Northumbria Research Link

Transcription

1 Northumbria Research Link Underwood, C. P. (2005) 'A fuzzy logic controller for temperature control of air handling plant', CSTB Matlab / Simulink Building and HVAC Simulation Workshop, 3-4 November, SIMBAD. This presentation is available on the Northumbria Research Link site: Northumbria Research Link: University Library, Sandyford Road, Newcastle-upon-Tyne, NE1 8ST lr.openrepository@northumbria.ac.uk

2 CSTB3 rd Matlab/SimulinkBuildingand HVAC Simulation Workshop 3-4 th November2005,Paris A Fuzzy Logic Controller for Temperature Control of Air Handling Plant CPUnderwood Northumbria University, United Kingdom

3 Summary of This Work Inthisworkthedevelopmentofathree-channelfuzzytemperature controller for air handling plant is considered Sugenoinferencemechanismsareusedandtheiradvantagesforcontrol type problems are highlighted Thefuzzycontrollerisbenchmarkedagainstconventionally-tunedPID control ItisdemonstratedthatthefuzzycontrolleriseasiertosetupthanthePID controller whilst offering superior control tracking performance TheworkhasmadeuseofMatlab/SimulinkaswellastheMatlabFuzzy Logic Toolbox

4 Background: Fuzzy Inference Fuzzify inputs apply rules fuzzy consequent DoM Input membership functions Inference Mechanism Output membership functions Input Other inputs Output Other outputs Aggregate & defuzzify

5 Background: Fuzzy Inference Methods Mamdaniinference(themost common method) outputs to fuzzy sets Examples(T=temperature). If(Tishigh)Then(Valveislow) If(Tishigh)And(dT/dtis negative) Then(Valve is medium) Combinatorial rule setting examples: AND=minorproduct OR = max or probabilistic Sugeno(orTakagi-Sugeno) inference outputs to numerical values Examples(u=valvesignal) If(Tishigh)Thenu=0 If(Tislow)Thenu=1 If(Tishigh)And(dT/dtis negative) Then u = 0.5*input i.e. Sugeno inference can output to some mathematical function of the fuzzy input(s)

6 Simple Comparison of Inference Methods: Heating Coil Control Resulting in sub-saturated valve positioning Can alleviate by working on the MFsoruse Sugeno inference Sugeno case: If(Tislow)Then:Valve=1 If(TisOK)Then:Valve=Constant*Input If(Tishigh)Then:Valve=0

7 Results: Alternative Inference Methods(Simple Heating Coil Test Case) T ao ( o C) T ai ( o C) Mamdani Inference Sugeno Inference Time(mins) u Mamdani Inference Sugeno Inference Time(mins) Outlettemperaturesetpoint:10 o C T ao =airoutlettemperature T ai =airinlettemperature

8 The Problem: Sequencing Temperature Control of AHP 1 Signal h fresh >h recirc Min 0 Heating Free Cooling Cooling - + Temperature

9 Vehicle Simulink Air Handler Model 4 Ta_inlet RHf Ta RHa ha_out1 Enthalpy u(d) Mixing Dampers Valve-h u(h) Heating Coil Valve-c Cooling Coil u(c) 5 Tcwf 1 Ta_outlet Twc 20 Tai 50 Ta ha_out2 RHa Enthalpy1 50 Thwf S hao Thwo Maf Terminator RHai

10 PID Control: Ziegler-Nichols Tuning Parameters Fitted to Open Loop Step Response Tests 18 T ao ( o C) Damper PID tuning parameters: Heating Coil CoolingCoil Loop K c K i Damper Heating Coil Cooling Coil Time(secs)

11 Fuzzy Heating Controller 1. Input MFs 2. Output MFs 3. Rules 4. Surface

12 Fuzzy Cooling Controller 1. Input MFs 2. Output MFs 3. Rules 4. Surface

13 Fuzzy Damper Controller 1. InputMFs(2inputs) 2. Output MFs 3. Rules 4. Surface

14 Simulink Model Adapted for Controller Comparisons PID u(h) ErrorPID Heating Controller u(d) u(c) Ta_outlet PID Ta_inlet -C- SetPoints Ta_inlet Damper Controller PID AHU_PIDControl Tas Cooling Controller u_pid Random Number Rate Transition u_fuzzy u(h) Ta_outlet ErrorFuzzy Saturation1 Fuzzy Heating Controller u(d) u(c) Ta_inlet har_outlet haf_outlet Rate Transition1 Results To Workspace AHU_FuzzyControl Saturation2 Fuzzy Damper Controller Saturation3 Fuzzy Cooling Controller Saturation deltah

15 Comparative Results: Controlled Variable T ao ( o C) T ai ( o C) PID Control Fuzzy Control Time(mins) Setpoints:11 o C(heating);12 o C(freecooling);13 o C(cooling) (50%minF/A)

16 Comparative Results: Signal Fuzzy Control u PID Control Heating Control Damper Control Cooling Control u Time(mins)

17 CONCLUSIONS Fuzzycontroloverairhandlingplanttemperaturesrequiresno tuning and offers better tracking performance than conventionallytuned PID control SugenoinferencehasgreaterflexibilitythanconventionalMamdani inference for fuzzy controllers especially at signal saturation and requires no experience/intuition to apply it Thewellknown chatter aroundthesetpointthatcanarisewith fuzzycontrolhasbeennotedinthepresentworkandneedsrobust procedures to remove it(conventionally, introducing an additional rate variable can help)