热值的神经网络模型研究

Study on the Neural Network Model of Calorific Value

在钢铁冶炼过程中,若能实时有效地检测使用煤气的热值、及时设计合理的供气方案,将有效提高煤气利用率、降低废气排放。这不仅能降低企业成本,而且对环境保护具有重要意义。国内热值检测设备的检测方式和手段,在热值检测的准确性和快速性方面有着很大的弊端。为改善燃气热值仪存在的大滞后,提高精准度,提出了一种由小波分析进行数据处理,应用BP神经网络建模并用遗传算法优化的学习算法,用来检测燃气热值。小波分析主要用来对过程数据进行降噪变换,提高源信号的信噪比;BP神经网络用来辨识过程模型;遗传算法用来优化神经网络的初始权值。小波分析可以克服源信号的噪声干扰。小波变换可以变换初始数据并进行特征提取,变换后的数据具有更高的信噪比。采用Matlab软件进行仿真,结果表明:该神经网络模型具有良好的逼近能力和泛化性能。该研究方法同样适用其他相关领域的研究。

In the process of iron and steel smelting, effectively detecting the heat value of gas in real time and making reasonable gas supply scheme can greatly improve the gas utilization and reduce the exhaust emission. This reduces the costs of enterprises, and is significant for environment protection. Some of the conventional detection equipment of heat value is available in China, but the traditional detection methods and measures features disadvantages in correctness and rapidity of the detection of heat value. In order to improve the accuracy and overcome the large delay of these instruments, the learning algorithm is proposed for detecting heat value of gas. With this algorithm,the data processing is conducted by using wavelet analysis,and modeling is conducted by BP neural network,and the genetic algorithm is used for optimization. The wavelet is mainly used for denosing transform of the process data and enhancing the SNR of the source signal; the SP neural network is used for recognizing process model; and the generic algorithmis used for optimizing the initial weights of the neural network. Wavelet analysis can overcome the noise interference of source signal, wavelet transform can concert initial data and conducts feature extraction, the data after transform possesses high SNR. Matlab software is adopted for simulation; the result of simulation indicates that the neural network model has excellent approximation ability and generalization performance. The research method of this model is suitable for the research in other related areas.