基于改进Lyapunov指数的煤矿井下瓦斯浓度预测研究

发布时间：2018-05-18 04:18

本文选题：混沌理论 + 相空间重构　；参考：《山西大学》2015年硕士论文

【摘要】：煤炭为我国国民经济的发展提供了有力的能源支持。但是煤炭生产一直被安全问题所困扰,其中瓦斯事故是威胁我国煤矿井下安全生产的主要灾害之一。因此对瓦斯浓度进行科学准确的预测具有重要意义。煤矿井下系统受到多种因素的影响,各种因素相互作用形成了具有混沌性质的复杂煤岩瓦斯动力系统。因此,可以基于混沌理论建立煤矿井下瓦斯浓度预测模型,进而达到瓦斯预警和事故防控的目的。本文首先针对基于混沌理论的最大Lyapunov指数预测模型存在的符号选择问题,引入加权一阶局域的思想推导出新的预测公式,并在单瓦斯传感器数据预测中进行模型验证及分析。然后,针对多传感器数据预测模型存在的多变量选择问题,引入相关性分析的方法分析变量之间的相关性强弱。接着,采用联合考虑欧式距离和夹角余弦的方法对基于最大Lyapunov指数预测模型的邻近点选择问题进行改进。最后建立基于多传感器数据的改进最大Lyapunov指数瓦斯浓度预测模型并通过实验分析和验证。本文研究内容主要有：(1)探讨了混沌理论的相空间重构技术,并通过参数相关法和参数不相关法分别对重构参数进行了求取分析；研究了混沌性的识别,特别是通过最大Lyapunov指数是否大于零来验证系统的混沌性。(2)通过引入加权一阶局域法的思想,对基于最大Lyapunov旨数的单瓦斯传感器数据预测模型进行推导,消除预测时符号选择的问题,并用鹿台山煤矿的实时瓦斯传感器数据验证模型。通过与传统预测模型的对比分析得出,改进模型符号确定且均方根误差为2.61%较传统模型4.27%低,改进模型在瓦斯浓度预测上较优。(3)首先引入相关性分析法得出对瓦斯影响大的因素作为多变量预测模型的输入。然后提出考虑欧式距离和夹角余弦的思路对基于最大Lyapunov指数的多变量预测模型进行改进。接着用采集自霍尔辛赫煤矿的多传感器数据验证模型,并分别与传统预测模型和BP神经网络预测模型进行实验对比。结果得出改进模型的预测精度较后两者模型都有提高,说明改进模型在多变量瓦斯浓度预测上是有效的。
[Abstract]:Coal provides powerful energy support for the development of our national economy. However, coal production has always been troubled by safety problems, among which gas accident is one of the main disasters that threaten the safety of underground coal production in China. Therefore, it is of great significance to predict gas concentration scientifically and accurately. The underground coal mine system is influenced by many factors, and various factors interact to form a complex coal-rock gas power system with chaotic properties. Therefore, the gas concentration prediction model can be established based on chaos theory, and the purpose of gas early warning and accident prevention and control can be achieved. In order to solve the problem of symbol selection in the maximum Lyapunov exponent prediction model based on chaos theory, a new prediction formula is derived by introducing the idea of weighted first order local area, and the model verification and analysis are carried out in the prediction of single gas sensor data. Then, aiming at the problem of multivariable selection in multisensor data prediction model, the correlation analysis method is introduced to analyze the correlation between variables. Then, the method of combining Euclidean distance and angle cosine is used to improve the selection of adjacent points based on the maximum Lyapunov exponent prediction model. Finally, an improved maximum Lyapunov exponent gas concentration prediction model based on multi-sensor data is established and verified by experiments. In this paper, the phase space reconstruction technology of chaos theory is discussed, and the parameter correlation method and parameter independent method are used to obtain and analyze the reconstruction parameters, and the identification of chaos is studied. In particular, the chaos of the system is verified by whether the maximum Lyapunov exponent is greater than zero.) by introducing the idea of weighted first order local method, the prediction model of single gas sensor data based on the maximum Lyapunov number is derived. The problem of symbol selection in prediction is eliminated and the real time gas sensor data of Lutaishan Coal Mine is used to verify the model. By comparing with the traditional prediction model, the results show that the root-mean-square error of the improved model is 2.61% lower than that of the traditional model 4.27%. The improved model is superior in predicting gas concentration. Firstly, the correlation analysis method is introduced to obtain the factors that have a great influence on gas concentration as the input of multivariate prediction model. Then the idea of considering the Euclidean distance and the angle cosine is proposed to improve the multivariable prediction model based on the maximum Lyapunov exponent. Then the multi-sensor data collected from Holchingham coal mine are verified and compared with the traditional prediction model and BP neural network model. The results show that the prediction accuracy of the improved model is higher than that of the latter two models, which shows that the improved model is effective in predicting multivariable gas concentration.
【学位授予单位】：山西大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TD712

【参考文献】