The Research On Coal-fired Units Combustion Optimal Algorithm Based On Support Vector Machine

The production and the progress of human civilization is based on energy.Be accompanied by the rapid development of modern economy,human’s demand for energy grows more and more fast.China’s energy structure determines that coal resources are used the most. It’s the thermal power by using coal as fuel dominants the electric productions. In order to adapt to the rhythm of economic development, and to meet the increasing life demands of human, the consumption of coal resources has increased dramatically in recent years and the burning process has produced a lot of pollutants which pollute our environment and destroy the ecological system. How to improve the efficiency of coal-fired units combustion, save coal resources, reduce environmental pollution effectively is one of the important problems we are facing. Considering the energy conservation and emission reduction policy, composed by our country, a dual optimization considering both thermal efficiency of coal combustion and pollutant emissions is needed.,or to optimize the combustion system by a fixed thermal efficiency which add constraints of pollutant emissions into it.The paper’s work is divided into several parts as follows.First, to establish the object function. The combustion process of a coal-fired units is a complex physical and chemical process which involves a lot of subjects. There are many factors affecting the combustion process and they are not independent at all which brings a lot of difficulties to the analysis of the combustion process. Obviously,it’s a complex and multivariate nonlinear system. Advanced control and optimization technique would be used to optimize the combustion process by using the fixed thermal efficiency as the objective function. The analysis of the thermal efficiency shows that components of the fixed thermal efficiency including the basic thermal efficiency,the heat loss caused by CO and the differential pressure between furnace box and secondary bellow,and correction term of pollutants,are affected by the oxygen content in hearth at the same time. So,the uncontrolled boiler load and controllable oxygen content in hearth can be chosen as inputs to optimize the boiler combustion system.The second step, to compose the identification algorithm. Analysis and establish of traditional control systems are based on the precise mathematical model of the control object and the intelligent control system represented by the neural network is the study of the asymptotic theory based on infinite training samples. The statistical learning theory is based on the structural risk minimization principle which can find the optimal solution of small sample situation. The support vector machine (SVM) control in this paper based on the statistical learning theory can find the global optimal solution by solving a convex quadratic optimization problem and shows great advantage solving the pattern-recognition problems of nonlinear and high dimension systems. So the least square support vector machine (LSSVM) algorithm derivation is given in this paper and it is chosen to identify the boiler system. The Matlab simulation results shows that the LSSVM identification algorithm can follow the model output exactly and it owns a good precision.The third step is to optimize the identification model.The analysis of the optimal algorithms’ process and the Matlab simulation result show that the BFGS Quasi-newton algorithm is a powerful method to solve nonlinear equations, with simplified calculations and the super-linear convergence rate at the same time.At last,the paper represents the simulation research of combustion optimal control algorithm online by oxygen setting. Through the Matlab simulation, we can see that the combustion optimal control algorithm composed in this paper can improve the combustion efficiency practically and verify the feasibility and effectiveness of the proposed algorithms.