Application of Firefly Algorithm-Based Optimization of SVR Parameters in Short-term Power Load Forecasting

Because firefly algorithm has such advantages as good global performance and high convergence precision, it is used to optimize the SVR penalty coefficient C and kernel parameter σ. A random disturbance is applied to the position of the brightest firefly in the iterative process to improve the original firefly algorithm, and a higher convergence rate and optimization accuracy are obtained. Optimized parameters are used for a short-term load forecasting to improve the prediction accuracy. This method is used to find the optimal parameters and make the regression forecast. Compared with grid search method, genetic algorithm and particle swarm optimization algorithm, the prediction results demonstrate modified firefly algorithm better than other several algorithms for parameter optimization.