WP2 Benchmarking: performance evaluation

1
WP2Benchmarking performance evaluation
HYCON Kick-off meeting

• Eduardo F. Camacho
• Univ. of Sevilla
• Spain

2
Objectives

• To use benchmarking for testing control
  technologies of hybrid systems developed in HYCON
  NoE.
• To help integration by the execution of joint
  activities within the NoE centred on common
  benchmark problems .
• To foster excellence through the identification
  and then adoption of best practices.

3
WP2 Main tasks
a) Development and implementation of a
Benchmark Model-Guide b) Preparation of
benchmarking cases c) Implementation of
Benchmark exercises d) Maintenance of the
Benchmarking Service e) Dissemination of
Benchmarking and results
4
Tasks (M1-M18)
Task 2.1 Development and implementation of a
Benchmark Model-Guide Definition of the
Model-Guide requirements Design and
implementation of an interactive environment to
assist the designers of a benchmark Task 2.2
Preparation of benchmarking cases Preparing the
installation and models for the following
benchmark exercises Preparing the benchmark
documentation Task 2.3 Implementation of
Benchmark exercises (first phase) Announcing the
call for the Benchmark exercises Selection of
the accepted proposals and scheduling experiment
time slots.
5
(No Transcript)
6
Deliverables Milestones (M1-M18) Deliverables
D2.1.2 - Web based Interactive Environment D2.2.2
- Documentation for Benchmark exercises M1 and
I1 D2.3.2 - List of accepted experiments Milesto
nes M2.1.1 - Set qualitative and quantitative
criteria for benchmarking activities. M2.2.1 -
Benchmark installation I1 and model M1 available
for remote experimentation M2.3.1 - A list of
experiment proposals from the research Community
for M1 and I1
7
WP2 Man-month allocation
8
WP2-Benchmark

• Benchmark proposers
• Industry- academia
• Internal-external NoE

Benchmark specs.
Interactive tool (model guide)
Benchmark preparation

• Accepted experiments
• Industry- academia
• Internal-external NoE

Benchmark exercise
Benchmark analysis
9
WP2-Benchmark

• Model-Guide
• Facilitate the proposal and preparation of
  benchmark exercises and
• Provide a common format for the description of
  benchmarks
• Description of the system (including control
  objectives. operating constraints, etc.)
• Performance indexes How the benchmark exercises
  are going to be measured
• .
• Preparation of benchmarking cases
• Preparing the system (real or virtual)
• Web-based (remote experimentation, publishing
  data and results).
• hybrid models of increasing levels of complexity
  to provide scalability.
• Implementation of Benchmark exercises
• Call for the BE, collecting and selecting
  proposals. (Not only open to NoE)

10
WP2 Experiment
Client 2
Client n
Client 1
Internet
SERVER (matlab server, EJS, CUBE www server, OPC,
)
BD -user data - Resources information -
Experiments Definition Results,
COMMAND SERVER AND QUERIES
Local platform
camera
plant
11
Benchmarking
INET
12
(No Transcript)
13
(No Transcript)
14
Functionality of the Server

• Monitorization and supervision
• Interaction between the user (remote) and the
  plant some remote actions (depending on the
  experiment), camera, microphones
• BD repository
• Several models of the plant with different
  complexity
• Experiment results (publishing experimental data
  of the tests carried out)
• Control specifications (inputs, outputs,
  controller, set points, operating constraints,
  performance indexes, time,)
• Users data
• Scheduling of experiments.

15
WP2- I1 Solar cooling plant

• Two control levels
• Low level control
• Security
• Start-up and close-down sequences
• Single loops (PID)
• High level compute the operation regime of each
  element in real-time in order to fulfil demand
  and minimize energy. Optimization strategy with
  operational constraints.

16
(No Transcript)
17
Control System
Fieldbus
18
(No Transcript)
19
Control problems

• Solar collector field.
• Solar radiation
• Inlet temperature
• Cooling machine
• Disturbance heater temperature (on/off)
• Lab air conditioning
• optimization

20
Tests

• Solar field
• Start-up and shut-down
• Setpoint tracking and disturbances rejection
  during normal operation
• Accumulators on/off policy Safety (overheating)
  policy
•  Cooling machine
• Start-up and shut-down
• Setpoint tracking and disturbances rejection
  during normal operation
• Gas heater on/off policy
• Constrained control of the cooling machine inlet
  temperature
•  Whole plant solar field and cooling machine
• Start-up and shut-down
• Calculation of the optimum strategy accumulator
  on/off and gas heater on/off
• Tracking of load demand (room temperature in
  laboratory) minimizing auxiliary energy
•  

21
WP2 M1-Automotive engine idle speed control

• Activation condition
• Gear idle and accelerator pedal released
• Objectives
• drive and keep the crankshaft speed in the idle
  speed range, preventing engine stall
• reject crankshaft torque load disturbances with
  smooth transients
• minimize fuel consumption and emissions
• Classical control synthesis techniques applied to
  controller synthesis
• µ-synthesis Hrovat Bodehimer, 1993
• L1 control Butts, et al., 1999, Hinf control
  Carnevale Moschetti, 1993
• Sliding-modes control Kjergaard et al, 1994
• LQ-based optimization Abate Di Nunzio, 1990
• Hybrid system approaches
• Maximal safe set Balluchi et al., 2000,
  Assume-guarantee approach Balluchi et al.,
  2002, Control-to-facet Balluchi et al., 2004
• Multivariable control Onder Geering, 1993

22
Problem formulation for MPFI SI engines

• Control Problem
• Given a value of n0 , ? and TLM, determine
  whether there exist
• an ignition control spark and
• a throttle valve control ?(t)
• that maintain the crankshaft speed n(t) in the
  given range n0 ? under
• any drivers action on clutch pedal,
• any load torque TL(t) in 0,TLM .

?(t)
MPFI Multi-point FueI Injection SI Spark
Ignition
23
Throttle valve and spark ignition actuators
24
Other possible benchmarks

• UCAM
• Air-Traffic Control benchmark. Stochastic hybrid
  system. Problem definition and simulator (Java).
  (In collaboration with UPAT)
• IC Exhaust Catalyst Purging. NOx trap and diesel
  particulate filter. 2-3 modes of operation
  occasional purging Realistic models, software,
  public-domain specs. (Subject to terms of NDA
  with Ford.)
• UNIDO
• Learn2Control A System for Project-Oriented
  Learning
• UNISI
• Telelabs.
• Football
• US-UV
• Four tanks plant
• Detailed models of sugar refineries

25
Univ. of Dortmund Learn2Control A System for
Project-Oriented Learning
26
UNISI Automatic Control Telelab
playback access to acquired data, from remote
computer linked to a collection site through the
Internet.
supports real-time configuration and observation
of experiments
27
UNISI Different laboratory remore experiments
Remote experiments through a web based interface
On-line camera for equipment inspection
Predefined or user provided controllers
Downloadable templates and models of controller
and processes
28
WP2 - UNISI

• Share UNISI expertise with web-based remote
  experiments

http//www.dii.unisi.it/control/act

• Will build a truly "hybrid" lab experiment an
  automatized table soccer (continuous dynamics
  with impacts logic decisions) that may become a
  benchmark for hybrid control strategies.

• Contribution to various tasks in WP2(research
  collaboration with Sevilla already ongoing)

29
US Telerobot
30
WP2 US The four-tank process
31
Four-tank plant
32
Integrates Labview Matlab/Simulink
33
WP2 What comes nextshort term

• Questionaire circulated for inputs (Nov)
• Definition of the Model-Guide requirements
• Interactive environment
• Contributions
• Meeting CDC04 ?
• Draft circulated for corrections (Xmas-Jan05)
• Deliverable (Jan05)
• Model guide (starts Jan05) ....

34
WP2 What comes nextlong term

• Model guide finished May05
• Benchmarks finished by Dec05
• Call for Benchmark exercises Dec05
  (CDC-ECC05)

35
CDC-ECC05