Gas Turbine Outlet Temperature Prediction Model
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for determining the washing timing of gas turbines rely heavily on experienced maintenance crews, leading to variability and inefficiency, as the prediction of future turbine outlet temperature is often subjective and difficult to share, resulting in potential over-washing and increased maintenance costs.
Innovation Solution
A method and apparatus that predict turbine outlet temperature using a model based on environmental and operational conditions, along with rotating speed, to identify coefficients for accurate future temperature estimation, allowing for automated determination of optimal washing timing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If washing timing is determined based on future outlet temperature prediction, then maintenance efficiency is improved, but prediction accuracy deteriorates due to reliance on subjective experience
Solution Approach 1:
The patent replaces the mechanical/subjective method of temperature prediction (relying on maintenance crew experience) with an information-processing system (prediction device using machine learning). The prediction device collects operational data, environmental conditions, and maintenance history, then uses a learning model to objectively predict future outlet temperatures and determine optimal washing timing, thereby improving both efficiency and accuracy.
Solution Approach 2:
The prediction device enables the gas turbine system to determine its own maintenance timing autonomously based on predicted temperature trends. Instead of relying on external expert judgment, the system uses its own operational data and the prediction device to self-determine when washing is necessary, improving maintenance efficiency while maintaining high accuracy through automated analysis.
2Reliability
If washing timing is determined by experienced maintenance crews, then prediction capability is improved, but standardization deteriorates due to individual variability
Solution Approach 1:
The prediction device creates a standardized model that copies and codifies the knowledge of experienced maintenance crews into an automated system. The machine learning model learns from historical data including maintenance records and operational parameters, replicating expert judgment capabilities in a standardized, reproducible form that can be consistently applied across different operators and situations.
Solution Approach 2:
The prediction device serves multiple functions: it predicts outlet temperature, determines optimal washing timing, and provides maintenance recommendations. This universal system replaces the need for individual expert judgment with a single standardized tool that provides consistent, reliable predictions across all users, eliminating variability while maintaining high prediction capability.
3Power
If fuel combustion amount is increased to maintain output power, then output power is maintained, but turbine outlet temperature rises above upper limit
Solution Approach 1:
The prediction device performs preliminary analysis of future outlet temperature trends before the temperature actually rises above the upper limit. By predicting temperature increases based on current deposits and operational conditions, the system allows maintenance to be scheduled in advance, preventing the temperature from reaching dangerous levels while maintaining output power through timely cleaning rather than continuous fuel increase.
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
In a method for predicting a turbine outlet temperature at a future use based on a past use of a gas turbine, the turbine outlet temperature at a future use that is an objective variable is predicted by a turbine outlet temperature model by using a parameter in environmental and operational conditions planned for the future use and a rotating speed of a fan planned for the future use as explanatory variables, and coefficients with respect to the explanatory variables are identified through a learning every time when a learning trigger is satisfied. In learning of the coefficients, the coefficients with respect to the explanatory variables are identified based on a result of a regression learning of the explanatory variables and the objective variable of the turbine outlet temperature model that is made by using the parameter, the rotating speed of the fan and the turbine outlet temperature at the past use of the gas turbine.