Discrimination of mine inrush water source based on PCA -CRHJ model
-
摘要: 针对传统矿井突水水源判别模型存在非线性能力较差、模型稳定性较差、判别精度低等问题，基于主成分分析（PCA）法和确定性分层跳跃循环网络（CRHJ）构建了PCA-CRHJ矿井突水水源判别模型。引入PCA对多元时间突水序列进行降维并提取关键特征，重构突水数据,获得主成分突水序列，对CRHJ进行模型训练，将训练完成的模型应用到张集煤矿和新庄孜煤矿突水水源判别中进行有效性验证。结果表明：① 通过与CRHJ、确定性循环跳跃网络（CRJ）、回声状态网络（ESN）模型进行对比，表明PCA-CRHJ模型的实际判别效果最优，准确率可达100%；② PCA-CRHJ模型有5类主要参数，分别为储备池规模、输入连接权重、单向连接权重、分层双向跳跃权重、跳跃步长，对该5类参数进行敏感性分析，表明输入权重参数对模型判别结果的影响最大；当3类权重参数取得最优值且保持不变时，储备池规模对模型误差影响最大，而跳跃步长的影响则较小。
-
关键词:
- 矿井突水 /
- 突水水源判别 /
- 确定性分层跳跃循环网络 /
- 主成分分析 /
- 多元时间序列
Abstract: Aiming at problems of traditional discriminant model of mine inrush water source, such as poor nonlinear ability, poor model stability and low discrimination accuracy, PCA -CRHJ discriminant model of mine inrush water source is constructed based on principal component analysis (PCA) method and cycle reservoir with hierarchical jumps (CRHJ). PCA is introduced to reduce dimension of multivariate time water inrush sequence and extract key features, the water inrush data is reconstructed to obtain principal component water inrush series, and the CRHJ model is trained by reconstructed sequence. The model completed by the training is applied to water inrush source discrimination in Zhangji Coal Mine and Xinzhuangzi Coal Mine for validity verfication. The results show that: ① By comparing with CRHJ、cycle reservoir with regular jumps (CRJ) and echo state network (ESN) models, the results show that PCA -CRHJ model has the best actual discriminant effect and the accuracy can reach 100%. ② The PCA -CRHJ model has five main types of parameters, namely, reserve pool size, input connection weight, one -way connection weight, hierarchical two -way jump weight and jump step size, the sensitivity analysis of these five types of parameters shows that the input weight parameters have the greatest impact on the model discrimination accuracy. When three kinds of weight parameters obtain the optimal value and remain unchanged, the reserve pool size has the greatest impact on the model error, while the jump step size has less effect. -
-
期刊类型引用(6)
1. 刘伟韬，李蓓蓓，杜衍辉，韩梦珂，赵吉园. 基于改进的SSA-BP神经网络的矿井突水水源识别模型研究. 工矿自动化. 2024(02): 98-105+115 . 
本站查看
2. 刘艳冬，刘滢，卢兰萍，白峰青，王铁记，卫皓皓. 基于ZOA-CNN-GRU模型的煤层底板突水等级预测. 中国煤炭. 2024(06): 44-51 . 百度学术
3. 姜子豪. 基于Weka平台的矿井突水水源判别. 地下水. 2023(04): 26-29+121 . 百度学术
4. 孙钧青，王皓，杨建，尚宏波，王甜甜，乔伟. 无机-有机综合指标在煤层顶板涌水水源判别中的应用. 煤矿安全. 2023(12): 182-190 . 百度学术
5. 陈向俊，傅军平，李科，陈栋栋，李黎苹. 基于AOA-CRHJ的曳引电梯钢丝绳动态张力预测方法. 起重运输机械. 2022(04): 36-41 . 百度学术
6. 邢馨予，赵雪花. 基于LMD-QPSO-CRJ模型汾河上游月径流预测方法研究. 中国农村水利水电. 2021(10): 15-20 . 百度学术
其他类型引用(4)
计量
- 文章访问数: 63
- HTML全文浏览量: 16
- PDF下载量: 14
- 被引次数: 10
下载:
