中文摘要：

针对大量分散的光伏发电单元接入配电网后引起的过电压问题，提出了一种基于BP神经网络预测模型的过电压抑制策略。该策略利用气象预报的光照、温度数据信息以及检测的光伏发电系统的输出功率和馈线电压等历史数据，预测出需要削减的光伏发电有功功率，并控制光伏板的功率输出以避免过电压的发生。对提出的预测模型作了介绍，提出了不同日类型（晴天、多云）的分类预测模型。最后通过MATLAB仿真软件对算例进行分析，结果验证了该预测模型抑制过电压的可行性和有效性。

英文摘要：

In the light of the over-voltage resulted from large scale photovoltaic grid-connected power system, a new over-voltage suppression strategy based on BP neural network is proposed. The strategy forecasts the active power of photovoltaic power generation needed to be cut, and control the power output of photovoltaic panels to avoid the occurrence of over-voltage by using historical data, such as the solar irradiance and temperature data of meteorological forecast, output power and feeder voltage of photovoltaic grid- connected power system. The forecast model is introduced, the prediction models of different types （sunny and cloudy） are presented. Finally, the analysis results of MATLAB simulation software confirm the feasibility and effectiveness of predictive model.