Intelligent Autonomous Setting Method Of Municipal Solid Waste Incineration Process Parameters

The waste treatment method is mainly incineration,it has the advantages of reduction,harmlessness and resourcefulness.In the waste incineration process control field,because the process has the problems of strong coupling,large lag and nonlinearity,it is difficult to establish the mechanism setting model of key parameters(grate speed and air flow).In this case,it is usually the operator who judges the operating conditions based on experience,and then the key parameters setpoints are adjusted.This method is subjective and inefficient,so it is very difficult to guarantee the complete incineration,and resulting in substandard pollutants emission.Therefore,it is of great practical significance to study the automatic setting method of waste incineration process key parameters.Taking a waste incineration power plant in Beijing as an example,the research on setting method of incineration process key parameter is carried out,an intelligent autonomous setting modeling strategy of "perception-setting-evaluation-correction-learning" is proposed,artificial intelligence technologies such as neural network and case-based reasoning are adopted to design the structure of intelligent autonomous setting method with the functions of working condition perception,indexes prediction,dynamic compensation,evaluation and learning,etc,the algorithm of each model is implemented,and the operation effect of the proposed method is tested by experiments.The main contents are as follows:(1)Fault detection model of incineration process based on learning pseudo metric method.Aiming at the problem that Euclidean distance similarity measurement method in case-based reasoning model has problems of weight distribution and easy to fall into distance trap,a learning case similarity pseudo measurement method is designed by combining stochastic configuration network with the pseudo measurement criterion.On this basis,fault detection model of incineration process is established by the case-based reasoning process model,namely case retrieval,case reuse,case correction and case storage.Experiments show that the proposed learning pseudo metric method applied to case-based reasoning process model has good comprehensive performance for the indicators such as the accuracy of fault detection in the incineration process,good robustness and stability for noise and interference.It can realize the intelligent perception function of operating conditions in the incineration process,and provide working condition information for the adjustment of key parameter setpoints.(2)Fuzzy neural network waste heat value grade prediction model.Aiming at the problems that the waste components cannot be measured and the exact value of waste heat value is difficult to measure during the incineration process.Waste heat value grade prediction model is studied by using fuzzy neural network which has advantage of dealing with fuzziness problem.Firstly,the first screening of feature variables is realized by calculating the mutual information between waste heat value and related variables;Secondly,the adaptive mutation strategy is introduced into the particle swarm optimization algorithm,and it is integrated with fuzzy neural network,the feature variables are screened again,and fuzzy neural network waste heat value grade prediction model is obtained at the same time.The effectiveness of the twice feature selection method is verified,and the waste heat value grade prediction model trained by it has high accuracy.It can provide accurate heat value information for the adjustment of key parameter setpoints.(3)Stochastic configuration network prediction model of gas oxygen content and CO concentration.Be directed against problems of large delay and nonlinearity during the change of CO concentration and gas oxygen content,the prediction model of the above two process indexes are studied by using stochastic configuration network which has universal approximation property and fast learning ability.The feature variables are selected by using the feature selection method of waste heat value,that is,the mutual information and particle swarm optimization algorithm with adaptive mutation ability,and the stochastic configuration network prediction model of gas oxygen content and CO concentration is obtained at the same time.Experiments show that the training efficiency and prediction accuracy of the stochastic configuration network model have obvious advantages over BP network and radial basis function network.It can meet the demand of rapid and accurate prediction of process indexes,and can provide accurate index change information for the dynamic compensation of key parameter setpoints.(4)Intelligent autonomous setting method of incineration process key parameters.Aiming at the problem of low efficiency and strong subjectivity of manual setting based on experience,an intelligent autonomous setting modeling strategy of "perception-setting-evaluation-correction-learning" is proposed,and the structural principle of intelligent autonomous setting with the functions of working condition perception,key parameter pre-set,indexes prediction,dynamic compensation,evaluation and learning is designed.Among them,the working condition perception function is realized by the fault detection model based on learning pseudo metric method and the waste heat value grade prediction model based on fuzzy neural network;The key parameters pre-set function and the evaluation and learning function of setpoints are realized by case-based reasoning,where,evaluation,correction and learning of setpoints are realized by production rules;The indexes prediction is obtained through the prediction model of gas oxygen content and CO concentration,which is used as the input of intelligent compensation model,and the dynamic compensation of setpoints is completed through radial basis function network and fuzzy reasoning,so a set of intelligent autonomous setting method of the incineration process key parameters is obtained,and the algorithm implementation steps of the setting method are given.(5)Software development and experiments of intelligent autonomous setting method.According to the algorithm implementation of the intelligent autonomous setting method obtained above,the structure and function of the intelligent autonomous setting method software are designed based on the demand analysis.The developed software mainly includes the background model layer,communication layer,human-computer interaction interface and so on.Among them,the background model layer includes fault detect,waste heat value grade prediction,indexes prediction and key parameter setting modules,which are developed in MATLAB programming environment;The communication layer includes the communication among C#,the model algorithm program in MATLAB and My SQL database management system;The human-computer interaction interface includes the user login and the interface of each software module corresponding to the background model layer,which is developed in the C# programming environment.The test experiments show that the communication and operation functions of the software are normal.On this basis,the performance of the intelligent autonomous setting method is tested in the software.The results show that the models cooperate with each other and the key parameters setpoints can be given effectively.The fluctuation of the setpoints is significantly reduced,and the qualification rate of process indexes is significantly improved under various working conditions.It lays a foundation for the operation optimization and control of waste incineration process.