超短期风力发电量预测技术及其比较分析

Ultra Short Term Wind Power Generation Forecasting Technology and Its Comparative Analysis

风力发电作为清洁绿色的新能源,是实现“双碳”目标的主力军之一,但是其对新能源消纳系统提出了新要求,故对风力发电量的科学分析和精确预测研究具有现实意义。首先,对风电多维历史数据属性、特点和离群值、噪声平滑等进行分析与预处理,再通过2种回归树集成和4种回归神经网络及其超参数优化算法对不同机组数据进行回归分析,超参数优化运行时间代价较高。回归拟合效果通过5个评价指标进行对比与分析,经过大量仿真实验,证明了三层神经网络的回归模型拟合和预测效果均较好。

As clean and green new energy,wind power generation is one of the main forces to achieve the“double carbon”goal.However,it puts forward new requirements for the new energy consumption system,so it has practical significance for the scientific analysis and accurate prediction of wind power generation.Firstly,the attributes,characteristics,outliers and noise smoothing of multi-dimensional historical data of wind power are analyzed and preprocessed,and then the data of different units are regressed and analyzed through two regression tree ensemble,four regression neural networks and their super parameter optimization algorithms.The cost of super parameter optimization is high.The regression fitting effect is compared and analyzed through five evaluation indexes.A large number of simulation experiments proved that the regression model fitting and prediction verification effect of the threelayer neural network are good.