基于最小二乘支持向量机和负荷密度指标法的配电网空间负荷预测

Spatial Load Forecasting of Distribution Network Based on Least Squares Support Vector Machine and Load Density Index System

传统的负荷密度指标的求取方法通常采用经验法或简单类比法,难以满足精度要求,从负荷密度与其影响因素存在着某种非线性关系的角度出发,提出了一种基于最小二乘支持向量机(least squares support vector machine,LS-SVM)的配电网空间负荷预测方法。该方法首先引入模糊C–均值算法把各类用地性质负荷聚类为几个等级,建立比较精确的负荷密度指标体系;然后根据待预测地块的规划属性,在体系中为LS-SVM预测模型选出与预测样本特征更为相似的样本进行训练,提高LS-SVM的泛化能力和预测精度;采用遗传算法对LS-SVM预测模型的参数进行自动优化,进一步提高预测模型的适应性和预测精度,实例验证了该方法的实用性和有效性。

Empirical method or simple analogy method are often adopted in traditional methods to obtain load density index, however it is hard to meet the demand of precision. From the viewpoint that there is a certain nonlinear relation between load density and its impacting factors, a spatial load forecasting （SLF） method based on least squares support vector machine （LS-SVM） for distribution network is proposed. Firstly, the fuzzy C-means （FCM） method is led into the proposed method to cluster the land usage loads into several grades and a more precise load density index system is built; then according to the planning attribute of the plot to be forecasted, in the index system the samples that is more similar to the samples to be forecasted are selected for LS-SVM forecasting model to improve the generalization ability of LS-SVM and forecasting accuracy; and then by use of genetic algorithm （GA） the parameters of LS-SVM model are automatically optimized to further enhance the adaptability of LS-SVM model and improve forecasting accuracy. SLF results of actual case verify the practicality and availability of the proposed method.