论文部分内容阅读
针对支持向量机在建模中的参数选取问题,提出一种二阶振荡和带斥力因子的粒子群优化算法优化支持向量机参数。采用非线性递减权重平衡算法的全局和局部搜索能力,二阶振荡因子保持种群多样性,提高全局搜索能力。斥力因子使粒子在搜索空间均匀分布,避免陷入局部最优。针对电力负荷的非线性、时变性、受多因素影响的复杂特点,提出一种基于数据的支持向量机预测模型,综合考虑天气、时间因素、历史负荷对预测结果的影响。仿真表明该方法可以建立短期电力负荷的有效高精度预测模型。
Aiming at the parameter selection of SVM in modeling, a particle swarm optimization algorithm with second order oscillation and repulsion factor is proposed to optimize SVM parameters. The global and local search ability of nonlinear descent weight balance algorithm is adopted. The second order oscillation factor maintains the population diversity and improves the global search ability. The repulsive force causes the particles to be evenly distributed in the search space and avoids falling into local optima. According to the nonlinear and time-varying characteristics of power load and the complicated characteristics influenced by many factors, a data-based SVM forecasting model is proposed, which considers the influence of weather, time and historic load on the prediction results. The simulation shows that this method can establish an effective high-accuracy prediction model for short-term power load.