摘要
在梯级水电站短期优化调度的水力联系描述中,目前实际应用的固定滞时方法未考虑水流滞时与上游电站出库流量大小和河道状态等因素的动态关系,且有关动态滞时方法的研究亦常忽略各种输入因子的时间因次。因此,引入长短期记忆网络(LSTM)作为描述站间动态滞时关系的工具,同时针对数据挖掘模型参数难以确定的问题选取模拟退火算法(SA)作为网络超参数的选取方法,建立了一种模拟退火-长短期记忆网络(SA-LSTM)模型。以梯级水电站实际运行数据进行测试,同时与BP神经网络模型及固定滞时模型进行对比分析。结果表明,与固定滞时和BP神经网络模型相比,SA-LSTM模型能更准确地描述水电站站间的动态滞时关系,是分析计算梯级水电站站间水力联系的一种新的简明实用的方法。
In the description of hydraulic connection for short-term optimal operation of cascade hydropower stations,the current fixed flow lag time method does not consider the dynamic relationship between the flow lag time and the discharge of upstream hydropower stations and the state of river channel,while the study of dynamic lag method often ignores the time dimension of input factors.Therefore,this paper introduced long short-term memory(LSTM)network to describe the dynamic flow lag time relationship between hydropower stations.And then simulated annealing(SA)algorithm was used as the hyperparameters selection method to solve the problem that the hyperparameters of data mining model were difficult to be determined.Finally,the SA-LSTM model was established and tested with the actual operation data of cascade hydropower stations.The results show that the SA-LSTM model can more accurately describe the flow lag time relationship between hydropower stations than the fixed lag time and BP neural network model,and is a new concise and practical method to analyze the hydraulic connection between cascade hydropower stations.
作者
魏勤
陈仕军
谭政宇
黄炜斌
马光文
WEI Qin;CHEN Shi-jun;TAN Zheng-yu;HUANG Wei-bin;MA Guang-wen(College of Water Resource and Hydropower,Sichuan University,Chengdu 610065,China;State Key Laboratory of Hydraulics and Mountain River Engineering,Sichuan University,Chengdu 610065,China)
出处
《水电能源科学》
北大核心
2021年第6期16-19,共4页
Water Resources and Power
基金
国家重点研发计划(2018YFB0905204)
四川大学专职博士后研发基金(2018SCU12062)。
关键词
梯级水电站
动态滞时
长短期记忆网络
模拟退火算法
cascade hydropower stations
dynamic flow lag time
long short-term memory network
simulated annealing algorithm
