基于改进相似日算法的光伏电站功率预测

Power forecasting of photovoltaic plant based on improved similar day

针对传统光伏电站功率预测方法精度不高的问题,给出一种基于改进相似日算法的广义回归神经网络功率预测方法。在该方法中,提出一种对日相似度与前趋势相似度进行多目标优化来选取相似日集的改进相似日算法,并建立了基于此算法的广义回归神经网络功率预测模型,以太阳辐照度、温度和湿度作为模型输入,光伏电站发电功率作为模型输出,提前一天预测间隔为1 h的输出功率。利用甘肃某光伏电站的实测数据进行仿真,结果表明,该方法较传统功率预测方法有更高的预测精度。

A method of power forecasting of the photovoltaic plant based on improved similar day and GRNN was introduced aiming at the problem of the low precision obtained from the traditional method.According to this method,an improved similar day algorithm was proposed,taking the multi-objective optimization on the day-similarity and the previous-trend-similarity to select the similar day set.And then,the GRNN power forecasting model was created based on this algorithm,taking the solar irradiance,temperature and humidity as the input and the power of the photovoltaic plant as the output.The output power was forecasted with 1 hour’s resolution 1 day ahead finally.The simulation result coming from the history data from the PV plant in Gansu province shows that the GRNN model can obtain a higher forecasting accuracy than the traditional method and this method can use in the PV plant with the numerical weather prediction system.