Statistical estimation of representative weeks over a year [power system operation]

In this article, we present a method to determine five weeks in the year (and their corresponding coefficients), that estimate with a good accuracy the effective annual value of important parameters, and for a representative sample of investment studies. This is aimed at reducing the number of data and time needed to perform real time operations. The method is based on optimization first, and then statistical sorting. The optimization step consists in finding out optimal coefficients for each combination of five weeks selected. The final step sorts the combinations to find the best set of weeks by using statistical tools such as principal components analysis and mobile centers classification. As a result, two sets of weeks answer the requirements, and prove that the method works well.