The conventional wind farm(WF)power generation modelling method highly relies on wind hindcast produced by record time-series data or numerical weather modelling.However,estimating production at future sites is challe...The conventional wind farm(WF)power generation modelling method highly relies on wind hindcast produced by record time-series data or numerical weather modelling.However,estimating production at future sites is challenging in the absence of local wind monitoring.To address this,a data-driven WF modelling and model transfer strategy is proposed in this work.It considers the challenge of how to transpose metered data from existing operational WFs to sites that might feature as a prospective site for a new WF.By modelling 14 WFs distributed across Scotland using a machine learning(ML)approach,this study proved it was possible to effectively model metered production at a site using modelled wind speed and direction.In addition,this study also found when the latitude difference between two WFs is less than 0.2 degrees and the distance is less than 5o km,two WFs in non-mountainous areas can share an ML model.The results of the shared ML model remain superior to the results of the given power curve from manufacturers,after adjusting the results by the ratio of the power curve in these two WFs.The WF model transfer strategy investigated in this work offered a novel approach to transposing WF production estimates to new sites and appeared to offer better value than simple power curves,which is of importance at the early planning stage for site selection,although it would likely not fully replace detailed micro-siting modelling which are well established in the industry.Index Terms-Machine learning,model transfer strategy,power curve,power output estimation,wind farm.展开更多
In this paper a procedure is established for solving the Probabilistic Load Flow in an electrical power network, considering correlation between power generated by power plants, loads demanded on each bus and power in...In this paper a procedure is established for solving the Probabilistic Load Flow in an electrical power network, considering correlation between power generated by power plants, loads demanded on each bus and power injected by wind farms. The method proposed is based on the generation of correlated series of power values, which can be used in a MonteCarlo simulation, to obtain the probability density function of the power through branches of an electrical network.展开更多
为实现风电出力时间序列的高性能模拟,文中提出了一种基于SAGA-KM(simulated annealing and genetic algorithms-K-means)算法实现典型风电场景分类和基于Copula函数进行风电日过程马尔可夫过程建模的风电模拟方法。SAGA-KM算法将传统K...为实现风电出力时间序列的高性能模拟,文中提出了一种基于SAGA-KM(simulated annealing and genetic algorithms-K-means)算法实现典型风电场景分类和基于Copula函数进行风电日过程马尔可夫过程建模的风电模拟方法。SAGA-KM算法将传统KM算法与遗传算法和退火算法相结合,能显著提高风电场景分类效果;基于Copula函数建立的马尔可夫链精细概率模型,以去粗粒化方式实现马尔可夫过程蒙特卡洛模拟,克服了粗粒化引起的概率分布偏差。针对甘肃省某风电场数据进行实际模拟,结果表明文中方法生成模拟序列的统计分布特性、自相关函数特性和日均功率的分布特性与实测数据都非常接近,该方法能很好地保留风电序列的概率分布特性和随时间变化的波动特性,具有重要的工程实用价值。展开更多
风电和抽水蓄能电站联合运行可以平抑风电随机波动、提升风电消纳率。文章针对多个风电场的出力不确定性,采用概率性序列方法进行处理,提出了一种基于厂网协商机制的风电-抽水蓄能联合调度模式,并建立了基于机会约束规划的风电-抽水蓄...风电和抽水蓄能电站联合运行可以平抑风电随机波动、提升风电消纳率。文章针对多个风电场的出力不确定性,采用概率性序列方法进行处理,提出了一种基于厂网协商机制的风电-抽水蓄能联合调度模式,并建立了基于机会约束规划的风电-抽水蓄能互补系统短期优化调度模型。模型以风电-抽水蓄能互补系统总发电收益最大为目标,综合考虑了抽水蓄能电站的水力约束、机组运行约束和系统功率平衡约束。为提高模型求解效率并获得全局最优解,文章将原模型转换为混合整数线性规划(Mixed Integer Linear Programming,MILP)模型,最后使用商业化求解器LINGO进行求解。优化调度结果表明,抽水蓄能电站能够很好的补偿风电出力的波动性,并显著提升互补系统的总发电收益。展开更多
基金supported by the EPSRC through the National Centre for Energy Systems Integration(EP/P001173/1)。
文摘The conventional wind farm(WF)power generation modelling method highly relies on wind hindcast produced by record time-series data or numerical weather modelling.However,estimating production at future sites is challenging in the absence of local wind monitoring.To address this,a data-driven WF modelling and model transfer strategy is proposed in this work.It considers the challenge of how to transpose metered data from existing operational WFs to sites that might feature as a prospective site for a new WF.By modelling 14 WFs distributed across Scotland using a machine learning(ML)approach,this study proved it was possible to effectively model metered production at a site using modelled wind speed and direction.In addition,this study also found when the latitude difference between two WFs is less than 0.2 degrees and the distance is less than 5o km,two WFs in non-mountainous areas can share an ML model.The results of the shared ML model remain superior to the results of the given power curve from manufacturers,after adjusting the results by the ratio of the power curve in these two WFs.The WF model transfer strategy investigated in this work offered a novel approach to transposing WF production estimates to new sites and appeared to offer better value than simple power curves,which is of importance at the early planning stage for site selection,although it would likely not fully replace detailed micro-siting modelling which are well established in the industry.Index Terms-Machine learning,model transfer strategy,power curve,power output estimation,wind farm.
文摘In this paper a procedure is established for solving the Probabilistic Load Flow in an electrical power network, considering correlation between power generated by power plants, loads demanded on each bus and power injected by wind farms. The method proposed is based on the generation of correlated series of power values, which can be used in a MonteCarlo simulation, to obtain the probability density function of the power through branches of an electrical network.
文摘为实现风电出力时间序列的高性能模拟,文中提出了一种基于SAGA-KM(simulated annealing and genetic algorithms-K-means)算法实现典型风电场景分类和基于Copula函数进行风电日过程马尔可夫过程建模的风电模拟方法。SAGA-KM算法将传统KM算法与遗传算法和退火算法相结合,能显著提高风电场景分类效果;基于Copula函数建立的马尔可夫链精细概率模型,以去粗粒化方式实现马尔可夫过程蒙特卡洛模拟,克服了粗粒化引起的概率分布偏差。针对甘肃省某风电场数据进行实际模拟,结果表明文中方法生成模拟序列的统计分布特性、自相关函数特性和日均功率的分布特性与实测数据都非常接近,该方法能很好地保留风电序列的概率分布特性和随时间变化的波动特性,具有重要的工程实用价值。
文摘风电和抽水蓄能电站联合运行可以平抑风电随机波动、提升风电消纳率。文章针对多个风电场的出力不确定性,采用概率性序列方法进行处理,提出了一种基于厂网协商机制的风电-抽水蓄能联合调度模式,并建立了基于机会约束规划的风电-抽水蓄能互补系统短期优化调度模型。模型以风电-抽水蓄能互补系统总发电收益最大为目标,综合考虑了抽水蓄能电站的水力约束、机组运行约束和系统功率平衡约束。为提高模型求解效率并获得全局最优解,文章将原模型转换为混合整数线性规划(Mixed Integer Linear Programming,MILP)模型,最后使用商业化求解器LINGO进行求解。优化调度结果表明,抽水蓄能电站能够很好的补偿风电出力的波动性,并显著提升互补系统的总发电收益。