Turbine performance diagnostic system and methods

• US 8,751,423 B2
• Filed: 11/30/2010
• Issued: 06/10/2014
• Est. Priority Date: 11/30/2010
• Status: Active Grant

• Current Assignee: GE Infrastructure Technology, LLC (General Electric Company)
• Original Assignee: General Electric Company
• Inventors: Pandey, Achalesh Kumar, Nanda, Subrat, Mazzaro, Maria Cecilia, Talukdar, Atanu
• Primary Examiner(s): Chen, Alan
• Assistant Examiner(s): Bharadwaj, Kalpana
• Application Number:
  US12/956,689
• Publication Number:
  US 20120136819A1
• Time in Patent Office: 1,288 Days
• Field of Search: 706/14
• US Class Current: 706/14
• CPC Class Codes: F02C 9/00 Controlling gas-turbine plants; Controlling fuel supply in air- breathing jet-propulsion plants controlling air intakes F02C7/057; controlling turbines F01D; controlling compressors F04D27/00; controlling in general G05
• CPC Class Codes: F05D 2260/80 Diagnostics
• CPC Class Codes: F05D 2270/44 active, predictive, or anticipative
• CPC Class Codes: F05D 2270/70 Type of control algorithm
• CPC Class Codes: F05D 2270/71 synthesized, i.e. parameter computed by a mathematical model
• CPC Class Codes: G06N 7/01 Probabilistic graphical models, e.g. probabilistic networks

Abstract

A turbine performance diagnostic system creates a performance report for one or more turbines and includes an assessment module that receives operating data from at least one turbine and produces a performance report from the operating data. The assessment module includes a change detection module configured to determine when at least one parameter has changed beyond an associated threshold and generate an alarm and a root cause analyzer coupled to the change detection module that predicts a root cause of the alarm utilizing a Bayesian Belief Network (BBN). The performance report includes an indication of the predicted root cause.

Claims

1. A turbine performance diagnostic system including one or more central processing units (CPUs) and memory, the one or more CPUs configured to execute software stored in memory to control the turbine performance diagnostic system to receive operating data from at least one turbine and create a performance report for one or more turbines, the system comprising:

an assessment module that receives the operating data from the at least one turbine and produces a performance report from the operating data, the assessment module including;

a change detection module configured to determine when at least one parameter has changed beyond an associated threshold and generate an alarm; and

a root cause analyzer coupled to the change detection module that predicts a root cause of the alarm utilizing a Bayesian Belief Network (BBN), the BBN based on an evidence matrix including a plurality of entries corresponding to previously-detected alarm states in which the at least one parameter changed beyond the associated threshold and, for each of the plurality of entries, an indication of states of additional system parameters relative to additional predetermined thresholds at the times at which the previously-detected alarm states were triggered;

wherein the performance report includes an indication of the predicted root cause.

2. The system of claim 1, wherein the change detection module performs at least one of:

a Hotelling T² transform test, a Likelihood Ratio test, a combination of the Hotelling T² transform test and a Likelihood Ratio test, a z score calculation, a two sample t test, an entropy base test, an offline water detection test, a mean or base value shift test and a high slope detection test to generate the alarm.

3. The system of claim 1, wherein the BBN is formed based on prior operating data related to the at least one turbine.

4. The system of claim 1, wherein the evidence matrix is at least partially formed from operating data from turbines other than the at least one turbine.

5. The system of claim 1, wherein the evidence matrix further includes an indication of the actual cause of the alarm for each portion.

6. The system of claim 1, further including:

a database for storing the BBN; and

a learning tool coupled to the database that modifies the BBN based on the accuracy of BBN root cause predictions.

7. The system of claim 1, further comprising:

a maintenance analyzer that creates a maintenance recommendation based on the report.

10. The method of claim 1 wherein the evidence matrix is at least partially formed from operating data from turbines other than the at least one turbine.

8. A method of determining a root cause of a turbine alarm for at least one turbine, the method comprising:

determining at a computing device that the turbine alarm exists;

determining at the computing device which, if any, other alarm conditions exist when the alarm conditions exist to create an event portion; and

predicting the root cause of the event from the event portion by comparing the event portion to a Bayesian Belief Network (BBN), the BBN based on an evidence matrix including a plurality of entries corresponding to previously-detected alarm states in which the at least one parameter changed beyond the associated threshold and, for each of the plurality of entries, an indication of states of additional system parameters relative to additional predetermined thresholds at the times at which the previously-detected alarm states were triggered.

9. The method claim 8, wherein the BBN is formed based on prior operating data related to the at least one turbine.

11. A system including one or more central processing units (CPUs) and memory, the one or more CPUs configured to execute software stored in memory to control the turbine performance diagnostic system to create a scorecard for one or more turbines, the system including:

an assessment module that receives operating data from at least one turbine and produces the scorecard from the operating data, the assessment module including;

a change detection module configured to determine when at least one parameter including a heat of the one or more turbines has changed beyond an associated threshold and generate an alarm; and

a scorecard generator that generates the scorecard, the scorecard including a status indication, the status indication being based on a characteristic of the parameter that generated the alarm, the characteristic including a change in heat rate of the one or more turbines after a compressor water wash.

12. The system of claim 11, wherein the scorecard is a unit scorecard related only to one of the one or more turbines.

13. The system of claim 11, wherein the scorecard generator generates a fleet scorecard that combines a plurality of unit scorecards.