【作者】 张伟；

【导师】 林湘宁；

【作者基本信息】 华中科技大学 ， 新能源科学与工程， 2016， 硕士

【摘要】 进入新世纪以来,由于能源危机的加重和生态问题的恶化,新能源领域受到越来越多的研究者的关注。作为技术和商业上比较成熟的发电形式之一,风力发电已经在包括我国在内的全球范围内拥有成功应用的众多工程和商业实例。我国的风力资源丰富,但是受到地理及气象条件的限制,风能的利用多采取集中式的大规模风电场模式。风能具有随机性的特点,这主要体现在风电场功率输出的波动性上。风电场的接入给电网的正常安全运行带来很大的挑战。我国曾发生过多起由于风电场故障引发大量风机脱网、继电保护装置误动作,进而威胁电网安全运行的案例。经典的电网静态稳定性评估方法能够根据历史数据考虑电力负荷的波动、发电机和输电线路的故障等运行情况,并可以给出电网运行风险水平。但是,这些结果是面向过去的、“静止”的评估结果。风电场的出力在短时间内可能产生较大幅度的波动,并且在不同的时间尺度下,风电场的出力会呈现不同的特性,传统的的电网风险评估方法由于自身的局限性并不能及时地反映上述情况。针对这一问题,本文以风电场出力预测与电网风险评估相结合的技术方案为研究对象,围绕含风电电网多时间尺度风险评估及其加速算法展开系统性的研究工作。在含风电电网多时间尺度风险评估研究方面,为研究未来不同时间尺度下风电场出力对电网运行风险造成的影响,本文基于经典电网静态稳定性分析理论和风电场出力预测理论,并借鉴金融风险测度领域的未来资产损失分析模型(Va R)的理论和方法,设计了三类不同的风险评估指标:第一类指标反映电网负荷削减的情况,第二类指标表征电网母线电压、支路潮流的越限情况,第三类指标则能给出未来多时间尺度下风电接入给电网造成的影响。在风电出力预测模型方面,采用经济学中的时间序列预测模型根据某风电场的实际出力数据进行风电场有功功率的预测。同时,在MATLAB软件平台上编写了考虑风电的电网多时间尺度风险评估程序代码。此项工作不仅能够给出未来某一时刻电网风险评估的静态数据,而且能够反映基于风电功率预测数据的未来电网整体风险水平的动态变化过程;进而可为风电场接入电网后,运行控制人员针对不同时间尺度下未来风电场可能出现的波动情况提前采取预防性措施提供参考。在加速算法的研究内容方面,主要完成了两个方面的工作:硬件加速算法和软件加速算法。在硬件加速方面,本文首先基于多核CPU计算平台讨论了不同计算任务的分配方式和通信模式对并行算法的影响,通过设计不同的算法场景讨论最佳的并行策略。然后引进图形处理器(GPU)来加速计算,研究通过不同的计算量分配指标来优化综合异构硬件加速算法的效率。在软件加速方面,针对电网风险评估方法中耗时最长的随机潮流计算部分,采用改进等分散抽样法来减少对电力系统运行状态的抽样;同时,采用启发式就近负荷削减模型对风险评估程序中第一类评估指标计算的代码进行加速。最后,将软件加速算法与硬件加速算法相结合,对电网风险评估技术方案进行加速。文章采用的算例分析表明,本文所提的加速算法能够在确保计算结果准确的约束下大幅度缩短程序的计算时间,从而为本文所提风险评估系统的实时在线应用打好基础。更多还原

【Abstract】 With the energy crisis and environmental problems becoming increasingly severe in recent years, renewable energy has attracted more attention than before.As one of the mature technologies for renewable energy power generation, wind farms have been widely established globally.Wind resource is abundant in C hina, but it is restricted by geographical and meteorological conditions.In China,we prefer centralized large-scale wind farms.Wind energy has the character of randomness, which is mainly showned as the fluctuation of wind power output.After the wind power integration, the operational risk of the power grid is greatly affected.There were a lot of accidents in China, which caused the trip-off of wind turbines and relay protection device malfunction. All these problems could threaten the safe operation of the power grid.The classical power grid static stability assessment method can calculate the fluctuation of the power load, the fault of generators and transmission lines.Then it can illustrate the level of power grid operational risk.However, these results are evaluations which are past-oriented and static.In short time, the output of the wind farm may have a large range of fluctuations, and at different time scales, the output of the wind farm will show different characteristics.The classical risk assessment method of power grid can not reflect the above conditions timely and accurately.Aiming at this problem, this paper researches the combination of wind farm output forecasting and power grid risk assessment.The main topics of this paper are multi-time scale risk assessment of power grid and its acceleration algorithm considering wind power integration.On the topic of multi-time scale risk assessment of power grid considering wind power integration, in order to study the impact on power grid operation risk of wind farm output on different time scale in future, this paper will focus on the following works. Firstly, based on the classical power grid static stability analysis theory, wind power output prediction technology, and The theory and method of future asset loss analysis model(Va R) in the field of financial risk measurement, three different types of risk assessment indexes are designed.The first type of indexes reflect the level of power grid load reduction. And the second type of indexes illustrate the overload condition of the bus voltage and the power flow of transmission line. Then the third type of indexes can show the influence of wind power integration in multi- time scale in future. In wind farm output prediction, the time series forecasting model is used to test and forecast the power data of a wind farm.Meanwhile,the code of this program is designed and comp iled on MATLAB software.The results of this work can not only give the static data of the future power grid risk assessment, but also can reflect the dynamic change process of the overall risk level of the power grid based on wind power forecast data.And t hen, after the wind farm is connected to the power grid, the operational control personnel can take preventive measures in advance to provide the reference for the possible fluctuation of wind farm in different time scales.On the topic of accelerating algorithm, the work of this paper is mainly focused on two aspects: hardware acceleration algorithm and software acceleration algorithm.Firstly, based on the multi-core CPU computing platform, this paper discusses the effects of different tasks allocation strategy and communication mode.Then, the graphics processor unit(GPU) is introduced to accelerate the computational efficiency. And different workload distribution strategies are designed to optimize the integrated heterogeneous hardware accelerated algorithm.Secondly,in the aspect of software acceleration, the study is focused on the most time-consuming part of the power grid risk assessment method, which is power flow calculation.New improved sampling method is designed to reduce the number of operating states sampling, which is combined with the Latin hypercube sampling method and the average scattered sampling method.At the same time, the heuristic approach to local load shedding scheme method is used to accelerate the calculation process of the first type of evaluation indexes.Finally, the software acceleration algorithm is combined with the hardware acceleration algorithm to accelerate the program. Case study shows that the proposed acceleration algorithm can guarantee the accuracy of the results and shorten the calculation time.This work is a useful exploration for the on- line application of power grid risk assessment system.更多还原