Data-driven Intelligent Identification Algorithm And Application Research For Ultra-Supercritical Unit Coordinated Control System

In recent years,despite the vigorous development of new energy power generation represented by wind power and photovoltaic power generation,thermal power generation still plays an important role in China’s electricity industry.Compared with traditional supercritical and subcritical units,ultra-supercritical units have the characteristics of high steam parameters,large power capacity,and wide power generation range.They have significant advantages in improving power generation efficiency,enhancing the economy and stability of power grid peaking and frequency regulation,and reducing CO2 emission pollution.The deployment of China’s thermal power industry has entered the "ultra-supercritical" era,and ultra-supercritical units have been planned as a key component of China’s thermal power unit construction.To guarantee the optimal operation of ultra-supercritical units,it is necessary to ensure the safety and stability of the core segment coordinated control system.For the coordinated control system of ultra-supercritical units,an accurate boiler-turbine coupled process model is the basis for further advanced control strategy design,satisfactory control performance and energy-saving optimization operation.However,there is a strong coupling between the important variables involved in the boiler-turbine coupled process,and the overall boiler-turbine coupled process presents serious nonlinearity and complexity,which causes great obstacles to the model identification of boiler-turbine coupled process,and is also the main difficulty in establishing the model of boiler-turbine coupled process in this paper.In view of the above modeling problem of the boiler-turbine coupled process in the coordinated control system of ultra-supercritical units,this paper considers that the transfer function model is still widely used in power plants.Therefore,it is hoped that the explicit transfer function model of the boiler-turbine coupled process is constructed by using the swarm intelligent algorithm.In addition,the deep learning method is suitable for analyzing the key information contained in the big data.Therefore,stacked denoising autoencoder is introduced into the boiler-turbine coupled process of coordinated control system in the ultra-supercritical unit,and the deep learning implicit model of the process is constructed to realize the large-scale operation condition modeling of the unit.Firstly.this paper expounds the overall framework of the coordinated control system of ultra-supercritical units and elaborates the principle and key variables of the boiler-turbine unit of the coordinated control system to be identified.By analyzing the influence of key variables on the process,a simplified conceptual model of boiler-turbine coupled process in the ultra-supercritical unit is determined.Secondly,the transfer function model structure of the boiler-turbine coupled process is established,and the data-driven multivariable model parameters intelligent identification scheme is developed to determine the transfer function model parameters.To make the model parameters identified by the scheme accurate and effective,cloud adaptive chaotic bird swarm algorithm combining sheep optimization and lion swarm optimization(SO-LSO-CACBSA)is proposed,and test functions are used to validate the effectiveness of SO-LSO-CACBSA algorithm from multiple perspectives.On the basis of the unit actual operation data,the model identification and verification of the boiler-turbine coupled process of a 1000 MW ultra-supercritical unit in China are carried out.The experimental results show the effectiveness of the data-driven multivariable model parameters intelligent identification scheme and the SO-LSO-CACBSA identification algorithm.Finally,in order to obtain the boiler-turbine coupled process model under large-scale operation conditions of ultra-supercritical units,the stacked denoising auto-encoder(SDAE)in the deep learning method is introduced into the model identification of boiler-turbine coupled process in ultra-supercritical units.Based on the actual operation data of a 1000 MW ultra-supercritical unit in China,the model identification and verification of the boiler-turbine coupled process are carried out.The experimental results show that the stacked denoising autoencoder model can basically reflect the characteristics of the boiler-turbine coupled process of coordinated control system in the 1000 MW ultra-supercritical unit under large-scale operation conditions.In order to make the stacked denoising auto-encoder more convenient to be applied in the actual ultra-supercritical unit,the GUI function in MATLAB is used to design the stacked denoising auto-encoder model identification software in the boiler-turbine coupled process,which can better realize human-computer interaction.