论文部分内容阅读
由统计学习理论发展的通用学习方法——支持向量机,在解决小样本、非线性及高维数等问题中表现出许多特有的优势。提出了采用最小二乘支持向量机建立负荷预测模型,它是对标准的支持向量机的一种扩展,降低了问题的复杂性,使得计算速度相对加快。在选取最小二乘支持向量机的训练样本时,采用加权的灰色关联度方法来选择相似日,对不同样本根据其重要性赋予不同的权重,同传统的关联度相比更具客观性。另外,对于最小二乘支持向量机的参数选择问题,针对目前尚无统一有效方法的现状,尝试采用了一种基于蚁群种群的新型优化算法———蚁群算法来优化选择,在很大程度上减少了人为选择参数的主观影响。最后通过实例验证了该模型的有效性,取得了比较满意的预测效果。
The general learning method developed by statistical learning theory, support vector machine, has many unique advantages in solving the problems of small sample, nonlinearity and high dimensionality. A load forecasting model based on least square support vector machine is proposed. It is an extension of standard support vector machine, which reduces the complexity of the problem and makes the calculation speed relatively faster. When choosing training samples of least square support vector machine, weighted gray relational degree method is used to select similar days, and different samples are given different weights according to their importance, which is more objective than traditional correlation degree. In addition, for the parameter selection problem of LS-SVM, aiming at the current situation that there is no uniform and effective method, a new type of optimization algorithm based on ant colony swarm optimization algorithm --- Ant Colony Optimization To a certain degree, the subjective influence of anthropogenic selection parameters is reduced. Finally, the validity of the model is verified by an example, and a satisfactory prediction result is obtained.