Deep Learning Model-based Optimization Of Target Values For Power Plant Boiler Operation

As a basic industry for the construction and development of China,the power industry plays an extremely important role in the overall development of the national economy.Thermal power generation still dominates the power generation industry in China.Large thermal power units participate in automatic power generation control and deep peaking of the power grid,and the reduction of power generation efficiency during variable operating conditions and high emissions of pollutants such as nitrogen oxides(NO_x)from boiler combustion are still prominent problems faced by the thermal power generation industry.Based on a large amount of historical data obtained from boiler operation in power plants,it is of great practical significance to combine advanced optimization algorithms with prediction models to perform intelligent optimization search for controllable parameters in boiler operation in order to improve the economy and environmental protection of boiler operation.Based on the introduction of the basic principles of deep learning,the paper discusses the stacked denoising autoencoder(SDAE)and analyzes the main factors affecting the prediction accuracy of the self-coding network model.For the proposed sparrow search algorithm(SSA)in this paper,the basic principle and the search process are introduced in detail.To improve the global search capability of SSA,a chaotic optimized sparrow search algorithm with the introduction of firefly perturbation(FCOSSA)is proposed.Six classical functions are used to test the optimization performance of FCOSSA and compared with genetic algorithm(GA)and particle swarm optimization(PSO),and the results show that the improved sparrow search algorithm has higher search accuracy and stronger global search capability.Based on the detailed analysis of operational parameters affecting boiler efficiency and NO_x emission concentration of thermal power units,the paper established prediction models for boiler efficiency and NO_x emission concentration using SDAE based on the real historical operational data of a 1000MW thermal power unit,and also considered the influence of factors such as the initial weight threshold of the network,the degree of damage and the activation function,and the performance of the model was performed by the grid search algorithm.The performance of the model is optimized by the grid search algorithm.Based on the above model,an optimization method of boiler oxygen target value and furnace air and powder distribution based on SDAE prediction models and improved sparrow search algorithm is proposed with boiler efficiency improvement and NO_xemission concentration reduction as the optimization objectives.Pycharm software is used to program the optimization algorithm,and based on the above prediction models,the sensitivity of the model output to each input variable is analyzed by artificial perturbation to determine the search range of each input variable;the optimal target values of operating parameters such as oxygen quantity,secondary air combustion air baffle opening and coal feed to the coal mill corresponding to the current operating conditions are determined by using the unit operation data.The results show that the optimal target values of oxygen quantity,secondary air combustion air baffle opening and coal feed to the mill can effectively improve the boiler efficiency,reduce the NO_x emission concentration,ensure the economy and environmental protection of the unit operation,and provide the operation guidance for the operators.