Research On Advanced RBF Control Strategies Of Internal Thermally Coupled Energy Saving Process

Distillation is one of the most important operation units in chemical and petroleum industries. Internal Thermally Coupled Distillation saving technology so far is the greatest potential for energy conservation programs which attract the attention of many internationally renowned scholars, the study is now in the distillation energy forefront, many experts are concerned about energy-saving technology in the field of energy-saving technologies and mainstream scholars of distillation of all research in this area. However, due to the internal mechanism of thermally coupled distillation special hair complex, which has a strong pathological nonlinear dynamic processes, as well as strong coupling and highly sensitive to disturbance variables, brought great distress control of the internal thermally coupled distillation column, study abroad so far are model-based control of the mechanism, but due to the mechanism of modeling run for a long time, online correction is not suitable for online operation. Therefore, this paper attempts from the perspective of machine learning to model control, consider the RBF machine learning to solve this control scheme. RBF advanced control research focuses on the internal thermally coupled distillation column energy-saving process, try to find one or more of the outstanding RBF control solutions, to provide support for this new technology to control industrial. The main work and contributions are listed as follows:(1) According to the internal thermal coupling distillation which is the most energy saving potential, I research about the internal thermal coupling distillation control strategy both at home and abroad, and puts forward the strategy based on the RBF neural network of the internal model control. Take the benzene-toluene system as a research case and compared with the international public reporting results, the RBF neural network is more reliable than PID algorithm and the international public reporting results.(2) Considering the research of the neural network model, then a hybrid learning algorithm is presented for the radial basis function neural network, which is based on particle swarm optimization, K-means clustering and subtractive clustering algorithm. Compared with the international public reporting results, the PSO-RBF-IMC is more reliable and can better tracking output and get better accuracy.(3) Based on the research of the internal model structure at home and abroad, a feedback filter for accuracy and robustness of the balance is considered to join the traditional internal model structure, to overcome the model mismatch under the disturbance of RBF-IMC, the results of the study show that the new control strategy that the TDOF-PSO-RBF-IMC has better control affection and preponderance.