一种基于双层人工神经网络的多时间尺度区域光伏功率预测方法

A Multi-time Scale Regional PV Power Forecasting Method Based on Double-layer Artificial Neural Network

由于具有间歇性、波动性和随机性的特点,光伏发电系统的大规模并网运行会严重影响电力系统的稳定与经济运行。因此,开展区域光伏功率预测能够为调度部门提供电源出力参考信息,以合理规划调度计划及安排备用容量。提出了一种基于双层人工神经网络的多时间尺度区域光伏出力预测方法,基于选取的基准光伏电站实现预测分辨率为1 min、5 min和15 min的多时间尺度出力预测,解决了部分光伏电站因设施不完善导致的历史出力预测数据缺失与失准造成的区域功率预测精度低的问题,并降低了所需数据量。除了考虑相关系数之外,引入第一层人工神经网络的预测精度作为选择基准光伏电站的指标,有效地提高了区域功率预测精度。此外,针对基准光伏电站因云团遮挡或电网故障导致其出力波动进而影响预测精度的难题,提出了基于相邻光伏电站出力的基准电站出力修正方案,并通过山东某地市的光伏历史出力数据对所提方法进行了分析验证。

Due to the intermittent,fluctuating and random of photovoltaic(PV)output,massive access of PV system will threaten the stability of system operation with economical loss.Therefore,it is significant to develop regional PV power forecasting technique to provide reference information for dispatching center to properly make dispatching plan and arrange spinning reserve.This paper proposed a multi-time scale regional PV power forecasting method based on double-layer artificial neural network.The regional PV power in the prediction time resolution of 1 minute,5 minutes and 15 minutes were realized based on historical data of selected reference PV plants,improving the poor prediction accuracy caused by incomplete facilities and lack of historical output data at some PV power plants.The forecasting accuracy of first-layer artificial neural network was chosen as the indicator for the selection of reference PV plant,effectively improving the forecasting accuracy of regional output power.This paper also proposed a method to improve regional output power forecasting accuracy based on neighboring PV plants when the output power of reference PV plant is fluctuating due to unexpected cloud shading or grid fault.The feasibility of the proposed method is verified by the historical output data of PV plants of a city in Shandong Province.