煤矿瓦斯灾害稳健预警的多智能体化建模与控制

Little attention is paid for the coupling among sensors in contemporary mine monitoring system, especially the correlation between the temporal consensus of the individual agents in time series and the spatial consensus induced among the agents which interact with each other because of their spatial deployment on some certain time or among some time windows; the sensors are hybrid, which can be for gas, co, temperature and wind tacometer, so they are both heterogeneous in chemical and physical characteristics and in dynamics. And on the account of insufficient emphasis on consensus of the time-space correlation, which results in time-space isolation for information processing among sensors, and sensitivity to both inner and outer disturbances turns into effect, and then the wrong and miss alarming in gas survillence can occur. In order to improve the robustness of the gas survillence, the sensors for mine monitoring system can be viewed as multi-agent system, in helps of iterative learning, algebraic topology, internal model principle, simultaneous invariant subspace as well as output disturbance decoupling theory, time-sapce consensus and invariance to outer disturbances are touched upon to the output information of hybrid agents, which can have parameter perturbation,aging as well as outer disturbance, and information transmission among them can have distributed time-delays, data packet dropout as well as switches. If they are availabe, then the non-fragile shielding ways on disturbance information are gotten, and even the further, the new rules and laws for sensors deploment and data management can be determined.

现有煤矿安全生产监控系统较少考虑多个传感器信息之间的耦合作用，特别是对瓦斯、一氧化碳、温度和风速等异质混杂多传感器信息在自身演化序列上的时间一致性和同一时刻或同一观察窗下同质不同位或既不同质又不同位各传感器之间协同演化的空间一致性的相关性，即时空关联一致性重视不够，造成对传感器信息加工的时空分离和传感器对内外扰动表现出敏感性，从而引发煤矿瓦斯预警的漏报和误报。为改善瓦斯预警的稳健性，将煤矿安全生产监控子系统中的传感器系统等效为多智能体系统，基于迭代学习、代数图论、内模原理、同时不变子空间和输出干扰解耦理论；在同时考虑多智能体系统个体本身参数漂移、老化等内部扰动和外界干扰的情况下；以信息传输存在分布时滞、丢包、切换的异质混杂多智能体系统为研究对象；探讨多智能体输出信息的时空一致性和对外界干扰的不变性，得到满足扰动信息非脆弱屏蔽的条件与方法，并尝试建立与此相对应的传感器设置原则和数据管理规则。

在“煤矿瓦斯灾害稳健预警的多智能体化建模与控制”项目支持下，课题组成员分别在多智能体系统的基础理论研究，网络化系统的理论探讨，煤矿安全的新理论与新装备三个子方向进行了有益的探索；其中，多智能体系统的基础理论研究涉及到基于不变子空间理论的多智能体系统的干扰解耦，基于自适应定位的传感器频率的对比研究，一种锚点选择的容错机制和基于细化的社会个体角色的领导者与跟随者多智能体系统的汇聚控制四个方面的内容；网络化系统的理论探讨涉及到非连续间断测量下2-D非线性系统的H无穷控制和丢包与时变扰动同时发生情况下的H无穷迭代学习控制两个方面的内容；煤矿安全的新理论与新装备涉及到基于云模型的煤矿瓦斯涌出系统混沌特征提取研究，基于V/S分析的瓦斯涌出量分形特性研究和矿用红外一氧化碳浓度测试仪的设计研究三个方面的内容，取得了一定的成果。