Data Mining in Distribution Consumer Database using Rough Sets and Self-Organizing Maps

The objective of this paper is to describe an artificial intelligence based methodology to estimate the daily demand curve of low voltage electrical energy consumers of a electricity distribution company. This methodology uses SOM-self-organizing maps and rough sets to do the estimation. The SOM is used to find a set of curves prototypes, which represents the space of possible curves of the consumers, and also to find the clusters of such curve space. The curves inside of each cluster are then statistically aggregated resulting in a unique curve, the so called typical curve, which is used to represent such cluster. In its turn, rough sets is used to classify each consumer to a typical curve based in some of its features existing in the electricity distribution company database. Consumers' features such as monthly bill, consumer type, number of phases, and so on are used as inputs in order to do the classification. The methodology validation was achieved through out transformer samples. The results found were satisfactory and demonstrate that the proposed methodology is applicable to this type of problem. It was also developed a computational system to the classification of the consumers to the typical curves and to estimate the load curves of the transformers, which can be used by the company as a decision support systems for investments and also for electrical losses analysis