Application of Firefly Algorithm to Gaussian Process-based Prediction of Electric Power Damage Caused by Typhoons

Firefly algorithm (FA) is an optimization technique inspired by an intelligent behavior of firefly swarms. This paper presents an application of FA to Gaussian process (GP)-based prediction of electric power damage caused by typhoons. The GP prior model for prediction is trained by a separable least-squares (LS) approach combining the linear LS method with FA. The proposed prediction system yields not only the predicted amount of damage but also its confidence measure. Simulation results based on actual data of typhoons that hit or came close to the Amami archipelago in Japan are shown to illustrate the effectiveness of this prediction system. Keywords—Damage caused by typhoon, electric power system, firefly algorithm, Gaussian process model, prediction.