基于LSTM-Adma模型的采矿业电力消耗量研究  

Research on Electricity Consumption in Mining Industry Based on LSTM-Adma Modeling

作　　者：李晨昊 汪运辰 郑泽辉 Li Chenhao;Wang Yunchen;Zheng Zehui(School of Electrical and Control Engineering,Liaoning Technical University,Huludao Liaoning 125000,China)

机构地区：[1]辽宁工程技术大学电气与控制工程学院,辽宁葫芦岛125000

出　　处：《现代工业经济和信息化》2024年第7期255-258,共4页Modern Industrial Economy and Informationization

摘　　要：基于某地2000—2021年采矿业电力消耗量历史数据,分别进行时间序列预测(ARMA)和LSTM神经网络预测,在现有模型预测结果的基础上,通过数据插值处理,引入Adam算法,利用梯度的一阶矩的估计为每个电力消耗量参数更新方向增加一个惯性项,最终使得预测结果的迭代方向更加的稳定和平滑。结合两参数网格搜索算法经过Matlab平台进行验证,对比模型的RMSE和R2,发现Adam算法使得该地电力消耗量预测避免进入局部最优解,同时加速LSTM收敛速度,使得优化的LSTM-Adma预测模型RMSE减小和R2得到提高,避免进入局部最优解且预测值符合2022年辽宁省统计公报的调查趋势,为2024年以及未来该地区电力生产和能源节约利用政策规划提供参考。Based on the historical data of electricity consumption in mining industry from 2000 to 2021 in a certain place,time series prediction(ARMA)and LSTM neural network prediction were carried out,respectively.On the basis of the prediction results of the existing model,Adam algorithm was introduced through data interpolation,and the estimation of the first-order moments of the gradient was utilized to add an inertia term for the updating direction of each parameter of electricity consumption,which finally made the prediction increase by An inertia term is added to each power consumption parameter update direction using the first-order moment estimation of the gradient,which finally makes the iterative direction of the prediction results more stable and smooth.Combined with the two-parameter grid search algorithm verified by matlab platform,comparing the RMSE and the model,it is found that Adam's algorithm makes the power consumption prediction of the place avoid entering the local optimal solution,accelerates the convergence speed of LSTM,and at the same time,makes the optimized LSTM-Adma prediction model RMSE reduced and improved,avoiding to enter the local optimal solution and the predicted value conforms to the trend of the survey of Liaoning Provincial Statistical Bulletin in 2022.The survey trend of Liaoning Provincial Statistical Bulletin in 2022 provides a reference for the policy planning of electric power production and energy conservation and utilization in 2024 and in the future in the region.

关 键 词：电力消耗量 LSTM-Adma预测模型 网格搜索算法 

分 类 号：TM715[电气工程—电力系统及自动化]