Abstract: Due to the nonlinear and high fluctuation of short-term load data, an improved least squares support vector machine(LSSVM) is proposed for short-term load forecasting. Firstly, the parameters of LSSVM are optimized by particle swarm optimization(PSO) to obtain the optimal LSSVM, which is used to predict and analyse the future load of power system. The improved support vector machine reduces the space complexity and increases the computing speed. An empirical example proves the validity of this prediction method.