动量自适应学习速率梯度下降法神经网络电力负荷预测

电力系统负荷预测的精度将直接影响电力系统的经济效益和用电的安全和稳定,是电力负荷预测的重要组成部分。利用人工神经网络可以任意逼近非线性系统的特性,将其用于短期负荷预测。在标准的BP网络中加入了动量项和自适应学习速率,预测结果表明比标准BP算法具有更好的性能。在相同的情况下,连续预测六天的负荷和一年的负荷,结果都证明了研究方法具有一定的实用性。

BP Neural Network Power Load Forecasting Based on Momentum and Adaptive Learning Rate

The accuracy of the forecast of power system load,which is an important part of the forecast of power system load,will directly affect the economic of the power systems and its security and stability. The use of artificial neural network could get the similar feature like nonlinear system and use it on the short term forecast. Researches add momentum and adaptive learning rate into the improved BP network and combinate the same type of vague and mapping results when building input networks shows that it has better performance than standard BP algorithms. In the same circumstances,the result of forecasting a six day load and one year load shows that the method of this study has a certain practicality.