Measurement And Trend Prediction Of Soot Concentration In Multi Burner Furnace

The main cause of energy loss during furnace combustion in thermal power plant is incomplete combustion of fuel,and incomplete combustion of fuel will produce soot,so the soot concentration in furnace reflects the combustion efficiency of combustion equipment.Studying the change of soot concentration in flame combustion stage is very helpful to improve the combustion efficiency of fuel,so the research of this topic is of great significance.A method for measuring soot concentration in multi-burner industrial furnaces based on radiation inversion is presented in this paper.In this method,multi-images of furnace flame in four directions need to be collected synchronously.Because the resolution of CCD camera is not high,the collected images can not describe the combustion situation in the furnace accurately.Therefore,a convolution neural network based super-resolution reconstruction algorithm for flame image is proposed to improve the resolution of the acquired flame image.It provides an accurate calculation basis for the subsequent calculation of soot concentration field based on vision.After super-resolution reconstruction by convolution neural network,high-resolution synchronous flame images at different wavelengths can be obtained.According to the high-resolution synchronous flame images at different wavelengths,a method of calculating three-dimensional soot concentration field based on inversion calculation is proposed to obtain three-dimensional soot concentration field in a single direction.The three-dimensional soot concentration field calculated in a single direction is not very accurate,so we can calculate the soot concentration field in the furnace accurately by fusing the soot concentration field calculated from the inversion in four directions.Then the fused soot concentration is compared with the soot concentration measured by thermocouple particle density method,which verifies the validity and accuracy of the proposed algorithm.After obtaining the time series data and other parameters of soot concentration,a trend prediction algorithm based on convolution neural network and long-term and short-term memory model is proposed.The algorithm is used to predict the trend of s soot concentration,and compared with several common trend prediction algorithms to verify the accuracy of the prediction algorithm.