Knowledge Evoluation Algorithm And Its Application In Chemical Dynamic Optimization

The significant and cumulatively increase of concerns about environment protection from the modern society, the steady growth of competition between chemical enterprises and the drastic decrease of trust of chemical industries'security have brought enormous pressure on the chemical industry. As a consequence, it is obviously for our society and researcher to improve the chemical control and optimization technology. The dynamic process models, which own characteristics such as strong nonlinearity, high dimension and complex reaction mechanism, cannot be perfectly solved by traditional optimization technique based on steady-state model. Consequently, it is necessary to develop dynamic optimal methods for the dynamic process analysis. Therefore, studying and devising Intelligent Optimization Algorithms for Chemical Dynamic Optimization Problems (DOP) are meaningful and meritorious.With the developments of Intelligent Optimization Algorithms, it is found that such methods are powerful and efficient to solve the aforementioned issues. To be specific, the emergence of Knowledge Evolution Algorithm (KEA) has provided a more efficient algorithm framework than traditional bionic algorithms. In addition, the applied researches on this algorithm are rarely present so far. Therefore, in this dissertation, the research key point would be focus on the construction of an Improved Knowledge Evolution Algorithm (IKEA) which is catering to the characteristics of process optimization.The literature reviews of research achievements in the field of process optimization have been presented before the merits and demerits of Intelligent Optimization Algorithms (when they are apply to chemical process optimization) were analyzed in this dissertation. The intelligent optimization algorithms have been widely adopted in solving dynamic optimization problems for their ability to converge to global optimum at a relative high probability without being trapped in local optimums. However, the practical industrial application fields of such random mechanism-based algorithms are restricted due to their demerits such as slow convergence speed and low searching efficiency. Therefore, an improved knowledge-based evolutionary algorithm structure is presented for improving the efficiency of intelligent optimization algorithms which are used to solve dynamic optimization problems. The structure includes:1) A discretization method of candidate solutions in time and control domains. By using this method, the initial candidate solutions can be properly generated in the Population Space (PS) according to practical process characteristic.2) A candidate-solutions cuts method based on online clustering. This method enables the candidate population to adjust its size in accordance with the distribution of candidate solutions in the evolution.3) Two knowledge extract and evolutionary strategies. These strategies are designed on the basis of the characteristics of the framework of knowledge evolution algorithm and can extract the implicit evolution information efficiently from the Population Space during the evolutionary process.4) Three evolutionary strategies of Population Space (PS) under the influences of knowledge which are extracted by aforementioned knowledge exact and evolutionary strategies. These evolutionary strategies allow the Population Space (PS) evolves efficiently and can cut down the calculating cost of IKEA.To conclude, in terms of application in the dynamic optimization problems of four typical chemical processes such as the batch reactor, which owns distinguishing control feature, this algorithm demonstrates its competitive optimal searching ability, meanwhile verifying its satisfied convergence efficiency by using a tiny scale of population guided by knowledge and consuming less computational cost.