改进的BP算法在电力系统短期负荷预测中的应用

针对电力负荷短期预测问题,提出了一种基于相似日的神经网络预测方法,分析传统BP算法的不足,提出一种基于Levenbery-Marquardt优化法的BP模型学习算法,在建立具体模型时,对于24点负荷预测,采用24个单输出的神经网络来分别预测每天的整点负荷值,具有网络结构较小,训练时间短的优点,考虑了不同类型的负荷差异,并对四川省电力公司某区一条线路的供电负荷进行短期负荷预测仿真,仿真结果表明其具有较好的预测精度。

Application of improved BP Algorithm in power system short- term load forecast

This paper discusses a neural network algorithm based on similar days for power system short-term load predicting. After analyzing the lack of the traditional BP algorithm, a new BP learning algorithm based on Levenbery-Marquardt optimized method was brought up. While establishing model for 24 points load forecast, we adopt 24 single output neural network to forecast every hour load per day prediction, which has simple structure , short training time and better forecast precision. Considered the difference load type in hours, for some area power supply load of Power Corporation in Province SiChuan, a short-term load forecast is simulated. The simulation shows it has better forecast precision.