The Steel Enterprise Consumption Forecast Based On Interval Neural Network

The topic is based on steel enterprise companies. For the purpose of energy conservation, the article analyzed energy consumption and identified key energy-saving of the iron and steel enterprises to conduct energy forecasting of steel enterprises and tap the energy potential, which is the national industrial energy content specified priority themes highly consistent. This thesis has a strong research and promotion of versatility in the process industry, the economy and society of immense significance.This thesis aimed at conducting energy forecasting of steel enterprises. The work of it is showed as follows:First, a brief introduction of the iron and steel production processes and energy consumption is proposed and then the Fe-containing material inflow and outflow affect is analyzed to the production process of the comprehensive energy consumption per ton of steel. The results show that the inflow of Fe-containing substances can reduce energy consumption, and outflow will increase energy consumption. The later of the operation, the more significant of the result will be.Comprehensive energy consumption per ton of steel is a very important indicator of iron and steel enterprises, which can provide enterprises with the production guidance and the advice on policy-making decisions. Hence, how to reasonably predict the energy consumption per ton is a serious problem. In response to these problems, this thesis proposed the use of interval neural network to predict the comprehensive energy consumption per ton of steel.This thesis introduced the interval neural networks derived from gradient descent method, and conducted network testing. Experimental results achieve the expected results. Because interval neural networks is easy to fall into local minimum, adaptive quantum particle swarm is proposed and algorithm performance is tested. At last, the AQPSO is combined with interval neural networks, weights and thresholds of interval neural network are optimized with the help of AQPSO. After the experimental data test, the performance of this interval neural network has been significantly improved.During the energy consumption forecast of iron and steel enterprises, based on the model of AQPSO and interval neural networks, the thesis selected the affect energy consumption per ton of iron-containing substance as a neural network input range. Before predicting, we used PCA for data processing, which will affect the energy consumption per ton of 11 kinds of ferrous material dimension reduction to three new complex variables, and the new variable predictive model is used as input. Using adaptive quantum particle swarm of initial weights and thresholds for processing, the optimum particle dimension information PSO output range of neural networks as initial weights and thresholds, and then use the interval neural network to predict the energy consumption. The result reached the desired effect.