During this analysis, the aspect of the scalability of the GENFIS3 and NFTOOL tools was evaluated. Model generation time for NFTOOL was always 54 shorter than Results

55 Conclusion Conclusion This paper describes the steps necessary to use Big Data acquired using instrument stations to predict received solar power using Data Analytics tools to support a micro-grid The use of Fuzzy C-Means Clustering and Feed Forward Neural networks in conjunction with Principal Component Analysis successfully yielded predictions of GHI, DHI, and DNI The best performing model used a Neural Network trained using preprocessed input data shrunk to a usable dimension using PCA Both non-parametric models performed better than a simple, sub-optimal predictor The training time was significantly better for the neural network Future work is planned to include Optimal dataset reduction using Genetic Programming Using cloud computing to train models using large datasets Design and evaluate a controller to buy or sell energy to the grid based on 55 demand and predictions of received solar energy.

56 SIMULATION OF SoS

57 A Discrete Event XML based Simulation Framework for System of Systems Architectures Ferat Sahin Rochester Institute of Technology and UTSA Mo Jamshidi and Prasanna Sridhar University of Texas San Antonio

58 An XML based System of systems architecture <!--Created 11/8/2006 Ferat Sahin --> <?xml-stylesheet type="text/css" href="genericxml.css"?> <systemofsystem> <id> Id of the System of systems </id> <name> The name of the System of System</name> <system> <id>id of the first system</id> <name> The name of the first system </name> <description> The description of the 1st system </description> <dataset> <Output> <id>id of the first output</id> <data>data of the first output</data> </Output> <Output> <id>id of the second output</id> <data>data of the second output</data> </Output> </dataset> <subsystem> <id>id of the subsystem of the 1st System</id> <name>the name of the subsystem</name> <description>this is a subsystem in an SoS</description> <dataset> <Output> <data> Data of the subsystem </data> </Output> </dataset> </subsystem> </system> </systemofsystem>

59 Discrete Event Simulation Specification (DEVS) Created in University of Arizona Interactions in a SoS are generally asynchronous in nature Robust solution is to throw an event at the receiving end to capture messages Basic (atomic) Models Coupled Models

60 Discrete Event Simulation Specification (DEVS) DEVS Simulation Example Coupled Model Atomic Model Coupled Model Atomic Model Atomic Model System-A DEVS Coupled Model System-B DEVS Coupled Model <system-a> <data1> 10</data1> <data2> 20</data2> <data3>0.2</data3> </system-a> <system-b> <data1> 5</data1> <data2> 0.9</data2> </system-b> System of Systems DEVS Coupled Model System-C DEVS Coupled Model <system-c> <system-a> <data1> 10</data1> <data2> 20</data2> <data3>0.2</data3> </system-a> <system-b> <data1> 5</data1> <data2> 0.9</data2> </system-b> </system-c> Message in XML based language

61 System of Systems Simulation Framework Case study: System Name Vector Data Length Base Robot Base Robot 5 (X, Y, Threat, Xt, Yt) Swarm Robot Swarm Robot 5 (X, Y, Threat, Xt, Yt) Sensor Sensor 1 3 (X, Y, Threat) Sensor Sensor 2 3 (X, Y, Threat) Threat Fire 2 (X, Y)

62 System of Systems Simulation Framework The DEVSJAVA atomic and coupled modules for XML base SoS simulation.

63 Simulation Runs System of Systems Simulation Framework

64 Simulation Details Threat is detected successfully Communication was efficient during the operation Visualization is done by the plot module of the DEVSJAVA Each message has a destination, source, and data vector XML like data wrapping

65 SoS Simulation Details Scenario: Fire in a field Field Sensors (SN) Base rover senses fire remotely Scout (swarm) rover verifies fire existence

66 SoS Simulation Details Fire (red dot) appears 2. Two sensors in the area 3. Scout rover (green dot) is within vicinity 4. Base rover is out of picture

67 SoS Simulation Details Threat (red dot) has moved into the sensor field 2. Base rover (out of picture) sends scout rover (green dot)

68 SoS Simulation Details Scout rover verifies the existence of fire. 2. Base Rover seeks help

69 A Discrete Event XML based System of Systems Simulation for Robust Threat Detection and Integration Ferat Sahin and Carmen Parisi Rochester Institute of Technology Mo Jamshidi University of Texas San Antonio

70 System of Systems Simulation Simulation Runs Framework Initial Condition Sensor 1 Detects a Threat Swarm Robots Move to Verify Threat Swarm Robots Move to Neutral Location

71 System of Systems Simulation Framework Swarm Robot 1 Breaks off Verification Run

72 Simulation Details Threat is detected and verified successfully Communication was efficient during the operation Visualization is done by the plot module of the DEVSJAVA Each message has a destination, source, and data vector XML like data wrapping

73 SoS Simulation Details Scenario: Fire in a field Field Sensors (SN) Base rover senses fire remotely through sensors Swarm Robots verify the existence of a fire

74 SoS Simulation Details Fire (red dot) appears 2. Five sensors in the field 3. Swarm Robots (green and orange) are within vicinity 4. Base rover (Black dot) is stationary

75 SoS Simulation Details Threat (red dot) has moved into a sensor field 2. Base Robot signals Swarm Robots (green and orange dot) 3. Sensor detects the fire

76 SoS Simulation Details Threat (red dot) has moved into a sensor field 2. Base Robot signals Swarm Robots (green and orange dot) 3. One of the swarm robot is verifying the threat

77 SoS Simulation Details Orange Swarm Robot does not continue to move towards the sensor field. 2. Both Swarm Robots (green and orange dot) moves back to initial positions after threat is verified and handled.

78 A Discrete Event XML based System of Systems Hardware-in-the-Loop Simulation for Robust Threat Detection Hosking & Sahin, 2009

79 DEVS Models Plot Module Provides GUI for simulation Mobile Swarm Agent Mobile Threat Sensor Base Station XML Messages Wraps Java String into DEVS entity basic implementation of SoS tags

80 Simulation Results 7 sensors Blue/Red 3 swarm agents 2 Virtual 1 Groundscou t Command Center Enemy tank

81 XML Messages Sensor to Base Station (location and threat indication) <y><i>sr05</i><s> <i>$</i><d>509,384</d> <i> </i><d>1</d> </s></y> Base Station to Swarm Agents (destination) <y><i>bs09</i><s> <i>+</i><d>225,225</d> </s></y> Swarm Agent to Swarm Agent (location) <y><i>gs02</i><s> <i>$</i><d>509,384</d> </s></y>

82 Simulation Results: Threat Detected Groundscout leader goes to location Virtual leader goes to location

83 Mobile Agents successfully investigate and verify potential threats and cooperate among real and virtual agents Simulation Results: Swarm Returns Home XML SoS HIL simulation framework verified by successful XML data communication

84 Future Work Agent in the loop simulation of SoS

85 SoSE Research at the University of Texas, San Antonio and Open Cloud Paradigm

86 UTSA Smart Grid Project Distribute UTSA Sites - $ 2.4 ML North Campus Downtown Campus UTSA Campus View UCIII EB Building Durango Building Durango Parking Garage

87 SECO-2 Vehicle Electrification 1 Level 3 Fast DC Charging Stations 1 Level 2 Charging Stations

88

89 Sensor Network Architecture at University Center III

90 OPENSTACK AT A GLANCE Software to build a cloud anywhere An ecosystem devoted to innovation Flexibility in deployment/features Standards for broad deployment No fear of lock-in Open Source Cloud computing will meet the needs of public and private Cloud providers by being simple and massively scalable.

91 A common platform is here. OpenStack is open source software powering public and private clouds. Private Cloud: Run OpenStack software in your own corporate data centers OpenStack enables cloud federation Connecting clouds to create global resource pools Public Cloud: OpenStack powers some of the worlds largest public cloud deployments. Common software platform making Federation possible Washington Public Cloud Texas Private Cloud California Private Cloud Europe Public Cloud Server Virtualization 1. Virtualization 2. Cloud Data Center 3. Cloud Federation Automation & Efficiency

92 Research Institutes and Academia A short list of research institutes and universities: San Diego Supercomputer Center Argonne National Laboratory CERN Natioanl Security Agency NSA Brookhaven National Laboratory MIT CISAL Massachusetts' open cloud (MOC) University of Texas at San Antonio University of Victoria University of Melbourne Australian Government UCLA Purdue University Industrial Technology Research Institute (ITRI) Taiwan University of Southern California USC Unleashes the Value of its User Data by leveraging SWIFT Building High Performance Clouds Overlay opportunistic clouds in CMS/ATLAS at CERN OpenStack at the National Security Agency US Department of Energy OpenStack for scientific research HPC Cloud for Scientific Research and BigData Build Advanced OpenStack-based Software Open Cloud Platform for Academic Research Clouds in High Energy Physics Building a Cloud for National Collaborative Research NeCTAR High Performance Computing Cloud Cloud Infrastructure Cloud Infrastructure Information Sciences Institute

93 OpenStack Clouds for Scientific Research Create a scalable and collaborative hybrid cloud environment accessible for scientific applications to enable Open Science: OpenStack-based Private Cloud among universities for hosting, deploying and sharing research applications Enable their OpenStack-based Private Cloud to bursting to Open Cloud through Internet2 Sharing code, data, and platform openly for research empowerment Internet2 University of Texas at San Antonio Massachusetts' Open Cloud (MOC) Argonne National Lab Notre Dame Facebook Mellanox Dr. Stephen Wolff, Dr. Khalil Yazdi Dr. Mo Jamshidi, Dr. Raj Boppana, Daniel Pack Chris Hill (MIT), Dr. Orran Krieger (Boston), Michael Goroff (UMass) Narayan Desai Dr. Paul Brenner (Notre Dame) John Kenevey, Charlie Manese Arik Kol, Eli Karpilovski, Marina Lipshteyn 80/20 foundation Lorenzo Gomez, Jeff Prevost

94 Trans-Atlantic Research and Education Agenda in System of Systems T-AREA-SoS U.S. WORKSHOP 5 th 7 th November 2012

95 T-AREA-SoS Partners 95

96 Creating the Agenda - Process Consultative approach Literature Review Expert workshops Panel sessions at IEEE and INCOSE Community of Experts Survey 96

97 Sectorial needs for SRA Energy Sept 2011 power outage Arizona & Southern California. 2.7M people affected. Due to loss of one 500 kv line: complexity + legacy SMART grids for integration of heterogeneous sources Manufacturing Virtualisation of structural specifications New ways of managing diverse stakeholders, new approaches to open architectures, system security improvements, migration of legacy systems 97

98 Sectorial needs for SRA Transport Single European Sky ATM Research (SESAR) Air traffic control for: double traffic over Europe within 20 years; reduce environmental impact Interoperability and data quality are key requirements Other sectors Healthcare, defence, infrastructure projects, geospatial monitoring, e-commerce,. 98

99 Global Drivers (1) Population demographics Europe will increase from 500M to 526M by 2040 BUT those aged 65+ will increase from 25% to 52% Also redistribution within Europe. UK from 62M to 79M; Poland from 38M to 32M Must focus on efficient services to non-working Food security Europe imports cereals/exports processed foods Changes in populations elsewhere in world affect availability to Europe Must improve efficiency of harvest and reduce waste (100kg per person per year rich or poor) 99

100 Global Drivers (2) Energy sector SMART grids Reduce environmental impact Emissions and global climate Need to take whole system view SoS perspective Security and Safety E-banking, e-voting, e-solutions IT-OT integration Normal accidents (e.g. Ariane 5) 100

101 Global Drivers (3) Transportation Integrating networks Environmental impact More autonomous decision making Globalisation of economic and social activity Rapid increase in share of economic activity taking place across national borders Jobs move from Europe to Far East Consumers get lower prices 101

102 Outline T-AREA-SoS project background information SoS(E) Basics Creating a Research Agenda Motivation Strategic Research Agenda in SoS (E) 102

103 Strategic Research Agenda 12 themes emerged as priority areas for research Questions within themes must be resolved Problems may be short, medium, or long term 12 priority areas Selected by survey for immediate attention More detailed questions within some themes 103

104 12 Main themes Characterisation and Description of SoS Theoretical Foundations for SoS Emergence Multi-level Modelling of SoS Measurement and Metrics for SoS Evaluation of SoS Definition & Evolution of SoS Architecture Prototyping SoS Trade-off in SoS Security in SoS Human Aspects of SoS Energy Efficient SoS 104

105 Groups that contributed INCOSE SoS WG Consulted Test emerging themes against pain points survey IEEE SoS TC Consulted Expert Community Generated by sign-up through groups and links EU SoS Projects Variety of EU projects either sponsored as SoS, or retrospectively identified by EC project officers as SoS INCOSE IS and IEEE SoS Panel sessions Rome and Genova conferences 105

106 Discrete-event simulation has promise to handle SoS Simulation. Renewable energy sources will continue to contribute to the energy picture, currently at only 5% of total energy, but in 20 years it could reach 20%. Application areas of SoS will soon be enhanced towards reality via Big Data Analytics. CONCLUSIONS SoS (CPS) has been with us for some time Soon System int will be a matter of necessity and not Choice Modeling via Big Data Analytics has the most promise, as no Scientific principles can easily be used to model constituent systems.

107 Amazon.com 2008 CRC 2009 Wiley 2010 CRC Publication link : Contact for ID and PW

108 QUESTIONS? Thank You תודה לך! Please send your questions to

109