摘要
针对电力系统短期负荷预测中,高维大样本环境下支持向量机算法面临的耗时增大与维数灾问题,将序列最小优化算法(SMO)和粗糙集(RS)理论相结合,提出了一种新的算法——RS-SMO算法。该算法主要是用粗糙集理论进行负荷预测属性的约简,然后用其生成的边界集作为SMO的训练子集,从而使训练集的维数和规模有所减少。采用河北省某市的实际负荷数据进行算例分析,并对RS-SMO和SMO算法的预测结果进行了比较。结果表明,提出的RS-SMO算法有较高的预测精度。
For Short-term load forecasting, high-dimensional environment, large sample support vector machine algorithm facing time-consuming and increased number of disaster-dimensional problem, this article will be the smallest sequence optimization （SMO） and the rough set （RS） theory, put forward a new algorithra, RS-SMO algorithm. The algorithm was used on rough set theory attributes the load forecast reduction, and then set the borders of their generation as a subset of SMO training, so that the training set of dimensions and size of the reduction. One city in Hebei Province using the actual load data analysis examples, and RS-SMO and SMO algorithm results were compared, the results showed that the RS-SMO algorithm for a higher forecast accuracy.
出处
《电力科学与工程》
2010年第2期32-35,共4页
Electric Power Science and Engineering
