Boiler exhaust emission prediction model establishing method, prediction method and device

Abstract

The invention discloses a boiler exhaust emission prediction model establishing method, a prediction method and a prediction device. The model establishing method comprises the following steps: acquiring historical data, and performing preprocessing and canonical correlation analysis on the historical data to obtain a training sample; establishing a deep belief network, and training the deep belief network through the training sample to obtain a training sample feature matrix; extracting a plurality of training sample sets from the training sample feature matrix, and performing decision tree modeling according to feature data of the training sample sets so as to form a random forest model and train the random forest model; and combining the trained deep belief network and the random forestmodel to generate an exhaust emission prediction model. According to the invention, the deep learning algorithm is utilized to extract the deep features of the input data, and the random forest algorithm with less parameter setting and high convergence rate is adopted, so that the waste gas emission prediction model has high convergence rate and high prediction precision, and is suitable for a fire grate waste incineration boiler with a more complex and difficult-to-control combustion process.

Description

technical field [0001] The invention relates to the technical field of waste gas emission prediction of waste incineration boilers, in particular to a method for establishing a boiler waste gas emission prediction model, a prediction method and a device. Background technique [0002] The main treatment methods of urban domestic waste in my country include landfill, incineration, composting and stacking. Among them, the domestic waste incineration treatment technology has been developed and matured for nearly 30 years, and has become the most effective way to realize the harmlessness, reduction and resource utilization of waste. way. According to the different forms of waste incineration, domestic and foreign waste incinerators are mainly divided into the following three categories: circulating fluidized bed incinerators, rotary kiln incinerators and mechanical grate incinerators. Among them, compared with other incinerators, grate furnaces have greater advantages in terms of...

AI Technical Summary

Problems solved by technology

However, unlike the actual garbage components in waste incineration boilers, which are complex and changeable, have large calorific value fluctuations, and are difficult to predict NOx in real time, coal-fired boilers in power plants have constant coal powder and small calorific value fluctuations, and the specific implementation of NOx emission prediction is relatively simple.

[0006] All in all, the existing research on NOx emission prediction of waste incineration boilers mostly uses shallow learning algorithms. The initial parameters of the model are obtained by random initialization, which is easy to fall into local optimum, poor generalization ability, and convergence speed in the time series prediction under large data volume. Slow, low prediction accuracy

Compared with circulating fluidized bed boilers, grate incineration boilers have a large processing capacity, but in terms of waste pretreatment, they are not classified or broken except for bulky waste, which leads to large fluctuations in the calorific value of the combustion process, which brings great impact on NOx emission prediction. greater difficulty

[0007] In actual production, the NOx measurement of domestic waste grate incineration boilers mainly relies on hardware measurement based on various flue gas analyzers. The investment cost is high, and regular maintenance or replacement is required, and the fly ash in the flue gas is easy to adhere and deposit on the analysis instrument. surface, affecting measurement reliability

Images