A methane concentration monitoring method based on deep Q network

By constructing a methane concentration monitoring method based on deep Q-networks, the problems of discontinuous monitoring data and noise interference in existing systems are solved. This enables accurate identification of methane concentration change trends and stable risk warning, thereby improving the stability and risk identification capabilities of the monitoring system.

CN122409952APending Publication Date: 2026-07-17
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Filing Date
2026-04-02
Publication Date
2026-07-17

AI Technical Summary

Technical Problem

Existing methane concentration monitoring systems are prone to abnormal fluctuations and time misalignments in monitoring data when faced with environmental disturbances, equipment errors, and inconsistent sampling frequencies. This results in discontinuous monitoring results and noise interference, making it difficult to accurately describe the trend of methane concentration changes. Furthermore, they are prone to false alarms or missed alarms in complex environments.

Method used

An environmental sequence is constructed by collecting methane concentration, temperature, humidity, and airflow data. Time alignment, anomaly removal, and normalization are performed. The SINDy algorithm is used to identify the dynamic structure of the environmental state. A state trajectory is generated by combining a neural ODE network. A state evolution direction mapping mechanism is introduced into the improved Rainbow DQN decision network to generate a monitoring strategy. Finally, a stable risk result or alarm signal is output.

Benefits of technology

It enables continuous modeling of environmental state change processes and stability analysis of monitoring strategies, improving monitoring accuracy and risk warning stability, and enhancing the ability to characterize environmental state evolution trends and identify risks.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122409952A_ABST
    Figure CN122409952A_ABST
Patent Text Reader

Abstract

The application discloses a kind of methane concentration monitoring methods based on deep Q network, comprising the following steps: S1, the methane concentration of acquisition monitoring area, temperature, humidity and air flow;S2, construct state feature;S3, the dynamic identification processing of state sequence is executed using SINDy algorithm, and the dynamic structure of environmental state is identified according to state sequence, and environmental state dynamic structure is input neural ODE network and executes continuous modeling processing, generates state trajectory;S4, construct evolution feature;S5, construct improved Rainbow DQN decision network, introduce state evolution direction mapping mechanism, generate monitoring strategy;S6, generate stable risk result;S7, stable risk result is compared with preset risk threshold, and output alarm signal or methane concentration monitoring result.The application improves the accuracy of methane concentration monitoring and the reliability of risk early warning.
Need to check novelty before this filing date? Find Prior Art