Film Thickness Prediction of Photovoltaic Cells Based on Improved Deep llarning
To solve the problems of high cost,difficulty and low efficiency in the PECVD coating process of photovoltaic cells,a film thickness prediction method based on improved deep learning was proposed based on BP neural network.Firstly,in the image information acquisition stage,the color depth value is de-signed to build a functional relationship between hue H and brightness L.In the data processing stage,the image data is normalized to turn the dimensional dataset into a pure quantity,which simplifies the calcula-tion and improves the performance of the model.In the neural network training stage,the Tanh function is used as the activation function,which makes the optimization process easier and improves the convergence speed of the function.Using the LM algorithm as the training function of the network,the convergence speed can be adaptively adjusted,the stability of the training process can be improved,and the model can complete the regression task more accurately.Increasing the momentum term reduces the oscillation trend during deep learning training and improves the stability of the model prediction process.Experimental re-sults show that compared with the traditional BP neural network,the number of iterations of the improved model is reduced by 8,the accuracy of film thickness prediction is increased by 12.9%,and the average er-ror is reduced by 1.558 nm.The maximum error is controlled within 4 nm.
film thickness predictionphotovoltaic cellsneural networksdeep learningcolor depth value
万方数据
维普
