【作者】 雷潇；

【导师】 李剑；

【作者基本信息】 重庆大学 ， 电气工程， 2014， 博士

【摘要】 风能作为一种成熟的可再生能源技术，近年来在世界范围内得到快速发展。然而，由于长期处于恶劣的自然环境中，风电机组的停运率较高，降低了风电场的运行经济性。同时，在电网的约束下，风电场也会因为风电功率预测误差和风电机组强迫停运产生一定程度的发电量损失。本文以提高风电场运行经济性为目标，对风速与风电功率预测误差分布特性、风电机组状态参数异常辨识方法、风电机组短期可靠性预测方法以及风电场有功功率控制策略进行了研究。论文主要包括以下内容：①提出了风速与风电功率预测误差的核密度分布模型，研究了预测误差分布特性。采用多种典型方法进行风速预测，对比分析了单机风速和风电场风速的预测精度。建立了基于实测数据的功率曲线，结合风速预测方法进行了风电功率预测。研究了单机风速与风电功率预测误差分布与风速、预测时间间隔之间的关系。结果表明，在三种预测方法下，单机风速的预测误差均明显大于风电场风速的预测误差，不同风速区间内的预测误差分布差异明显，预测误差随预测时间间隔的增大而增大。②提出并建立了风电机组状态参数广义模糊异常辨识模型。研究了风电场SCADA系统提供的风电机组状态参数与自然环境和机组工作特征的关联关系，建立了状态参数的预测模型，分析了影响预测模型精度的主要因素。对比了本机近期数据模型、本机历史数据模型和其他机组近期数据模型这三类预测模型的预测精度，在此基础上提出了预测模型的选择方法和预测残差的异常程度量化方法，最终采用模糊综合评判进行状态参数的异常辨识。结果表明，广义模糊异常辨识模型综合了多个预测模型的异常辨识结果，具有更高的准确度。③提出了风电机组短期可靠性预测模型。计及风速与风电机组停运率的相关性，建立了考虑风速的风电机组统计停运模型。对与自然环境密切相关的状态参数，提出了基于状态参数概率预测的保护动作模型；对具有保护动作整定时间的状态参数，提出了基于越限时间的保护动作模型。最后，综合考虑各状态参数的越限保护动作概率和统计停运概率，提出了计及状态参数越限的风电机组短期可靠性预测模型。结果表明，通过对状态参数越限概率的计算，风电机组短期停运模型的准确性得到大幅提高。④提出了考虑机组短期可靠性和功率预测误差的有功功率控制方法。分析了风电场在无电网约束和限功率运行两种情况下的发电量损失原因，在无电网约束情况下，通过对低可靠性机组进行降功率控制以减小其短期停运概率。针对风电场限功率运行情况，提出采用功率损失风险对可能损失的功率进行量化，并提出了多台机组总功率损失风险的蒙特卡洛模拟计算方法，获取了单机功率损失风险和风电场总功率损失风险的关系，提出了限功率情况下的风电场有功功率协调控制策略。结果表明，本文提出的方法有效降低了风电场的发电量损失。上述工作是为解决目前风电场运行维护成本偏高的问题所做的积极探索，不仅丰富了相关领域的研究成果，而且为大规模风电的安全高效利用提供了解决方案。更多还原

【Abstract】 As one of the most important and mature renewable energy,wind energy developedrapidly in the word range in recent years. However, economic operation of wind farmsis affected by the reliabilityof WTs and constraint of power grid. On one hand, theoutage rate of wind turbines is much higher than tranditional electrical components dueto the harsh natural environment. On the other hand, part of energy loss under theconstraint of power grid due to the error of wind speed prediction and WT outages.Although related technologies are developed in recent years, there are still manyproblems to be solved. The distribution of forecast error of wind speed and wind power,anomaly identification method of condition monitoring parameters of wind turbines,short-term reliability assessment method for wind turbines, and the economicdispatching method of wind turbines in wind farms are presented in this thesis. Themain contents are shown as follows:①A Kernel density distribution model of forecasting error was proposed, and therelationship between among forecasting error distribution characteristics,wind speedand forecastingtime interval was studied. A variety of typical methods were utilized toforecast wind speed and the forecasting accuracy of single wind turbinewas analyzed incomparison with that of wind farms. Then a power curve based on actual measurementwas established and wind power was forecasted with the wind speed forecastingmethods. Results show that with three forecasting methods the forecasting error in thesingle wind turbine level is significantly greater than that in the wind farm level; theforecasting error increases when the forecasting time interval increases; the differenceamong forecasting error distributions in different wind speed ranges is quite significant.②A generalized fuzzy abnormal identification model for wind turbines wasproposed. The correlation relationship between wind turbine parameters provided by theSCADA system and the natural environment or the wind turbine operationcharacteristics was studied. Then a forecasting model for parameters was establishedand the main factors influencing the accuracy were analyzed. The forecasting accuracyof the native recent data model of the local wind turbines, the history data model of thelocal wind turbines and the recent data modelof other wind turbines were analyzed andcompared to each other. A selection method for the forecasting model and a quantitativemethod for the degree of abnormality of forecasting residual error were put forward. Finally, the fuzzy comprehensive evaluation was utilized to identify the abnormalparameters. The results show that theaccuracy of identifying abnormal parameters ishigher when the identifying results of abnormal state of multiple forecasting models aretaken into account.③Considering the correlation between wind speed and wind turbine outage rate,a wind turbine outage model with the consideration of wind speed was established. Forthe parameters closely related to the natural environment, an off-limitprotectionactionmodel based on parameters probability forecastwas presented.Accordingly, for the parameters which have setting time of protection action, anoff-limit protectionactionmodelbased on the time of off-limit was present. Finally,considering the limit protection probability and the statistical outage probability of eachstate parameter comprehensively, a wind turbine reliability evaluation modelconsidering the parameters limit was put forward. The results show that the accuracy ofthe short-termoutage model for the wind turbine model is much higher by calculatingthe parameters off-limit probability.④An effective controlling method for active power was presented consideringpower forecasting error andshort-term reliability of wind turbines. Without the gridconstrains, the short-term outage probability can be reduced by implementing thepower-reduced control measures on the wind turbines with low reliability. For thelimited power operating condition, the risk of power loss was proposed to quantifypossible power loss. Finally, based on the concept of monte carlo simulation, thecalculating method for the risk of multiple wind turbines power loss was presented, thusthe relation between the risk of single wind power loss and the wind farm total powerloss was obtained. Besides, the coordinated control strategy of wind farm active powerunder limited power conditionwas presented. The results show that the methodsmentioned in this thesis effectively reduce the power loss of wind farms.The above work is active exploration for reducing the high cost of operation andmaintenance of wind farm. The methods proposed in this thesis not only enrich researchin related fields, but also offer feasible solution to efficient use of large-scale windpower.更多还原