摘要
可再生能源具有较强的波动性和随机性,其大规模接入配电网将严重威胁系统安全稳定运行。基于风-光-储多能协同的动态优化方法展现出显著优势,但此方法因涉及随机优化理论与深度强化学习等前沿领域,存在内容复杂、多学科交叉性强等教学难点。为了有效改善传统教学实验的不足,激发学生创新思维和多学科交叉应用能力,该文搭建了基于风-光-储多能协同的主动配电网动态优化实验教学平台,构建了含可再生能源主动配电网双层随机动态优化模型,提出了一种嵌入数学模型的双深度Q网络算法,并从多个角度详细分析和比较了仿真实验结果。通过该实验平台建设和案例设计,有助于提升学生对主动配电网动态优化方法的认知与理解,显著提升其多学科知识融合能力与复杂工程问题解决能力,为新工科背景下培养高水平人才提供了有效实践载体。
[Objective]The advancement of dual carbon goals has driven rapid expansion in renewable energy installations,particularly wind and photovoltaic systems.However,the inherent intermittency and stochasticity of these renewable sources substantially challenge power system stability when integrated at scale into distribution networks.While dynamic optimization strategies leveraging wind–solar–storage synergy demonstrate considerable operational advantages,their implementation presents notable pedagogical complexities due to the interdisciplinary integration of stochastic optimization theory and deep reinforcement learning frameworks.[Methods]To address the limitations of conventional pedagogical experiments and cultivate students’innovative thinking in interdisciplinary applications,this paper develops an experimental platform for active distribution network(ADN)dynamic optimization with wind–solar–storage multienergy coordination.The formulated ADN dynamic optimization model,characterized by extensive binary variables and multidimensional uncertainties,inherently constitutes a high-dimensional,nonlinear stochastic optimization problem.Employing a divide-and-conquer methodology,we design a bilevel stochastic optimization framework for ADNs with distributed energy resources.To optimize the distribution network topology,the upper-level model is designed as an ADN dynamic reconfiguration model with binary variables.The lower-level model is formulated as an ADN operational optimization model incorporating wind,solar,and storage systems,ensuring economic efficiency and operational security of the distribution network.An innovative embedded mathematical model–double deep Q-network(EMM–DDQN)algorithm is proposed to efficiently solve this stochastic system.The upper-level dynamic reconfiguration model for ADNs,challenged by substantial uncertainties and integer variables,is addressed through a DDQN algorithm specifically designed for computationally efficient and precise solutions.When the grid topology is finalized through this process,the lower-level model executes operational optimization for renewable energy and energy storage-integrated ADNs,effectively reducing power losses,mitigating voltage deviations,and minimizing renewable curtailment rates.[Results]The experimental results of the dynamic optimization for ADNs based on wind–solar–storage multienergy coordination are as follows:1)The proposed EMM–DDQN algorithm can rapidly learn by dynamically interacting with the ADN environment.It converges at approximately 650 episodes and yields high-quality solutions.2)The ADN topology can be dynamically adjusted on the basis of load curves and renewable energy output.3)The dynamic operation strategy of the ADN,under the coordination of wind,solar,and storage,effectively mitigates voltage deviations,reduces network losses,and enhances the accommodation of renewable energy.4)In Scenario 1(without network reconfiguration),network losses and voltage deviations increase by 12.40%and 7.52%,respectively,compared to the proposed model.In Scenario 2(without energy storage),the renewable energy utilization rate decreases to 88.27%.In Scenario 3(without network reconfiguration and energy storage),the performance is the worst,with the average reward function value decreasing by 6.93%compared to the proposed model.In Scenario 4,the renewable energy utilization rate approaches its theoretical maximum,while network losses and voltage deviations reach optimal levels,ultimately yielding the highest average reward value.[Conclusions]This paper establishes a dynamic optimization experimental platform for ADNs based on wind–solar–storage multienergy coordination.A three-stage teaching framework—“key equipment–model formulation–system optimization”—is designed to effectively deepen students’understanding of ADN dynamic optimization methods.Using the simulation platform,students must integrate multidisciplinary knowledge,including power system analysis,optimization computation,and machine learning,to complete the full process from theoretical modeling to optimization solving.This hands-on practice enhances their comprehension of ADN dynamic optimization methods,considerably strengthens their ability to integrate interdisciplinary knowledge,and improves their capability to solve complex engineering problems.Ultimately,the platform serves as an effective educational tool for cultivating high-level talent under the framework of emerging engineering education.
作者
江昌旭
庄鹏威
林俊杰
郑文迪
邵振国
JIANG Changxu;ZHUANG Pengwei;LIN Junjie;ZHENG Wendi;SHAO Zhenguo(College of Electrical Engineering and Automation,Fuzhou University,Fuzhou 350108,China)
出处
《实验技术与管理》
北大核心
2025年第5期105-114,共10页
Experimental Technology and Management
基金
福州大学研究生教育教学改革项目(00489449)
福州大学研究生教育教学改革重点项目(FYJG2023001)
国家自然科学基金(72401069)。
关键词
可再生能源
风-光-储多能协同
主动配电网
实验平台
renewable energy
wind–solar–storage multienergy coordination
active distribution network
experimental platform
