沙坪二级水电站短期水位预测与实时调控策略  

Short-term Water Level Prediction and Real-time Control Strategy for ShapingⅡHydropower Station

作　　者：郭爽 龙岩 王孝群 李有明 何滔[4] 汪广明 GUO Shuang;LONG Yan;WANG Xiao-qun;LI You-ming;HE Tao;WANG Guang-ming(School of Water Resources and Electric Power,Hebei Universily of Engineering,Handan 056038,China;Hebei Key Laboratory of Smart Water Conservancy,Hebei Universily of Engineering,Handan 056038,China;Bgi Engineering Consultants LTD,Bejing 100038,China;Guoneng Daduhe Shaping Power Generation Co.,LTD.,Leshan 614300,China)

机构地区：[1]河北工程大学水利水电学院,河北邯郸056038 [2]河北工程大学河北省智慧水利重点实验室,河北邯郸056038 [3]北京市勘察设计研究院有限公司,北京100038 [4]国能大渡河沙坪发电有限公司,四川乐山614300

出　　处：《水电能源科学》2022年第8期83-87,共5页Water Resources and Power

基　　金：河北省高等学校科学技术研究项目(BJK2022038);邯郸市科学技术研究与发展计划项目(21422012319)。

摘　　要：针对沙坪二级水电站来水不确定、闸门负荷动作频繁的问题,提出了一种基于神经网络的水电站坝前水位预测新方法,建立了一个基于长短期记忆神经网络(LSTM)的水位预测模型,并应用于沙坪二级水电站的坝前水位预测,与BP神经网络预测结果进行对比分析。结果表明,LSTM预测结果具有更高的精度,平均绝对误差为0.1347,均方根误差为0.1950,纳什系数为0.9337,能很好地预测短期水位;提出了基于负荷调整余量与水位预测模型的水电站实时调控策略,根据预测的水位超上限、水位超下限、水位不超限3种情况进行决策分析,实现了减少沙坪二级电站的闸门动作次数,保障电网的安全稳定运行。In view of the uncertainty of water inflow and frequent gate load action of Shaping II hydropower station,a new method for water level prediction in front of the dam was proposed based on neural network.A water level prediction model based on long-term and short-term memory neural network(LSTM)was established to forecast the water level in front of the dam of Shaping II hydropower station.The prediction results were compared with those of BP neural network.The results show that the LSTM prediction results have higher accuracy,the average absolute error is 0.1347,the root mean square error is 0.1950,and the Nash coefficient is 0.9337,which can well predict the short-term water level.The real-time regulation strategy of hydropower station based on load adjustment allowance and water level prediction model was proposed.The decision analysis was carried out according to the predicted three cases of water level exceeding the upper limit,water level exceeding the lower limit and water level not exceeding the limit.The gate action times of Shaping II hydropower station are reduced to ensure the safe and stable operation of power grid.

关 键 词：LSTM神经网络 BP神经网络 短期水位预测 实时调控策略 

分 类 号：TV742[水利工程—水利水电工程]