基于神经网络算法的旱季排水管网泵站液位预测方法

Prediction of Pump Station Level of Drainage Pipe in Dry Season Based on Neural Network Algorithm

由于排水管网水源的不确定性、非线性和滞后性,增加了运行管理和优化调度的难度,在运用水力模型计算时,无法准确获得实液位这一重要参数。建立排水管网泵站水池液位的BP神经网络预测模型作为替代方法,以上下游的水池液位及泵的排出量作为输入参数,未来的液位值为输出参数。根据现有数据对比,平均学习误差为1.65%,预测30min、60min、180min、360min时间段后液位的平均误差分别为0.61%、0.82%、2.70%、5.15%。更准确地利用液位预测值,可以为泵的启停提供有效的方法支持。

Because of the uncertainty,non-linearity and lag of the water source of the drainage network,the difficulty of operation management and optimal operation is increased.When using the hydraulic model to calculate,the important parameters of the real liquid level cannot be obtained accurately.In this paper,BP neural network prediction model of water level in pumping station of drainage network is established as an alternative method.The water level in the upper and lower reaches of the pump and the discharge of the pump are input parameters,and the future water level is output parameters.According to the comparison of existing data,the average learning error is 1.65%,and the average error of predicting the liquid level after 30min,60min,180min and 360min is 0.61%,0.82%,2.70%and 5.15%,respectively.Using more accurate liquid level prediction value can provide effective method support for pump start and stop.