Solar Plant Startup Optimization

Transcription

1 Solar Plant Startup Optimization Predictive controller to optimize the startup cost in Solar Power Plants. Presentation by: Prabir Purkayastha Co-Authors: K.V. Lakshmi, V. Agrawal, R. Talwar

2 Overview INTRODUCTION OPTIMISATION OF OPERATION OF CSP WITH TES FIRST PRINCIPLES MODELING OF CSP WITH TES DYNAMIC SIMULATOR OVERVIEW SOLAR PLANT START UP OPTIMIZATION CONVENTIONAL PLANT STARTUP OPTIMIZATION REFERENCES

3 Operating Environment Why Optimise the operation of a CSP? To decide when to switch the turbine on (or off) To decide when to draw heat from storage and supply to the power block To adapt to a time-of-day tariff regime These are binary choices that are required to be made at all times during the chosen operating horizon an 8-hour shift or a 24-hour day or a week Freitag, 27. Dezember 2013 Fußzeilentext 3

4 Modeling Approach The very nature of these operational decisions calls for an approach that can deal with integer and real variables that represent the operating choices in discrete time blocks of 15 or 60 minutes duration The decision problem can be modeled adequately by means of a mixed integer linear programming (MILP) model Freitag, 27. Dezember 2013 Fußzeilentext 4

5 Model Framework Objective Maximise the net margin between the tariff earning from net power despatched and the following costs Start-up cost O&M cost Constraints A number of constraints are required to capture adequately the techno-economics of a CSP plant with thermal storage Freitag, 27. Dezember 2013 Fußzeilentext 5

6 Modeling Approach The major data requirements are Hourly solar insolation Plant capacity Solar field Thermal energy storage Power block Performance curves These are obtained from STEAG s in-house firstprinciples model for CSP with TES (presented after the modeling approach). Freitag, 27. Dezember 2013 Fußzeilentext 6

7 Model Framework Decision Variables The following non-negative variables are relevant to the problem. The time block is denoted by the index T. X(T) GROSSE(T) NETE(T) STORE(T) INSOLE(T) INSOLES(T) INSOLEP(T) USTORE(T) STARTE(T) a binary variable denoting whether the power block is running in T gross generation in T net generation in T stored energy at start of T energy captured by solar block in T portion of energy captured by solar block sent to storage portion of energy from solar block sent directly to power block stored energy used in T energy used for power block start-up in T Freitag, 27. Dezember 2013 Fußzeilentext 7

8 Model Framework Constraints A number of constraints are required to capture adequately the techno-economics of a CSP plant with thermal storage. (In the following constraints f and g denote functions a. Energy captured by solar block depends on the insolation. The latter data will be an input to the model. INSOLE(T) = f(insolation_data) Freitag, 27. Dezember 2013 Fußzeilentext 8

9 Model Framework Constraints (continued) b. Gross generation would depend on energy drawn from the energy store and the energy directly drawn from the solar block and the characteristic curves of the turbine and of the storage system. GROSSE(T) = f(ustore(e),insolep(t),starte(t)) c. Net generation is a function of gross generation. NETE(T) = f(grosse(t) Freitag, 27. Dezember 2013 Fußzeilentext 9

10 Model Framework Constraints (continued) d. Gross generation would be limited to a given capacity and whether the unit is available in the time block or not. GROSSE(T) <= captg*x(t) e. Inter-temporal changes in the gross generation would be limited by the given ramp rates applied to gross generation in the preceding time block. f(grosse(t-1)) <= GROSSE(T) <= g(grosse(t-1))) Freitag, 27. Dezember 2013 Fußzeilentext 10

11 Model Framework Constraints (continued) f. Energy balances and limits on energy that can be drawn from storage. INSOLE(T) = INSOLES(T) + INSOLEP(T) STORE(T+1) = STORE(T) + INSOLES(T) USTORE(T) STARTE(T) f(ustore(t-1)) <= USTORE(T) <= g(ustore(t-1)) g. Initial conditions at start of first time block T1. X(T1) = x1 STORE(T1) = store1 Freitag, 27. Dezember 2013 Fußzeilentext 11

12 Ebsilon Model- 1 st Principle Thermodynamic Modeling The Solar-Thermal plant behavior can be examined using the Ebsilon tool Freitag, 27. Dezember 2013 Fußzeilentext 12

13 INTRODUCTION: Solar Plant Start-Up Time Optimization Optimal Start-up strategies vary with the Type and Construction of the Solar Plant: HTF Based/DSG based With or without Storage With or without backup Boiler. Start-Up time computed using predictive controller Validate using dynamic Solar Simulator Reduction in Start Up time Additional Revenue

14 INTRODUCTION: Solar Plant Start-Up Time Optimization Start-up procedure realized at any specific day depends on : Irradiation profile 1. Time of the day 2. The season. Thermal State of HTF in the morning i.e. Initial Condition 1. Number of night hours 2. Night-time ambient temperature 3. Storage usage during the night Other meteorological conditions like wind speed, presence of clouds etc. Reduction in Start Up time Additional Revenue

15 Dynamic Simulator 50 MW Parabolic Trough Solar Thermal Freitag, 27. Dezember 2013 Fußzeilentext 15

16 Dynamic Simulator 50 MW Parabolic Trough Features: Indirect steam generation - HTF Based. Solar Field : 1. Total 112 Loops; 2. Four solar subfields having 28 loops each; 3. Collector operation Modes: Solar Tracking; 4. Complete defocus & specific position; 5. Control for HTF Outlet Temperature. Solar Steam Generation : 1. Three HTF Pumps, 2.Two streams of solar super heaters, steam generator and a preheater; 3. Two streams of reheaters; 4.Steam generator level control. Turbine : 1.Turbine with HP and LP Stage; 2.Single reheat and Governing Controls. Condensate and Feedwater System: Two MDBFPs, Three LP and Two HP heaters. PROCESS + CONTROL MODELS (In TRAX) SHARED MEMORY HUMAN MACHINE INTERFACE (In INTOUCH)

17 , Solar Plant Start-Up Time Possible Optimization Options : Options for Start Up time Reduction 1. Quick Heating of HTF/Steam(using thermal storage) 5. Optimizing inlet temperature and amount of HTF to solar heat exchangers 2. Quick Heating of HTF/Steam (using quick start backup boiler) 4. Better Control System 3. Better Design for Maximum Heat Capture and Minimum Loss

18 Solar Plant Start-Up Time Optimization - Formulation Objective Function is defined to minimize the start up time: where, x[n] are the state variables, u[n] are the control variables, and are the optimizing weights corresponding to state and control variables, N is the prediction time horizon For Nonlinear Plant, Partial Differential Model Equations: Subjected to Constraint Equations: Here, constraint are Thermal Stresses on heat exchangers walls, -σ < Differential Metal-wall Temperatures < +σ Optimized Set Points: a. Flow of HTF to Heat Exchangers b. Temperature at which HTF to be passed to Heat Exchangers c. Heat flow from Storage/Backup Boiler (if present).

19 Functional Block Diagram INPUTS: NON-LINEAR MODEL PREDICTIVE CONTROLLER Predicted DNI Wind Speed HTF Amount HTF Temperature Heat from Storage/Backup Boiler PROCESS VARIABLES FROM SOLAR SIMULATOR SOLAR HEAT EXCHANGERS MODELLING: Preheater, Steam Generator, Economizer OPTIMIZED VARIABLES: SQP OPTIMIZER Iteration Main Steam Flow Drum Pressure Metal and Steam Temperatures OPTIMIZED SET POINTS OPTIMIZATION GOALS Cost Function = Min. start-up time Constraint: Thermal Stresses

20 Dynamic Simulator 50 MW Parabolic Trough Solar HTF Path Freitag, 27. Dezember 2013 Fußzeilentext 20

21 Dynamic Simulator : Parabolic Trough Heat Exchanger Path Freitag, 27. Dezember 2013 Fußzeilentext 21

22 Conventional Boiler Startup Optimization Nonlinear Model Predictive Controller (NMPC) to minimize the start-up time for Subcritical Suratgarh 250MW Thermal Power Plant. Constraints: 1. Fuel Cost 2. Thermal stresses on critical walls of Heat Exchangers. Heat Exchangers are modeled in Modelica Language using OpenModelica Compiler, as the simulation environment. 1.Highly Non Linear Solver available 2.Multi-variable, Time Based Partial Differential Equations 3.Component Based Modeling, using connectors 4.An Open Source & commercial versions available, Dymola etc. 5.Supports huge Multi-Domain Modelica Standard Library 6.External linking with Non-Modelica Environments

23 Conventional Boiler Startup Optimization Optimization Technique: Sequential Quadratic Programming. Optimizer computes the following optimal set-points: 1. Heat input 2. Control valve opening position of HP by-pass station. Objective Function is defined to match the desired pressure profile, for 2 to 60 bar drum pressure: Results: where, = tuning weights = pressure at each time step Start-up time, between 2 to 60bar drum pressure: Actual Start-up Time: 4 to 6 Hours With Predictive Controls : Reduction of mins.

24 Conventional Boiler Startup Optimization- Block Diagram OPTIMIZATION GOALS Inputs Process Inputs & Variables BOILER MODEL (OpenModelica Platform) Optimized Variables SQP OPTIMIZER Iteration Cost function = minimum Optimized Set points REAL BOILER / SIMULATOR MODEL

25 References K. Azizian, M. Yaghoubi, I. Niknia, and P. Kanan, Analysis of Shiraz Solar Thermal Power Plant Response Time, in Journal of Clean Energy Technologies, Vol. 1, No. 1, January Juergen H. Peterseim, Udo Hellwig, Manoj Guthikonda, and Paul Widera, Quick Start-up Auxiliary Boiler / Heater Optimizing Solar Thermal Performance, in SolarPACES 2012 conference Marrakech, September Tobias Hirsch, Heiko Schenk, Norbert Schmidt and Richard Meyer Dynamics Of Oil-based Parabolic Trough Plants - Impact Of Transient Behaviour On Energy Yields, Proc. of the 2010 SolarPACES conference, Perpignan, France (2010). Modelica home page:

26 References Hubert Thieriot, Maroun Nemer, Mohsen Torabzadeh-Tari, Peter Fritzson, Rajiv Singh, Towards Design Optimization with OpenModelica including Parameter Optimization with Genetic Algorithms, in 8th International Modelica Conference, Rüdiger Franke, and Bernd Weidmann, Startup optimization for steam boilers in E.ON power plants, in ABB Review 1/2008. R udiger Franke, B.S. Babji, Marc Antoine and Alf Isaksson Model-based online applications in the ABB Dynamic Optimization framework, in Modelica Association 2008, March 3rd - 4th, THANK YOU

27 REFERENCE SLIDES 1. NMPC CONTROLLER 2. OPEN MODELICA FEATURES 3. CONVENSIONAL SIMPLE BOILER MODEL 4. CONVENSIONAL SIMPLE BOILER EQUATIONS

28 , 1.Conventional Boiler Startup Optimization- Non-linear Model Predictive(NMPC) Controller Receding horizon philosophy A multivariable control algorithm that uses: an internal dynamic model of the nonlinear process. a history of past control moves an optimization cost function, P over the receding prediction horizon. Steps: At time t: solve an optimal control problem over a finite future horizon of N steps. Only apply the first optimal move. At time t+1: Get the new measurements, repeat the optimization. Advantage of repeated on-line optimization: FEEDBACK It will sample the current plant state and computes/predicts a cost minimal control strategy for relatively short time horizon in the future.

29 , 2. Conventional Boiler Startup Optimization- OpenModelica Platform Object-oriented Modeling Paradigm. Declarative type equation based textual language. Multivariable Time Based Differential Equation for simulation purposes. Component Based Modeling: using connectors that reduces modeling errors. Multi-Domain Hybrid Modeling that supports Modelica Standard Library (contains about 1280 model components and 910 functions, from different physical domains like electrical, mechanical, hydraulic domains, etc.). External linking of Library Functions developed in Non-Modelica Languages can be interlinked, like steam tables defined in Fortran Compiler. Modelica Simulation Environments are available as open source and also commercially, like CATIA Systems, CyModelica, Dymola, LMS AMESim, JModelica.org, MapleSim, SCICOS, SimulationX, Vertex and Wolfram SystemModeler. Reference:

30 3. Conventional Boiler Startup Optimization- Component Based Boiler Model INPUTS COMPONENTS In MODELICA OUTPUTS Fuel Flow HPBP VLV POSITION FURNACE BOILER DRUM SUPERHEATERS 1. LTSH 2. SHDP 3. SHPL ECONOMIZER REHEATER Main Steam Flow Drum Pressure Thermal Stresses Metal and Steam Temperatures TUNING PARAMETERS: Heat Transfer Coefficients Thermal and Flow Conductance Emissivity Constants

31 4. Conventional Boiler Startup Optimization- OpenModelica Platform The Simple Boiler Drum Equations : Steam temperature is a non-linear function of pressure inside drum, Feedwater flow is equal to the main steam flow, assuming no blowdown. Steam temperature inside drum is assumed to be equal to the metal wall temperature. Energy transferred from flue gas to riser water walls Q, is a function of the difference between flue gas temperature and metal temperature. where, Heat given to boiler drum Q, is also equal to the energy stored in the water walls and the heat given to the steam. Where, HPBP control linear valve equation: