基于混沌理论和模型识别神经网络样本预处理技术的负荷组合预测模型研究

Load forecasting models combining sample pretreatment techniques based on chaos theory and pattern recognition

针对神经网络在输入变量选择问题上存在的缺陷,提出不考虑影响负荷的所有因素,仅用混沌理论处理后的负荷样本作为神经网络的输入。但是这种不考虑影响因素的方法不能明显提高预测精度,因此引入一种只考虑影响因素的样本处理方法,并将这两种方法的预测结果进行组合预测。仿真结果表明将这两种方法组合是可行的和有效的。

To solve the defects of input variable selection for nueral networks, a new method is proposed which only uses the load sample treated based on chaos theory as the input of the neural network. The method dose not consider factors influencing the load and can not improve the forecast precision. Therefore, a sample treatment method which only considers the influencing factors is introduced. The prediction results of the two methods are combined. Simulation shows that combination of the two methods is feasible and effective.