基于主成分分析的短期负荷预测模型

电力负荷预测的准确性直接影响到电力系统的安全性和经济性,但在应用神经网络进行短期负测精度造成了显著的负面影响。针对这一问题,本文采用多元统计分析中的主成分分析,根据各主成分贡献率对输入空间进行约简,提取线性无关的输入变量,以此达到压缩变量维数的目的,然后利用考虑模型输入变量相互关系的递推合成BP网络进行预测,使之更符合电力短期负荷预测的特点,提高模型的预测精度。仿真实验的结果表明,该简化模型用于短期负荷预测建模速度快、预测精度高,是一种行之有效的方法。

Short-term Load Forecasting Model Based on Principal Component Analysis

The accuracy of short-term load forecasting in power system has directly influence on its safety and economy.When neural network forecasting model is used to perform short-term load forecasting,its input dimension number is commonly too great as well as its input varianes have heavy self-correlation to have a negative effect on network training efficiency and forecasting accuracy of neural network.Focusing on solving this problem,a principal component analysis in multiple method of data processing is developed.The original input space is reduced by the contribution rate of each principal component to eliminate the multicollinerity of input variables,thus reducing the variable dimension.Then next the recurrent composite BP neural network in use of the correlations between input variances is used to forecast.All these improvement can make forecasting model in accordance with real condition,and improve the accuracy of load forecasting.The results of simulation experiments form a real power system to indicate that the simplified model is effectual.With this method the high modelling speed and high forecasting accuracy can be obtained.