机构地区: 广东电网公司
出 处: 《广东电力》 2012年第11期16-19,共4页
摘 要: 针对标准反向传播(back propagation,BP)神经网络负荷预测精度不高的缺点,提出利用贝叶斯正则化算法来改善模型的泛化能力,根据河源电网负荷容易受天气影响等特点,给出一种分层的贝叶斯神经网络预测模型,预测结果表明,新的模型具有更好的泛化能力,应用效果良好,提高了负荷预测准确率。 Aiming at disadvantages of low load prediction precision of back propagation neural network, this paper proposes to improve generalization ability of the model by means of Bayes regularization algorithm. According to characteristics of that Heyuan power grid load is easy to be affected by weather it provides a layered Bayes neural network prediction model. The prediction result shows that this new mode is provided with better generalization ability and application effectiveness Improving the accuracy of load forcasting.
关 键 词: 神经网络 贝叶斯神经网络 短期负荷预测 泛化能力
领 域: [电气工程]