基于滚动的BP神经网络的光伏发电功率预报

• 作 者: 梁允 许沛华 孙芊 周宁 陈正洪
• 机 构: 国网河南省电力公司电力科学研究院,湖北省气象服务中心,湖北省气象能源技术开发中心,
• 出 处: 水电能源科学
• 发表日期: 2017年09期
• 关键词: BP神经网络 辐射滚动订正 发电功率 短期预报

鉴于准确预测光伏发电功率可减少大规模光伏并网发电对电网造成的冲击,以河南省某并网的光伏发电站7~11月共5个月的数据为例,通过建立非滚动的和滚动的BP神经网络模型,分别进行数值预报辐射订正和发电功率预报,并对预报的72h结果分为第1、2、3d分别进行检验。结果表明,滚动的BP神经网络对辐射订正和功率预报均具有较好的泛化能力,方法简便、实用,能够有效降低光伏发电功率的预报误差。

In view of the accurate prediction of photovoltaic power generation can reduce the impact of large-scale photovoltaic grid-connected power generation on the grid, a photovoltaic power plant in Henan Province, a total of five months from July to November data as an example, by establishing non-rolling and rolling BP neural network model, respectively, the numerical prediction of radiation correction and power generation power forecast, and the forecast of 72h results were divided into 1, 2, 3d were tested. The results show that the rolling BP neural network has good generalization ability for radiation correction and power forecasting. The method is simple and practical, and can effectively reduce the forecast error of photovoltaic power generation.