基于深度学习的风光场群功率预测方法研究

Research on Power Forecasting Method for Wind Farms and Photovoltaic Stations Based on Deep Learning

随着风光场站集群化发展,大规模的风光电力接入电力系统会威胁电力系统的安全稳定运行。精准的风光功率预测能有效缓解这一问题,但是现有的风光功率预测方法多集中在场站级别,区域总出力对电力系统制定调度计划、安排旋转备用容量具有重要的意义。为此,提出了基于堆叠降噪自编码器的风光功率预测模型实现场站-区域风光功率预测。以分布在我国某省的风光场群的运行数据为例,验证所提模型的有效性。结果表明:与场站原有预测系统精度相比,平坦地形风电场功率预测精度平均提高了15.56%,光伏场站功率预测精度平均提高了21.75%;复杂地形风电场功率预测精度平均提高了3.28%。

With the development of clusters of wind and solar power stations,large-scale wind and solar power access to the power system will threaten the safe and stable operation of the power system.Accurate power forecasting can effectively alleviate this problem,but the existing power forecasting methods are mostly concentrated at the station level and the total regional output is of great significance for the power system to formulate dispatch plans and arrange spinning reserve capacity.Therefore,a regional wind and solar power forecasting model based on Stacked Denoising auto-encoder is proposed to realize station-region wind and solar power forecasting.Taking the operating data of a group of wind and solar farms distributed in a certain province of our country as an example,the validity of the proposed model is verified.The results show that,compared with the accuracy of the original station prediction system,the prediction accuracy of wind farms in the flat terrain has increased by 15.56%on average,the prediction accuracy of solar power stations has increased by 21.75%on average,and the prediction accuracy of wind farms in the hilly terrain has increased by 3.28%on average.