基于同伦神经网络的短期负荷预测研究

将同伦方法引入到神经网络的学习训练中,提高了神经网络的学习效率,较好地解决了BP神经网络收敛速度慢和局部极小值等问题.同伦神经网络具有稳定性强,收敛性好的特点,用这种神经网络进行了电力系统短期负荷预测,算例表明了网络收敛快、预测精度高的优越性.

Power system short-term load forecasting based on homotopy method of neural networks

In this paper, we present a learning algorithm of BP neural network based on homotopy method, simulations of which show that the results can be obtained faster than by other methods of BP neural networks. Based on the homotopy method of neural networks, generalization ability and learning efficiency for network can be improved, and the method is robust and has very good convergence property. Applying this homotopy method of neural network on power system short-term load time series forecasting, the results of calculation show that advantages of higher forecasting accuracy, smaller forecasting error, and quick convergence. The method should be popularized.