Application of Evolutionary Computation for Identification of Parameters of Sky Luminance Distributions

Abstract

The International Commission on Illumination has adopted a set of fifteen standard sky luminance distributions (SSLD). One of our previous papers uses a least squares method to sort sky luminance measurements from a station located north of Bangkok into fifteen slots, each corresponding to each type of SSLD, to appraise the statistics of occurrence of each sky type at the location. The paper also uses Kittler's method (Kittler; and others, 1997) to prepare paired values from the measurements sorted and obtains parameter values of the gradation function and those of the indicatrix function separately from measurements in each slot. This paper applies evolutionary computation to obtain a set of parameter values for both gradation function and indcatrix function simultaneously under least squares criterion. This method enables all measurements in each slot to be used. The resulting luminance distribution function thus obtained for each sky type possesses lower bias error and lower mean square error than those of the corresponding type of SSLD when both are evaluated against measurements.