Modeling And Optimization For Combustion System Of The Boiler In Power Plants Based On Hadoop Big Data Platform

At present,the thermal power still occupies a dominant position in China's power industry,accounting for about 70% of the total power generation.The coal-fired boiler is the largest energy consumption point in the power plant.However,due to the boiler equipment and management of the operation of the reasons,the boiler combustion system often cannot reach the requirements of high efficiency and low emission.It can help the power plant to reduce energy consumption and reduce personnel operations and improve the accuracy of forecasting using the machine learning algorithm to optimize the boiler combustion system.However,the training efficiency and accuracy of modeling and optimization methods of boiler combustion system based on the traditional machine learning need to be further improved.With the explosive growth of the boiler combustion system data,the traditional machine learning algorithm has been difficult to meet the needs of massive high-dimensional data processing.The proposal of the extreme learning machine opens up an effective way to improve the efficiency of the boiler combustion system.At present,the Hadoop technology has a strong adaptability for dealing with massive high-dimensional data,in which the MapReduce programming framework uses distributed computing,which provides an effective means to solve the problem of large data.Therefore,how to implement the ELM algorithm on the Hadoop platform to achieve distributed computing,and its application to the optimization model of power plant boiler combustion has important theoretical significance andpractical value.In this paper,we use the MapReduce programming framework in the Hadoop platform to realize the distributed computing of the extreme learning machine,and to improve the shortcomings of distributed extreme learning machine.The specific work of this paper is as follows:(1)In view of the traditional single ELM algorithm may face slow calculation and single computer resources shortage in processing the massive high-dimensional data.Using the MapReduce distributed computing framework in Hadoop to optimize the large matrix operations in the ELM algorithm,a distributed extreme learning machine,which is able to deal with large scale data,is proposed,which is called ELM*I algorithm.(2)Analysis of ELM*I algorithm,the intermediate result storage problem of map and reduce methods,which leads to an increase in the calculation and transmission cost between the data.An improved distributed limit learning machine is proposed,and the performance of IDELM is better than that of the ELM*I is verified by experiment.(3)The data collected from the power plant boiler combustion system were pretreated,and then the NOx emission and Boiler combustion thermal efficiency prediction model was established by using the IDELM algorithm respectively.The influence of the two parameters of the hidden layer node L and the regular term A on the prediction accuracy and generalization ability of the model is analyzed in detail by experiment.The optimal combination of the two parameters is selected according to the experimental results to determine the optimal model,and then use this model to predict the NOx emissions and Boiler combustion thermal efficiency.(4)The multi-objective combustion optimization problem of boiler is studied.The multi-objective combustion optimization model of the boiler was established based on the previously established NOx emission and Boilercombustion thermal efficiency prediction model.The model used boiler NOx emission and Boiler combustion thermal efficiency as the output of the model.The multi-objective optimization problem,which minimizes the NOx emission and maximizes the Boiler combustion thermal efficiency,is transformed into a single-objective optimization problem by the weighting coefficient method,and the distributed particle swarm optimization(MR-PSO)algorithm is used to optimize the input parameters of the model,and to found the optimal combination value of each input parameter.In summary,this paper studies the distributed extreme learning machine based on MapReduce and its improved method,which enhances the ability of ELM to deal with massive high dimensional data.The model is applied to the multi-objective combustion model of boiler,and the MR-PSO algorithm is used to optimize the model.The feasibility and effectiveness of the proposed method in solving the boiler combustion optimization problem is proved by a large number of experimental analyzes.