Method Research Of RBF Neural Networks Modeling For Nonlinear Systems

Nonlinear phenomena are general problems in every field of engineering technology, science research, natural world and human society activities. Nonlinear system identification is a hotspot that many savants are researching. RBF neural networks is a new tool for nonlinear system identification. It has important theoretical significance and application value how to find an effective learning algorithm of RBF neural networks. Seeing that the computation burden of many intrinsic algorithms are too heavy and learning speeds are too slow, in this paper we adopt a new hybrid learning algorithm based on system identification and RBF neural networks theories. The hybrid algorithm includes an optimal selection cluster algorithm and a second-order algorithm. Combining with the properties of well test interpretation in oilfield, we adopt new basis function networks models. By analyzing the principle and tendency of accumulative quantity of oil product in oil field, we adopt function connection networks models. We take these neural networks as modern describing method of oilfield system identification and it has high precision. Based on the data of different districts of several oil plants of an oilfield, the effect of training and prediction is excellent. The effect of simulating different nonlinear systems shows validity of the scheme in this paper.