在线自适应LASSO罚向量自回归模型的风电功率预测

An online adaptive LASSO penalty vector autoregressive model for wind power prediction

针对许多领域中普遍存在的非平稳多元时间序列的建模处理问题,提出了LASSO向量自回归模型的递推在线拟合方法,利用遗忘指数来实现模型的动态变化,并用循环坐标下降算法在线的对向量自回归模型进行系数估计。为证明模型的有效性,将其应用于风电场风电功率的预测,并以传统的向量自回归模型和分层向量自回归模型作为比较基准。根据实验结果表明,在线自适应LASSO向量自回归模型的预测精度高于传统的批量模型,通过系数矩阵图也可以看出,预测风电场临近的风电场对预测点存在一定程度的影响,但自身影响是最大的。将递归在线估计与LASSO向量自回归模型的结合应用于风电功率的预测,对于提高风电功率的预测精度以及改善风电系统工作效率有重要意义。

Aiming at the modeling problem of non-stationary multivariate time series prevalent in many fields,a recursive online fitting method for LASSO vector autoregressive model is proposed.The forgetting index is used to realize the dynamic change of the model,and the cyclic coordinate descent algorithm is used online.The coefficient estimation is performed on the vector autoregressive model.In order to prove the validity of the model,it is applied to the wind power prediction of wind farms,and the traditional vector autoregressive model and hierarchical vector autoregressive model are used as the benchmark.According to the experimental results,the prediction accuracy of the online adaptive LASSO vector autoregressive model based on the coordinate descent algorithm is higher than that of the traditional batch model.It can also be seen from the coefficient matrix diagram that the wind farm adjacent to the wind farm is predicted.The forecast point has a certain degree of influence,but its own impact is the biggest.The application of recursive online estimation and LASSO vector autoregressive model to wind power prediction is of great significance for improving the prediction accuracy of wind power and improving the efficiency of wind power system.