wide – wp5 achievementscse.lab.imtlucca.it/hybrid/wide/documents/im2011_07.pdf · wide...

WIDE – WP5 Achievements

WIDE

ist-wide.dii.unisi.it

2

WIDE

Agenda

• Introduction• Overview• Concluding Remarks• Discussion

WP5 Introduction

WIDE


4

WIDE

WP5 Overview

• Proof of concept: water network

• Objectives– The role of WP5 is to validate the

concepts developed in WP2, WP3 and WP4 on a large-scale water distribution system.

– It will also serve to motivate and circumscribe the class of large-scale systems WP2, WP3 and WP4 will focus on, providing useful engineering inputs to the theoretical developments.

– The concepts will be demonstrated in a software demonstration of the entire water network using a simulation environment based on MATLAB/SIMULINK and an experimental demonstration of part of the water network.

• Tasks– T5.1 Definition of the water network

demonstration – T5.2 Dynamical model of the water network

and model decomposition– T5.3 Control design for the water network– T5.4 Development of the water demo

demonstrator – T5.4 Testing and evaluation of the solution

T5.1 Definition of the Water Network Demonstration

WIDE


6

WIDE

Pilot Definition

• Definition of Pilot 1: DMPC of the Barcelona Water Transprot Network

– Definition of the part of Barcelona water network used to illustrate DMPC

– Definition of the control problem

• Definition of Pilot 2: Wireless Control of Valve in Barcelona Water Network

– Definition of the part of Barcelona water network used to illustrate wireless control

– Definition of the control problem

WIDE


7

WIDE

Test Pilot 1: Demo of DMPC in the Barcelona Network

c70PAL

c125PAL

CPIV_1

d110PAP_2

C110 PAP

CPII_2

d54REL_8

d100FCE_9 c100FCE

VSJD_29C100 LLO

d80GAVi80CAS_6 i 85CRO

c80GAVi 80CAS85CRO

c70LLO

VCA_28

CRE_8

CGIV

CCA_3

d115CAST

C115 CAST VCR_27

CB_4

dPLANTA_7

ApotLL 1

CPLANTA70_6

CPLANTA50_7

d10COR_10

c10COR

PLANTA10_10

CC50_9

CC70_12c70FLL

VZF_33

VCT_34

VT_35

c100 BLLsud

VRM_32

VCO_37

CCO_16

VS_36

CE_13

VE_31 c130BAR

CF200_14CF176_15

d200BLL_11c200BLL

VF_30

d176 BARsud_13

C176BARsud

c200 BARs-c

VB_38

VP_39

VMC_44d200ALT _15

c200ALT

VBSLL _43 d200 BARnord_17c200BAR

nordd101 MIR_18

c101MIR

CA_17

c100BLLcentre

VPSJ _41

d100BLLnord _16

c100BLLnord

VCOA_40

d70BBE

c70BBE

CC100_11CC130_19

CRO_20

aMS bMS_21

aPou CAST1, 5 , 6 i 7

bPou CAST1, 5, 6 i 7

aPousB

aPousE

bPousB _26

bPousE _23

d130BAR

c140 LLO

d125PAL_1

nAportA1_19

nAportA2_21

n70PAL_20

n100LLO_22

n70LLO_23

n140LLO_24

n100BLLsud_25

n70FLL_26

n200BARs-c

n100BLLcentre_29

AportT

vAdd_45

vAdd

vAdd_47

vAdd_48

vAdd_53

vAdd _54

vAdd_55vAdd_56

vAdd_57

nAportT_32

ApotLL2

c176BARcentre

n176BARcentre

ApotA

vAdd_60

vAdd_61

VBMC_42

vAdd_64

vAdd_50

vAdd _55

vAdd_309vAdd_308

d205FON d320FON

CPI CPII CPIII

c205 FON c320FON

c356 FON

d175PAP

C175PAP i 135PAP

CP_ATLL

d82PAL_1

c82PAL

d130LSE C LSE

cLSE

A CAST 8 b CAST 8

d145 MMA

C MJ

c145MMA

C175 BVI

C MMA II

C SB C VIL 1 C SC 1 C BEG 1

d150 SBO

d175LOR

C LOR

d184ESP

d255 BEG 2

C BEG 2

d114 SCL 1

d190 SCL 2

C SCL 2

d135 VIL

d185 VIL

C VIL 2

d369BEG 3

C BEG 3

d450 BEG 4

C BEG 4

c175LOR

c185VIL

c190 SCL2

c255BEG

c150SBO

c135VIL

c114SCL

C184 ESP

C369 BEG3

C450 BEG4

d195TOR

C CRc195TOR

d147SCC

C SCC

c147SCC

d205 CES

C CES 1

C205 CES

d263CES

C CES 2

C263 CES

d252 CGL

d313 CGL

C CGLL2 (D3)

d246CGYd200 CGY

d268 CGY

c200CGY

c246CGY C252

CGL

C CGY1 (D2)

C CGY1 (D5)

C CGLL1

C CGY2

c268CGY

d361 CGY

c361CGY

C CGY3

d374 CGL

C CGLL2 (D5)

C374CGL

C313CGL

d200BSO

C BSO

c200BSO

d300 BAR

d437 VVI

C300BAR

C F300

d320 MGB

C PAP1

c320MGB c437

VVI

C MG1 C TIB

d400 MGB

C MG2

c400MGB

C MG3

c475MGB

C VVI

c541 TIB

d255CAR

C CAR

c255 CAR

d260 SGE

d328SGE

c260SGE

c328 SGE

C SG1

C SG2

d250 TBA

C250 TBA

C TBA 2

C C UAB C CS C CM

d197 BET

C UAB

d215VALL d184SMM

C SMM

d132 CMF

d200 FDM

C FDM

c238UAB

c200 FDM

c255BEG

c150SBO

c132CMF

c215 VALL

c250 VASAB

C VALL 1

C VALL2

c260VALL

c275BEV

C BEV

d169CMECTBA1

c169 CME

aPous G 3 i 4 bPou s G 3 i 4

c55BAR

d120POMC_MO

C_TP

C_SA

d171SAM

d144TPI

d190TCA

c144TPI

c144TPI c190TCA

V a 55BAR

c120POM

d117MTG

c117MTG

v117MTG

n135SCGc135SCG

d151 BON d197GUI

c151BON

c197GUI

d225GUI

c225GUI

C_BONC_GUI

1

C_GUI2

c70CFE

V_70CFE

n100BESV_BSC

c100BES

V_FoAl

C_SA

C_SCM

C-PR

VFIN176 A 140

V 200a150 AL

T

d150ALT _15

c150ALT

V -Hc176 BARn

d202CRUc202CRU

CCRU

Transport Water Network

WIDE


8

WIDE


Supervisory MPC Control Architecture

WIDE


9

WIDE

9

Interface Simulator/DMPC Controller


WIDE


10

WIDE

Test Pilot 2: Demo of Wireless Control Functionalities

CABLE CONNECTION

WIRELESS CONNECTION

• The overall idea is to check the use of wireless connection between the sensors and the remote station.

• Each signal will be doubled (cable connection and wireless connection) as a protection to the overall control system. Both sets of data will be compared and contrasted afterwards.

• Double aim: a) Transference of information through wireless. b) Operational feasibility/constraints regarding the overall valve control (delays,

etc.)

Closed-loop Control of a Valve at the Regulatory Level

WIDE


11

WIDE

Phase 1: Lab Test


AGBAR Labs UPC Labs

WIDE


12

WIDE

Valve and Servomotor

Flow / pressure meters

Remote stationCurrent working situation

• Pressure and flow data are stored every 0.4 s in the remote station.• Input data + desired pressure downstream the PID controller in the remote station changes valve position.• The remote station sends flow and pressure data every 4-5 s to the Control Centre.• Set points of pressure or flow can be sent from the Control Centre to the remote station when necessary.• Alarms due to values out of the acceptable range also pop up at the alarms panel in the Control Centre.

Phase 2: Real Test


WIDE


13

WIDE

Demonstration Objectives

• Demonstrate the benefits of distributed MPC by comparing overall costs for legacy and distributed MPC control strategies.

• Demonstrate the ability of the proposed distributed MPC solution to replace the legacy control in multiple steps.

– In each step, the APC will be implemented on a selected network part only.

– Important for practical application and advanced process control life cycle; standard requirement by process operators.

– Demonstrating the ability of APC to work with some controllers switched to a backup strategy based on the legacy control

• Demonstrate control over wireless network on the lowest control level.

– Robust algorithms assuming communication delays, packet dropouts, etc. will be demonstrated on flow or tank level control with wireless sensors and actuators.

WIDE


14

WIDE

Benefit Demonstration

• As a baseline, the price of the current PID-based control strategy will be evaluated from 3 days historical process data.

• Two distributed advanced control strategies will be evaluated:

– Distributed MPC with the central coordinator

– Distributed MPC without the central coordinator

– In either case, local MPC controllers will control groups of tanks.

• All strategies will use identical initial tank level settings and demand data.

• Demonstration Scale:– Full model of the Barcelona Water

network• Demonstration Platform

– MATLAB/Simulink

1

2 5

3 4 6

7

1

2 5

3 4 6

7

Coordinated Distributed MPCControllers communicate with coordinator only

+ low number of iterations

Distributed MPCControllers communicate with neighbors only.

+ modularity, robustness

1

2 5

3 4 6

7

1

2 5

3 4 6

7

C12 C345 C67C12 C345 C67 C12 C345 C67

COORDINATOR

1

2 5

3 4 6

7

11

22 55

33 44 66

77

Water network for the demonstration of the different distributedcontrol strategies

WIDE


15

WIDE

Sequential Replacement of Legacy Control• Replacement of legacy control in

successive steps.• Demonstration will be done on a part

of the network model.• The part of the network is divided

into 3 groups of tanks– At the beginning, the network section

under consideration is controlled by the PID-based legacy control.

– Nominal trajectories are pre-computed for all flows.

• Group #1 is switched to advanced control while the rest is controlled by the legacy control strategy.

• Groups #2 and #3 are sequentially switched from the legacy to the advanced control.

• Simulated communication failures can test switching from the advanced to legacy control used as a back-up.

• Demonstration scale– Part of the network

• Demonstration platforms– Model – Simulink– Backup control and backup/APC switching:

Simulink, using blocks emulating Honeywell Experion DCS

– Advanced Control – Honeywell URT platform communicating with Simulink over the OPC bridge

WIDE


16

WIDE

Wireless Control

• Applied to the base control layer with fast sampling (1 second)

– Network effects will not show up at the advanced control layer with sampling interval 1 hour.

• Demonstrating Integration of WSN and DCS.

• PID controller will be given, in place of raw measurement, a process value estimated by a network-aware Kalman filter.

– Kalman filter will receive measurement data from the network with time stamps provided by the sensor.

– Kalman filter will further receive time stamps of manipulated variables provided by the actuator, to account for network delays in the MV channel.

• Network delays and data losses will be simulated using network models developed in WP2.

• Demonstration scale– Single flow/tank level controller

• Demonstration platform– Simulink– Real Site

WIDE


17

WIDE

Distributed MPC

• Local controller controls water level for a group of tanks by defining set-points of internal flows

– Internal flows – between tanks of the same group, controlled by PID controllers at the DCS level

– External flows – between tanks of different groups. Determined by a consensus negotiated between neighboring controllers.

– Set-point for an external flows is imposed by the predefined controller

• The implementation of the local controller allows its use in 2 strategies

– With a global coordinator defining shadow prices of external flows

– Without a global coordinator: shadow prices are defined locally by one of the neighboring controllers

Distributed MPC Controller

Distributed MPC Controller

Demands (DV)Internal Flows (MV)

External Flows (MV)

(Controlled) External Flows Set Points

Communication Network

Tank Levels

Internal Flows Set Points

Global Coordinator

Interface

Neighboring Controllers Interface

External Flows Coordinator Local Optimizer

Local Demands Predictor

Constraints and Prices

coordinated / non-coordinated

WIDE


18

WIDE

Backup Strategy

• Legacy PID-based solution guarantees operational safety, and a baseline operation.

• Required as a back-up control strategy with advanced control

• Back-up control uses pre-computed nominal flow/level trajectories.

• Flow controllers either– follow nominal flow trajectories– track flow trajectories of the master level

controller.

• Master level controller (one per tank)– Tracks the nominal level trajectory– Typically controls the inflow with largest

capacity for the tank.

• Advanced control (MPC), if available, provides set-points for flow controllers

pump or valve

PID

PID

Level controller

Nominal level trajectory

Flow controller

Tank water level

SP from Distributed MPC Remote CAS

Master Level Controller

CAS

Flow Controller

Nominal flow trajectory

CommunicationNetwork

SW1

SW2

• In the case of a level control on a tank being in the back-up mode due to the communication failure, MPC can still control other tanks

– Manipulated variable in the back-up mode can be treated as a disturbance by MPC

T5.2 Dynamical model of the water network and model decomposition

WIDE


20

WIDE

Network Model

121 control variables

67 system states (tank volumes)

88 perturbations (water demands)

15 additional constrains (related to the nodes equalities)

Matlab implementation

WIDE


21

WIDE

Graph Analysis

WIDE


22

WIDE

Subsystem Decomposition using Graph PartitioningPROPOSED SOLUTION – Graph Partitioning Algorithm

• Novel algorithm for automatic LSS subsytem decomposition based on a graph partitioning approach

• Step 0: Graph representation of the LSS system

• Step 1: Subsystem decompostion based on identifyng loosely coupled subsystems using graph theory

• Step 2: Solution refinement using some auxiliary criteria

PUBLICATIONSPaper submitted to IFAC World Congress 2011 and Journal of Process Control

WIDE


23

WIDE

Daily forecast

Hourly forecast

Demand Forecast Time Series Model

WIDE


24

WIDE

Electricity Cost Model

- An electricity cost model that takes into account the price of electricity depending on the day, hour and period of the year has been developed and taken into account in the MPC formulation.

T5.3 Control design for the water network

WIDE


26

WIDE

1. Energy/Production Costs

2. Operation-safety Costs

3. Stability Costs

( )1

0( )

N

i i ii k

FCP a t q k C−

=

= ∆ ⋅ ⋅

∑ ∑

1sec,

sec,0

( )1 Nj j

jjj k

V V kFCS c

N V

−

=

−=

∑ ∑

( ) ( ) 21

max,0

11 Nj j

jjj k

q k q kFCE d

N q

−

=

− − = ∑ ∑

MPC Objective Function Formulation

WIDE


27

WIDE

Distributed MPC

SOLVED PROBLEM• Design of Distributed MPC for truly large scale

systems

• Class of target systems (Barcelona WN):• Centralized optimization problem impossible to

solve due to high number of variables (~5000) and constraints (~10000)

• Long sampling period (1 hrs) iterative solution with nearly optimal performance

• Two controller types:• WITH Central Coordination – fast coordination,

useful also for large scale optimization on a single computer

• WITHOUT Central Coordination – local controllers coordinate flows between neighbors only (modular architecture)STATE OF THE ART

• Well known methods based on dual decompositions; however, with problematic speed of convergence

• No application of MPC to full-scale water distribution network

• LS system partitioning algorithms for DMPC are missing

Distributed MPC

Coordinated Distributed MPC

WIDE


28

WIDE

Hierachical Decentralized MPC PROPOSED SOLUTION - Decentralized MPC

• A hierarchy of a local MPC controllers (one per each subsystem) is designed.

• Only one optimization per MPC controller is necessary. Coordination is established thanks to the unidirectional flow of shared variables.

PUBLICATIONSPapers presented at ACC’10, LSS’10 and submitted to IEEE Journal

WIDE


29

WIDE

Wireless PID Control in Lab Environment (1)

WIDE


30

WIDE

Wireless PID Control in Lab Environment (2)

0 2 0 0 4 0 0 6 0 0 8 0 0 1 0 0 0 1 2 0 0 1 4 0 00

0 . 2

0 . 4

0 . 6

0 . 8

1

1 . 2

1 . 4T a n k L e v e l

T i m e ( s )

Lev

el (

mm

)

S e t P o i n t

L e v e l

0 1 0 0 2 0 0 3 0 0 4 0 0 5 0 0 6 0 0 7 0 0 8 0 0 9 0 0 1 0 0 00

2

4

6

8

1 0

1 2S a m p l i n g T i m e E S e n z a

S a m p l e s ( s )

Tim

e (s

)

S a m p l i n g T i m e E S e n z a

0 1 0 0 2 0 0 3 0 0 4 0 0 5 0 0 6 0 0 7 0 0 8 0 0 9 0 0 1 0 0 00

5

1 0

1 5

2 0

2 5T i m e R e c e p t i o n M a t l a b

S a m p l e s ( s )

Tim

e (s

)

T i m e R e c e p t i o n M a t l a b

0 1 0 0 2 0 0 3 0 0 4 0 0 5 0 0 6 0 0 7 0 0 8 0 0 9 0 0 1 0 0 00

2 0 0

4 0 0

6 0 0

8 0 0

1 0 0 0T i m e D e l a y

T i m e ( s )

Del

ay (

s)

E S e n z a s a m p l i n g t i m e

M a t l a b r e c e p t i o n T i m e

0 1 0 0 2 0 0 3 0 0 4 0 0 5 0 0 6 0 0 7 0 0 8 0 0 9 0 0 1 0 0 0- 5

0

5

1 0

1 5

2 0T i m e D e l a y d i f e r e n c e

T i m e ( s )

Del

ay d

ifer

ence

(s)

D i f e r e n c e b e t w e e n E S e n z a a n d M a t l a b t i m e s

T5.4 Development of the water demo demonstrator

WIDE


32

WIDE

Network Simulator

WIDE


33

WIDE

Simulated volume (blue) compared to real volume (red) for some tanks

2 0 4 0 6 0 8 0 1 0 0 1 2 0

2

3

4

5

6

7x 1 0

4 d 1 0 0 C F E

t i m e ( h )

m3

2 0 4 0 6 0 8 0 1 0 0 1 2 0

2 0 0

3 0 0

4 0 0

5 0 0

d 1 1 4 S C L

t i m e ( h )

m3

2 0 4 0 6 0 8 0 1 0 0 1 2 0

1 0 0 0

2 0 0 0

3 0 0 0

4 0 0 0

d 1 1 5 C A S T

t i m e ( h )

m3

2 0 4 0 6 0 8 0 1 0 0 1 2 0

0 . 5

1

1 . 5

x 1 04 d 1 3 0 B A R

t i m e ( h )

m3

2 0 4 0 6 0 8 0 1 0 0 1 2 05 0 0

1 0 0 0

1 5 0 0

2 0 0 0

2 5 0 0

3 0 0 0

d 1 3 2 C M F

t i m e ( h )

m3

2 0 4 0 6 0 8 0 1 0 0 1 2 0

4 0 0

6 0 0

8 0 0

1 0 0 0d 1 3 5 V I L

t i m e ( h )

m3

2 0 4 0 6 0 8 0 1 0 0 1 2 02 0 0

4 0 0

6 0 0

8 0 0

1 0 0 0

d 1 7 6 B A R s u d

t i m e ( h )

m3

2 0 4 0 6 0 8 0 1 0 0 1 2 0

5 0 0

1 0 0 0

1 5 0 0

2 0 0 0

2 5 0 0

3 0 0 0

d 4 5 0 B E G

t i m e ( h )

m3

2 0 4 0 6 0 8 0 1 0 0 1 2 0

1 0 0 0

2 0 0 0

3 0 0 0

d 8 0 G A V i 8 0 C A S 8 5

t i m e ( h )m

3

S i m u l a t o r

R e a l d a t aU p p e r l i m i t

S a f e t y l e v e l

Network Simulator Validation

WIDE


34

WIDE

34

Interface Simulator/DMPC Controller

WIDE


35

WIDE

Electrical and Software Interfaces for Wireless Test

WIDE


36

WIDE

MATLAB Toolbox for E-Senza Wireless Devices

Gra

phica

l Reu

lts

C

ontro

l Pro

cess

Dat

a Ac

quis

ition

Con

figur

atio

n

Port Set-Up

Network

Module

Variables Declaration

Is data OK?

Read Port

no

Check module data source

yes

Store data

PID

Write Port

Plot Data

Acquisition Time end?

END

yes

no

Network Management

Process Controller

Matlab

Real Tank System

Sensor Environment

Data

Esenza Coordinator

SC130

Esenza BlockSB130

AO AI

T5.5 Testing and evaluation

WIDE


38

WIDE

Centralized MPC Results

• Centralized MPC Control of Barcelona water network has been implemented by means of PLIO tool.

• To test and adjust the MPC controller some different scenarios have been studied. Parameters to take into account in the calibration of the model are:

– Initial and security levels in tanks– Objective function weights: economical, safety and maintenance factors. – Working with different sources operation:

• Llobregat source set at constant flow (Scenario 1)• Fixed sources at real flow (Scenario 2)• Source optimization. The optimizer calculates the flow for each time step

inside the operational limits of each source (Scenario 3)

WIDE


39

WIDE

Centralized MPC: Scenario 1 (1)

Flow(m3/s)

Case 1 Case 2

Llobregat surface source 3 0

Llobregat underground source 2 2

• Barcelona’s average input flow is about 7.5 m3/s.• In case 1 an important part of the total demand is taken from Llobregat. • In case 2 only a 25% of the total demand is taken from Llobregat. It is expected that

an important part of the network consumption is going to be taken from Ter.• These two scenarios are interesting from the point of view of the behaviour of the

economical cost.

Scenario 1: Llobregat source set at constant flow

WIDE


40

WIDE


• Conclusions• It exists a strong and linear dependency between economical cost and the

operation of this two sources.• In order to reduce the total cost it is necessary to maximise the quantity of

water taken from Llobregat.

Electrical cost Water cost Total cost

Day 1 52,42 47,58 100,00

Day 2 46,65 53,35 100,00

Day 3 48,10 51,90 100,00

Day 4 47,57 52,43 100,00


Day 1 -50,27 +91,34 +17,11

Day 2 -47,94 +72,77 +16,47

Day 3 -48,37 +78,27 +17,36

Day 4 -47,67 +71,06 +14,58

Case 1

Case 2

Increase/decrease % in comparison to case 1

2 2.5 3 3.5 4 4.5 5 5.520

30

40

50

60

70

80

Llobregat flow (m3/s)

Elec

tric

al/w

ater

cos

t (%

)

2 2.5 3 3.5 4 4.5 520

30

40

50

60

70

80

Ter flow (m3/s)

Elec

tric

al/w

ater

Cos

t (%

)

Electrical cost regressionWater cost regressionqLl=4qLl=3

Optimised sources case

WIDE


41

WIDE

Centralized MPC: Scenario 2

Sources flow is imposed by using real data obtained from AGBAR historical database.

It is an interesting case study in order to compare centralised MPC control and current control applied regarding to transportation cost.

It is a previous step before comparing centralised and decentralised MPC control.

Important improvement in electrical cost, which represents between 10% and the 25 % of the real operation cost.

Total cost using MPC control is between 4 and 8 % lower than the real one.

Scenario 2: Sources set at real flow

Current control

MPC

Increase/decrease % in comparison to current control


23/07/2007 33,13 66,87 100,00

24/07/2007 34,66 65,34 100,00

25/07/2007 32,00 68,00 100,00

26/07/2007 31,29 68,71 100,00


23/07/2007 -23,27 +0,00 -7,71

24/07/2007 -10,56 +0,00 -3,66

25/07/2007 -20,61 +0,00 -6,59

26/07/2007 -18,58 +0,00 -5,81

0 2 0 4 0 6 0 8 0 1 0 00

0 . 5

1

1 . 5

2U n d e r g r o u n d s o u r c e s f l o w

t i m e ( h )

flow

(m

3/s)

i S J D S u bi P o u s C a s t

i C a s t e l l d e f e l s 8

i M i n a S e i x

i E s t r e l l a 1 2

i E s t r e l l a 3 4 5 6i E t a p B e s o s

0 2 0 4 0 6 0 8 0 1 0 00

1

2

3

4

5

6S u r f a c e s o u r c e s f l o w

t i m e ( h )

flow

(m

3/s)

i S J D S p f

v A b r e r a

v T e r

0 1 0 2 0 3 0 4 0 5 0 6 0 7 0 8 0 9 0 1 0 05

6

7

8

9

1 0

1 1T o t a l i n p u t f l o w

t i m e ( h )

flow

(m

3/s)

WIDE


42

WIDE


42

In this case electrical and water costs are minimised, so it is expected a higher improvement in the total cost referring to the scenario with fixed sources.

Taking into account results obtained in the first case study (constant fixed flow in Llobregat source) a solution with maximum average flow from Llobregat source is expected.

In the optimization results shown the term that guarantees stability in control elements (pumps and valves) is on.

Underground sources’ water cost is penalized to avoid its over-exploitation.

Scenario 3: Flow optimization

WIDE


43

WIDE



23/07/2007 33,13 66,87 100,00

24/07/2007 34,66 65,34 100,00

25/07/2007 32,00 68,00 100,00

26/07/2007 31,29 68,71 100,00

Current control


23/07/2007 18,92 -50,70 -27,63

24/07/2007 14,04 -32,56 -16,41

25/07/2007 26,29 -43,91 -21,45

26/07/2007 26,09 -44,43 -22,36

MPC improvement in comparison to current control case

MPC improvement in comparison to fixed sources to real flow case (Scenario 2)


23/07/2007 54,99 -50,70 -21,59

24/07/2007 27,51 -32,56 -13,23

25/07/2007 59,08 -43,91 -15,91

26/07/2007 54,86 -44,43 -17,57

– Big water cost savings, between 30% and 50 %.– Electrical cost has increased regarding to

current control case ([+18,+27]%) and MPC case with fixed sources ([+27,+60]%).

– Total cost has decreased between 13% and 22 % regarding to MPC results obtained with fixed sources.

– Sources flow distribution is the expected one. Llobregat’s source flow is maximized.

wide – wp5 achievementscse.lab.imtlucca.it/hybrid/wide/documents/im2011_07.pdf · wide...

Documents