基于PSO-SVR的瓦斯浓度预测方法研究

发布时间：2018-03-25 11:45

本文选题：瓦斯浓度预测　切入点：PSO-SVR　出处：《中国矿业大学》2017年硕士论文

【摘要】：瓦斯灾害是煤矿生产过程中最严重的灾害之一,研究和分析瓦斯数据的特征及其内在联系,设计有效的算法模型实现对瓦斯浓度精确地实时预测,对瓦斯灾害的预警和防治具有重要意义。本文在对PSO-SVR预测模型研究的基础上,针对瓦斯浓度序列特征提取、瓦斯浓度序列时空建模和瓦斯浓度趋势预测提出三种瓦斯浓度预测方法,具体研究内容如下:一、研究瓦斯浓度数据的采集、处理及分析过程。利用插值等方法对采集到的数据进行初步清洗处理,将处理后的瓦斯浓度序列作为一种离散时间序列利用其高阶统计量对其高斯性和线性特征进行分析,为后续研究做铺垫。二、深度信念网络能够更好地表征样本的深层次特征,本文提出了DBN-PSO-SVR的瓦斯浓度预测方法。算法利用深度信念网络对瓦斯浓度训练集和预测集进行特征提取,将提取的特征作为PSO-SVR预测网络的输入来获取最终的预测值。实验结果表明,该方法使瓦斯浓度预测精度得到提高,算法性能优于常用的传统预测方法。三、研究瓦斯浓度运移扩散规律及其数字特征,证明瓦斯浓度序列具有时空特性,提出一种瓦斯浓度时空建模方法。该方法建立了基于样本数据驱动的空间权矩阵,根据空间权矩阵以及时空延迟量建立瓦斯浓度的时空训练集和预测集,利用PSO-SVR模型进行训练和预测,得到预测值。实验结果表明,该方法对比于只依赖时间维度的传统方法,在预测精度与泛化性能上都有很大的提高。四、为了掌握瓦斯浓度的变化趋势和变化范围,本文提出一种基于模糊信息粒化的瓦斯浓度趋势预测方法。通过构造三角模糊粒子并利用PSO-SVR模型对瓦斯浓度变化趋势和范围进行预测。结合第四章的时空建模方法对时空化的趋势预测方法进行研究。实验结果验证了算法的有效性。
[Abstract]:Gas disaster is one of the most serious disasters in the process of coal mine production. The characteristics of gas data and their internal relations are studied and analyzed, and an effective algorithm model is designed to realize accurate and real-time prediction of gas concentration. It is of great significance to early warning and prevention of gas disaster. Based on the study of PSO-SVR prediction model, this paper extracts the characteristics of gas concentration sequence. Three methods of gas concentration prediction are put forward in this paper. The main contents are as follows: firstly, the collection of gas concentration data is studied. The process of processing and analysis. Using interpolation and other methods to carry on the preliminary cleaning treatment to the collected data, taking the treated gas concentration series as a discrete time series, using its high-order statistics to analyze its Gao Si and linear characteristics. To lay the groundwork for further research. Second, the deep belief network can better represent the deep-seated characteristics of the sample. In this paper, a gas concentration prediction method based on DBN-PSO-SVR is proposed. The algorithm uses depth belief network to extract the feature of gas concentration training set and prediction set. The extracted feature is used as the input of the PSO-SVR prediction network to obtain the final prediction value. The experimental results show that the proposed method improves the accuracy of gas concentration prediction, and the performance of the algorithm is better than that of the conventional prediction method. The law of gas concentration migration and diffusion and its digital characteristics are studied. It is proved that the gas concentration sequence has space-time characteristics. A spatio-temporal modeling method of gas concentration is proposed. The method establishes a spatial weight matrix based on sample data driven. According to the spatial weight matrix and the space-time delay, the space-time training set and prediction set of gas concentration are established, and the predicted values are obtained by using PSO-SVR model. The experimental results show that this method is compared with the traditional method, which only depends on the time dimension. The prediction accuracy and generalization performance have been greatly improved. Fourthly, in order to grasp the change trend and range of gas concentration, In this paper, a gas concentration trend prediction method based on fuzzy information granulation is proposed. By constructing triangular fuzzy particles and using PSO-SVR model to predict the change trend and range of gas concentration, a spatio-temporal modeling method based on chapter 4 is presented. The experimental results show that the algorithm is effective.
【学位授予单位】：中国矿业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TD712.5

【参考文献】