基于改进的BP神经网络管输能耗预测模型

Prediction model of pipeline energy consumption based on improved BP neural network

输油管道能耗指标受多种因素影响,为准确预测输油管道的能耗值,选择具有自组织、自适应能力且能逼近任意非线性连续映射的BP神经网络创建能耗预测模型;为提高模型的泛化能力,在传统的BP神经网络计算过程中加入误差控制公式,最终建立了基于改进的BP神经网络原油管道能耗预测模型。选用某输油管道运行能耗数据作为样本,为提高计算收敛速度和精度,对样本数据进行预处理,对建立的能耗预测模型进行训练和验证,得到该模型的模拟误差在2.77%以内,且模拟值能够真实反映真实值的变化趋势。将该模型推广至某天然气管道进行能耗预测,结果表明:其能够准确预测天然气管输能耗情况,预测误差不超过4.06%。因此,该模型适用于油气管输能耗预测,为管输能耗提供了一种新的预测方法。(图3,表2,参16)

Energy consumption of oil pipeline is affected by many factors. To predict the energy consumption accurately, the BP neural network, which is capable of self-organizing, self-adaptation, and approximating any nonlinear continuous mapping, is adopted to build a prediction model. In order to make the model more generalized, the error control formula is added in the computing process with traditional BP neural network. Finally, a prediction model of crude oil pipeline is obtained based on improved BP neural network. This model is applied to the samples of energy consumption taken from a pipeline, which are pre-processed in order to improve the convergence speed and accuracy. The results show that the simulation error of the model is within 2.77%, and the analog value can truly reflect the true value. Then, the model is promoted to a natural gas pipeline, indicating that the model can accurately predict the energy consumption of natural gas pipeline, with error not more than 4.06 % . Therefore, this model, as a new method, is applicable to predicting the energy consumption of oil and gas pipelines. （3 Figures, 2 Tables, 16 References）