一种新的天然气管网负荷预测方法

A NEW PREDICTIVE METHOD OF GAS PIPELINE LOAD

针对天然气管网负荷变化的特点 ,提出了用模糊逻辑和RBF神经网络模型 (FL RBFNNM)来预测天然气管网的负荷。即首先利用模糊逻辑系统预测出负荷误差及误差变化率 ,从而实现了天然气负荷的在线修正 ;再利用改进的RBF神经网络进行天然气管网负荷的预测。在数据的处理上 ,应用了数据分类处理以及“近大远小”原则 ,并且在RBF网络模型中采用了最新邻聚类算法 ,实现了网络结构和参数的双重调节 ,大大提高了训练的速度和预测的精度。最后将此模型应用于实际中 ,并和单纯的RBF神经网络模型进行了比较 ,结果证明该模型可以快速准确预测出天然气管网的负荷值。

Aiming to the features of the load variation gas pipeline, it is suggested Fuzzy Logic system and RBF Nerve Network Model (FL - RBF NNM) is used to predict the load of gas pipeline. At first, the fuzzy logic system is applied to predict the load error and the error variation rate. So, the modification of gas pipeline load can be conducted on line. Then, the modified RBF nerve network is used to predict the load of gas pipeline. As for data processing, the data are processed by classification, and the principle of 'the nearer, the bigger and the farther the smaller' is applied. Besides, the latest adjacent clustering algorism is used in the RBF nerve network model to achieve the double adjustment for the network configuration and parameters. So, the training progress and the predictive accuracy are improved greatly. The model has been used in real cases, and compared with the pure RBF nerve network model. The results indicate the new model can predict the load of gas pipeline network speedily and accurately.