摘要
随着电力系统中新能源比例不断增加,区域间的能量传输变化愈加剧烈,研究大电网断面功率调整方法十分必要。然而,传统算法由于存在潮流不收敛问题以及依赖专家经验等局限性,无法很好地克服在调整目标变化幅度较大时收敛性不佳的困难。为此,提出一种将人工经验和深度强化学习相结合的断面功率调整方法。首先,介绍了深度强化学习的基本概念,并提出了发电机提前筛选和功率补偿机制;然后,设计了深度强化学习模型的状态、奖励函数以及神经网络的结构,并通过在模型训练过程中引入知识经验,有效缩减了智能体的动作空间。最后,利用IEEE 39节点系统和东北电网的实际算例验证了所提方法的有效性。
With the increasing proportion of renewable energy in power systems,the energy transmission change between regions is becoming more and more violent.Therefore,it is necessary to study the power adjustment methods for transmission sections in large power grids.However,due to the limitations of the traditional algorithms,such as the problems of the non-convergence of power flow and relying on expert experiences,it cannot well overcome the difficulties of poor convergence when the adjustment target changes greatly.Therefore,a power flow adjustment method for transmission section combining artificial experiences and deep reinforcement learning is proposed.Firstly,the basic concepts of deep reinforcement learning are introduced,and the generator pre-selection and power compensation mechanism is proposed.Secondly,the reinforcement learning state,action space,reward function,and deep neural network framework are designed.The knowledge experience is introduced in the model training process,and the action space of the agents is effectively reduced.Finally,the effectiveness of the method is verified by practical cases of IEEE 39-bus system and Northeast China power grid.
作者
杨晓东
严剑峰
刘佳霖
YANG Xiaodong;YAN Jianfeng;LIU Jialin(China Electric Power Research Institute,Beijing 100192,China)
出处
《电力系统自动化》
EI
CSCD
北大核心
2023年第15期133-141,共9页
Automation of Electric Power Systems
基金
国家电网有限公司科技项目(5100-202155455A-0-0-00)。
关键词
潮流
输电断面
深度强化学习
近端策略优化
功率调整
power flow
transmission section
deep reinforcement learning
proximal policy optimization
power adjustment