Anticipatory Monitoring and Control of Complex Systems using a Fuzzy based Fusion of Support Vector Regressors
This paper places itself in the realm of anticipatory systems and envisions monitoring and control methods being capable of making predictions over system critical parameters. Anticipatory systems allow intelligent control of complex systems by predicting their future state. In the current work, an intelligent model aimed at implementing anticipatory monitoring and control in energy industry is presented and tested. More particularly, a set of support vector regressors (SVRs) are trained using both historical and observed data. The trained SVRs are used to predict the future value of the system based on current operational system parameter. The predicted values are then inputted to a fuzzy logic based module where the values are fused to obtain a single value, i.e., final system output prediction. The methodology is tested on real turbine degradation datasets. The outcome of the approach presented in this paper highlights the superiority over single support vector regressors. In addition, it is shown that appropriate selection of fuzzy sets and fuzzy rules plays an important role in improving system performance.
- Research Organization:
- Idaho National Lab. (INL), Idaho Falls, ID (United States)
- Sponsoring Organization:
- USDOE
- DOE Contract Number:
- DE-AC07-05ID14517
- OSTI ID:
- 1166025
- Report Number(s):
- INL/CON-13-30245
- Resource Relation:
- Conference: Future of Instrumentation International Conference,Orlando, FL,10/07/2013,10/09/2013
- Country of Publication:
- United States
- Language:
- English
Similar Records
Predictive based monitoring of nuclear plant component degradation using support vector regression
Online Support Vector Regression with Varying Parameters for Time-Dependent Data