Research On Intelligent Optimization Of Power Plant Operation Target Value Based On Neural Network Prediction Model

In recent years,the share of large-capacity high-parameter supercritical and ultra-supercritical units continues to increase in the total installed capacity of thermal power plants.In the context of the automatic generation control(AGC)and in-depth peak load shaving condition,accurate and reasonable determination of the optimal target values of the boiler operation parameters based on advanced intelligent modeling and optimization strategy,giving reasonable operating guidance to the operators,is of great significance to improve the economy and environmental protection of boiler operation.Based on the introduction of the principle and training method of neural network,2modeling algorithms,double hidden layer BP neural network(d BPNN)and regularized limit learning machine(RELM),were discussed,and the main factors affecting the prediction performance of neural network were analyzed.The principle and flow of sine cosine algorithm(SCA)were introduced in detail,and a simplified sine algorithm(SA),which adaptively adjusts its inertia weight was proposed to improve the local search ability of the algorithm.10 classical functions were used to test the optimization performance of SA,and the test results were compared with the simplified particle swarm optimization(s PSO)algorithm.The results show that the SA has higher search accuracy and execution efficiency.The operation parameters affecting boiler efficiency and NOx emission of thermal power unit were analyzed in detail.Based on the real historical operation data of a1000 MW thermal power unit,the prediction models of boiler efficiency and NOx emission were established with d BPNN and RELM respectively.At the same time,considering the initial weight threshold of the network,the activation function and the number of hidden layer nodes,the performance of the model was optimized by simulation experiments.Based on the above model,a boiler oxygen target value and furnace air distribution optimization method based on neural network prediction model and SA algorithm was proposed to optimize the boiler efficiency and NOx emission.Based on the operating data of the unit,the optimal target values of operating parameters such as oxygen content and secondary air door baffle openings were determined by MATLAB,and the optimization results of SA and s PSO algorithm were compared.The results show that the target values of operating parameters such as oxygen content and opening of secondary air door baffle obtained by the optimization scheme based on sine algorithm are more accurate.Using this value to optimize and adjust the operating parameters can effectively improve boiler efficiency and reduce NOx emission,and ensure the economy and environmental protection of unit operation,and provide operation guidance for operators.