Research On Data Rectification And Multi-Objective Optimization In Thermal Process

Data rectification and multi-objective optimization in thermal process are studied in the thesis. The main contents and achievements can be described as follow:1. Data rectification algorithm in thermal process is studied in the thesis. The fundamental theoretical model of data rectification is developed. Data rectification is made up of data reconciliation and gross error detection and identification. Data reconciliation algorithm based on matrix projection and digital filtering is analyzed in detail in the thesis. In order to detect and identify the gross error in the measure data, Measurement Test based on measurement residual, Constraint Test and Global Test based on constraint residual are studied in the thesis. The application of data rectification algorithm in on-line performance calculation in power station is briefly presented in the thesis.2. Multi-objective optimization algorithm is studied in the thesis. Multi-objective optimization problems are essentially different from single objective optimization problems. In order to solve multi-objective optimization problems, the general mathematic model of MOPs(multi-objective optimization problems) is developed and then the definition of solution to MOPs is made. Traditional approaches to solve MOPs are introduced, such as weighting method. Genetic Algorithm and its improvement is studied and Multi-objective Evolutionary Algorithms are analyzed in detail in the thesis, such as VEGA, NSGA/NSGA-II, NPGA, SPEA/SPEA2, etc.3. Application of multi-objective optimization algorithms in boiler combustion optimization is studied in the thesis. The creation mechanism, influential factors and control approaches of NOx emission are analyzed. A improved simplified on-line calculation boiler efficiency model is developed and a BP neural network is built to forecast the volumes of NOx, carbon in fly ash and oxygen in smoke based on the analysis of influential factors of NOx emission and boiler combustion experiment. The test of experimental data proved that the BP neural network had good generalization. Based on the BP neural network and the improved boiler efficiency model, the NOx emission and efficiency response characteristics model of coal-fired boiler was built. Based on this response characteristics model the optimization mode with two objectives(boiler efficiency,volumes of NOx) was built. The Pareto Front was obtained using multi-objective evolutionary algorithms and then analysis and discussion was made on the Pareto Optimal Sets.