基于优化LSTM超参数的短期空调负荷预测模型
Short-term air conditioning load prediction model based on optimized LSTM hyperparameters
摘要：
提出了一种基于反向学习(OBL)策略改进的信息获取优化算法(IIAO)来优化长短期记忆(LSTM)网络的混合预测算法。首先采用Spearman(斯皮尔曼)相关系数法筛选出与空调负荷高度相关的特征,同时利用IIAO算法对LSTM模型的学习率和L2正则化系数等超参数进行优化,得到最优超参数组合,构建了IIAO-LSTM空调负荷预测模型,最后将该模型应用于西安市某高校实验室的空调负荷预测中,并与其他预测模型进行了对比。实验结果表明,IIAO-LSTM模型预测空调负荷的平均绝对百分比误差MAPE和均方根误差RMSE分别为1.05%和3.71 kW,模型运行时间为23.33 s,具有更高的预测精度和更短的运行时间,泛化能力较强,适用于具有强时序性特征的空调负荷预测。
Abstract：
This paper proposes a hybrid prediction algorithm with a long short-term memory (LSTM) network optimized by the improved information acquisition optimizer (IIAO) based on the opposition-based learning (OBL) strategy. Firstly, the Spearman correlation coefficient method is used to select features highly correlated with air conditioning loads. Subsequently, the hyperparameters of the LSTM model, including learning rate and L2 regularization coefficient, are optimized using the IIAO algorithm to obtain the optimal combination of hyperparameters, and the IIAO-LSTM air conditioning load prediction model is constructed. Finally, this model is applied to the air conditioning load prediction of a university laboratory in Xi’an city, and is compared with other prediction models. Experimental results show that the mean absolute percentage error (MAPE) and root mean square error (RMSE) of the IIAO-LSTM model in predicting air conditioning load are 1.05% and 3.71 kW, respectively, and the model running time is 23.33 seconds. It has higher prediction accuracy and shorter running time, and has strong generalization ability. It is suitable for predicting air conditioning loads with strong temporal characteristics.
Keywords:air conditioning load; prediction model; long short-term memory (LSTM) network; hyperparameter; opposition-based learning (OBL); improved information acquisition optimizer (IIAO); Spearman correlation coefficient method