Research On Gas Mixture Modeling Method

As an essential department in the iron and steel factory, gas mixing and pressuring station plays an important role on the quantity and quanlity of the iron and steel products, which have significant influnce on other process. Meanwhile, as calorific value and pressure are the two basic control targets, they make it very important to build an accurate model of the gas mixing and pressuring station. But it is a challenge for modeling the nonlinear mixture process. In addition, the mechanism model is build with many ideal hypotheses which do not fit actual situations.To solve the problems metioned above, we study the problem of controlling the calorific value and pressure stability of the output gas and take the "Qian Gang No. 2106 Gas Mixturing and Pressuring System" as example. Based on the data collected from the real production line, we build a model using the statistical method and optimize the parameters by differential evolution algorithm.First, BP Neural Network is used to build a basic model for the system, i.e "the BP Neural Network coalgas blended model". Compared with the actual data, it illustrates that the model can basicly describe the actual situation of the gas mixing and pressuring station. But at the sametime, we can find out that the model is relied on the initial weight, which makes it has a low convergence speed and bad control accuracy, sometimes even falls into local optimum.Second, we propose an improved differential evolution algorithm based BP neural network, which is different from common differential evolution algorithm raised by Kenneth Price & Rainer Storn. In this method, improved DE algorithm is used to generate and optimize the weight of BP algorithm. After a lot of simulations, we draw a conclusion of that the improved algorithm shows effectiveness on improving the control accuracy and output fitness, and it can solve the problems of common BP algorithm metioned defore.The experiment results indicate that model based on production data is effective on the controlling of gas mixture process.At last, a gas mixture system is proposed in this paper to guide the actual producing process, which will certainly offer convenient to the engineers and operators. And the system using improved DE algorithm will be helpful with retaining the stability of the mixed gas.