基于BLSTM-随机森林的短期光伏发电输出功率预测

Output power prediction of short-term photovoltaic power generation based on BLSTM-random forest

光伏发电功率具有不确定性和波动性,准确预测光伏发电功率对提高光伏并网效率和保持电网安全运行具有重要作用。对江苏某地区光伏发电站的功率特性进行分析,使用小波降噪处理历史功率曲线,并对各气象条件使用灰色关联分析筛选出强相关影响因素,减少输出功率噪声和无关气象条件对功率预测的影响。将小波降噪处理后的历史输出功率及强相关特性构建数据集,建立基于双向长短期记忆网络(BLSTM)与随机森林的短期光伏发电功率预测模型,并与其他模型的预测误差进行比较。仿真结果表明,提出的BLSTM-随机森林的短期光伏功率预测模型具有较高的预测精度。

The power of the photovoltaic power generation has uncertainty and volatility,and the accurate power prediction of photovoltaic power generation plays an important role in improving photovoltaic grid-connection efficiency and maintaining the safe operation of the grid.The power characteristics of PV power generation stations in a certain area of Jiangsu province were analyzed,the wavelet denoising was used to process the historical power curve,and the strong correlation influential factors were selected by grey correlation analysis for each meteorological condition,so as to reduce the influence of output power noise and irrelevant meteorological conditions on power prediction.The data was constructed by using historical output power after wavelet denoising and strong correlation characteristics to build the power prediction model of the short-term photovoltaic power generation based on bidirectional long and short term memory network(BLSTM)and random forest,and the prediction error was compared with the prediction error of other models.The simulation results show that the proposed short-term PV power prediction model based on BLSTM-random forest has high prediction accuracy.