The Visual Feedback Neural Network Detection Method For Orientation Of Heliostats

In the social development,solar energy as a permanent clean energy has received widespread attention,and how to improve the utilization efficiency of solar energy has gradually become the focus of current research.Tower heliostat collector power has gradually become research focus of solar thermal utilization due to its unique advantages of being able to establish a large-scale grid-connected power generation system and high thermal storage performance.In this paper,the calibration problems of heliostat in tower heliostat collector power are studied,which mainly including the modeling and effective correction of heliostat attitude.The main research contents and contributions are as follows:The first part: a modeling method of heliostat attitude based on spherical coordinate system is proposed.The method uses the spherical sine and cosine law to solve the attitude of heliostat.In addition,a strategy of correcting heliostat attitude based on the shape of the reflected spot is proposed,aiming at the existing situation that the method of correcting the deviation based on the center point of the reflected spot can not fully use the information of the reflected spot.The second part: aiming at the heliostat attitude proposed based on heliostat reflected speckle shape,the heliostat reflected speckle image is collected and the characteristic information of the reflected speckle is obtained by image processing.We combine the BP neural network algorithm to mine the hidden function relationship between the spot features and the heliostat attitude.Firstly,a simple image geometric invariant method is used to identify the target image.Secondly,this paper analyzes the shape of the solar reflected light spot and proposes to use the least squares method to fit the sun spot by elliptic curve fitting,and gets five basic features by elliptic curve fitting.Finally,based on the BP neural network algorithm,the heliostat reflected speckle features and corresponding heliostat attitude angles are theoretically analyzed,and the selection of important parameters such as training BP neural network and the number of hidden layers are involved.The third part: In this paper,a small biaxial experimental platform and a CCD camera are performed as physical experiment platform to verify the proposed heliostat attitude strategy based on heliostat reflected speckle shape.In this process,the heliostat reflected speckle features are used as the input values of training BP neural network,and the corresponding heliostat actual attitude angle is used as the output value to get threshold matrix and threshold parameters and verify the effect of BP neural network.At the same time,comparing the proposed strategy with the traditional scheme based on the center of the reflected speckle,we verify the effectiveness and superiority of the proposed strategy.