GDGA-BP模型及其在干熄焦烧损控制中的应用
GDGA-BP Model and Its Application in Control of Dry Quenching Loss

文章提出了一种基于决策基因的改进遗传算法(Gene-Decision Genetic Algorithms, GDGA)，同时优化BP神经网络的隐含层节点数及初始权值、阈值，构建了GDGA-BP模型。首先，在进行遗传算法编码时采取了两种不同的编码方式，设计了决策算子以解决隐含层节点数对编码染色体长度的影响；然后在遗传算法流程中增加了染色体的锁存及调用，保证在迭代时遗传信息不会丢失，并对该算法做出自适应改进；最后通过对干熄焦烧损率的仿真及工艺实验，验证了该模型的优良性能及实用性。
In this paper, an improved gene-decision genetic algorithm (GDGA) based on Decision Gene is proposed, the number of hidden layer nodes, initial weights and thresholds of BP neural network is optimized, and the GDGA-BP model is constructed. Firstly, two different encoding methods were adopted in the genetic algorithm encoding, and decision operators were designed to solve the in-fluence of the number of hidden layer nodes on the length of encoding chromosomes. Then, the latch and call of chromosome are added to the genetic algorithm process to ensure that the genetic information will not be lost during iteration, and the algorithm is improved adaptatively. Finally, the simulation and technological experiment on the loss rate of dry quenching coke prove the ex-cellent performance and practicability of the model.