基于RIME-VMD-SSA-LSTM法研究非生态因素影响的来水预报模型
Research on Inflow Forecasting Model of Non-Ecological Factors Based on RIME-VMD-SSA-LSTM Method
段宇 1黄君 2杨关友 1洪国仁 1段娟2
作者信息
- 1. 南方电网调峰调频发电有限公司鲁布革水力发电厂,云南 曲靖 655800
- 2. 昆明理工大学,云南 昆明 650093
- 折叠
摘要
对于上下游电站群落较多的来水预测问题,使用传统水文模型预测水库来水流量误差较大,流量测点信息包含大量非线性影响因素.以鲁布革电站汛期来水数据为例,采用RIME-VMD-SSA-LSTM组合算法研究上游具有较强非生态因素干扰的水库短期来水流量预测模型.结果显示,基于该算法提出的流量预测模型,4 项评价指标(RMSE=8.8743、MAE=6.3193、MAPE=3.5335%、R2=0.98631)较好,最大预测误差控制在 50%以内,相对于LSTM法及VMD-LSTM算法,在有较强非生态因素干扰下,可更精确地预测短期来水情况.
Abstract
For the water inflow prediction problem with a large number of upstream and downstream power plant communities,using traditional hydrological models to predict the inflow flow of reservoirs has a significant error.The flow measurement point information contains a large number of nonlinear influencing factors.Taking the flood season inflow data of Lubuge Power Station as an example,RIME-VMD-SSA-LSTM combination algorithm is used to study the short-term inflow prediction model of reservoirs with strong non ecological interference in the upstream.The results show that the traffic prediction model proposed based on this algorithm has good performance in four evaluation indicators(RMSE=8.8743,MAE=6.3193,MAPE=3.5335%,R2=0.98631),and the maximum prediction error is controlled within 50%.Compared to LSTM and VMD-LSTM algorithms,it can more accurately predict short-term water inflow under strong non ecological interference.
关键词
深度学习/短期来水预测/VMD-LSTM/预测精度Key words
deep learning/short term water inflow prediction/VMD-LSTM/prediction accuracy引用本文复制引用
基金项目
云南省基础研究计划(202201AUO70114)
出版年
2024
国家科技期刊平台
NSTL
万方数据