基于云团-辐照度时空关联模式挖掘的光伏发电功率超短期预测方法研究

Ultra-short-term forecast of PV power generation is the basic key technology to improve the control and dispatching performance of PV power plants and ensure the safe and stable operation of the power grid with high-rate PVs. The change in irradiance caused by cloud movement is the root cause of fluctuations in PV power output. While traditional ultra-short-term PV power forecasting is mostly focused on local meteorological and power information of a single PV plant. These techniques are unable to break through the current bottleneck of ultra-short-term PV power forecasting as they ignore the influence of the spatial distribution of cloud in different time-space scales..Facing the power forecasting requirements of large-scale and wide-area distributed PV power plant cluster in China, this project research focuses on two major issues based on synchronous orbit satellite observation data: the mathematical description and prediction method of cloud cluster spatial distribution and its non-linear motion process, and the spatio-temporal correlation between the spatial distribution of clouds and the temporal distribution of irradiance. First, we build a cloud tracking and prediction model based on the optimal continuous mapping of the cloud motion matrix to achieve an accurate selection of the associated cloud image area of the PV power plant; Then, based on the cloud-irradiance spatio-temporal correlation pattern mining, we establish a mapping model from regional satellite images to the irradiance time series, to realize ultra-short-term forecasting of the surface irradiance and even the power generation of PV power plants.

光伏发电功率超短期预测是提高光伏电站控制与调度性能，保障高比率光伏发电接入电网安全稳定运行的基础性关键技术。云团运动引起的辐照度改变是导致光伏发电出力波动的根本原因，而传统光伏功率超短期预测技术多以单一光伏场站为中心，着眼局部气象与功率信息，忽略了不同时空尺度下的云团空间分布影响，无法突破当前超短期预测精度瓶颈。.面对我国大规模、广地域分布的光伏电站集群功率预测需求，本项目基于同步轨道卫星观测数据，围绕云团空间分布及其非线性运动过程的数学描述与预测方法，以及天空云团空间分布变化规律与地表辐照度时间分布规律间的时空关联模式两大科学问题展开研究。首先构建基于云团运动特征矩阵最优连续映射的云团追踪与预测模型，实现光伏电站关联云图区域的准确选择；然后基于云团-辐照度时空关联模式挖掘建立从区域云团卫星图像到地表辐照度时间序列间的映射模型，从而实现对光伏电站地表辐照度乃至发电功率的超短期预测。

光伏发电功率超短期预测是提高光伏电站控制与调度性能，保障高比率光伏发电接入电网安全稳定运行的基础性关键技术。云团运动引起的辐照度改变是导致光伏发电出力波动的根本原因，而传统光伏功率超短期预测技术多以单一光伏场站为中心，着眼局部气象与功率信息，忽略了不同时空尺度下的云团空间分布影响，无法突破当前超短期预测精度瓶颈。面对我国大规模、广地域分布的光伏电站集群功率预测需求，本项目基于同步轨道卫星观测数据，围绕云团空间分布及其非线性运动过程的数学描述与预测方法，以及天空云团空间分布变化规律与地表辐照度时间分布规律间的时空关联模式两大科学问题展开了研究工作。.项目利用先进卫星遥感与地基天空成像多源观测数据和太阳辐射传输理论，揭示了不同气象条件下光伏电站天空云团的运动规律及其生消、形变演化机理，发明了基于位移矩阵能量函数的云团运动矢量计算方法，建立了基于“天空图像-地表辐照度”映射的光伏功率分钟级预测模型；针对广域地理范围内太阳辐射的时空关联分布特性，创建了基于出力波动模式识别与卫星云图特征的光伏功率超短期预测方法；基于图神经网络建立了融合临近场站观测数据的光伏发电短期预测模型；构建了“空-天-地-网”四位一体的新型光伏功率预测技术体系，破解了分钟级快速剧烈波动与非平稳转折性天气条件下的光伏功率准确预测难题。