Research on thermal effect of power LED devices based on PCA and neural network

This paper proposes a method to predict the thermal effect of the power LED based on principal component analysis (PCA) and BP neural network. This method adopts PCA for processing the sample data of the power LED, and then the processed data samples are considered as the input of BP neural network. Through this data processing the number of the input variables can be reduced and the BP network structure is simplified. The predicting experiment show that this method has strong adaptability and robustness. The method is used to predict the thermal effect of the power LED devices produced by APT Electronics LTD., and the prediction error is less than 5%.