超短期负荷的聚类分析及其变化情况预测

电力系统的超短期负荷值及其变化情况对系统调度具有重要意义,提出了一种优化的聚类算法对超短期负荷在一天中的变化情况进行归类并预测。由于模糊C均值聚类对初始聚类中心敏感,不能准确收敛于全局最优解,加入蚁群聚类,从而自动获得最佳聚类数目,采用模糊神经网络对聚类结果（负荷值的变化情况）进行预测。通过对相似日（非负荷因素如经济、气象等相似）的历史数据仿真实验,验证算法的合理性、有效性,为日后负荷调度提供决策依据。

Clustering Analysis of Super Short-term Load and Forecasting of its Variations

Super short-term load value and its variations are of great importance to the system scheduling. This paper introduces an optimized clustering algorithm to classify and forecast the super short-term load within one day. The fussy c-means clustering can not converge at the general optimal solution due to its sensitiveness to the initial clustering center. The ant clustering is added for automatic acquisition of the best number of clustering. Fussy neural network is used to forecast the clustering result（ variation of the load value）. Simulation experiments on historical data of similar days（ similar economical and meteorological factors other than load factor） are made to verify the rationality and validity of the algorithm,thus providing a basis for decision-making on future load scheduling.