基于相似日与加权马尔可夫模型的风力发电功率区间预测

Wind power interval prediction based on similar day and weighted Markov model

为了提高风力发电功率预测精度,提出一种基于相似日与加权马尔可夫模型的风力发电功率区间预测方法(SWMQ)。风电功率数据与风速数据直接相关。首先对于数据中的异常值和缺失值,通过线箱图法和相关性填补法对数据进行预处理,提高数据的关联性,通过卷积神经网络(CNN)对风速进行预测;然后由预测到的风速数据在历史数据中通过皮尔逊相关系数法寻找相似日,以相似日功率数据为数据集进行加权马尔可夫模型预测;最后通过分位数回归原理对预测区间进行求取,同时建立基于CNN模型、相关性填补、CNN模型和加权马尔可夫模型,以西北某风电场数据进行仿真对比。实验表明该模型在风力发电功率预测上有较高的精度,能更好地体现数据变化的阈值。

A wind power interval prediction method based on the similar day and weighted Markov model(SWMQ)is proposed to improve the wind power prediction.The wind power data is related to wind speed data directly.The abnormal and missing values in the data are preprocessed by the boxplot method and correlation filling method,so as to improve the data correlation.The wind speed is predicted by convolutional neural network(CNN).And then,the predicted wind speed data is used to find out the similar day in the historical data with the method of Pearson correlation coefficient(PCC),and the power data of the similar day is used as the data set for weighted Markov model prediction.The prediction interval is obtained by the principle of quantile regression.The CNN-based model,the model based on correlation filling and CNN,and the weighted Markov model are established.The simulation and comparison of the data of a wind farm in northwest China show that the proposed model is more accurate in wind power prediction,and can better reflect the threshold of data change.