Combining Mechanism Modeling And Deep Learning,the Output Prediction Of Concentrated Solar Power Plants Is Studied

With the depletion of fossil fuels and the consequent environmental pollution,the demand for renewable energy is increasing,which requires researchers to improve the utilization of renewable energy.Solar energy is one of the renewable energy sources,with its high universality,pollution-free and richness,it has been highly valued by many countries.In this context,a new power generation method that uses solar radiation energy for power generation-solar thermal energy storage power generation was born.For solar thermal power plants,due to the uncertainty of solar energy,output fluctuations are difficult to avoid and unpredictable.In view of the fact that the above reasons will adversely affect the CSP,after summarizing the factors that limit the output prediction of the CSP station,a method for predicting the output of the CSP energy storage power station is proposed.The research content of this thesis mainly includes the following points:(1)The main classification and working principle of solar thermal energy storage power generation system are summarized.Through the analysis and understanding of the working principle of the solar thermal energy storage power station,it is known that the direct radiation of the sun is the main factor affecting its final output.Then,on the basis of the above research,with the purpose of power station output prediction,the unequal equation and equation constraint of the energy flow between subsystems are obtained by using the steady-state difference equation,and finally the mathematical model of the energy flow of the CSP energy storage power station is constructed.The mathematical model established by the simulation results can accurately describe the energy flow of the CSP energy storage power generation system.(2)The basic network structure and training algorithms of several deep learning are briefly described,which are convolutional neural networks,long short-term memory networks,and convolutional long-short-term memory neural networks.In order to solve the disadvantages of the low prediction accuracy and insufficient convergence of these neural networks in practice,combined with the research background and existing research results,sparse operations and adaptive ideas are introduced,and the above neural network structure and training algorithm are optimized and improved.Simulation comparison results show that the improved and optimized network effectively improves the shortcomings of the original network.(3)In view of the fact that the output data of CSP energy storage power station is difficult to meet the research requirements,a method for predicting the output of CSP energy storage power station combining mechanism modeling and deep learning is proposed.Firstly,the improved and optimized deep learning network model is used to predict the direct solar radiation of the concentrating system,and then the prediction model of direct solar radiation is established to screen out the model with the best prediction effect and accuracy.Secondly,the predicted value of direct solar radiation is brought into the mathematical model of the solar thermal energy storage power station to obtain the predicted value of the final output of the power station.Simulation results show that compared with the other two neural networks,the improved convolutional long-term memory neural network can obtain more accurate prediction results,which proves the feasibility of the proposed CSP output prediction method.