Model for CHP operation optimisation

Transcription

1 Jožef Stefan Institut, Ljubljana, Slovenia Energy Efficiency Centre Model for CHP operation optimisation Stane Merše, M.Sc., Andreja Urbančič, M.Sc. Workshop Cogeneration Operation in Competitive Markets January 2003, St. Veit a.d. Glain, Carinthia, Austria 2/12/2003 Industriepark, Greenonetec 1 Outline 1. Problem formulation 2. Model definition on TETOL case study 3. Example of modeling approach for Steam boiler characteristics definition 4. Conclusions 2

2 Problem formulation Key project task: long-term & short-term optimization tool for efficient CHP plant operation in new market conditions Optimization objectives (short term) Optimal CHP operation Optimization criterion - economic Considering market environment and uncertainties 3 Optimization results - deterministic model The optimization tool will enable to: by: reduce operation cost (and/or) increase incomes from electricity sales computation of optimal operation load by production units (Load dispatching - Linear programming) computation of optimal start up - times (Unit commitment - Genetic alghoritem - GA) computation of optimal charging/discharging dynamic of heat storage 4

3 Model definition objectives to build the system model simple enough to enable application of faster optimisation algorithms, to reach the accuracy of the model high enough, to have robust answers on the system optimal response to the changes of varying external parameters in ranges of their expected values, to express the model relations in the most general form in order to enable designing the description of the search space which suites the most for GA needs. This is intended particularly to open other modelling options, besides the most known, MILP form for the unit commitment problem. time accuracy 1h - static model (dynamic Heat storage) 5 Modeling approach on TETOL case study Scope of the model The TE-TOL (CHP plant Ljubljana) system Deterministic non-linear model Accurate and simple for short term optimization: daily; weekly Extension to the long-term model Heat demand - planned Status of TE-TOL on the electricity market: fixed contracts, preference dispatching, efficiency constraints 6

4 MODELLING ISSUES ADDRESSED Non-linearity Boiler characteristics non-linearity's Heat sub-stations capacities/dependent on external parameters (distribution heat system return temperature, which can not be optimised) Sensitivity analysis to start-up costs (lack of appropriate data - constant estimate) End of interval values Heat storage boundary conditions Qualified electricity producers status compliance (efficiency constraint) 7 TE-TOL CHP system Boilers (B) 3 coal steam boilers (B1, B2, B3) 2 fuel oil peak-load steam boilers (B6, B7) 2 fuel oil peak-load hot water boilers (B4, B5) Steam turbines (T) T1, T3 - Extraction condensing T2 - Back pressure extraction 2 steam extraction levels (8.5 bar and 2.5 bar) Heat storage unit (SU) Heat exchangers (HE) high & low pressure (LP1, LP2, HP1, HP2) Steam demand (8.5 bar) and hot water demand 8

5 HEAT DIAGRAM OF THERMO POWER PLANT LJUBLJANA F11 P11 T11 T2 O2 F21 P21 T21 T3 O3 F31 P31 T31 F32 T1 O1 C1 F12 E1 C2 E2 C3 E3 F10 P10 T10 P15 F20 P20 T20 F30 F16 P25 P13 P23 T33 Q92 Q01 Q02 Q102 A4 A5 A6 F01 P01 T01 F02 P02 T02 A10 A11 T81 F81 Q4 F4 Q5 F5 Q91 Q101 Q8 F61 F71 E8 A7 A12 T80 F80 A1 A2 A3 A8 A9 F00 P00 T00 Q03 F03 P03 T03 F04 P04 T04 F05 P05 T05 EAST WEST EAST LEGEND: AKU ČD ČGN ČK ČKT ČKO ČN ČO - Water pump KPV - Chemical water treatment ZDV - Make-up water collector of hot water boiler - PIPING TO HEAT STORAGE TANK 9 ČST ČZK - Heat storage tank - Make-up water pump - Condensate pump - Hot water boiler pump - Turbine condensate pump - Heater condensate pump - Feedwater pump - Static pressure pump - Condensate pump of start-up condenser G GN GO GV HE E K KT PL - Generator - Low-pressure heater - Water heater - High-pressure heater - Ejector cooler - Ejector - Boiler - Turbine condenser - Check valve R RL RV RZ T ZD ZDO ZN ZK - Cold reduction station - Control flap - Control valve - Start-up reduction station - Turbine - Make-up water collector - Raw make-up water collector - Feed-water collector - Start-up condenser - LIVE STEAM PIPE - STEAM LINES FROM 0 TO 50 BAR - HEATING WATER PIPES - CONDENSATE OR FEED-WATER PIPES Reference energy system TE-TOL 10

6 TE-TOL model elements Equations in optimization: I/O characteristics: boilers, turbines dynamic model: heat storage constraints - unit bounds (min/max) heat balances by pressure level Parameters: data from available measurements (on line, unit s documentation, etc.) were used 11 Mathematical models/ devices Identification and selection of approximation options: Boiler characteristics constant (auxiliary boilers) / linear (main boilers)/ piece-wise linear (tested)/ quadratic Turbine characteristics linear/ piece-wise linear/ quadratic sum of extraction flows/ separate flows by extraction level (for steam turbines) Boiler start-up costs constant/ linear/ piece-wise constant/ exponential Heat station bounds constant/ linear/ piece-wise linear/ other: constant - varying by time interval with temperature forecast 12

7 Variables Q... flows s ON/OFF status Characteristics and bounds STEAM BOILERS I/O characteristics: Q fuel = f ( Q technical minimum/maximum gradients minimal on and off time In cost function fuel costs variable O&M (non-fuel) costs start-up costs (fuel related) Approximation options L /PWL, fuel costs, C, start up costs Technical parameters efficiencies were determined by an indirect method from measured data gradients from operation manual off-time: TE-TOL data Cost function parameters fuel, O&M costs approximated from annual values start costs - current estimate 13 out ) BOILER EFFICIENCY data collection and modelling Efficiency determined by indirect method from measurements Heat output [%] y = -0,2186x 2 + 0,4625x + 0,6787 1) 2) Efficiency [%] Influencing parameters - outdoor temperature - flue gases temperatures -O 2 in flue gases -output mass flow 70 Fuel input [%] I/O curves derived from efficiency 1) point-wise 2) from quadratic fit of efficiency 14

8 BOILER I/O CHARACTERISTICS linear approximation (L): 110% 105% 94% 93% Q FUEL ( t ) = a Q( t ) + bo( t ) Heat output [%] 100% 95% 90% 85% 80% 75% 93% 92% 92% 91% 91% Efficiency [%] 70% 90% 80% 85% 90% 95% 100% 105% 110% 115% 120% Fuel input [%] piece-wise linear approximation (PWL) Q N ( t) = k Q i i ( t) c0o( t) FUEL + i= 1 Heat output [%] 110% 105% 100% 95% 90% 85% 80% 75% 94% 93% 93% 92% 92% 91% 91% Efficiency [%] 70% 90% 80% 85% 90% 95% 100% 105% 110% 115% 120% Fuel input [% max power] 15 The impact of L & PWL differences on optimisation results (MILP) Heat demand/production [MW] Heat and electricity production Heat storage Heat demand Heat production Heat storage contents [MW] Electricity [MW] Hours L PWL 16

9 Comparison of results Optimal solutions were equivalent for both cases: same optimal cost function value, electricity production, amount of heat storage in-outflow over period, some differences between hours: similar operation status have changed the time of occurrence, some difference in heat contents at the end of period (due to shift of production between hours). Conclusion: Optimisation results as well show that linear fit is accurate in a sense that, the improved fit by a piece wise function do not affect the optimal result. 17 HEAT SUB-STATIONS - capacity limits Capacity limits at heat sub-station: Low pressure heat exchanger capacity limit depends on district heating return water temperature (varying by heat demand, outdoor temperature) and on heat storage water temperature Modelling options: Constant/ linear/ piece-wise linear/ other: constant - varying by time interval with temperature forecast - outside the model (model input data) 18

10 TE-TOL model MILP formulation For each hour: 15 independent real variables 8 integer (binary) variables 44 inequality equations 2+1 equality equation (DH, Steam, + Electricity) 19 Long term model approximation less detailed description - reduction of model complexity: linear relations, constant efficiencies (independent from load) difference from 2 to 10% 20

11 Conclusions Linear modelling gives comparable results, but with significant reduction in computational time compared to more detailed non-linear modelling. CHP optimisation problem was defined as a linear mixed-integer problem (MILP), divided into two subproblems: unit commitment (UC) - solved using a GA economic dispatch (ED)- by linear programming (LP) Dynamic extensive process - going on with further model testing and validation, future changes, etc. For more details about modeling (equations, etc.), look in Project Deliverables (D 1.4, D 2.1, D 2.2, ) 21 Thank you for your attention! Stane Merše, M.Sc., Andreja Urbančič, M.Sc. stane.merse@ijs.si, Workshop Cogeneration Operation in Competitive Markets January 2003, St. Veit a.d. Glain, Carinthia, Austria Industriepark, Greenonetec 22