基于支持向量区间回归的供热负荷概率预报

Probabilistic prediction approach of heat load based on support vector interval regression

提出了一种基于支持向量区间回归(SVIR)的概率预报方法,克服了点预报法无法确定预报结果波动范围的缺点。该方法利用支持向量回归确定SVIR模型的初始参数,再利用两个径向基网络分别辨识SVIR的上限和下限,可以同时给出置信区间和点预报。最后,以某热力站实际供热负荷数据与BP神经网络点预报方法进行比较,验证了该方法的有效性和实用性。

A probabilistic prediction approach was proposed based on the support vector interval regression（SVIR） .The initial parameters of the SVIR model are determined by the support vector regression,and its upper and lower bounds are identified by two radial basis function networks.The confidence intervals and the point prediction outputs can be obtained at the same time by the proposed approach.The validity and practicability of the proposed approach were proved by comparative simulations by the proposed model and the BP network model using the heat load data collected from a practical heat supply station.