基于小波变换的混沌神经网络在陕西电网短期负荷预测中的应用

Application of Wavelet Transform Based Chaotic Neural Network in Short-term Load Forecasting for Shaanxi Power Grid

对陕西电网历史负荷数据采用小波变换进行预处理,其中分解后的低频部分作为分析数据。基于混沌理论基础,分析陕西电网负荷的混沌特性,计算延迟时间和嵌入维数,重构系统相空间。计算Lyapunov指数,确定其混沌状态。根据神经网络强大的非线性映射能力,建立合适的混沌神经网络模型,利用BP算法训练网络结构。应用混沌神经网络模型对陕西电网短期负荷进行预测分析,结果证明建立的预测模型比纯BP网络算法具有更高的预测精度和计算速度。

The paper makes the data preproeessing for the historic load data of Shaanxi power grid based on wavelet transform, in which the low-frequency part of the decomposition acts the analysis of data. Based on chaos theory, the paper analyzes the chaotic characteristics of Shaanxi power grid loads to calculate the delay time ＆ embedding dimension and reconstruct the phase-space, and calculates the maximal Lyapunov exponent to determine the chaotic state. According to the powerful nonlinear mapping ability of chaotic neural network, the paper establishes an appropriate chaotic neural network model, and trains each model by BP algorithm. Finally, the chaotic neural network model is used to forecast and analyzes Shaanxi power grid short-term loads, the results show that the prediction accuracy and computing speed of proposed model is superior to that of pure BP network.