论文部分内容阅读
蒸气管网是具有典型大时滞特点的非线性网络系统结构,提高管网运行预测能力,对管网的安全高效运行有很好的指导意义。贝叶斯神经网络具有良好的泛化能力和准确计算能力,在网络目标函数中引入表示网络结构复杂性的惩罚项,以便能够在训练优化过程中降低网络结构的复杂性,达到避免网络过拟合的目的。实例验证表明,模型计算结果和泛化能力均有良好表现,优于传统BP算法计算性能,可提高企业蒸气管网运行管理水平,对流程工业节能减排建设有一定的帮助。
The steam pipe network is a non-linear network system structure with the typical characteristics of large time lag. It can improve the pipe network operation forecasting ability and has a good guiding significance for the safe and efficient operation of the pipe network. Bayesian neural network has good generalization ability and accurate calculation ability, and introduce penalty items in the network objective function to represent the complexity of the network structure so as to reduce the complexity of the network structure in the training optimization process and to avoid the network overhead The purpose of the The example verification shows that both the model calculation results and the generalization ability have good performance, which is better than the traditional calculation performance of BP algorithm, which can improve the operation and management level of enterprise steam pipe network and help to some extent in the industrial energy saving and emission reduction.