论文部分内容阅读
发光二极管(LED)太阳光模拟器的设计需要对LED光谱建立精度高且稳定性好的数学模型。针对LED光谱数学模型非线性的特点,提出利用一种经改进遗传算法(GA)优化的反向传播(BP)神经网络对LED光谱模型进行辨识。通过改进GA的算子,提高算法收敛效果和辨识精度,利用改进GA对BP神经网络初始和权值阈值进行优化,用于建立可靠的LED光谱模型。选取不同驱动电流条件下的白色、红色LED光谱进行实验验证,实验结果表明该算法拟合的LED光谱模型与实际测量光谱分布非常接近,相比其他模型精度更高,普适性更好。
The design of a light-emitting diode (LED) solar simulator requires the establishment of a mathematical model of the LED spectrum with high precision and good stability. In view of the non-linearity of mathematical model of LED spectrum, this paper proposes to use a backpropagation (BP) neural network optimized by improved genetic algorithm (GA) to identify the LED spectral model. By improving the operator of GA, the convergence effect and recognition accuracy of the algorithm are improved. The improved GA is used to optimize the initial and weight threshold of BP neural network to establish a reliable LED spectral model. Experimental results show that the LED spectrum model fitted by this algorithm is very close to the actual measurement spectrum, which is more accurate and universal than the other models.