Intelligent Prediction And Coordinative Optimization Of Comprehensive Production Targets In Coking Process

As the main raw material and fuel in the metallurgy, machinery, and chemical industry, coke has been widely used in blast furnace iron making, calcium carbide, gasification, casting and nonferrous metal smelting, etc. The effective coordination and stability of each local process in coking production could directly influence the performance of the quality and quantity of coke, and energy consumption of coke oven. In order to overcome the strong nonlinearity and uncertainty in the coking production process and realize the optimization objective of high productivity, good quality and low consumption, intelligent prediction and coordinative optimization of comprehensive production targets have been studied in this thesis. The main achievements in this thesis include the following aspects:(1) The analysis of relationship between production indices and process parameters, and the coordinative optimization structureThe status of coking production infects the quality and quantity of coke, and energy consumption. There are many process parameters and they have different effects on production indices. Based on the analysis of the coking production technology and mechanism, the production indices and state parameters in coking process are determined according to the production requirement of coking enterprises and the control request of each local process. Moreover, the relationship among production indices, state parameters and operation parameters have been deeply analyzed qualitatively, and then quantitative analysis is also conducted based on the grey relational analysis.Using the method of layered optimization and dividing-and-rule, an intelligent modeling and coordinative optimization strategy of multi-level hierarchical structure is proposed for the purpose of reaching the comprehensive production targets in coking process. This structure has three levels including coordinative optimization level, optimization control level, and basic automation control level. The function and relationship between each level are analyzed and designed. This method provides a new idea for intelligent modeling and coordinative optimization of coking production process. (2) Intelligent hybrid prediction models of the coking production indicesIn order to implement the coordinative optimization of coking production process, it is necessary to make online prediction of the production indices, including the quality and quantity of coke, and energy consumption of coke oven. According to the features of strong nonlinearity and uncertainty of coking process, BP neural network models for the prediction of quality and quantity of coke and energy consumption of coke oven are proposed based on the analysis of relationship between production indices and process parameters. The learning process and convergence speed of traditional BP training algorithm is very slow and it tends to fall into local optima. Hence, here a niching differential evolution based on density clustering is applied to train the BP neural network.Considering the fluctuations of production status, the time variation of the production process, and the imperfection of the model, an evaluation strategy of the prediction performance of BP model with sliding time window is proposed from improving the prediction performance and adaptive ability of model. A segmental correct method based on the effective integration of production indices deviation correction, the short-time compensation correction of just-in-time learning, and the BP neural network model parameter correction are also proposed. The improved weighted LSSVM local modeling method based on just-in-time learning and the dynamic weighted hybrid prediction method are studied. The proposed hybrid prediction model can have good prediction performance through effective evaluation of the prediction performance of BP model and segmental correction based on performance evaluation, predicting the coking production indices real-timely and effectively in production with fluctuating status.(3) Coordinative optimization strategy for comprehensive production targetsIn order to make the production has high quantity, good quality, and low consumption, a multi-objective optimization model is established with quantity and energy consumption as objective function, quality of coke as restrictions, state parameters in each local process as decision variable. Based on the multiplier penalty function method, the original objective function and multiplier penalty function are normalized and universally extended. The extended feasible region is introduced to make full use of the information of good infeasible solution. The coordinative optimization strategy, which combines kernel fuzzy C-means clustering and multi-population differential evolution, is proposed to obtain the optimal set value of state parameters in each local process. Thus the global optimal control problem is transformed into the local optimal control problem of each local process.Due to the non-normal feature of the coke quality data, capability analysis of the multivariate process with non-normal data is applied to obtain the multivariate process capability index of coke quality. The evaluation factors include the multivariate process capability index of coke quality, economic index and productivity index of comprehensive production indices. The comprehensive evaluation of coking production status is obtained using variable weight fuzzy comprehensive evaluation method. When the evaluation result does not meet the requirements of the given level, the coordinative optimization algorithm is utilized to adjust the optimal set value of state parameters, thus realize the real-time optimization for operation performance of coking production process.The implementation steps of coordinative optimization strategy are interpreted. The delay compensation model of flue temperature and optimal setting model of suction in flue in the combustion process are also investigated. The actual operation data in coking production process is used to confirm the efficiency of the proposed coordinative optimization strategy. Experimental results show that the proposed coordinative optimization strategy can realize the optimal operation of coking production process and realize the enterprise production targets of high quantity, good quality and low consumption under the prerequisite condition of stabilization production.