Online measurement method for NOx emission of coal-fired boiler

Abstract

A NOx online measurement calculation model is established based on a deep belief network and a partial least squares support vector machine. The DBN network is formed by stacking two layers of restricted Boltzmann machines, each RBM comprises a visual layer v and a hidden layer h, the units of each layer are not connected, and the layers are fully connected; the method comprises steps of extracting deep feature pairs of the sample data set by using a DBN network model for training, completing establishment of a prediction model by minimizing an error between the output of a regression device and a real value of NOx; packaging and publishing DBN-LSSVM model as a dynamic link library, compling an interface program to obtain boiler side input parameters in real time, calling the dynamic linklibrary from a database, outputting a calculation result to a monitoring system, and measuring the NOx emission value online in real time. According to the NOX online soft measurement method providedby the invention, no extra measurement device needs to be added, the input cost is low, the operation reliability is high, the single measurement time is 0.05 s, and the real-time performance is good.

Description

technical field [0001] The invention belongs to the field of electric power and provides an X Real-time online measurement technology. Background technique [0002] Coal burning in power plant boilers is an air pollutant NO X One of the main sources of NO emitted into the atmosphere X It will cause acid rain and photochemical smog pollution, destroy the ozone layer, threaten the ecological environment, and damage human health. The air pollutant emission standard for thermal power plants (GB13223-2011) requires that NOx emissions from coal-fired power plants be controlled at 100mg / m 3 Below, the ultra-low emission requirement is controlled at 50mg / m 3 the following. With the reduction of pollutant emission limits of coal-fired power plants, the on-line monitoring technology of flue gas pollutants needs to be continuously improved to monitor pollutant emissions in real time and accurately. [0003] Power plant NO X There are two main types of measuring devices: one is a...

AI Technical Summary

Problems solved by technology

But they have the following disadvantages: (1) The amount of sample data is limited and it takes time and effort to obtain

Only limited sample data can be obtained through several field experiments, and it is difficult to meet the sample size requirements of machine learning modeling. More sample data in the modeling process means a model with higher accuracy; (2) This type of method uses a large number of Empirical formula and empirical value, the parameters contained in the empirical formula are limited, but the combustion process is complex, changeable and non-linear, so that the calculated value often deviates greatly from the actual value

(3) Existing NO X The modeling methods are all shallow machine learning methods

It is difficult for shallow learning algorithms to obtain the deep features of the data, which makes this type of measurement device still have some shortcomings in measurement accuracy

Images