基于相似日聚类和VMD-LTWDBO-BiLSTM的短期光伏功率预测

Short-term PV Power Prediction Based on Similar Day Clustering with VMD-LTWDBO-BiLSTM

针对光伏发电功率波动性、间歇性和预测精度较低问题,提出一种基于改进蜣螂优化算法(LTWDBO)对变分模态分解(VMD)中惩罚系数和分解层数参数以及双向长短期记忆网络(BiLSTM)中关键性参数进行寻优的组合光伏发电功率预测模型。首先,采用斯皮尔曼相关系数选取主要因素作为K-means++相似日聚类的输入,将历史数据划分为不同天气的相似日样本;其次,将不同天气下的强相关性特征和光伏发电功率数据进行VMD分解,并对各子分量构建LTWDBO-BiLSTM预测模型;最后将预测值进行叠加重构得到最终预测结果。仿真结果表明,所提VMD-LTWDBO-BiLSTM组合模型相较于其它几组模型,在不同天气下的平均绝对误差均显著降低,验证了其在不同天气条件下仍具有较好的精度和鲁棒性。

Aiming at the problems of fluctuation,intermittency and low prediction accuracy of photovoltaic power,the paper proposes a combined photovoltaic power prediction model based on the improved dung beetle optimization algorithm(LTWDBO)to find the optimality of penalty coefficients and decomposition layer parameters in the variational modal decomposition(VMD)as well as key parameters in the bidirectional long and short-term memory network(BiLSTM).Firstly,the Spearman correlation coefficient is used to select the main factors as inputs for K-means++similar day clustering and the historical data are divided into similar day samples of the different weather condition.Then the strong correlation features and PV power data under different weather conditions are decomposed by VMD,and the LTWDBO-BiLSTM prediction model is constructed for the sub-components.Finally,the predicted values are reconstructed by superposition to obtain the final prediction results.The simulation results show that the proposed VMD-LTWDBO-BiLSTM combined model has a significantly lower average absolute error under different weather conditions compared with other groups of models,verifying the better accuracy and robustness of the model under different weather conditions.