Soft sensor for coal mill primary air flow based on LSSVR

In the power plant coal-fired units, the primary air for boiler combustion and coal powder conveying is directly related to the actual combustion chamber conditions. Therefore, the appropriate primary air flow is very important for the normal operations of the coal mill and even the whole units. Coal mill primary air flow soft sensor model was established based on least squares support vector regression machine algorithm. Gaussian radial basis function is selected as the kernel function and training performance is used to select model parameters. Reasonable choices of the variables that are closely related to the primary air flow are used as input features. The historical data of a selected power plant DCS system is used as training samples and testing samples. Gross error, random error and normalization of samples are dealt with by using a statistical discriminant method and a sliding average method. Experimental verification shows that this soft sensor model method can achieve higher accuracy than the existing flow meter. The soft sensor technology has a good application prospect in the detection process of a thermal power plant.