基于混合模型的原油管道运行电耗预测研究

Prediction for Power Consumption of Crude Oil Pipeline Based on Hybrid Model

电耗预测是原油管道运行能耗管理的重要依据,有助于输油企业制定批量调度与负荷分配等运行方案。由于采集的数据通常包含噪声和信息冗余,传统预测模型在小样本时预测精度较低。为提高在小样本时训练学习的预测能力,提出一种基于分解技术、改进粒子群算法(IPSO)和支持向量机(SVM)结合的原油管道电耗预测模型。利用自适应噪声的完整集成经验模态分解(CEEMDAN)对原始数据进行降噪和去信息冗余,提高预测精度;使用改进的粒子群算法优化支持向量机的超参数,增强拟合能力,对原油管道运行电耗进行预测。以国内3条原油管道为例,对建立的混合预测模型进行准确性评价。3条管道的决定系数值(R~2)分别为0.9109、0.9091和0.9534,与传统方法预测结果相比,在小样本时上述混合模型预测准确度最高。

The energy consumption prediction is important for energy management of crude oil pipeline. Accurate energy consumption prediction is helpful for oil enterprises to make decisions on batch scheduling and load distribution. The collected data usually contain noise and information redundancy, so the prediction accuracy of the traditional prediction model is often insufficient in the case of small samples. In order to improve the prediction ability of the model in the case of small samples, the paper proposed a crude oil pipeline energy consumption prediction model based on CEEMDAM-IPSO-SVM. The complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN) was used to reduce the noise and information redundancy of the original data. The improved particle swarm optimization(IPSO) was used to optimize the hyperparameters of support vector machine(SVM) and improve the fitting ability, so as to predict the energy consumption of crude oil pipeline. Taking three domestic crude oil pipelines as examples, the accuracy of the model was evaluated. Compared with the prediction results of traditional methods, the mixed model has the highest prediction accuracy in small samples, and the coefficients of determination of the three pipelines are 0.9109,0.9091,and 0.9534.