ABSTRACT Solution prediction can accelerate unit commitment (UC) for repeatedly solved scheduling instances, but prediction errors may cause infeasible or economically inefficient schedules whose impact is revealed only after dispatch evaluation. Existing post‐prediction repair strategies usually rely on confidence thresholds, handcrafted rules or one‐shot re‐optimization, and therefore do not explicitly learn from delayed dispatch consequences. This paper proposes a prediction‐and‐repair framework for UC in which repair decisions are trained using dispatch feedback. The framework contains four replaceable modules, including commitment prediction, reliability‐based fixed/repair partition, reinforcement‐learning‐based repair policy and linear programming (LP)‐based dispatch evaluation. A behaviour‐cloning model first provides an initial commitment schedule. Historical root‐relaxation records and prediction confidence are then used to restrict the repair space, and the repair policy refines repairable variables using operating‐cost and feasibility feedback from the LP evaluator. Proximal policy optimization (PPO) is used as one implementation of the reinforcement‐learning repair module, but the framework is not restricted to a specific learning algorithm. Case studies on IEEE 300‐bus and French 1888‐bus systems show that, when root‐relaxation records are available for related UC instances, the framework improves prediction‐only and relaxation‐rounding schedules while maintaining low online evaluation cost.

Aiming at the shortcomings of traditional motor scheduling methods in modeling accuracy, constraint processing and real-time, this paper proposes a combined modeling and scheduling framework of safety constrained motors based on swarm intelligence algorithm. Firstly, the hybrid modeling strategy based on the combination of physical model of electromagnetic equation and LSTM network is adopted to realize high-precision dynamic modeling, effectively integrate mechanism knowledge and measured data, and significantly improve the prediction accuracy of the model under complex working conditions. Secondly, a hierarchical processing mechanism of security constraints is designed, which manages hard constraints, soft constraints and optimization objectives in different levels, and improves the robustness and practicability of the scheduling algorithm by dynamically adjusting the penalty coefficient. Finally, the adaptive weighted particle swarm optimization (AW-PSO) algorithm is proposed, and the dynamic inertia weight mechanism and adaptive penalty function are introduced to balance the global exploration and local search capabilities and effectively deal with security constraints. The experimental results show that compared with traditional methods, this method has obvious advantages in modeling accuracy, scheduling safety, energy saving effect and load balance, and can realize efficient and safe multi-motor collaborative scheduling, which provides new ideas and methods for the intelligent development of motor systems.

The cross-sectional control of the power system is the core link to ensure cross regional power balance and safe and stable operation. However, the uncertainty of output caused by the high proportion of new energy access, the multi section coupling effect, and the real-time limitations of traditional control methods pose serious challenges to the existing regulation system. This article proposes a cross-section control agent system based on multi-agent deep reinforcement learning, which achieves precise power control and economic improvement in complex scenarios through layered collaborative architecture design, algorithm optimization, and integration of source network load storage resources. The improved Deep Deterministic Policy Gradient (DDPG) algorithm introduces a high reward experience pool and stock judgment mechanism to accelerate training convergence and enhance policy generalization ability, combined with a convex relaxation power flow calculation module to ensure real-time performance. Establish a refined time series simulation model that includes the characteristics of new energy fluctuations and demand side response capabilities, and achieve multi section collaborative optimization through probabilistic power balance methods. The verification of the IEEE 118 node system shows that the system stabilizes the cross-sectional power control error within $\pm 0.8 \%$ per unit value, shortens the adjustment time to 2.1 seconds, reduces the adjustment cost by $31.3 \%$ compared to traditional methods, and increases the new energy consumption rate to $96.5 \%$. The research results indicate that the multi-agent collaborative mechanism effectively responds to the dynamic changes and uncertainties of the power grid, providing a datadriven intelligent solution for the cross-sectional safety control of the new power system. Its hierarchical control framework and resource collaboration strategy have important reference value for the extended application of multi energy interconnection systems.

The multithread deterministic replay debugging is a foundation problem in concurrent software programing, analysis and testing. Since there are many non-deterministic factors, realizing multithread deterministic replay debugging have been a challenge. This paper proposes a new multithread deterministic replay debugging method Memento. The main contributions include: Firstly, in the replay framework, the record and replay functions are realized independently as a shared library, which can be loaded to GDB dynamically during running. That can greatly reduce the complex of using debugging. Secondly, for system calls, which is a kind of important non-deterministic factor, its record and replay are solved by dynamic code stub using GDB, then corresponding record log is optimized to reduce the size of log file. This method supports almost all system calls. Thirdly, for thread schedule, another kind of important non-deterministic factor, its record and replay are solved by transferring schedule to access order of shared memory, and then corresponding record log is also optimized to reduce the size of log file and improve the debugging performance. Memento has the features including without need of supporting of special hardware and programing language and modifying the OS kernel and codes. The experimental results show that our method not only provides correct record and replay functions, but also perform well on log file size and running time.

In the virtual power plant scheduling system based on cloud edge collaboration, the edge computing effectively alleviates the computing pressure. And the effective computing offloading strategy can provide better analysis services. A virtual power plant analysis task offloading strategy based on delay and task requirements in cloud edge collaboration system is proposed in this paper. Analysis tasks in the edge cluster are realized by Docker containers. The completed information is sent up after the container calculation is completed. The same type of analysis tasks are unloaded to the container that has just completed the calculation according to the transmission delay. For tasks that need to build a new container, it is unloaded to the edge cluster server with the most abundant resources according to the different memory, storage and CPU requirements of the task. The strategy can improve the efficiency of analysis task execution. Its feasibility and advantage are verified by application example.

为深入贯彻落实电力体制改革要求,南方电网公司积极推动南方区域调频辅助服务市场建设.设计了面向区域调频辅助服务市场的南方电网统一调频控制架构,通过技术和管理手段相结合,解决了区域电网多调度机构自动发电控制(automatic generation control,AGC)协调控制、与区域调频市场衔接、电网运行调峰调频解耦、省区间联络线控制等问题.为兼顾电网运行安全与市场公平,对统一调频控制策略及参数进行了优化,以满足大扰动下电网频率恢复的控制要求.运行结果表明,面向区域调频辅助服务市场的统一调频控制模式可有效提高电网的频率质量,加快大扰动下电网频率的恢复速度,能够充分发挥南方电网大平台对调频资源的优化配置作用.

China's resources and load centers are distributed in reverse, and high voltage direct current (HVDC) technology is utilized to optimize the allocation of resources in the maximum range. Multi-terminal HVDC power can be freely controlled, and its characteristics of bidirectional transmission, network losses in the transmission process, power transmission restrictions, and power change restrictions in adjacent periods bring challenges to the spot market clearing model in the new power system environment. So, this paper proposes a HVDC tie-line model considering the characteristics including bidirectional transmission and transmission network loss. On this basis, this paper further proposes a regional spot market clearing model that also considers the transmission price of the tie-line in the spot market. With the goal of minimizing the overall cost of the region, the DC tie-line power, unit on and off state, and unit power are used as decision variables in the proposed model, and integer and nonnegative continuous variables are introduced to optimize and determine the tie-line transmission direction. Then the DC tie-line transmission power schedule, the AC tie-line transmission power schedule, unit start-up and shutdown schedule, and unit power schedule of the regional power grid are obtained on the premise of ensuring the security of the regional power grid. The proposed model finally realizes the synergy of wind, solar, water, and fire resources in the region. The three-region IEEE RTS-96 system is used for calculation, and the results verify the validity of the proposed model.

高比例新能源电网中,功率与频率变化存在很强的非线性,自动发电控制(AGC)作为电网调节频率的主要控制手段,目前的控制方式无法很好地适应强非线性特性电网的调频需求.鉴于此,提出了基于极限学习机(ELM)预测模型的高比例新能源电网改进频率控制策略.其特点在于通过ELM算法和历史运行数据,建立电网功率变化与频率变化的实时频率预测模型,进一步基于预测模型分析AGC调节机组的调频能力,按照调频能力优化AGC的区域功率控制需求功率分配.其优势在于通过机器学习拟合频率非线性调节规律,优化AGC频率控制,提高系统频率调节的快速性和可靠性,从而提高含新能源电网稳定性.最后通过电网SCADA实际数据建立预测模型并验证其准确性和实时性,并通过应用实例证明所提策略可以实现快速稳定调频.

With the increasing penetration of renewable energy, more attention has been paid to the renewable energy clearing method in electricity market. Firstly, this paper deduces the robust counterpart of box robust optimization. Secondly, the box robust optimization model is proposed to deal with the uncertainty of renewable energy. Finally, in the simulation analysis, through the comparison of calculation efficiency and the comparison of different system scale and different penetration rates of renewable energy, it is verified that the proposed risk factor allocation method has a good effect in large-scale systems and has little impact on the calculation efficiency of the model.

当前电网正开展统一调频市场建设,在实际电网中开展自动发电控制(automatic generation control,AGC)功能测试的难度将进一步加大,调度自动化迫切需要有一个可真实模拟实际电网的仿真环境来支撑开展AGC功能测试.实时仿真在技术可行性上具备开展AGC闭环测试的能力,为验证其有效性,本文搭建了基于实时仿真的AGC闭环测试系统,反演了实际现场的AGC动作事件.试验结果表明建立的RTDS实时仿真模型具备仿真大电网系统频率变化的能力,同时RTDS系统具备与AGC应用闭环试验的能力,可应用于电网AGC功能测试工作.

目前针对综合需求响应的市场机制尚未成熟,上级电网无法准确获取用户的调控能力,对响应资源进行充分精准的管理.面向配网峰谷差降低及新能源消纳率提高,提出了一种电热综合需求响应市场出清机制.首先建立了考虑柔性负荷与采暖负荷热惯性的综合能源服务商模型,接着提出了综合能源服务商的调节域以刻画其提供可调资源的能力,然后设计了考虑综合能源服务商调节域的综合需求响应市场出清机制.最后基于近似动态规划算法,提出了综合能源服务商调节域的求解流程.算例分析表明该市场机制可以降低配网负荷峰谷差,并使得新能源消纳率得到提高.

为推进调频辅助服务市场的建设,南方电网在中东部主网和云南电网分别建成基于网省交互的统一调频控制区.云南电网统一调频控制区试运行期间,中标机组采用了"计划出力+带宽限制"的AUTOR模式,当中标机组响应频率的调节速率过慢,调峰速率过快,易导致调峰和调频任务的失配,进而使系统频率在49.95 Hz和50.05 Hz附近悬浮.为此提出了提高系统调频速率、优化分配日前调峰任务等优化策略,有效缓解了系统频率悬浮现象,为调频辅助服务市场的运行奠定基础.

The reform of China electricity marketization, like the international electricity reform, has broken monopolies, built a competitive electricity market, changed the vertically integrated organizational structure of power generation, transmission, distribution and sales, and has given market designers and market supervisors a new mission. At the same time, with the rapid development of information technology, it is of great significance to establish a market dynamic monitoring system on the existing power trading technology support system using modern information network technology. Based on the existing problems of the dynamic monitoring system in the current power market environment, this article designed a power market dynamic monitoring system architecture based on computer software development technology, big technology. The corresponding functional components and data mining, modeling analysis, and intelligent control supporting technologies were deployed. Finally, the application prospect of the system in typical scenarios of the power market was described. It is expected to provide technical and theoretical support for the cultivation and establishment of a healthy and orderly power market ecosystem.

自动发电控制(automatic generation control,AGC)是电网调度自动化系统中的一项重要功能,在统一调频控制和电力现货市场环境下AGC控制功能重要性愈加凸显.本文研究提出了基于实时仿真的网级和省级AGC联合闭环测试方案,进而研究开发了实时仿真与网省AGC异构系统的通信交互接口和实时闭环测试技术,提出了多GTNET卡并联同步以及"单端接口、数据转发"等数据交互接口方法,解决了网省AGC仿真联合实时闭环测试中的海量实时数据交互问题.最后通过接口测试验证了多GTNET卡并联同步高速通信能力及其接口延时特性.

在南方区域现货市场两级运作阶段,仅有部分省份开展现货市场交易,因此需要综合考虑不同省份市场化改革所处的不同阶段,在充分考虑全网新能源发电消纳、负荷预测、水电计划和检修计划的基础上,采用多目标发电控制方法对新能源发电的最大化消纳.在进行搜索多代理均衡解时间内,智能体个数的增长往往会导致DCEQ(入)算法呈几何数增加,其结果将会导致在电网里面的应用范围十分局限,同时也限制了控制性能的发展和提升.为此,为了解决该问题,采用了一种基于WoLF-PHC和资格迹构成的狼爬山算法.其算例结果表明了该算法的有效性,较其他多目标发电控制方法而言,具有快速收敛速率的狼爬山算法更能适应环境的变化.

电能量市场与备用辅助服务市场联合优化能够在满足系统安全运行约束的前提下实现经济效益最优,是国外典型电力现货市场运行的最佳实践.在我国现行的"分层分区"电力调度管理体制下,电网安全与备用容量管理在区域、省级调度中心之间各有权责,存在全系统及各分区的备用需求容量配置、分区间备用的有效调度等问题,不能直接照搬国外的电能量与备用联合出清机制与模型.分析了我国区域互联电网条件下的电能量与备用辅助服务联合优化市场模式,提出了计及多因素的系统及各分区的备用需求容量设置方法,分别构建了单区域、多区域互联两种情形下的联合优化出清数学模型.最后,基于三区域IEEE-RTS96系统进行算例分析,验证了所提分区备用需求容量配置方法及区域电网联合优化出清数学模型可充分考虑区域间运行约束对预留备用容量有效性的影响,从而极大提升系统运行的可靠性.

随着我国西电东送工程和高压直流输电技术的快速发展,高压直流输电线路在运行中对油气管道产生的影响日渐显著.研究表明,高压直流输电线路产生的接地极入地电流会在管道上产生干扰电位,对管道运行产生风险.为保证油气管道的安全运行,开展了高压直流线路单极闭锁后入地电流的多对象协调控制相关研究,以全网发电机组出力变化最小为优化目标,加入快速降低入地极电流为约束条件,将高压直流单极闭锁入地电流协调控制建模为非线性规划模型,采用不可行内点法对模型进行求解.并通过南方电网高压直流线路进行在线开环算例验证,通过算例验证了该方法的正确性和有效性.

对南方区域电力现货市场技术支持系统方案进行了研究。基于一体化、模块化、智能化、安全性、开放性、适应性原则,提出了包含中长期计划分解、日前市场、实时市场、辅助服务市场等电力现货市场主要业务的系统架构及功能。针对系统建设的关键问题,本文提出采用云平台架构实现系统的灵活弹性和高可用;建设现货市场运营数据中心实现数据全景建模和数据统一交互;构建完整边界防护、纵深防御体系,保障现货系统对外安全防护;研究统一系统架构、模块化建设方案,实现区域起步、区域两级、区域一级建设路径下系统平滑过渡。

The sub-synchronous torsional interaction (SSTI) behaviour between turbine-generator (TG) units and the HVDC systems has been a vital research content in power systems. In the past, the characteristics of SSTI with line-commutated-converter (LCC)-HVDC have already been well studied. With the development of voltage source converter (VSC)-based HVDC, some researchers have found that different control parameters of the converter have an impact on the electrical damping of the connected synchronous generator (SG) system. However, the existing results did not show the specific mechanism of SSTI. In this paper, SSTI with VSC-HVDC affected by voltage feed-forward (VVF) compensations in the AC current controllers is investigated. The effects of the asymmetric VVF in both d- and q-axes on electrical damping are paid special attention to. First, a general output impedance model of VSC-HVDC concentrated on AC current controllers is developed. Then the influence of compensation factor (k(d/q)) and filter bandwidth (a(fd/q)) on the bandwidth of converter and the electrical damping is studied. At last, a SSTI suppression scheme based on the former analysis is given. Simulation studies are carried out in PSCAD/EMTDC, and simulated results validate the analytical results and proposed suppression scheme.