高级检索

低浓度瓦斯直燃系统瓦斯混配浓度异常预测与调节

Prediction and regulation of abnormal gas mixing concentration based on low-concentration gas direct combustion system

  • 摘要: 低浓度瓦斯直燃系统常面临瓦斯浓度异常波动的问题,尤其是在瓦斯混配过程中,若未能及时发现并调节异常浓度,可能引发严重的安全事故。如何提高混配后异常瓦斯浓度预测的精确性与自适应性,成为亟待解决的问题。针对这一问题,提出了一种低浓度瓦斯直燃系统瓦斯混配浓度异常预测与调节策略。首先基于甲烷、温度传感器等获取主管瓦斯浓度等数据构建了归一化流规范化数据集,并对数据集进行了多源数据融合;进一步设计了一种异常瓦斯浓度自适应预测与调节模型,利用编码特征提取模块对试验数据进行特征提取,异常瓦斯浓度预测模块对异常瓦斯浓度进行实时预测,解码调节模块利用自适应上采样调节异常瓦斯浓度值,依据消融试验、对比试验、自适应试验对模型性能进行全面评估;最后将异常瓦斯浓度自适应预测与调节模型集成于图形交互(Graphical User Interface, GUI)界面联合开发的低浓度瓦斯直燃系统混配瓦斯浓度异常预测与调节平台。结果表明:模型的准确率(Accuracy, Acc)为95.70%、精确率(Precision, Pre)为95.83%、均方误差(Mean Square Error, MSE)为0.068、每秒帧率(Frames Per Second, FPS)为98.3 帧/s,优于Transformer、Mamba-2等模型,集成于低浓度瓦斯直燃系统混配瓦斯浓度异常预测与调节平台的最大误差值不超过3%,能够及时有效地调节瓦斯浓度异常波动,确保低浓度瓦斯直燃系统稳定运行。

     

    Abstract: Low-concentration gas direct combustion systems often face the problem of abnormal fluctuations in gas concentration, especially during the gas mixing process. If the abnormal concentration is not discovered and corrected in time, it may cause serious safety accidents. Improving the accuracy and adaptability of the prediction of abnormal gas concentration after mixing has become an urgent problem to be solved. To address this problem, a strategy for the prediction and regulation of abnormal gas concentrations based on low-concentration gas direct combustion system is proposed. Firstly, a normalization flow normalized dataset is constructed based on the data of main gas concentration obtained from sensors such as methane and temperature, and multi-source data fusion is performed on the dataset; further, an adaptive prediction and regulation model for abnormal gas concentration is designed, and the coding feature extraction module is used to extract features from the experimental data, the abnormal gas concentration prediction module is used to predict the abnormal gas concentration in real time, and the decoding regulation module is used to regulate the abnormal gas concentration value using adaptive upsampling, the performance of the designed model is comprehensively evaluated based on ablation experiments, comparative experiments, and adaptive experiments; finally, the adaptive prediction and regulation model for abnormal gas concentration is integrated into a low-concentration gas direct-combustion system mixed gas concentration abnormality prediction and regulation platform jointly developed with a graphical user interface (GUI). The results show that the model’s accuracy (Acc) is 95.70%, precision (Pre) is 95.83%, mean square error (MSE) is 0.068, and frame rate per second (FPS) is 98.3 fps/s, which is better than Transformer, Mamba-2, and other models. The maximum error value of the mixed gas concentration anomaly prediction and regulation platform integrated into the low-concentration gas direct combustion system does not exceed 3%, which can timely and effectively regulate abnormal fluctuations in gas concentration and ensure the stable operation of the low-concentration gas direct combustion system.

     

/

返回文章
返回