首页|基于DQN算法的水电站站内负荷优化分配研究

基于DQN算法的水电站站内负荷优化分配研究

扫码查看
水电站负荷分配是自动发电控制体系的基础,既往方法难以适应新时期水电站大容量、多机组、复杂振动区的特性,以及流域复杂多样的多目标调控要求.本文分析电站不同运行时期的综合利用要求,构建了兼顾水利-电力需求的站内负荷分配模型,研发了基于深度Q网络(DQN)的高效求解算法,可在完成电网负荷指令的前提下,灵活满足不同时期和工况下各类流量、水位的调控目标.以大渡河枕头坝水电站为例,枯水期代表月内发电耗水量减少0.14%,实现节水增发,提高了电站运行经济性;汛期代表月内水位变幅的方差减小0.876 m2,有效平抑了高强度调峰调频下电站水位大幅波动,保证电站运行的安全性.同时,计算精度从既往静态负荷分配表的3~5MW提升至1 MW,计算效率提高88.2~106.4倍,可为新时期复杂水电系统调度运行提供技术支撑.
Research on optimal load allocation in hydropower station based on DQN algorithm
Load allocation of hydropower station is the basis of automatic power generation control system,and the conventional method is difficult to adapt to the complex and diversified multi-objective regulation and control re-quirements.This paper constructs a load distribution model that takes into account the demand of water resources and electric power,and develops an efficient solution algorithm based on the DQN(Deep Q-Netwoek)algorithm.Taking Zhentouba Hydropower Station as an example,the water consumption of power generation in the dry period is reduced by 0.14%,which improves the operation economy of the power station;and the variance of water level variation in the flood season is reduced by 0.876 m2,which effectively suppresses the water level fluctuation of the power station.At the same time,the accuracy of load distribution calculation is improved from 3-5 MW to 1 MW,and the calculation efficiency is improved by 88.2-106.4 times,which can provide technical support for the sched-uling and operation of complex hydropower systems in the new period.

load allocationreinforcement learningeconomic operationwater-electricity regulationefficient decision making

谭乔凤、宋嘉伟、闻昕、曾宇轩、王浩

展开 >

河海大学水利水电学院,江苏南京 210098

河海大学水安全与水科学协同创新中心,江苏南京 210098

中国水利水电科学研究院,北京 100038

负荷分配 强化学习 经济运行 水利-电力调控 高效决策

2024

水利学报
中国水利学会

水利学报

CSTPCD北大核心
影响因子:1.778
ISSN:0559-9350
年,卷(期):2024.55(11)