Deterministic and non-deterministic methods for power sector performance ranking based on machine productivity

Abstract

The objective of this paper is to present a framework for ranking of power sector's performance based on machinery productivity indicators. To rank this sector of industry, the combination of a non-deterministic method, genetic algorithm (hereunder GA), and two deterministic methods, principle component analysis (hereunder PCA) and numerical taxonomy (hereunder NT) are efficiently used for all branches (sub sectors) of the power sector. In other words, all of useful and influential points of the mentioned methods are utilized to measure the power sector's performance. In this study, PCA and NT verify validity of the GA. Furthermore, two non-parametric correlation methods, Spearman correlation experiment and Kendall Tau, are used to determine the correlation among the findings of GA, PCA and NT. As a result, a great degree of correlation is shown. To achieve the objectives of this study, a comprehensive study was conducted to recognize all economic and technical indicators (indices), which have great influences upon machine performance. These indicators are related to machine productivity, efficiency, effectiveness and profitability. Standard factors such as down time, time to repair, mean time between failure, operating time, value added and production value were considered as shaping factors. According to ISIC (International Standard Industrial Classified) codes, all of economic activities in this industry are identified to 2, 3 and 4- digit codes. By these codes, all of branches in the power sector are classified from 2 to 4-digit codes hierarchically. In this study, the data-base used to measure the 10 indicators are formed based on ISIC codes and collected from power sector in a developing country. Then through GA, the best array of branches (DMUs, decision making units,) among the generations produced is selected. This array is the rank of power sub sectors, which optimizes the fitness function in GA. Moreover, by PCA the major impacts of each 10 indicators on the performance are identified. Finally, the result is analyzed to promote the total system performance.