【摘要】 针对电力负荷预测中对参考临域的选取不精确以及传统神经网络外延能力弱的特点,文中提出一种运用改进BP神经网络来进行电力负荷混沌多步预测的方法,在重构负荷数据相空间的基础上利用夹角余弦取代欧氏距离寻找参考临域,找出训练样本并用改进的BP神经网络来进行负荷的多步预测。通过对南方某城市的负荷数据进行实例分析,利用Matlab软件进行仿真,取得较为理想的预测效果。算例分析结果证明了此方法在电力负荷预测中的可行性。更多还原

【Abstract】 Aimed at inaccurate choose of reference field and weak extension capability of tradition neural network at present,this article introduces a method for chaotic multi-step forecasting of power load using improved BP neural network which uses angle cosine to replace Euclid distance to find out the reference field after restructuring load phase space,then employs BP neural network to help load multi-step forecasting.At last,this algorithm is proved to have better forecasting effect through analyzing the load data of a city of south and the simulation’s curve obtained by using Matlab.The result of case study indicates the feasibility of this method in power load forecasting.更多还原