基于RAN网及小波神经网电力系统短期负荷预测

Short-term Load Forecasting in Power System Based on RAN Network and Wavelet Neural Network

介绍电力系统负荷预测研究现状,将小波分析与神经网络相结合,构造了一种适用于非线性系统建模预测的小波神经网络。讨论运用小波神经网进行电力系统短期负荷预测的算法及在预测过程中对电网负荷数据进行预处理的方法。首次提出了RAN网新型网络结构并探讨了在电力系统短期负荷预测中的应用。分别应用2种方法对东北电网进行了72h短期负荷预测仿真。仿真结果表明,用小波神经网和RAN网进行建模预测比BP网训练步数大大减少,缩短了网络训练时间,提高了预测精度。

Current status of load forecasting system is described. A wavelet neutral network suitable to non-linear modeling is set up combining wavelet analysis and neutral network. Load forecasting algorithm with wavelet neutral network and pretreatment methodology of power system load data is then discussed. RAN network structure is firstly put forward and its application to short-term load forecasting is introduced. Two methods are used separately in simulating 72-hour short-term load forecasting in Northeast Power Grid. The result shows the training period of wavelet neutral network and RAN Network is greatly reduced than BP network, shortening training period and increasing its accuracy.