基于相空间重构和Chebyshev正交基神经网络的短期负荷预测

Short-term load forecasting based on phase space reconstruction and Chebyshev orthogonal basis neural network

电力系统短期负荷数据具有明显的混沌特性。在讲述混沌中相空间重构的相关理论后,计算了算例中需要用到的延迟时间和嵌入维数。根据正交多项式优越的泛化和预测性能,在简单介绍Chebyshev正交基函数后,构建了单输入Chebyshev正交基神经网络预测模型。由于重构后的相空间中每个相点的分量个数不止一个,故所构建的单输入预测模型无法满足要求。为此,在单输入的基础上,设计了基于相空间重构的多输入Chebyshev正交基神经网络动态预测模型。将该模型运用到短期负荷预测中,取得了很高的精度和很好的预测效果。

The electric power system short-term load data has obvious chaos characteristics. After talking about the related theory of phase space reconstruction in chaos, this paper calculates the delay time and embedded dimension needed in later example. According to orthogonal polynomial prediction＇s superior generalization and forecast performance, the paper constructs a single input neural network forecast model which is based on Chebyshev orthogonal basis after introducing Chebyshev orthogonal basis briefly. Because the point of every phase point in phase space reconstructed is more than one, the foregoing model can not meet the requirements. Therefore, the paper designs a multi input dynamic prediction model of Chebyshev orthogonal basis neural network based on phase space reconstruction. Through applying it to short-term load forecasting, the model gets a high precision and good prediction effect.