An integration of computational intelligence techniques in engineering data processing for improving forecast accuracy using cubic-spline interpolation and ANN model

Puteh, Saifullizam and Buhari, Rosnawati (2017) An integration of computational intelligence techniques in engineering data processing for improving forecast accuracy using cubic-spline interpolation and ANN model. ARPN Journal of Engineering and Applied Sciences, 12 (8). pp. 2496-2502. ISSN 18196608

Full text not available from this repository.

Official URL: http://www.arpnjournals.org/jeas/research_papers/r...

Abstract

In engineering operations and maintenance, the system failures and wrong decisions cause critical effects; and subsequently large economic losses. Therefore, designing a model that is able to predict the future trends of an engineering operation system has become an important issue. This paper investigates how to solve prediction problems that have a limited amount of data. In order to demonstrate the efficiency of the proposed integration of computational intelligent techniques, Dissolved Gas Analysis (DGA) was chosen as a case study. DGA is the standard technique used in power transformer condition monitoring and fault diagnosis. Many computational intelligences and statistical techniques have been proposed to develop a forecasting model to predict the future condition of a transformer in transmission system using DGA analysis. Dissolved gasses (e.g., hydrogen (H2), methane (CH4), acetylene (C2H2), ethylene (C2H4) and ethane (C2H6)) in transformer oil are used to detect the type of electrical faults. However, because the dissolved gas’s data, collected from the oil-insulated transformer, require expensive laboratory analysis costs, only limited data has been received. The limited number of data points in time-series data collection is a cause of significant accuracy problems for analysis and prediction results. In this paper, integrations of computational intelligence techniques using Cubic-Spline Interpolation (CSI) and Artificial Neural Network (ANN) are proposed to improve training data set, in order to obtain better results from intelligent prediction models. The cubic-spline interpolation technique is applied to enhance the limited data of dissolved gases, by fitting smoothly to the limited data points and thus generate new and sufficient data. This generated data is used as training data to re-train a Focused Time Delay Neural Network (FTDNN) model, to predict dissolved gases in oil-insulated transformers. Experiments have shown that even using only 10 data points (generated by the CSI technique) can lead to a significantly improved accuracy of forecasting dissolved results.

Item Type:Article
Uncontrolled Keywords:Computational intelligence; soft computing techniques; neural network; cubic spline interpolation; dissolved gas analysis
Subjects:Q Science > QA Mathematics > QA76 Computer software
Divisions:Faculty of Technical and Vocational Education > Department of Engineering Education
ID Code:9363
Deposited By:Mr. Mohammad Shaifulrip Ithnin
Deposited On:13 Aug 2018 11:14
Last Modified:13 Aug 2018 11:14

Repository Staff Only: item control page