Research On Thermal Process Identification Method Based On Field Data

In general, when designing the controlling system, the mathematical model of the controlled process is needed. Traditional model identification of the thermal process is mainly based on the test of dynamic characteristics of the process. Because of the characteristics of the power generation unit, the dynamic characteristic test in the field is often difficult to implement. Even though the dynamic characteristics test is carried out, the results are not very ideal. Therefore, research on thermal process identification based on field data has important theoretical significance and application value.The neural network can approximate continuous function with arbitrary precision and it has strong adaptability and learning ability, so the neural network is widely used in nonlinear modeling of thermal process. But in the identification these problems are not well resolved, such as how to determine the structure of the neural network and how to improve the generalization ability of the network. In addition, the neural network model is not intuitive, not easy to understand, difficult to combine with classical control methods.According to the above problems, the identification of the neural network of thermal process based on the field data is studied, and the method of extracting the transfer function from the neural network is put forward. The main research contents and results are as follows:1. Because of the problems of ordinary sensitivity pruning algorithm, a pruning optimization algorithm based on RBF neural network on the basis of original sensitivity pruning algorithm is proposed. And the pruning strategy is illustrated. The simulation results verify the effectiveness of the algorithm;2. The feasibility of neural network model identification based on the field data is analyzed. Then, a kind method of model identification of neural network for thermal process based on RBF neural network pruning algorithm is put forward. And the application steps of the algorithm are described in detail. In final, the simulation results indicate that the algorithm is effective;3. The characteristics of superheated steam temperature of the boiler is analyzed. Then, a kind method of model identification of neural network for thermal process based on RBF neural network pruning algorithm is applied in model identification of superheated steam temperature of the boiler, In final, the simulation experiment based on the field data indicate that the means is effective;4. In view of the difficulty of the field thermal test, the method of extracting transfer function model from neural network model is proposed. Combined with genetic algorithm, the application steps of the algorithm are described in detail. In final, the simulation results verify the effectiveness of the proposed method.