Application Of Lssvm Based On Afsa To Short-Term Load Forecasting

Raising the accuracy of short-term load forecasting is extraordinarily crucial for the security, stability and the economy of the power grid. So significant is the short-term load forecasting known to all, it pays to find a proper method for application.This paper introduced the theory of load forecasting method firstly, and then analyzed the principle and advantages and disadvantages of the Lease Squares Support Vector Machine (LS-SVM). Due to kernel width and regularization parameters in LS-SVM have a great impact on prediction results, improper selection will not be able to guarantee the optimality of solution. While the artificial fish swarm algorithm has good adaptive ability, parallelism and global, it can solve the optimization problem of LS-SVM parameters. This paper improves the basic artificial fish swarm algorithm, solves the basic algorithm to reduce the convergence efficiency and process problem, accelerate the search speed and accuracy. Finally the improved artificial fish swarm algorithm to optimize the LS-SVM model is applied to the short-term load forecasting of each link, the prediction process is discussed in detail, including the historic data preprocessing, using a similar day method to choose forecasting model input load samples. As for the geographical features of A region in Guangxi, the regional load characteristics and influencing factors are dissected and the predicting of load that contains96nodes in A area following. In the end, the feasibility and effectiveness of LS-SVM optimized by AFSA are validated via practical examples.