Research On Forecasting Of Caustic Soda Production Capacity In Chemical Industry

Energy is not only one of the important materials for human’s survival and development but also the driving force of economic development. Doing a good job for energy planning and utilizing resources properly is the foundation to improve the market competitiveness for a enterprise, and finding out the relationship between production capacity and energy prediction is one of the important components of energy planning for chemical enterprises. Through energy consumption to predict production capacity not only can grasp the trend of energy consumption but also can control energy reserves, reduce energy waste and the cost of chemical products, besides, it can improve the market competitiveness of chemical products and the economic benefits of the enterprise. Scientific forecasting is the suggestion and guarantee for producer to make production strategic, and at the same time, safe and reliable operation of energy systems depend on accurate prediction. It is important to find an efficient forecasting method of energy, which can improve predicting accuracy.This thesis makes a short-term prediction of chemical enterprises in caustic soda production capacity. Firstly, the analysis about the influence factor of caustic soda production capacity prediction is done, and then, it carries on a pretreatment to the data. Mahalanobis-Taguchi Gram-Schmidt(MTGS) is used to calculate the weight of influence factors on caustic soda production prediction, and conclude to find the factors that affecting the prediction effect of caustic soda production capacity through combining the weight with the trend analysis of influencing factors. In order to verify the result correctness of MTGS, two kinds of different structure of BP neural network prediction model are established. By using BP neural network prediction mode to train the production data for seven months of one chemical factory, and then predicting the production capacity, conclusion gets a good prediction effect. Considering that BP neural network prediction model has some defects, the thesis adopts the cuckoo search algorithm to optimize BP neural network prediction model(CS-BP). In order to verify the superiority of this model, a prediction model of BP neural network based on particle swarm optimization algorithm is established. The test shows that CS-BP prediction model has a better prediction effect when it compared with BP neural network prediction mode and the particle swarm optimization algorithm to optimize BP neural network prediction model(PSO-BP). The proposing of this algorithm provides a reliable and scientific method for energy management.