基于VMD-CIGWO-BP-DTA算法的空调负荷预测

Air Conditioning Load Forecasting Based on VMD-CIGWO-BP-DTA Algorithm

摘要: 提出一种Circle混沌化灰狼算法(CIGWO)优化BP神经网络与变分模态分解(VMD)结合的预测模型(VMD-CIGWO-BP-DTA)，对蓄能空调负荷进行预测分析。采用CIGWO算法对BP神经网络模型寻优得到最优神经元阈值和权值，将其与多种单一模型进行实验比较，CIGWO-BP模型预测精度最高。采用变分模态分解(VMD)对单一模型的预测残差进行分解，利用决策树(DTA)模型对分解量预测，将其与原模型预测值合并为最终预测结果，预测精度均有较大提升，其中VMD-CIGWO-BP-DTA模型的MAE、MAPE和RMSE相较于CIGWO-BP模型分别降低了20.79%、45.58%、55.12%。

Abstract: A prediction model (VMD-CIGWO-BP-DTA) combining a Circle chaoticised grey wolf algorithm (CIGWO) optimised BP neural network with variational modal decomposition (VMD) is proposed for prediction analysis of storage air conditioning loads. The CIGWO algorithm is used to find the optimal neuron thresholds and weights for the BP neural network model, and the CIGWO-BP model has the highest prediction accuracy when compared with various single models. The prediction residuals of the single model were decomposed using variational modal decomposition (VMD), and the decomposed quantities were predicted using a decision tree (DTA) model, which were combined with the predicted values of the original model to form the final prediction results, and the prediction accuracies were all greatly improved. Compared with CIGWO-BP model, MAE, MAPE and RMSE of VMD-CIGWO-BP-DTA model were reduced by 20.79%, 45.58% and 55.12%, respectively.