扬程特性关联下矿井主泵排水智能控制方法研究

Research on Intelligent Control Method for Drainage of Mine Main Pump Under the Correlation of Head Characteristics

矿井下常常存在地下水渗积问题,通过控制主泵出水压力,能够有效地排除积水,保持井下环境相对干燥,增强煤矿地下环境的安全性。但由于矿井下的透水后压力负荷通常是不确定的,单纯以压力为传感信号的主泵排水控制的稳定性会受到较大影响,当前排水多以人工观测为主,排水控制智能化实现难度较大。提出一种扬程特性关联下矿井主泵排水智能控制方法。分析矿井下大功率变频式主泵的扬程特性(H-Q),获取变频式主泵排水压力和扬程特性(H-Q)之间的关系。利用这一关系作为传感信号,采用解耦补偿器和神经网络,利用神经网络在线调整比例、积分和微分(proportional integral derivative,PID)控制参数,组建大功率变频式主泵出水压力智能控制器,实现泵出水压力智能控制。通过实验分析证明,研究方法应用后在22 s左右矿井主泵达到最佳状态,过程中产生的相对误差均在0.5%以下。在多次迭代控制下,所提方法的控制延误始终低于0.1 ms。

There is often a problem of groundwater accumulation in underground mines.By controlling the outlet pressure of the main pump,water accumulation can be effectively removed,keeping the underground environment relatively dry,and enhancing the safety of the underground environment in coal mines.However,due to the uncertainty of the working load under the mine,the stability of the main pump outlet will be greatly affected,and water pressure control is difficult.To this end,an intelligent control method for drainage of mine main pumps in an unmanned environment is proposed.Analyze the working characteristics of high-power variable frequency main pumps in underground mines,and obtain the relationship between the outlet pressure and power consumption of the variable frequency main pumps.Combining the hydraulic model of the high-power variable frequency main pump in the mine,a decoupling compensator and neural network are used to adjust the proportional,integral,and differential(PID)control parameters online using neural networks.An intelligent controller for the outlet pressure of the high-power variable frequency main pump is established to achieve intelligent control of the pump outlet pressure.The experimental analysis shows that,after the application of the research method,the main pump of the mine reached its optimal state in about 22 s,and the relative errors generated during the process were all below 0.5%.Under multiple iterations of control,the control delay of the proposed method remains below 0.1 ms.