Hot Metal Temperature Forecast Research Based On Quantum Genetic Neural Network

The reasonable temperature of blast furnace is the key to ensure the blast furnace smooth, safe, efficient and energy-saving operation, and it is also an important guideline of estimating condition. Establishing the temperature forecast model which can guide ironmaking workers control the blast furnace temperature has not only important theoretical value but also practical value.First of all, the complexity of the production process of blast furnace is studied through analysis the furnace internal chemical reaction, the temperature curve of hot molten in iron ditch, and the temperature drop of hot molten in iron ditch. Meanwhile the main influence parameters related to the furnace temperature in blast furnace production and their data features is analyzed, and the correlation and time lag between the parameters and hot metal temperature is determined.Secondly, this thesis establish a hot metal temperature forecast model based on genetic neural network. This model makes use of global optimization of adaptive genetic algorithm and local optimization of BP neural network to modify connection weights and thresholds of network. The problem of slow convergence speed and being prone to converge to minimum are solved for the best neural network.Lastly, in view of the GABP model problems such as easy to premature, small scale processing, this thesis puts forward an improved methods of hot metal temperature forecast model using GABP based on quantum theory. Quantum theory is introduced into the GABP, the idea of superposition of quantum states is applied to the neural network and form a quantum neural network of multi-layer activation function firstly, then use the quantum genetic algorithm which is formed by combining quantum theory and genetic algorithm to optimize the quantum neural network, and then establish the hot metal temperature forecast model based on quantum genetic neural network. This model uses quantum coding characterize chromosome and can express the solution in a linear superposition state, so it has better population diversity, faster convergence and global optimization capability.This thesis uses data collected from 2# blast furnace of Liuzhou Steel Corporation as sample data to implement the algorithms, the sample contains 300 data pairs. Through the simulation analysis, when the relative error is within ±1%, the hit rate of genetic neural network model is 75.00%, and the hit rate of quantum genetic neural network model is 86.3%.The results shows that compared with the hot metal temperature forecast model based on genetic neural network, the forecast model based on quantum genetic neural network has a higher hit ratio and precision.