Research On Soft Sensor Application Of 600 MW Thermal Power Unit Flue Gas Oxygen Content

Electricity production is a large resource consumption and environmental pollution,economic development is the first army. And with the rapid development of national economy, the size of the power system is growing, so that primary energy consumption is amazing, therefore, improving boiler efficiency and reducing coal consumption of the power industry have became a much-needed task. To solve this problem, the State will be the implementation of the policy of competitive bidding in the power industry, the domestic thermal power plants will be do their best to improve the operation of the economy under the premise of safety in production. Power plant flue gas oxygen content is directly related to boiler efficiency as a parameter, the accuracy of measurement for guiding the production run has an important role. In this paper, the accuracy of measurement content difference, equipment investment and other issues for the power plant flue gas oxygen, the use of support vector machine technology to build a flue gas oxygen content of soft sensor model achieves an accurate measurement of the thermal power 600 MW oxygen content of the flue gas. Through actual operation and combustion process analysis of power plant boiler determine the flue gas oxygen content soft measurement model of alternative auxiliary variables, then according to the site DCS system, determines the sampling points modeling needs through DCS history database export construction mold required data.By analyzing the power plant DCS export data, determine the characteristics of flue gas oxygen content soft measurement model training and testing samples. Pre-process the training sample data using three principles and wavelet filtering, and auxiliary variables were further screened using the correlation coefficient method to determine build flue gas oxygen content soft measurement model requires auxiliary variables. Then using kernel principal component analysis model input variables of dimensionality reduction process to reduce the complexity of the model. Select the appropriate kernel function and parameters established flue gas oxygen content soft measurement model. Finally, the soft sensing model of the flue gas oxygen content was tested by using the 600 MW unit of Fuyang Huarun Electric Co., Ltd., the soft sensor model can accurately measure the oxygen content in the power plant, and the prediction error is less than 5.5%, and it canadapt to the change of the working conditions.Flue gas oxygen content soft measurement model established in this paper can measure the power plant flue gas oxygen content accurately under the situation of no increasing power plant equipment and measuring points, and it has the advantages of high accuracy, strong model generalization and good stability.