基于LSTM预测与泵阀协同的瓦斯抽采主管路智能调控研究

Intelligent control of gas drainage main pipelines based on LSTM prediction and pump-valve coordination

  • 摘要: 针对煤矿瓦斯抽采主管路中抽采泵与电动阀门强非线性耦合、瓦斯涌出剧烈波动导致纯流量跟踪滞后与控制目标失配的问题,构建了遵循“先阀后泵”原则的泵阀协同智能调控策略,设计了基于长短时记忆网络(LSTM)预测与泵阀协同的瓦斯抽采主管路智能调控系统。以抽采纯流量优化为目标,以抽采纯流量、阀门开度、泵频率、抽采负压为输入,通过LSTM模型预测未来抽采纯流量,对预测结果进行可信度校验:当预测结果通过可信度校验后,根据预测值与实测值的相对关系判断抽采纯流量变化趋势;当预测结果未通过可信度校验或前馈调节难以满足控制要求时,系统切换至基于实时偏差的反馈修正模式。设计了基于连续调节失效触发的目标值动态更新机制,实现了控制目标与工况变化的动态匹配。试验结果表明:泵阀协同智能调控的最大动态偏差为±6.2 m3/min,较人工经验调控和传统PID控制分别降低66.5%和50.8%;平均绝对百分比误差为6.8%,较人工经验调控和传统PID控制分别降低63%和45%;泵阀协同智能调控的响应时间最短,超调次数最少,泵变频动作次数显著低于PID控制,验证了“先阀后泵”策略的有效性。工程应用结果表明,智能调控系统运行过程中未出现明显振荡或高频调节现象,能够有效提升抽采泵站主管路抽采纯流量的动态维持能力,具有良好的调控稳定性和工程适应性。

     

    Abstract: This study aims to address delayed tracking of pure gas extraction flow and mismatched control targets caused by strong nonlinear coupling between drainage pumps and motorized valves and sharp fluctuations in gas emission in coal mine gas drainage main pipelines. An intelligent pump-valve coordinated control strategy following a valve-first, pump-second principle was developed, and an intelligent control system for gas drainage main pipelines based on Long Short-Term Memory (LSTM) prediction and pump-valve coordination was designed. To optimize pure gas extraction flow, the LSTM model predicted future pure gas extraction flow using pure gas extraction flow, valve opening, pump frequency, and drainage negative pressure as inputs. The predictions were subjected to reliability checks. When the predictions passed these checks, the trend in pure gas extraction flow was determined from the relationship between predicted and measured values. When the predictions failed the checks or feedforward adjustment could not meet the control requirements, the system switched to a feedback correction mode based on real-time deviations. A dynamic target update mechanism triggered by the failure of successive adjustments was designed, achieving a dynamic match between control targets and changing operating conditions. Experimental results showed that the maximum dynamic deviation under intelligent pump-valve coordinated control was ±6.2 m3/min, representing reductions of 66.5% and 50.8% compared with manual experience-based control and conventional Proportional-Integral-Derivative (PID) control, respectively. The mean absolute percentage error was 6.8%, representing reductions of 63% and 45%, respectively. Intelligent pump-valve coordinated control achieved the shortest response time and the fewest overshoots, with significantly fewer pump frequency adjustments than PID control, confirming the effectiveness of the valve-first, pump-second strategy. Field application results showed no obvious oscillations or high-frequency adjustments during operation. The intelligent control system can effectively improve the ability to dynamically maintain pure gas extraction flow in drainage pump station main pipelines, with good control stability and adaptability to field conditions.

     

/

返回文章
返回