Research On Combustion Stability Determination Based On Improving PSO To Optimize FWSVM

Reducing the emission of combustion pollution of power plants meets the requirements of environmental protection.Optimizing the combustion of the boiler according to the combustion state can effectively control the generation of pollution and can prevent accidents such as boiler flameout and explosion,so that the boiler can operate safely and economically.However,Some of the tools used to diagnose the combustion state of the boiler will have problems of instability and low recognition rate.This paper analyzed the existing methods for diagnosing the flame combustion state in furnace,and combined with current research directions at home and abroad,and a combustion stability determination model based on digital image and feature weighted support vector machine is proposed.The stability of the flame in furnace was quantified.The image processing technology was used to extract the five characteristic parameters which can represent the characteristics of the combustion image after simple processing the flame image.Based on this,a database for combustion judgment was established to provide training samples for the latter model.For the support vector machine(SVM)can only be used for the two-classification problem,and the optimization problem of the penalty parameters and the nuclear parameters,this paper used one-to-one method to improve it,so it can be used for multi-category classification.And the improved grid search is used to optimizing the two parameters.The large step size is used firstly for coarse searching in a large parameters range,and then the small step size is used for detailed searching,so as to find the optimal solution and shorten the calculation time.Finally,the improved SVM model is used to train and test the combustion database.The experimental results show that the model has high accuracy in the determination of the flame combustion stability,and the calculation speed is also faster.Support vector machine does not consider the impact of feature attributes on the decision results.However,in practice,the sample will contain some features that have little influence on the classification result,which will affect the generalization of the classifier.For this reason,a feature weighted support vector machine model is proposed based on the Relief-F algorithm in this paper,which is used for the judgment of the flame stability.Considering that the particle swarm optimization algorithm is easy to fall into the local minimum point,an annealing algorithm is used to improve the particle swarm.The improved particle swarm algorithm is used to optimize the penalty and kernel parameters of the feature weighted support vector machine,and this model is used to train and test samples from the combustion database.The experimental results show that the model is feasible and has a very good classification effect.Compared with the support vector machine model,the feature-weighted support vector machine model greatly improves the accuracy of furnace flame identification and is more suitable for combustion the stability of the flame.