Two-phase Flow Measurement Data Processing Based On Wavelet Neural Network

Online measurement of wet gas flow rate is significant for gas field. Two methods of two-phase flow measurement data processing are proposed based on wavelet neural networks(WNN). Firstly, this thesis makes syste- matic studies of structures of wavelet neural networks, it also compares nonlinear function approach properties of wavelet neural networks with BP networks in detail. Secondly, two kinds of wavelet neural networks are built. The first one, based on continuous wavelet transform theory, the thesis presents continuous parameter wavelet neural network, whose activation functions are continuous wavelet functions. Furthermore, the algorithm of network is given, and initial parameters are optimized by genetic algorithm (GA). The second one, according to multi-resolution analysis and ortho- normal wavelet decomposition theory, multi-resolution wavelet neural network is built, whose activation functions consist of orthonormal wavelet and orthonormal scaling functions, and hierarchical approximation algorithm is given. Lastly, use such both wavelet neural networks to process the two-phase flow measurement data respectively, following conclusion can be drawn: To use wavelet neural networks to forecast two-phase flow is feasible.