基于ARIMA-SVM方法的梯级泵站机组运行趋势预测

Operation Trend Prediction of Cascade Pumping Stations Based on ARIMA-SVM Method

针对多因素参与下梯级泵站机组运行趋势预测建模困难且预测准确度低、适应性差的问题,以宁夏盐环定扬黄工程为研究对象,引入时间序列分析法,提出了基于ARIMA与SVM模型组合的泵站机组运行趋势预测方法,即选择机组运行技术参数中的能源单耗和平均负荷作为试验样本,由ARIMA建模对处理后的数据进行线性拟合,通过SVM模型对残差进行预测处理,补偿机组运行中的非线性变化,综合二者预测结果得到组合模型预测值。结果表明,最优模型为ARIMA(1,1,3)、ARIMA(2,1,1),SVM模型最优参数分别为c=38、g=0.06和c=68、g=0.18;组合模型对试验样本的预测拟合优度分别达到0.9992、0.9984,均方根误差分别为1.67×10-5、3.9×10-8,平均绝对百分比误差分别为0.0361%、0.0747%,说明该组合模型预测泵站机组运行趋势精度较高、效果良好,可为泵站机组运行状态监测系统优化升级提供理论基础。

Aiming at the problems of difficult modeling,low prediction accuracy and poor adaptability of the operation trend prediction of cascade pumping station units under the process of multi-factor participation,this study took Yanhuanding Yellow River Project in Ningxia as the research object,introduced the time series analysis method,and put forward the operation trend prediction method of pumping station units based on ARIMA and SVM combination model.The energy consumption and average load in the operation technical parameters of the unit were selected as the test samples.The ARIMA model was used to linearly fit the processed data,and the SVM model was used to predict the residual error to compensate for the nonlinear change in the operation of the unit.The prediction results of the combination model were obtained by combining the two prediction results.The results show that the optimal models are ARIMA(1,1,3)and ARIMA(2,1,1),and the optimal parameters of SVM model are c=38,g=0.06 and c=68,g=0.18,respectively.The goodness of fit of the combined model for the test samples were 0.9992 and 0.9984,RRMSEwere 1.67×10-5and 3.9×10-8,the MMAPEwere 0.0361%and 0.0747%,indicating that the combined model has high accuracy and good effect in predicting the operation trend of pumping stations.ARIMA-SVM combination model can provide a theoretical basis for the optimization and upgrading of pumping station unit operation condition monitoring system.