Embedded Controllers for Increasing HVAC Energy EfficiencyIncreasing HVAC Energy Efficiency

by Automated Fault Diagnostics

J. Vass, J. Trojanová, R. Fišera, J. Rojíček, j , , j

Honeywell Prague LaboratoryAutomation & Control Solutions

GREEMBED Workshop, Stockholm, April 12th 2010

Outline

• Honeywell Building Solutions (HBS)

• JACE controller (embedded system)

• HVAC diagnostics- Air Handling Unit (AHU)

• AFDD algorithmD t Cl i- Data Cleansing

- Mode Detection- System Observationy- Fault Isolation

• Examples

Honeywell

Transport & Power12% Aerospace

Specialty 12% Aerospace

40%Materials

16%

Automation and Control SolutionsSolutions

32%

Automation & Control Solutions (ACS)

More than 50,000 employeesProfileProfile

HoneywellSecurity

HoneywellSecurity

HoneywellBuilding Solutions

HoneywellBuilding Solutions

HoneywellProcess Solutions

HoneywellProcess Solutions

HoneywellSensing & Control

HoneywellSensing & Control

HoneywellE & C Control

HoneywellE & C Control

HoneywellLife SafetyHoneywellLife Safety

Embedded System (JACE)

• JACE = JAVA® Application Control Engine• Integration Controller

Integrator- Integrator Integration of communication protocols (LON, BACnet, Modbus etc.)Graphical interface & Built-in web serverData processing capability (e.g. AFDD)

- Controller Freely programmable DDC controllerFreely programmable DDC controller

Multiple protocols

Graphical interface

JACE

interface(e.g. AHU schematics)

JACE: Integration ControllerI t ti f i b ildi b t• Integration of various building subsystems- HVAC; Lighting; Metering

• Multiple protocols to communicate with:p p- Plant controllers (Boiler, Chiller, Air handlers)- Room controllers (VAV terminals, Fan coil units)

JACE

I/O modules Sensors

ActuatorsWWW interface

AHU1 AHU2 AHU3

JACE

Energymeters

Bli dLi htiBoiler Chiller RTU

VAV1 VAV2XL10 XL12

BlindsLighting

FCU1XL10

Next Generation of JACE controllers

• HVAC Diagnostics in JACE- Automated fault detectionHVAC equipment - Boilers, Chillers, AHUs, VAV terminal units, Fan coils Auxiliary devices - Pumps, Fans, Valves, etc.

- Present AFDD results using:Present AFDD results using:Online diagnostic screens - GUIDiagnostic reports - Charts, tables, recommendations

• Challenge = Address practical aspects:Diff t l l f i t t ti- Different levels of instrumentation (Scenarios according to available sensors)

- Contextual information from BIM( A t ti ll dif di ti l b d i HVAC(e.g. Automatically modify diagnostic rules based on given HVAC configuration)

- Monetization(C d d d i f fi i l l )(Convert degraded equipment performance to financial loss)

HVAC Equipment

exhaust air

Air Handlers Fans

VAV Boxes

H Cfresh

coils

air

Pumps

compression tank Chillers

boilercombustion gas

evaporator

filter

Boilers

gas

drain

condenser

compressorcooling towercondenser

water pumpHW circulating

pumpfuelair

p p

Air Handling Unit (AHU)

from exhaust

Return air temperatureReturn air relative humidity

zones

Occupancy

to

p ystatusDamper position

to zonesoutdoor

air

H C

Supply air temperatureO td i Mixed air Supply air Supply air temperature- set point

Outdoor air temperature

Mixed air temperature

Supply air temperature

Outdoor air relative humidity

hot water chilled waterHeating/cooling coil valve control signal

y

Types of AHU Faults

• Hardware Faults- Abrupt Faults

• Control System Faults- Control Loop Performance

Damper faults - stuck damper Valve faults - stuck or leaking

valve (of heating/cooling coil)

Missed setpoint (permanent offset)Oscillating control signal - e.g. due

to wrong PID parametersHeating/cooling failure Sensor faults - frozen sensor,

noisy data, missing values,

Excessive overshoot Long settling time

- Control Strategyoutliers- Performance Degradation Air filter clogging

Control Strategy Simultaneous heating & coolingUsing mechanical cooling instead

of ventilation (free cooling)gg gHeating/cooling coil scaling Sensor drift/bias

of ventilation (free cooling)Wrong sequencing (Late/early

start & stop)

AFDD implementation in JACE

Module 1: Data Cleansing

Control Signals: Oscillations & Outliers

no oscillations (valid control strategy)

oscillations (control strategy fault)

outliers

zoomed view

outliers

oo ed e• oscillatory control signal causes:

- unnecessary opening & closing of the valves- unnecessary movement of the dampersy p

• this results in premature wear and failure of:- valves- actuators

Method 1: Oscillation Detector

higher standard deviation lower standard deviation

raw data

1 STD [ ]STD Ratio =

1 STD [ ]x ky k

smoothed data(low-pass filtered)

threshold = 3

Method 2: Outlier Detector

low-pass filtered

high pass filtered

[ ] [ ] [ ]z k x k y k

high-pass filtered

three-sigma limits:

[ ] 3 [ ]z zUCL k k [ ] 3 [ ]z zLCL k k

three sigma limits:

Module 2: Mode Detection

AHU Operation Mode

100%Mi i b

Control signal

Heating Cooling il l

Mixing box damper

coil valve coil valve

0%

min

Outdoor air temperature

Mode 1

Heating VentilationMode 2 Mode 3

Cooling gmode mode mode

Mode 4

Unknownmode

AHU Operation Mode

• AHU mode- Determined by controller using temperatures (or enthalpies)

AFDD algorithm detects AHU mode using control signals only- AFDD algorithm detects AHU mode using control signals only• Finding stabilized mode:

- Heating mode: ch > 30 minHeating mode: ch > 30 min- Cooling mode: cc > 30 min- Ventilation mode: cv > 30 min

Sampling period (e.g. 5 min)

Threshold on control signals

( 1) if ( ) 2% and ( ) 2%( ) s i iCCh HCtc UT t tU

Counter for

( g ) g

( ) ( ) ( )( )

0 otherwise

( 1) if ( ) 2% and ( ) 2%

s i i

s i i

CC

c CC

h HCh

HC

c

U

t

t T t tc U

Counter for Heating mode

Counter for ( ) ( ) % ( ) %( )

0 otherwise

( 1) if ( ) 2% and ( ) 2%

s i ic CCc

C

H

C

C

i iCHc

t

t T t tU

c

U

Counter for Cooling mode

Counter for ( 1) if ( ) 2% and ( ) 2%( )

0 otherwises C

vCi iCv Hc

ct T t t

tU U

Counter for Ventilation mode

Module 3: System Observation

Vector of Measurements: CCHCSASSAMA UUTTT ,,,,

Vector of Measurements:

Observation o2 Function of supply air temp & setpoint

Th ibl l 1 0 1

if1

Three possible values: 1, 0, -1

T

T

2 if 0if 1

)(

SASSA

SASSA

TTTT

o setpoint is met

above setpoint

T

T2

if 1)(

SASSA

SASSA

TTp

below setpoint

Observed State

• Mapping between observations & states- Each state defined by particular combination of o1, o2 & o3

• 45 possible states of the AHU (states are mutually exclusive)• 45 possible states of the AHU (states are mutually exclusive)- States are classified as normal and abnormal

Heating Mode Cooling Mode Ventilation ModeHeating Mode Cooling Mode Ventilation Mode

Module 4: Fault Isolation & Identification

Mapping Table

• Mapping between abnormal states & faults- Binary diagnostic matrix- Represents the expert knowledge

1 f lt1 = fault0 = don't know

State s12 implies faultsF2 F3 & F7F2, F3 & F7

State s1implies faultsF2 & F6

Fault Aggregation

• Fault Relevance is updated in time using CUSUM: )1(),(,0max)( tRksfmtR ijii )(),(,)( f ijii

Previousvalue

Currentvalue

Forgetting factor k=0 3

Mapping function output 0 or 1 valuevalue factor, k=0.3output, 0 or 1

vanc

et R

elev

Faul

t

Example 1: Stuck Heating Valve

• Air was heated despite UHC = 0 (heating control signal)• Controller compensates by increasing UCC (cooling control signal)• Simultaneous heating & cooling wasted energy financial losses• Simultaneous heating & cooling wasted energy financial losses

detected by d t l i

coolingdata cleansing

e

ventilation

Most likelyl

evan

ce

Most likely fault is F2(stuck heating

only one AHU state

at each time

ault

Rel heating

valve)instant

(confirmedFa

correct fault cannot be isolated

(confirmed by building technician)

Example 2: Oscilating control signalC With D t Cl iWithout Data Cleansing With Data Cleansing

Data cleansing needed to avoid detection of non-existing faults

Summary

• AFDD algorithm for HVAC equipment- HVAC system = Major energy consumer in buildings

Red ce energ asting b a tomated detection of HVAC fa lts- Reduce energy wasting by automated detection of HVAC faults Abrupt hardware faults; Performance degradation

- Performed by Honeywell JACE controller (embedded system)- Graphical visualization in JACE AHU scheme, Measured data, Observed state, Fault relevances

AHU diagnostics• AHU diagnostics- Rule-based diagnostics (based on APAR by Schein et al.)- Data cleansing moduleData cleansing moduleDetect raw data errors (outliers, oscillations, etc.) Protects the AFDD algorithm from wrong decision (hoax faults)

- Diagnostic mapping tableMeasurements → Observations → States → Faults

Fault aggregation- Fault aggregation Fault relevance (of each fault) is updated in time