Research On Parallel Layer Perceptron Fast Learning Network Algorithm And Its Application To Circulating Fluidized Bed Boiler

As the main form of power generation in China,thermal power generation brings two major problems: energy consumption and environmental pollution.It is an important research topic for scholars at home and abroad to optimize on the basis of accurate forecasting of operating parameters of thermal power plants and achieve the goal of high thermal efficiency and low pollutants.Artificial Neural Network(ANN),as a new and efficient modeling method,overcomes the shortcomings of the traditional mechanism modeling method in the prediction accuracy of strong coupled and nonlinear complex systems.However,early neural network models(such as stochastic gradient descent method)have the disadvantages of long training time and easy fall into local optimum.To address these deficiencies,the Fast Learning Network(FLN)overcomes the lengthy parameter iteration process with its unique learning mechanism,while possessing good learning and generalization performance.Therefore,in order to further improve the network performance,this article chooses to do relevant research and improvement on the basis of FLN.The relevant research content is described as follows:First,on the basis of FLN,a novel neural network model—the Parallel Layer Perceptron Fast Learning Network(PLP-FLN),is proposed for its fast learning ability and good processing ability for linear and nonlinear data.Through 12 sets of classical regression datasets,relevant experiments were done and the simulation results were compared with the results of four other classical ANN models.It was verified that the PLP-FLN model has better data fitting ability and generalization ability than other ANN.Next,a 300 MW Circulating Fluidized Bed Boiler(CFBB)combustion system in a thermal power plant was studied.In order to achieve the above objectives,a key step is to accurately predict the output parameter information.In this paper,the operating condition data collected by the thermal power plant are used,and the NOx emissions and thermal efficiency are modeled using PLP-FLN.The experimental results prove that it has good prediction accuracy,generalization ability and stability.Then,in order to improve the combustion thermal efficiency and reduce NOx emissions as the optimization goal,using the PLP-FLN model combined with the Improved Artificial Bee Colony(I-ABC),a single-objective optimization model and a multi-objective synthesis optimization model were established.After determining the respective objective function for each model's characteristics,the optimal combination of variables can be searched by adjusting the matching relationship between the variables.The CFBB system operates in an optimal state to improve the fitness value and achieve combustion optimization.