基于级联神经网络的短期负荷预测方法

针对常和BP算法预测速度慢、易陷入局部最优解的缺点，提出了基于RBF网络和BP网络的级联神经网络预测方法，把天气因素和历史负荷对负荷预测值的影响分开考虑，其中RBF子网络用于描述历史负荷的影响，BP子网络则对在RBF子网络中难以考虑的天气因素给出了较好的映射关系，最终将两个子网络组合为一个级联神经网络，一系列的研究算例证明该方法是快速，准确的。

A SHORT-TERM LOAD FORECASTING METHOD BASED ON CASCADE NEURAL NETWORK

To improve the defects in usually used BP algorithms, such as slow forecasting speed and easy to fall into local minimum, a Cascade Neural Network (CNN) load forecasting method is put forward in this paper. In this method the influences of weather factor and past load on load forecasting are separately considered, the RBF sub network is used to describe the relation between the past load and the forecasted load, the BP sub network is used to process the mapping relation between the weather factor and the forecasted load, and a cascade neural network is composed of the two above mentioned sub networks. The results of a series of experimental research show that this method is accurate and fast..