Predictive Intelligence Optimizing Performance & Reliability. Neeraj Agarwal Manish Khetrapal

Size: px
Start display at page:

Download "Predictive Intelligence Optimizing Performance & Reliability. Neeraj Agarwal Manish Khetrapal"

Transcription

1 Predictive Intelligence Optimizing Performance & Reliability Neeraj Agarwal Manish Khetrapal 1

2 Outline.. Introduction Basic Architecture The MPC Model Potential Areas Conclusion 2

3 Sustainable Growth Optimize performance Renewable Resources Enhance Efficiency Intelligent Control System Reduce Emission Improve Reliability Minimize Generation Cost 3

4 Prediction is. 4

5 Predictive Intelligence is. The science of forecasting the future. Predictive intelligence means bounding the space of future uncertainties within an estimative framework. 5

6 Similarity to Human Decision Making Sense Assess Predict Optimize Implement Repeat Collect new information Update the memories Forecast the outcome for a variety of possible decisions Determine the best decision Impose the action for this time Collect data, update and optimize 6

7 Predictive Intelligence Structure Data Technology Predictive Intelligence Predictive Analytics 7

8 Control Algorithm Era Conventional Cascade PID Split range Override Ratio Advanced Adaptive control Fuzzy logic control Optimal control Neural network control Model predictive control 8

9 Model Predictive Control 9

10 Real processes are multivariable and often involve significant interaction. Here is a good analogy : Driving a car When driving a car, we make coordinated use of: Steering wheel Brake Accelerator Gear Why? Now consider how well one can drive a car at speed using only steering wheel, with all other controls regulated by an intelligent predictive controller. 10

11 When? Processes are difficult to control with standard PID algorithm (e.g., large time constants, substantial time delays, inverse response, etc. There is significant process interactions between MV and CV. i.e., more than one manipulated variable has a significant effect on an important process variable. Constraints (limits) on process variables and manipulated variables are important for normal control. 11

12 Many organizations are already profiting from the predictive intelligence technologies : An international hospital group reduced time to resolution for critical incidents by 68% and achieved $1.2 million of annualized savings. An International bank uses predictive intelligence to increase campaign response rates by 600%, cut customer acquisition costs in half, and boost campaign ROI by 100%. An airline increased revenue and customer satisfaction by better estimating the number of passengers who won t show up for a flight. Petrochemical Industry Where? 12

13 Process Industry Scenario Source : ARC Advisory group, USA 13

14 Operational Hierarchy MPC Structure Global Steady-State Optimization (every day) Local Steady-State Optimization (every hour) Dynamic Constraint Control (every minute) Supervisory Dynamic Control (every minute) Basic Dynamic Control (every second) 14

15 Implementation Challenges Process Complexity Inconsistent Data Processing Expense Technical Expertise Interoperability Model Pricing 15

16 Project Definition The Predictive Intelligence Process Define the business objectives and desired outcomes for the project Avg. time spent per phase 9% Data Exploration Analyze source data to determine the most appropriate data 12% Data Preparation Select, extract, and transform data upon which to create models. 23% Model Building Create, test, validate & evaluate models, whether they will meet project goals. 25% Deployment Model Management Apply model results to business decisions or processes. Manage models to improve performance, standardize toolsets & minimize redundant activities. 18% 13% 16

17 Design Efforts 17

18 MPC Block Diagram SET POINT CALCULATIONS SET POINTS (TARGETS) PREDICTION PREDICTED OUTPUTS CONTROL CALCULATIONS INPUTS PROCESS PROCESS OUTPUTS INPUTS RESIDUALS MODEL MODEL OUTPUTS + - The MPC Model is build up as an exact replica of the process..

19 MPC Functionality Output Set point Past Output Past Control action 19

20 Integrating MPC with DCS 20

21 Conventional v/s MPC 21

22 Conventional v/s MPC Lagged response Set point Change Response Large overshoot Large settling time Process Disturbance Response 22

23 Conventional v/s MPC Conventional 23

24 Return on Investment 24

25 Advantages of MPC over Conventional Controller Integrated solution Automatic constraint handling No need for decoupling or delay compensation Efficient Utilization of degrees of freedom Can handle non-square systems Assignable priorities, ideal settling values for MVs Realized benefits Consistent, systematic methodology Higher on-line times Cheaper implementation Easier maintenance User Friendly Easier online implementation 25

26 MPC Vendors & Packages ASPENTECH ADERSA EMERSON HONEYWELL ABB INVENSYS DMC plus DMC plus-model HIECON GLIDE SMOC DELTA V Predict RMPCT 3D MPC Connoisseur

27 POTENTIAL APPLICATIONS 27

28 Future Potential Application Boiler Startup Time Optimization Predictive Soot Blowing Combustion Optimization Load - Frequency Control Improving operational Efficiencies Set Point Optimization Boiler Tube Metal Temperature Monitoring L E A D S T O On line Decision support Process Fine tuning Controllable losses identification Efficiency enhancement Guidance for overall improvement Maintenance cost Optimization Manpower Optimization Emission Reduction Thermo Economic Optimization

29 Optimizing Boiler Startup 29

30 Unit Start Up (Boiler Thermal Stress) FSH & HRH TEMP. PREDICTED FSH & HRH TEMP. 30

31 Unit Start Up (Long start up time) PREDICTED MW & STEAM FLOW ACTUAL MW & STEAM FLOW 31

32 Considering : Estimated Savings During Start-Up Average Start up delay of 1 hour during each start, Partial loading loss of 0.25 MU/start Average 25 nos. of starts/year (for 05 units) SAVING CALCULATIONS Revenue Loss due to each 210 MW unit delayed start-up Oil consumption loss due to delayed start (1/24) x 5.04x 10 6 x 0.60 Rs Lac 10 kl x 60,000 Rs Lac Partial loading loss / start 0.25 x 10 6 x Lac Net Saving /year due to timely start 25 x 8.76 Lac 2.19 Crore 32

33 Functional Principle MPC takes into account fuel costs & thermal stresses in critical thick walled tubes components & uses this data to compute optimal set points for boiler startup It is a closed control loop for defining the fuel and HP- Bypass control actions 33

34 Startup Time Reduction WITHOUT MPC WITH MPC 34

35 Response Time Improvement STEAM FLOW UNIT LOAD 35

36 Benefits Accrued Reduction in boiler start up power (10-20%) Reduction in boiler start up time duration Reduction in boiler stress loading Saving on account of APC reduction 36

37 Enhancing Efficiencies 37

38 Optimisation Results J. M/s M. Stuart Dayton Power Plant and near Light Aberdeen, (DP&L), Ohio. J. M. (M/s Stuart Dayton Station, Power Aberdeen, and Light (DP&L) Ohio OPTIMIZER MODE HEAT RATE IMPROVEMENT (%) NOX REDUCTION (%) 38

39 Optimal Real Time Operator Interface CONTROLLABLE LOSSES (Rs./Day) UNCONTROLLABLE LOSSES (Rs./Day) 39

40 Typical Monetary Loss Calculation Average HR loss per day /unit 10 Kcal/KwHr Average monetary loss per day/unit 42, Avg. monetary loss per year/unit Crore Net Monetary Loss per year for 5 units Crore 40

41 Estimating Cooling Tower Effectiveness 41

42 Estimating Cooling Tower Effectiveness RANGE OF CT APPROACH OF CT EFFECTIVENESS OF CT 42

43 PI SERVER TREND CT EFFECTIVENESS CT INLET TEMP. CT OUTLET TEMP. CT APPROACH WET BULB TEMP. CT RANGE 43

44 SET POINT OPTIMIZATION 44

45 Set point Optimization Model 45

46 Set point Optimization 46

47 MS/HRH temperature control 47

48 MS/HRH temperature control 48

49 Benefits Accrued Tighter Control of SH & RH temperatures Prevents temperature excursions Stabilizes Load & throttle pressure Reduced boiler stresses Improves Efficiency 49

50 PREDICTIVE SOOT BLOWING 50

51 The conventional way of boiler cleaning means to clean with steam, air or water. on scheduled timebasis The Traditional Way Depending upon the installation the cleaning system was operated with fixed parameters 51

52 NEED OF THE HOUR An intelligent heating surface cleaning system (Predictive Soot blowing) optimized with regard to the following parameter : Cost efficiency Optimized cleaning intensity Selection of cleaning area according to actual demand Predictive soot blowing means: AS EARLY AS NEEDED AS LESS AS POSSIBLE ONLY THE RELATED AREA 52

53 Predictive Soot Blowing Using Instrumentation Control Predictive soot blowing is the combination of modern-technology which features : 1. Smart pressure and heat flux sensors 2. Smart water cannon and steam lances 3. Smart models and controls 53

54 Furnace Performance Factor 54

55 Why Settle For Predictive Soot Blowing Reduce O&M cost Improve availability Improve thermal performance Meet environmental regulations To prevent accumulation of fireside deposits to remove existing deposits from same. Secondary benefits : Reduction of tube erosion Reduction of usage of water Reduction of use of soot blower Reduction of spray rates (extension of tube life) 55

56 CONCLUSION Just like each & every water drop makes an ocean, so does every bit of information counts. Information can be only few bits & bytes for someone while for others it can be a gold mine. The challenge lies in processing the information, drawing inferences & putting it to optimal use. In today's scenario when many power plants have achieved stability, its time to look ahead to future challenges of sustainability. Predictive Intelligence is one of the stepping stones to optimize performance & reliability. 56

57 The Best Way To Predict The Future Is To Invent It... (Alan Kay) 57

58 THANKS POWERING INDIA S GROWTH 58