Experimental And SVM Model Building Study On Coal Blending And Combustion Optimization Of Utility Boiler

Energy and Environment are crucial problems for our country's development. As coal is the main energy source and most coal was used to generate electric power in our country, how to achieve boiler combustion with high efficiency and low pollutants emission becomes the very key task for our country's sustainable development. This work was mainly involved in the technology about coal blend, boiler combustion optimization, a new coal blend approaches and combustion optimizations way were developed by means of experiments, models building and field tests.As a tool for model building, support vector machine(SVM) was introduced into coal blend calculation and combustion optimization.. The affection of parameter C and g on the SVM model was study under the condition of RBF function was selected as kernel function, and the conclusion that parameter C and g were only affect the properties of SVM model only in a certain range was draw, this becomes the basis of subsequent works.On the basis of experiments data of ash fusion temperature and coal ash compositions, a support vector machine model predicting the ash fusion temperature from the compositions of coal ash was developed and verified. Good predicting performance for both single coal ash fusion temperature and blended coal ash fusion temperature was achieved with the parameters optimized by genetic algorithm. On the basis of ash fusion temperature model, a model for blend coal calculation was developed. The blended coal calculated by this model for a boiler was test fired on that boiler, and the result shows the blended coal was meet that boiler's requirement to fire; A coal blend optimize system was developed based on the blend coal calculation model. This system took the coal manage process and the actual need of the boiler into consideration and modified its design accordingly. This coal blend system has been put into operation, and brought the power plant benefits.The affection of different operation parameters on aspects of NO_X emission, unburned carbon in fly ash, smoke temperature and CO concentration near the water wall of boiler filer, by means of round of parameters. Based on the experiment data, SVM models about boiler fire takes the boiler's main operate parameters as inputs and takes NO_X emission, unburned carbon in fly ash, smoke temperature and CO concentration near the water wall as output respectively, and these models were achieved good predicting performance. Combined with Genetic Algorithms these SVM models achieved good optimization of the boiler combustion that reduced the NO_X emission, unburned carbon in fly ash, smoke temperature and CO concentration near the water wall, the result of optimization was credibility compared with real operating conditions.The power boiler online combustion optimization system was developed with the tools of VB and SQL on the basis of combustion optimize SVM models. This system achieved boiler combustion online optimize by combine with the flue gas online monitor system and the DCS system, and also online self learning was achieved. The testing combustion optimization on a power plant boiler shows the online combustion optimization system achieved good performance on combustion optimization by enhanced boiler efficiency and reduced NO_X emissions.