技术领域Technical field
本发明属于电力系统安全稳定分析领域，更具体地，涉及一种统一调度计算资源的多时间尺度安全校核系统及方法。The invention belongs to the field of power system safety and stability analysis, and more specifically, relates to a multi-time scale safety check system and method for unified dispatching of computing resources.
背景技术Background technique
随着电力市场改革，跨区特高压交直流互联电网的发展，电网一体化特性日益凸显的同时，电力交易、新能源发电预测和负荷预测的不确定性给电网的安全稳定运行带来巨大的挑战。在现有在线安全稳定校核和辅助决策深化完善的同时，电力系统的风险防控向月度、日前、日内、实时等多时间尺度拓展，推进调度计划向精益化、智能化发展，提升跨区电网安全稳定风险的超前识别和防控能力已成为电网运行的迫切需求。With the reform of the power market and the development of cross-regional ultra-high voltage AC and DC interconnected power grids, the integration of power grids has become increasingly prominent. At the same time, the uncertainty of power transactions, new energy generation forecasts and load forecasts has brought huge challenges to the safe and stable operation of the power grid. challenge. While the existing online safety and stability verification and auxiliary decision-making are deepened and improved, the risk prevention and control of the power system is expanded to multiple time scales such as monthly, day-ahead, intraday, and real-time, promoting the lean and intelligent development of dispatch plans, and improving cross-regional The ability to advance the identification and prevention of power grid security and stability risks has become an urgent need for power grid operation.
我国省级以上电网目前大多已实现日前、日内及实时调度计划的安全稳定校核功能，但计算耗时较长，无法满足当前电网调度运行业务的需求。Most power grids at or above the provincial level in my country have now implemented safe and stable verification functions for day-ahead, intraday and real-time dispatch plans, but the calculations take a long time and cannot meet the needs of current power grid dispatch operations.
发明内容Contents of the invention
本发明提出一种统一调度计算资源的多时间尺度安全校核系统及方法，以解决目前电网调度计划安全校核应用中计算耗时较长的问题。The present invention proposes a multi-time scale safety check system and method for unified dispatching of computing resources to solve the problem of long calculation time in current power grid dispatch plan safety check applications.
第一方面，本发明提出一种统一调度计算资源的多时间尺度安全校核方法，包括以下步骤：In the first aspect, the present invention proposes a multi-time scale security check method for unified scheduling of computing resources, which includes the following steps:
根据获取的至少一项安全校核应用的计算内容，将各项安全校核应用分别划分为多个计算子任务，其中，每一项所述安全校核应用分别具有对应的时间尺度；According to the obtained calculation content of at least one security check application, each security check application is divided into multiple calculation subtasks, wherein each of the security check applications has a corresponding time scale;
将所述至少一项安全校核应用划分后的各计算子任务分配至n个计算节点，其中，所述n个计算节点为多时间尺度安全校核用的统一计算资源；Allocate each computing sub-task divided by the at least one security check application to n computing nodes, where the n computing nodes are unified computing resources for multi-time scale security verification;
根据预先设定的多应用调度策略，控制所述n个计算节点并行计算所述至少一项安全校核应用，以得到与所述至少一项安全校核应用的各计算子任务分别对应的计算结果；According to the preset multi-application scheduling policy, the n computing nodes are controlled to calculate the at least one security check application in parallel to obtain calculations corresponding to each calculation sub-task of the at least one security check application. result;
针对每一项安全校核应用，将由所述n个计算节点并行完成的计算结果合并，得到所述至少一项安全校核应用的安全校核结果。For each security check application, the calculation results completed in parallel by the n computing nodes are combined to obtain the security check result of the at least one security check application.
第二方面，本发明提出一种统一调度计算资源的多时间尺度安全校核系统，包括：In the second aspect, the present invention proposes a multi-time scale security verification system that uniformly schedules computing resources, including:
并行计算调度服务器；Parallel computing scheduling server;
与所述并行计算调度服务器通信连接的多个并行计算节点；A plurality of parallel computing nodes communicatively connected to the parallel computing scheduling server;
所述并行计算调度服务器用于执行权利要求1至8中任一项所述的多时间尺度安全校核方法；The parallel computing scheduling server is used to execute the multi-time scale security verification method according to any one of claims 1 to 8;
各所述并行计算节点，用于执行权利要求1至8中任一项所述的多时间尺度安全校核方法。Each of the parallel computing nodes is used to execute the multi-time scale security verification method described in any one of claims 1 to 8.
本发明提出的统一调度计算资源的多时间尺度安全校核系统及方法，统一地调度多时间尺度安全校核用计算资源；首先根据不同时间尺度安全校核应用的计算周期、计算内容和并行计算资源需求，提出计算资源统一管理时并行计算框架，生成了各类安全校核应用的并行计算任务分解方法，既提高了计算资源的利用效率，又缩短了各类安全校核应用的计算耗时，并进一步提高了并行计算响应时间满意度。The multi-time scale security check system and method proposed by the present invention uniformly dispatch computing resources for multi-time scale security check; firstly, according to the calculation cycle, calculation content and parallel computing of different time scale security check applications Resource requirements, a parallel computing framework for unified management of computing resources is proposed, and a parallel computing task decomposition method for various security verification applications is generated, which not only improves the utilization efficiency of computing resources, but also shortens the calculation time of various security verification applications. , and further improves parallel computing response time satisfaction.
附图说明Description of drawings
通过参考下面的附图，可以更为完整地理解本发明的示例性实施方式：A more complete understanding of exemplary embodiments of the invention may be obtained by reference to the following drawings:
图1为本发明一个实施例的统一调度计算资源的多时间尺度安全校核方法的流程示意图；Figure 1 is a schematic flowchart of a multi-time scale security check method for unified scheduling of computing resources according to an embodiment of the present invention;
图2为本发明一个实施例的统一调度计算资源的多时间尺度安全校核系统的组成示意图；Figure 2 is a schematic diagram of the composition of a multi-time scale security check system for unified scheduling of computing resources according to an embodiment of the present invention;
图3为本发明一个实施例中的一项安全校核计算的子任务分解示意图；Figure 3 is a schematic diagram of sub-task decomposition of a security check calculation in an embodiment of the present invention;
图4为应用本发明实施例的统一调度计算资源的安全校核方法与独立地调度各时间尺度的计算资源时多个时间尺度计划安全校核的计算耗时对比图。Figure 4 is a comparison diagram of the calculation time consumption of multiple time scale planned security checks when applying the security check method of unified scheduling of computing resources according to the embodiment of the present invention and independently scheduling computing resources at each time scale.
具体实施方式Detailed ways
现在参考附图介绍本发明的示例性实施方式，然而，本发明可以用许多不同的形式来实施，并且不局限于此处描述的实施例，提供这些实施例是为了详尽地且完全地公开本发明，并且向所属技术领域的技术人员充分传达本发明的范围。对于表示在附图中的示例性实施方式中的术语并不是对本发明的限定。在附图中，相同的单元/元件使用相同的附图标记。Exemplary embodiments of the present invention will now be described with reference to the accompanying drawings. However, the present invention may be embodied in many different forms and is not limited to the embodiments described herein. These embodiments are provided so that this disclosure will be thorough and complete. invention, and fully convey the scope of the invention to those skilled in the art. The terminology used in the exemplary embodiments represented in the drawings does not limit the invention. In the drawings, identical units/elements use the same reference numerals.
除非另有说明，此处使用的术语(包括科技术语)对所属技术领域的技术人员具有通常的理解含义。另外，可以理解的是，以通常使用的词典限定的术语，应当被理解为与其相关领域的语境具有一致的含义，而不应该被理解为理想化的或过于正式的意义。Unless otherwise defined, the terms (including scientific and technical terms) used herein have the commonly understood meaning to one of ordinary skill in the art. In addition, it is understood that terms defined in commonly used dictionaries should be understood to have consistent meanings in the context of their relevant fields and should not be understood as having an idealized or overly formal meaning.
目前，我国省级以上电网目前大多已实现日前、日内及实时调度计划的安全稳定校核功能。但因为针对调度计划多套计划方式和大规模预想故障集进行安全稳定评估和辅助决策的计算量极大，因此，计算耗时较长。At present, most power grids at or above the provincial level in my country have implemented the safety and stability verification functions of day-ahead, intraday and real-time dispatch plans. However, due to the huge amount of calculations required to conduct safety and stability assessment and auxiliary decision-making based on multiple planning methods and large-scale expected failure sets in the dispatch plan, the calculations take a long time.
近年来，并行计算机群管理技术有了飞跃式发展，可供利用计算资源的CPU核数由早年的单机四核发展至单机二十四核甚至更多，计算能力显著增强。随着智能电网调度控制支持系统的建设，分布式并行计算平台作为基本功能组件，在电力系统分析领域中起着关键、基本的平台支撑作用。In recent years, parallel computer cluster management technology has developed by leaps and bounds. The number of CPU cores available for utilizing computing resources has increased from four cores in a single machine in the early years to twenty-four cores in a single machine or even more, and computing capabilities have been significantly enhanced. With the construction of smart grid dispatch control support system, distributed parallel computing platform, as a basic functional component, plays a key and basic platform support role in the field of power system analysis.
本发明提出统一调度计算资源的多时间尺度安全校核方法，集中各时间尺度安全校核用计算资源，统一地对多个时间尺度的安全校核应用进行并行调度。解决了单个安全校核应用的并行调度方法；针对多个时间尺度安全校核应用的并行计算任务，提出了5种并行计算调度策略。该方法针对不同时间尺度的安全校核应用的响应时间需求分别采取针对性的算法，进一步提高并行计算的响应时间满意度。The present invention proposes a multi-time scale security check method that uniformly schedules computing resources, concentrates computing resources for security check at each time scale, and uniformly schedules security check applications at multiple time scales in parallel. The parallel scheduling method of a single security check application is solved; for the parallel computing tasks of multiple time scale security check applications, 5 parallel computing scheduling strategies are proposed. This method adopts targeted algorithms according to the response time requirements of security check applications at different time scales to further improve the response time satisfaction of parallel computing.
具体地，本发明提供的方法首先根据不同时间尺度安全校核应用的计算周期、计算内容和并行计算资源需求，给出多时间尺度安全校核应用的并行计算任务划分方案；然后，综合考虑不同安全校核应用的计算优先级和响应时间需求，生成基于计算资源统一管理的多时间尺度计划安全校核并行计算调度方法。Specifically, the method provided by the present invention first provides a parallel computing task division scheme for multi-time scale security verification applications based on the computing cycles, calculation content and parallel computing resource requirements of security verification applications at different time scales; then, comprehensively considers different The computing priority and response time requirements of the security check application are calculated, and a multi-timescale planned security check parallel computing scheduling method based on unified management of computing resources is generated.
如图1所示，本发明一个实施例的基于统一计算资源的多计划安全校核并行调度方法，包括：As shown in Figure 1, a multi-plan security check parallel scheduling method based on unified computing resources according to one embodiment of the present invention includes:
根据获取的至少一项安全校核应用的计算内容，将各项安全校核应用分别划分为多个计算子任务，其中，每一项所述安全校核应用分别具有对应的时间尺度；According to the obtained calculation content of at least one security check application, each security check application is divided into multiple calculation subtasks, wherein each of the security check applications has a corresponding time scale;
将所述至少一项安全校核应用划分后的各计算子任务分配至n个计算节点，其中，所述n个计算节点为多时间尺度安全校核用的统一计算资源；Allocate each computing sub-task divided by the at least one security check application to n computing nodes, where the n computing nodes are unified computing resources for multi-time scale security verification;
根据预先设定的多应用调度策略，控制所述n个计算节点并行计算所述至少一项安全校核应用，以得到与所述至少一项安全校核应用的各计算子任务分别对应的计算结果；According to the preset multi-application scheduling policy, the n computing nodes are controlled to calculate the at least one security check application in parallel to obtain calculations corresponding to each calculation sub-task of the at least one security check application. result;
针对每一项安全校核应用，将由所述n个计算节点并行完成的计算结果合并，得到所述至少一项安全校核应用的安全校核结果。For each security check application, the calculation results completed in parallel by the n computing nodes are combined to obtain the security check result of the at least one security check application.
进一步地，所述的方法，Further, the method,
不同时间尺度的安全校核应用具有对应的响应时间需求；Security verification applications at different time scales have corresponding response time requirements;
相应地，所述根据预先设定的多应用调度策略，控制所述n个计算节点并行计算所述至少一项安全校核应用，包括：Correspondingly, controlling the n computing nodes to calculate the at least one security check application in parallel according to the preset multi-application scheduling policy includes:
根据各项安全校核应用分别对应的时间尺度，确定各项安全校核应用分别对应的响应时间需求；According to the corresponding time scale of each security check application, determine the corresponding response time requirements of each security check application;
根据所述各项安全校核应用分别对应的响应时间需求，控制所述n个计算节点按照响应时间需求的高低顺序依次并行计算所述各项安全校核应用。According to the response time requirements corresponding to each of the security check applications, the n computing nodes are controlled to calculate the various security check applications in parallel in order of response time requirements.
应该理解为，具体实施时，还可以根据生成校核任务时，根据当前的校核需要，临时地人为设定的优先级参数；并根据所述各项安全校核应用分别对应的优先级，控制所述n个计算节点按照指定的优先级的先后顺序依次并行计算所述各项安全校核应用。It should be understood that during specific implementation, the priority parameters can also be temporarily set based on the current verification needs when generating the verification task; and based on the corresponding priorities of each of the security verification applications, Control the n computing nodes to calculate the various security verification applications in parallel in sequence according to the specified priority.
进一步地，所述的方法，Further, the method,
不同时间尺度的安全校核应用具有对应的响应时间需求；Security verification applications at different time scales have corresponding response time requirements;
相应地，所述根据预先设定的多应用调度策略，控制所述n个计算节点并行计算所述至少一项安全校核应用，包括：Correspondingly, controlling the n computing nodes to calculate the at least one security check application in parallel according to the preset multi-application scheduling policy includes:
根据各项安全校核应用分别对应的时间尺度，确定所述至少一项安全校核应用分别对应的响应时间需求及分别对应的抽签标识，其中，响应时间需求高的安全校核应用对应的抽签标识的权重大；According to the time scale corresponding to each security check application, the response time requirement and the corresponding lottery identification of the at least one security check application are determined, wherein the lottery corresponding to the security check application with high response time requirement is The weight of the logo is significant;
对所述至少一项安全校核应用划分后的各计算子任务分别设置与该项安全校核应用对应的抽签标识；Each computing sub-task divided by the at least one security check application is respectively set with a lottery mark corresponding to the security check application;
将所述至少一项各项安全校核应用划分后的各计算子任务组成共用计算任务队列；Each computing sub-task divided into at least one security check application is formed into a shared computing task queue;
以所述抽签标识的权重的大小为概率，控制所述n个计算节点轮转式地从所述共用计算任务队列中选择计算任务，以并行计算所述至少一项安全校核应用。Using the weight of the lottery identifier as a probability, the n computing nodes are controlled to select computing tasks from the shared computing task queue in a rotating manner to calculate the at least one security check application in parallel.
进一步地，所述的方法，Further, the method,
所述将所述至少一项安全校核应用划分后的各计算子任务分配至n个计算节点，包括：Allocating each computing sub-task divided into at least one security check application to n computing nodes includes:
根据划分后的各计算子任务对应的计算资源需求，确定所述各计算子任务对应的任务类型，其中，每类任务类型具有对应的权重；According to the computing resource requirements corresponding to each divided computing sub-task, determine the task type corresponding to each computing sub-task, wherein each type of task type has a corresponding weight;
获取各计算节点的负载当量：Get the load equivalent of each computing node:
负载当量lwi为各计算节点分配的各类任务的数量与任务类型的权重βj的加权之和：The load equivalent lwi is the weighted sum of the number of various tasks assigned to each computing node and the weight βj of the task type:
其中，1≤j≤m，Yji为第i个计算节点上所分配的第j类计算任务的数量，1≤i≤n；Among them, 1≤j≤m, Yji is the number of jth type computing tasks allocated on the i-th computing node, 1≤i≤n;
βj为第j类任务的权重；βj is the weight of the jth type of task;
记全部的计算节点的负载当量所形成的向量LW为：The vector LW formed by the load equivalents of all computing nodes is:
LW＝{lw1,lw2,lw3,...,lwi...,lwn}LW＝{lw1 ,lw2 ,lw3 ,...,lwi ...,lwn }
记最重载的计算节点的负载当量为lwmax＝max{LW}；The load equivalent of the most heavily loaded computing node is lwmax =max{LW};
记最轻载的计算节点的负载当量为lwmin＝max{LW}；The load equivalent of the least loaded computing node is lwmin =max{LW};
定义负载当量差为lwdiff＝lwmax-lwmin；Define the load equivalent difference as lwdiff =lwmax -lwmin ;
定义平均负载为Define the average load as
定义负载当量差阈值lwthresh＝α*lwaverage，α为阈值系数，0＜α≤1；Define the load equivalent difference threshold lwthresh =α*lwaverage , α is the threshold coefficient, 0＜α≤1;
当负载当量差lwdiff不小于阈值lwthresh时，按照从最轻载到最重载的顺序，依次将所述至少一项安全校核应用划分后的各计算子任务分配至n个计算节点。When the load equivalent difference lwdiff is not less than the threshold lwthresh , each computing subtask divided by the at least one safety check application is sequentially allocated to n computing nodes in order from the lightest load to the heaviest load.
进一步地，所述的方法，Further, the method,
所述根据获取的至少一项安全校核应用的计算内容，将各项安全校核应用分别划分为多个计算子任务，包括：According to the obtained calculation content of at least one security check application, each security check application is divided into multiple calculation subtasks, including:
根据每一项安全校核应用的计算内容，将所述安全校核应用划分为多个依次执行的步骤式的计算子任务，及According to the calculation content of each safety check application, the safety check application is divided into a plurality of step-like calculation subtasks that are executed sequentially, and
将步骤式的子任务划分为多个任务性的计算子任务；Divide step-based subtasks into multiple task-based calculation subtasks;
为各任务性的计算子任务设置与其对应的步骤式的子任务对应的步骤标识，其中，需要先执行的步骤式的子任务具有的步骤标识的值大。For each task-based calculation subtask, a step identifier corresponding to its corresponding step-like subtask is set, wherein the step-like subtask that needs to be executed first has a step identifier with a larger value.
进一步地，所述的方法，Further, the method,
每项安全校核应用为以下任一：Each safety check application is one of the following:
月度安全校核、日前安全校核、日内安全校核、实时安全校核；Monthly safety check, day-ahead safety check, intra-day safety check, real-time safety check;
所述月度安全校核的计算频率为每个自然月一次；The calculation frequency of the monthly safety check is once every calendar month;
所述日前安全校核的计算频率为每日至少两次；The calculation frequency of the day-ahead safety check is at least twice a day;
所述日内安全校核的计算频率为每日至少两次；The calculation frequency of the intra-day safety check is at least twice a day;
所述实时安全校核的计算频率为每小时不小于4次。The calculation frequency of the real-time security check is no less than 4 times per hour.
所述月度安全校核、日前安全校核、日内安全校核、实时安全校核的响应时间需求依次升高。The response time requirements of the monthly security check, day-ahead security check, intraday security check, and real-time security check increase in sequence.
进一步地，所述的方法，Further, the method,
所述多应用调度策略为以下任一：The multi-application scheduling strategy is any of the following:
顺序调度策略、共享式顺序策略、整体优先调度策略、轮转式调度策略、带优先级的轮转式调度策略。Sequential scheduling strategy, shared sequential strategy, overall priority scheduling strategy, round-robin scheduling strategy, round-robin scheduling strategy with priority.
进一步地，所述的方法，Further, the method,
在任一项安全校核应用中，控制校核断面潮流计算、安全稳定校核、辅助决策或稳定裕度评估依次执行；In any safety check application, control check section power flow calculation, safety and stability check, auxiliary decision-making or stability margin assessment are executed in sequence;
所述校核断面潮流计算、安全稳定校核、辅助决策或稳定裕度评估为依次执行的步骤式的子任务；The check section power flow calculation, safety and stability check, auxiliary decision-making or stability margin assessment are step-by-step subtasks executed in sequence;
所述辅助决策和所述稳定裕度评估为同一个步骤式的子任务下的两个功能性的子任务。The auxiliary decision-making and the stability margin evaluation are two functional subtasks under the same step-like subtask.
如图2所示，本发明一个实施例的基于统一计算资源的多计划安全校核并行调度系统，包括：As shown in Figure 2, a multi-plan safety check parallel scheduling system based on unified computing resources according to one embodiment of the present invention includes:
并行计算调度服务器10；Parallel computing scheduling server 10;
与所述并行计算调度服务器通信连接的多个并行计算节点20；A plurality of parallel computing nodes 20 communicatively connected to the parallel computing scheduling server;
所述并行计算调度服务器10用于执行前述的多时间尺度安全校核方法；The parallel computing scheduling server 10 is used to execute the aforementioned multi-time scale security verification method;
各所述并行计算节点20，用于执行前述的多时间尺度安全校核方法。Each of the parallel computing nodes 20 is used to execute the aforementioned multi-time scale security verification method.
进一步地，所述的校核系统，Further, the calibration system,
所述并行计算调度服务器具有至少一个CPU核；The parallel computing scheduling server has at least one CPU core;
各所述并行计算节点具有多个CPU核。Each parallel computing node has multiple CPU cores.
综上，本发明实施例的基于统一计算资源的多时间尺度计划安全校核并行调度方法和系统，统一管理不同时间尺度的计划安全校核用计算资源；在兼顾不同安全校核应用的响应时间需求的前提下，针对不同计算节点之间计算能力的差异配置并行计算任务，最大化地利用了计算资源的并行计算能力，缩短了电网调度计划安全校核应用的计算耗时，提高了电网调度计划安全校核应用的响应时间满意度。In summary, the multi-timescale plan security check parallel scheduling method and system based on unified computing resources according to the embodiment of the present invention uniformly manages the computing resources for plan security check at different time scales; while taking into account the response time of different security check applications Under the premise of demand, parallel computing tasks are configured according to the differences in computing capabilities between different computing nodes, which maximizes the use of parallel computing capabilities of computing resources, shortens the computing time of power grid dispatching plan safety verification applications, and improves power grid dispatching. Response time satisfaction of planned security check applications.
本发明另一个实施例的多时间尺度计划安全校核并行计算调度方法，包括以下步骤：A multi-time scale plan security check parallel computing scheduling method according to another embodiment of the present invention includes the following steps:
1)对多个时间尺度安全校核应用的并行计算任务进行划分，确定任务粒度划分方案；1) Divide the parallel computing tasks of multiple time scale security verification applications and determine the task granularity division plan;
2)实现单个时间尺度的安全校核应用的并行计算调度；2) Implement parallel computing scheduling for security verification applications at a single time scale;
3)实现多个时间尺度的安全校核应用的并行计算调度。3) Realize parallel computing scheduling of security check applications at multiple time scales.
其中，所述步骤1)中，对多时间尺度安全校核应用的并行计算任务进行划分，包括：Among them, in step 1), the parallel computing tasks of multi-time scale security verification applications are divided, including:
获取按照时间顺序到达的多项安全校核应用，确定对应的多项安全校核任务，这些安全校核任务可能具有相同或不同的时间尺度；Obtain multiple security check applications arriving in chronological order and determine the corresponding multiple security check tasks. These security check tasks may have the same or different time scales;
确定每一项安全校核任务包括的至少一个调度计划，确定每一个调度计划对应的校核断面，及每一校核断面对应的故障模式集合；Determine at least one scheduling plan included in each safety check task, determine the calibration section corresponding to each scheduling plan, and the set of failure modes corresponding to each calibration section;
依次按照断面潮流校核、断面故障校核、辅助决策和稳定裕度评估的计算内容，划分每一项安全校核任务的子任务集合，并确定各子任务的计算资源需求。According to the calculation content of section power flow check, section fault check, auxiliary decision-making and stability margin assessment, the sub-task set of each safety check task is divided, and the computing resource requirements of each sub-task are determined.
应该理解为，不同时间尺度的安全校核应用根据其计算周期，按照时间顺序依次到达并行计算调度服务器；一个安全校核应用对应一个安全校核任务；每一项安全校核任务的子任务集合构成了该安全校核应用的计算内容。It should be understood that security check applications at different time scales arrive at the parallel computing scheduling server in chronological order according to their calculation cycles; a security check application corresponds to a security check task; a set of subtasks for each security check task Constitutes the calculation content of the safety check application.
应该理解为，各子任务在多个计算节点上并行计算后，还包括将来自多个计算节点的计算结果合并的步骤。It should be understood that after each subtask is calculated in parallel on multiple computing nodes, it also includes the step of merging the calculation results from the multiple computing nodes.
应该理解为，安全校核应用对检修计划、发电计划等调度计划形成校核断面对应的潮流，然后进行全面的安全稳定校核，包括静态安全、暂态稳定、动态稳定和电压稳定等方面，并在校核完成后进行辅助决策和稳定裕度评估计算。It should be understood that the safety check application forms a trend corresponding to the check section of the maintenance plan, power generation plan and other dispatch plans, and then conducts a comprehensive safety and stability check, including static safety, transient stability, dynamic stability and voltage stability, etc. And after the verification is completed, auxiliary decision-making and stability margin evaluation calculations are performed.
具体地，对于每一项安全校核任务或应用，在进行任务调度与分派时，依次包括以下步骤：Specifically, for each security check task or application, the following steps are included in the task scheduling and assignment:
步骤(1)、任务分解。如，将计算量为S的计算任务J分解为N个子任务，其中，各子任务的计算量分别为S1，…，SN；Step (1), task decomposition. For example, the computing task J with a calculation amount of S is decomposed into N subtasks, where the calculation amount of each subtask is S1 ,..., SN respectively;
步骤(2)、并行计算。如，分派由P个计算节点并行地完成前述的N个子任务的计算，得到该N个子任务对应的计算结果：R1，…，RN；Step (2), parallel calculation. For example, P computing nodes are assigned to complete the calculation of the aforementioned N subtasks in parallel, and the calculation results corresponding to the N subtasks are obtained: R1 ,..., RN ;
步骤(3)、结果合并。如，基于前述N个子任务对应的计算结果：R1，…，RN，得到该计算量为S的计算任务J对应的安全校核计算内容的最终结果R。Step (3), merge the results. For example, based on the calculation results corresponding to the aforementioned N subtasks: R1 ,..., RN , the final result R of the security check calculation content corresponding to the calculation task J with the calculation amount S is obtained.
具体实施时，对一项安全校核任务来说，步骤(1)中的任务分解可以分为两种情况：In specific implementation, for a safety check task, the task decomposition in step (1) can be divided into two situations:
(1.1)按计算流程，将一项安全校核任务分为若干步骤式的子任务。这些步骤式的子任务之间有依赖关系，且计算次序有确定的先后顺序，因此，各子任务分别具有明确的执行优先级。图3中所示的某项安全校核任务按照执行的先后顺序，可以依次分为三个步骤：步骤一为校核断面潮流计算，步骤二为安全稳定故障校核，步骤三为辅助决策和稳定裕度评估。步骤二依赖于步骤一的计算结果；步骤三依赖于步骤二的计算结果。(1.1) According to the calculation process, a safety check task is divided into several step-like subtasks. There are dependencies between these step-like subtasks, and the calculation order has a definite sequence. Therefore, each subtask has a clear execution priority. A certain safety check task shown in Figure 3 can be divided into three steps according to the order of execution: step one is the power flow calculation of the check section, step two is safety and stability fault check, and step three is auxiliary decision-making and Stability margin assessment. Step two depends on the calculation result of step one; step three depends on the calculation result of step two.
这些依次执行的步骤之间既有依赖关系，也有数据交互。因此，图3中所示的该项安全校核任务可以划分为执行优先级依次递减的校核断面潮流计算子任务、安全稳定故障校核子任务、辅助决策子任务和稳定裕度评估子任务。There are dependencies and data interactions between these sequentially executed steps. Therefore, the safety check task shown in Figure 3 can be divided into check section power flow calculation sub-tasks, safety and stability fault check sub-tasks, auxiliary decision-making sub-tasks and stability margin assessment sub-tasks with decreasing execution priorities.
(1.2)将一项步骤式的子任务进一步分解为多个功能性的子任务。这些功能性子任务相互之间既没有依赖关系，也没有数据交互。如，对于某一项日前计划安全校核，其校核断面潮流计算子任务可以分解为96个时段的校核断面潮流计算，这96个时段的潮流计算各自相互独立，彼此之间没有相关性和依赖关系。(1.2) Further decompose a step-like subtask into multiple functional subtasks. These functional subtasks have neither dependencies nor data interactions with each other. For example, for a certain day-ahead plan safety check, the check section power flow calculation subtask can be decomposed into 96 time periods of check section power flow calculations. The power flow calculations of these 96 time periods are independent of each other and have no correlation with each other. and dependencies.
类似地，安全稳定故障校核子任务中的各个故障模式分析任务之间也是相互独立的；辅助决策各个计算和稳定裕度评估的各个计算之间也是相互独立的。Similarly, each failure mode analysis task in the safety and stability fault check subtask is also independent of each other; each calculation of auxiliary decision-making and each calculation of stability margin assessment are also independent of each other.
其中，所述步骤2)的实现单个安全校核应用的并行计算调度，包括：Among them, the implementation of parallel computing scheduling of a single security check application in step 2) includes:
采用与计算资源相关的任务调度算法对各子任务进行并行计算调度。与计算资源相关的信息包括工作机的可靠性、工作机的硬件资源信息及其他对并行计算调度有用的信息。A task scheduling algorithm related to computing resources is used to perform parallel computing and scheduling on each sub-task. Information related to computing resources includes the reliability of the work machine, the hardware resource information of the work machine and other information useful for parallel computing scheduling.
具体地，并行计算调度服务器使用这些与计算资源相关的信息来确定在各工作机上进行并行计算时的调度方法。Specifically, the parallel computing scheduling server uses the information related to the computing resources to determine the scheduling method when performing parallel computing on each worker machine.
应该理解为，工作机或计算服务器均可以作为计算节点。并行计算调度服务器则负责对全部的计算节点或工作机进行调度。It should be understood that either a work machine or a computing server can serve as a computing node. The parallel computing scheduling server is responsible for scheduling all computing nodes or work machines.
优选地，针对单个时间尺度的安全校核应用或任务，采取平均分配法完成并行计算调度。也即，考虑各计算节点的负载当量(又称负载率)，将并行计算任务尽量平均地分配到各计算节点。Preferably, for a single time scale security check application or task, an average allocation method is adopted to complete parallel computing scheduling. That is, considering the load equivalent (also called load rate) of each computing node, the parallel computing tasks are distributed as evenly as possible to each computing node.
优选地，定义负载当量为各计算节点所分配的各类计算任务的数量与任务类型的权重βj的加权之和：Preferably, the load equivalent is defined as the weighted sum of the number of various types of computing tasks allocated to each computing node and the weight βj of the task type:
其中，1≤j≤m，Yji为第i个计算节点上所分配的第j类任务的数量；βj为第j类任务的权重，其数值取决与任务的类型。Among them, 1≤j≤m, Yji is the number of j-th type tasks assigned to the i-th computing node; βj is the weight of the j-th type task, and its value depends on the type of the task.
应该理解为，不同任务类型需要的计算耗时是不同的；各节点的负载当量既和任务类型有关，也和各类任务的数量相关。It should be understood that different task types require different computing time; the load equivalent of each node is related to both the task type and the number of various tasks.
记全部的计算节点的负载当量所形成的向量LW为：The vector LW formed by the load equivalents of all computing nodes is:
LW＝{lw1,lw2,lw3,...,lwi...,lwn} (1)LW＝{lw1 ,lw2 ,lw3 ,...,lwi ...,lwn } (1)
记最重载的计算节点的负载当量为lwmax＝max{LW}，最轻载的计算节点的负载当量为lwmin＝max{LW}；定义负载当量差为lwdiff＝lwmax-lwmin；定义平均负载为设定负载当量差阈值lwthresh＝α*lwaverage；优选地，α为20％。The load equivalent of the most heavily loaded computing node is lwmax =max{LW}, and the load equivalent of the least loaded computing node is lwmin =max{LW}; define the load equivalent difference as lwdiff =lwmax -lwmin ;Define the average load as Set the load equivalent difference threshold lwthresh =α*lwaverage ; preferably, α is 20%.
具体地，在并行计算调度时，统计各个计算节点的负载当量；Specifically, during parallel computing scheduling, the load equivalent of each computing node is counted;
当负载当量差lwdiff小于阈值lwthresh时，按节点顺序分配计算任务；当负载当量差lwdiff不小于阈值lwthresh时，按照负载当量自大而小或自小而大将计算节点排序，并按照从最轻载到最重载的顺序依次分配任务类型及任务数量，将更多的计算任务分配给负载当量最小的计算节点。When the load equivalent difference lwdiff is less than the threshold lwthresh , the computing tasks are allocated in node order; when the load equivalent difference lwdiff is not less than the threshold lwthresh , the computing nodes are sorted according to the load equivalent from large to small or from small to large, and according to Task types and task quantities are assigned in order from the lightest load to the heaviest load, and more computing tasks are assigned to the computing nodes with the smallest load equivalent.
具体地，各计算节点的计算能力可以相同或不同；这些计算节点的节点顺序是预先设定的，如，基于坐落的空间位置或在通信网络中的拓扑关系依次编号。Specifically, the computing capabilities of each computing node may be the same or different; the node order of these computing nodes is preset, for example, numbered sequentially based on their spatial location or topological relationship in the communication network.
以上通过引入负载当量，在分配各安全应用校核的各子任务到多个计算节点时，可以针对不同计算节点之间计算能力的差异配置并行计算任务，最大化地利用统一计算资源并行计算调度系统的计算能力。By introducing the load equivalent above, when allocating each sub-task of each security application check to multiple computing nodes, parallel computing tasks can be configured according to the difference in computing capabilities between different computing nodes, maximizing the use of unified computing resources for parallel computing scheduling. The computing power of the system.
应该理解为，针对已经调度的某项安全校核任务或应用，并不会在执行该校核的过程中动态调整负载当量；负载当量均衡应用于在有新加入的安全校核任务时。具体地，每次有新校核任务分配时，统计各节点处于执行之前分配的安全校核任务而形成的负载当量；并针对新校核任务分配而采用负载均衡的操作。It should be understood that for a certain security check task or application that has been scheduled, the load equivalent will not be dynamically adjusted during the execution of the check; load equivalent balancing is applied when a new security check task is added. Specifically, each time a new verification task is assigned, the load equivalent formed by each node executing the previously assigned security verification task is counted; and a load balancing operation is adopted for the allocation of the new verification task.
其中，所述步骤3)的实现多时间尺度安全校核应用并行计算调度，包括：Among them, the implementation of multi-time scale security check application parallel computing scheduling in step 3) includes:
将步骤2)中单应用并行计算调度算法改进，提出如下多时间尺度安全校核应用的并行计算调度方法：The single-application parallel computing scheduling algorithm in step 2) is improved, and the following parallel computing scheduling method for multi-time scale security check applications is proposed:
(3.1)顺序调度法，即按不同应用的计算任务到达的先后顺序来安排调度。这时，并行计算调度系统中的所有计算资源都互斥地分配到当前应用中，执行当前应用的计算任务时，不执行其他应用；直到当前应用的所有子任务都已经计算完成后，再启动执行下一个到达的计算任务。顺序调度法实现简单，只需对单应用并行调度算法做简单的改进即可实现。(3.1) Sequential scheduling method, that is, scheduling is arranged according to the order in which computing tasks of different applications arrive. At this time, all computing resources in the parallel computing scheduling system are mutually allocated to the current application. When the computing tasks of the current application are executed, other applications will not be executed; they will not be started until all subtasks of the current application have been calculated. Execute the next arriving computing task. The sequential scheduling method is simple to implement and can be implemented by simply improving the single-application parallel scheduling algorithm.
(3.2)共享式顺序调度法。对顺序调度法进行改进，当新的应用到达后，并不互相排斥；一旦有空闲的计算资源时，其他任务可以随时进入计算。(3.2) Shared sequential scheduling method. Improve the sequential scheduling method. When new applications arrive, they are not mutually exclusive; once there are idle computing resources, other tasks can enter calculations at any time.
与顺序调度法一样，也是按照应用(如日前校核应用、日内校核应用和实时校核应用)到达的先后顺序，在工作机(也即计算节点)的CPU有空闲的时候分配其执行新的应用中的子任务。Like the sequential scheduling method, the CPU of the worker machine (that is, the computing node) is allocated to execute new tasks in the order in which the applications (such as day-ahead calibration applications, intra-day calibration applications, and real-time calibration applications) arrive when they are free. Subtasks in the application.
(3.3)整体优先调度法。考虑到不同任务类型的等待时间期限不同，针对不同的应用可设定优先级，不再按照应用到达的先后顺序分配计算资源，而是按应用的优先级分配计算资源。有空闲的计算资源时优先分配给计算优先高的应用；当高优先级应用的所有子任务计算完成后，再分配计算资源给低优先级应用。(3.3) Overall priority scheduling method. Considering that different task types have different waiting time periods, priorities can be set for different applications. Computing resources are no longer allocated according to the order in which applications arrive, but based on the priority of the application. When there are idle computing resources, they are first allocated to applications with high computing priority; when all subtasks of high-priority applications are calculated, computing resources are then allocated to low-priority applications.
(3.4)轮转式调度法。整体优先调度法和顺序调度法虽然排序原则不同，但都是排序靠前的先计算，排序靠后的只能等到排序靠前的计算完毕，或者任务分配完毕且有空闲计算资源时才能开始计算，可能造成排序靠后的应用等待时间过长，无法满足响应时间要求。(3.4) Round-robin scheduling method. Although the overall priority scheduling method and the sequential scheduling method have different sorting principles, they are both calculated first, and those at the bottom can only wait until the calculation of the top ones is completed, or the tasks are allocated and there are free computing resources before they can start calculation. , which may cause applications at the bottom of the ranking to wait too long and fail to meet response time requirements.
为此，提出轮转式调度法，将各个应用的所有子任务保存到统一的计算队列中，然后以一个固定的轮转的方式来分配计算资源，所有应用的计算子任务都以相同的概率被选择分配计算资源。To this end, a round-robin scheduling method is proposed, which saves all subtasks of each application into a unified computing queue, and then allocates computing resources in a fixed rotation manner. The computing subtasks of all applications are selected with the same probability. Allocate computing resources.
(3.5)带优先级的轮转式调度法。轮转式调度法考虑了各应用占用计算资源的公平性，但是无法考虑不同应用的优先级。为此进行改进，提出带优先级的轮转式调度法，在对计算任务队列进行轮转确定计算资源分配时，按优先级进行抽签，使得优先级高的应用抽中的概率大(也即抽签权重大)。这样，既能让各应用都能及时开始计算，又能适当考虑不同应用的优先次序。(3.5) Round-robin scheduling method with priority. The round-robin scheduling method considers the fairness of computing resources occupied by each application, but cannot consider the priorities of different applications. To improve this, a priority-based round-robin scheduling method is proposed. When the computing task queue is rotated to determine the allocation of computing resources, lots are drawn according to priority, so that applications with high priority have a high probability of being drawn (that is, the lottery right major). In this way, each application can start computing in time, and the priorities of different applications can be appropriately considered.
应该理解为，不同时间尺度的安全校核应用之间无依赖关系；对于任一项安全校核应用，最敏感的属性是任务响应时间(也即响应时间需求)。It should be understood that there is no dependency between security check applications at different time scales; for any security check application, the most sensitive attribute is the task response time (that is, the response time requirement).
一般情况下，由于需要和自动发电控制(Automatic Generation Control，简称AGC)等控制措施闭环运行，实时计划安全校核应用所需的任务响应时间最短，因此必须在指定的时间内完成才有意义。为了保证对实时计划安全校核任务的快速响应，具体实施时，针对实时计划安全校核设置较高的优先级，并采取带优先级的轮转式调度法并行地调度计算资源，从而在较短的时间内完成提交的多个时间尺度的调度计划的安全校核。Under normal circumstances, due to the need to operate in a closed loop with automatic generation control (AGC) and other control measures, the task response time required for real-time planned safety check applications is the shortest, so it must be completed within a specified time to be meaningful. In order to ensure a quick response to the real-time plan safety check task, during the specific implementation, a higher priority is set for the real-time plan safety check, and a priority round-robin scheduling method is adopted to schedule computing resources in parallel, so that in a short time Complete the safety check of the submitted scheduling plan for multiple time scales within the time limit.
以下对本发明实施例的统一调度并行计算资源的方法进行详细说明。The method for uniformly scheduling parallel computing resources according to the embodiment of the present invention will be described in detail below.
某电网的日前、日内、实时这三类不同时间尺度调度计划的安全校核配置的计算资源如表1所列。其中，实时计划校核的计算服务器型号较老，每个节点20个CPU核；日前计划校核和日内计划校核的计算服务器型号较新，每个节点40个CPU核。The computing resources for the security check configuration of three different time scale dispatch plans of a certain power grid: day-ahead, intraday, and real-time are listed in Table 1. Among them, the computing server model for real-time plan verification is older, with 20 CPU cores per node; the computing server model for day-ahead plan verification and intraday plan verification is newer, with 40 CPU cores per node.
表1不同安全校核应用的计算资源Table 1 Computing resources for different security check applications
日前、日内、实时这3种不同时间尺度的安全校核的计算量及统计数据如表2所列。在各自的计算资源配置下，日前、日内计划安全校核应用的平均耗时为16分钟，实时计划安全校核应用的平均耗时为10分钟。The calculation amount and statistical data of security verification at three different time scales: day-ahead, intraday, and real-time are listed in Table 2. Under the respective computing resource configurations, the average time-consuming for day-ahead and intra-day planned security check applications is 16 minutes, and the average time-consuming for real-time planned security check applications is 10 minutes.
表2不同安全校核应用的计算量Table 2 Computational amount of different security check applications
将该电网的日前、日内、实时这三类不同时间尺度调度计划的安全校核配置的计算资源统一调度，并依次选择共享式顺序调度法(即前述(3.2))、轮转式调度法(即前述(3.4))和带优先级的轮转式调度法(即前述(3.5))对在制定的考察周期内进行日前、日内、实时这3种不同时间尺度安全校核的应用的计算耗时进行分析。The computing resources for the safety check configuration of the three different time scale dispatch plans of the power grid, namely day-ahead, intraday, and real-time, are uniformly dispatched, and the shared sequential dispatch method (i.e., the aforementioned (3.2)) and the round-robin dispatch method (i.e., The aforementioned (3.4)) and the priority round-robin scheduling method (i.e. the aforementioned (3.5)) are used to calculate the time-consuming of applications that perform safety checks on three different time scales: day-ahead, intraday, and real-time within the specified inspection period. analyze.
具体地，日前、日内、实时这3种不同时间尺度安全校核的计算启动方式及计算时刻如表3所列。具体应用时，实时计划安全校核按照以每15分钟为一个周期启动一次，日前、日内计划安全校核则在设定的计算时间段内随机地确定计算时刻。Specifically, the calculation start methods and calculation times of security verification at three different time scales: day-ahead, intraday, and real-time are listed in Table 3. In specific applications, the real-time planned safety check is started every 15 minutes as a cycle, and the day-ahead and intra-day planned safety checks randomly determine the calculation time within the set calculation time period.
表3不同安全校核应用的计算周期Table 3 Calculation cycles for different security check applications
首先应用前述(3.2)算法在连续的10天内进行安全校核。在这10天内，日前计划安全校核共计算20次，日内计划安全校核共计算30次，实时计划安全校核共计算960次。各安全校核计算对应的平均耗时、最大耗时及最小耗时如表4所列。First, apply the aforementioned (3.2) algorithm to conduct security checks within 10 consecutive days. During these 10 days, a total of 20 daily planned safety checks were calculated, a total of 30 intraday planned safety checks were calculated, and a total of 960 real-time planned safety checks were calculated. The average, maximum, and minimum time-consuming corresponding to each security check calculation are listed in Table 4.
表4不同安全校核应用的计算耗时对比Table 4 Comparison of calculation time consumption of different security check applications
将表2和表4中各安全校核应用的平均耗时进行对比，结果如图4所示。可以看出，采用基于统一调度计算资源的并行计算模式，不同时间尺度安全校核应用计算的平均耗时都实现了大幅缩短，如日前计划安全校核计算耗时小于12分钟；日内计划安全校核耗时小于7分钟；实时计划安全校核耗时小于5分钟)。Compare the average time-consuming of each security check application in Table 2 and Table 4. The results are shown in Figure 4. It can be seen that by adopting the parallel computing model based on unified scheduling of computing resources, the average time-consuming calculation of safety check applications at different time scales has been greatly shortened. For example, the calculation time of the day-ahead planned safety check is less than 12 minutes; The verification takes less than 7 minutes; the real-time plan safety verification takes less than 5 minutes).
分别应用前述(3.4)算法和前述(3.5)算法按照表3的计算频率在连续的10天内进行安全校核计算。各安全校核计算的平均耗时、最大耗时及最小耗时如表5所列。可以看出，前述(3.4)算法实时计划安全校核最大耗时532秒，小于600秒；前述(3.5)算法实时计划安全校核最大耗时466秒，小于600秒。但相应地，日前计划安全校核、日内计划安全校核的最大耗时相比前述(3.2)算法有所增加，但仍小于如表2所列的原各自单独计算时的平均耗时。Apply the aforementioned (3.4) algorithm and the aforementioned (3.5) algorithm respectively to perform safety check calculations within 10 consecutive days according to the calculation frequency in Table 3. The average, maximum, and minimum time-consuming for each security check calculation are listed in Table 5. It can be seen that the maximum time for the real-time plan safety check of the algorithm (3.4) mentioned above is 532 seconds, which is less than 600 seconds; the maximum time of the real-time plan safety check of the algorithm (3.5) mentioned above is 466 seconds, which is less than 600 seconds. But correspondingly, the maximum time-consuming of day-ahead plan safety check and intra-day plan safety check has increased compared with the aforementioned (3.2) algorithm, but it is still less than the average time-consuming when calculated separately as listed in Table 2.
具体地，在前述(3.5)算法中，对实时计划安全校核设置了较高的抽签权重，即其分配到空闲计算资源的概率较大，因此实时计划校核最大耗时比前述(3.2)算法和前述(3.4)算法都少。Specifically, in the aforementioned algorithm (3.5), a higher lottery weight is set for the real-time plan safety check, that is, the probability of being allocated to idle computing resources is greater, so the maximum time consumption of the real-time plan check is greater than that of the aforementioned (3.2) There are few algorithms and the algorithm mentioned above (3.4).
表5不同安全校核应用的计算耗时对比Table 5 Comparison of calculation time consumption of different security check applications
综上，采用该统一调度计算资源的并行任务调度算法，通过将各时间尺度的校核任务划分为多个子任务，并在统一地调度计算资源时，针对不同计算节点之间计算能力的差异，采用负载均衡原则调整各节点的负载率，最大化地利用并行计算系统的计算能力，缩短了各时间尺度的校核任务的计算耗时。In summary, this parallel task scheduling algorithm that uniformly schedules computing resources is used to divide the verification tasks at each time scale into multiple subtasks, and when uniformly scheduling computing resources, the difference in computing capabilities between different computing nodes is The load balancing principle is used to adjust the load rate of each node, maximize the use of the computing power of the parallel computing system, and shorten the calculation time of verification tasks at various time scales.
以上已经通过参考少量实施方式描述了本发明。然而，本领域技术人员所公知的，正如附带的专利权利要求所限定的，除了本发明以上公开的其他的实施例等同地落在本发明的范围内。The invention has been described above with reference to a few embodiments. However, it is known to those skilled in the art that other embodiments than those disclosed above equally fall within the scope of the invention, as defined by the appended patent claims.
通常地，在权利要求中使用的所有术语都根据他们在技术领域的通常含义被解释，除非在其中被另外明确地定义。所有的参考“一个/所述/该[装置、组件等]”都被开放地解释为所述装置、组件等中的至少一个实例，除非另外明确地说明。这里公开的任何方法的步骤都没必要以公开的准确的顺序运行，除非明确地说明。Generally, all terms used in the claims are to be interpreted according to their ordinary meaning in the technical field, unless otherwise expressly defined therein. All references to "a/the/the [means, component, etc.]" are to be construed openly to mean at least one instance of the said means, component, etc., unless expressly stated otherwise. The steps of any method disclosed herein are not necessarily performed in the exact order disclosed unless explicitly stated.
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910334830.0ACN110196773B (en) | 2019-04-24 | 2019-04-24 | Multi-time-scale security check system and method for unified scheduling computing resources |
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910334830.0ACN110196773B (en) | 2019-04-24 | 2019-04-24 | Multi-time-scale security check system and method for unified scheduling computing resources |
| Publication Number | Publication Date |
|---|---|
| CN110196773A CN110196773A (en) | 2019-09-03 |
| CN110196773Btrue CN110196773B (en) | 2024-02-27 |
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910334830.0AActiveCN110196773B (en) | 2019-04-24 | 2019-04-24 | Multi-time-scale security check system and method for unified scheduling computing resources |
| Country | Link |
|---|---|
| CN (1) | CN110196773B (en) |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110874271B (en)* | 2019-11-20 | 2022-03-11 | 山东省国土测绘院 | Method and system for rapidly calculating mass building pattern spot characteristics |
| CN111429033A (en)* | 2020-04-21 | 2020-07-17 | 贵州新致普惠信息技术有限公司 | Workflow task scheduling system and method and service terminal |
| CN111711185B (en)* | 2020-05-25 | 2021-11-19 | 国网青海省电力公司 | Day safety checking and blocking management method based on linearized alternating current power flow |
| CN117172295A (en)* | 2022-05-23 | 2023-12-05 | Oppo广东移动通信有限公司 | Artificial intelligent accelerator, chip, data processing method and electronic equipment |
| CN118170923B (en)* | 2024-03-11 | 2024-08-20 | 万邦同和(北京)信息技术有限公司 | Distributed management method based on large-scale dictionary big data |
| CN119248499B (en)* | 2024-09-29 | 2025-05-27 | 上海稀宇极智科技有限公司 | Task processing load analysis method and device |
| CN119917290B (en)* | 2025-04-02 | 2025-06-10 | 国网浙江省电力有限公司宁波供电公司 | A large power grid security verification calculation method suitable for multiple time scales |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2010037297A1 (en)* | 2008-09-27 | 2010-04-08 | 国家电力调度通信中心 | Parallel computing method and computing platform for security and stability analysis of large power grid |
| CN102254245A (en)* | 2011-06-16 | 2011-11-23 | 国网电力科学研究院 | Method for checking safety and stability of scheduling daily plans of power system in time-interval sharing parallel mode |
| CN103325012A (en)* | 2013-06-18 | 2013-09-25 | 国家电网公司 | Parallel computing dynamic task distribution method applicable to grid security correction |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2010037297A1 (en)* | 2008-09-27 | 2010-04-08 | 国家电力调度通信中心 | Parallel computing method and computing platform for security and stability analysis of large power grid |
| CN102254245A (en)* | 2011-06-16 | 2011-11-23 | 国网电力科学研究院 | Method for checking safety and stability of scheduling daily plans of power system in time-interval sharing parallel mode |
| CN103325012A (en)* | 2013-06-18 | 2013-09-25 | 国家电网公司 | Parallel computing dynamic task distribution method applicable to grid security correction |
| Publication number | Publication date |
|---|---|
| CN110196773A (en) | 2019-09-03 |
| Publication | Publication Date | Title |
|---|---|---|
| CN110196773B (en) | Multi-time-scale security check system and method for unified scheduling computing resources | |
| Malawski et al. | Cost-and deadline-constrained provisioning for scientific workflow ensembles in IaaS clouds | |
| Gao et al. | An energy and deadline aware resource provisioning, scheduling and optimization framework for cloud systems | |
| Zhu et al. | Scheduling stochastic multi-stage jobs to elastic hybrid cloud resources | |
| CN111708627A (en) | Task scheduling method and device based on distributed scheduling framework | |
| CN108154317B (en) | Workflow group scheduling method based on example self-adaptive distribution integration in multi-cloud environment | |
| CN104657220A (en) | Model and method for scheduling for mixed cloud based on deadline and cost constraints | |
| CN112231081B (en) | PSO-AHP-based monotonic rate resource scheduling method and system in cloud environment | |
| CN104317658A (en) | MapReduce based load self-adaptive task scheduling method | |
| Hoseinyfarahabady et al. | Handling uncertainty: Pareto-efficient bot scheduling on hybrid clouds | |
| Chen et al. | EnergyQARE: QoS-aware data center participation in smart grid regulation service reserve provision | |
| Mastelic et al. | Predicting resource allocation and costs for business processes in the cloud | |
| Qiu et al. | Cost-minimizing preemptive scheduling of mapreduce workloads on hybrid clouds | |
| Li et al. | Scheduling microservice-based workflows to containers in on-demand cloud resources | |
| CN106708624A (en) | Adaptive adjustment method for calculation resources in multiple working areas | |
| Li et al. | A QoS-based scheduling algorithm for instance-intensive workflows in cloud environment | |
| Yaqub et al. | Metaheuristics-based planning and optimization for sla-aware resource management in paas clouds | |
| Xu et al. | Hybrid scheduling deadline-constrained multi-DAGs based on reverse HEFT | |
| Grounds et al. | Cost-minimizing scheduling of workflows on a cloud of memory managed multicore machines | |
| Kraemer et al. | Reducing the number of response time service level objective violations by a cloud‐HPC convergence scheduler | |
| CN117234733A (en) | Distributed system task allocation method, system, storage medium and equipment | |
| Alizadeh et al. | Scalable model predictive control of demand for ancillary services | |
| CN116401027A (en) | Heterogeneous cloud center task scheduling method based on prediction completion time matrix | |
| Shrivastava et al. | CRI: A Novel Rating Based Leasing Policy and Algorithm for Efficient Resource Management in IaaS Clouds | |
| CN115599557B (en) | A Scheduler System Considering the Dynamic Change of Task Importance |
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |