低碳住宅建筑日用电负荷多指标自适应预测

Multi Index Adaptive Prediction of Daily Electricity Load for Low-carbon Residential Buildings

考虑负荷数据具有差异性、周期性、波动性、多变性,极易受到周期性负荷、节假日效应与气象因素3个指标对负荷的影响,设计低碳住宅建筑日用电负荷多指标自适应预测模型。利用模糊C-均值聚类算法,聚类处理周期性负荷、节假日效应、气象因素3个指标的历史数据;通过结合长短期记忆与门控循环单元,组建多层双向递归神经网络模型;利用灰狼算法优化网络模型参数,建立成熟的日用电负荷多指标自适应预测模型;在成熟的模型内输入3个指标的特征数据,输出日用电负荷预测结果。实验证明:该模型可有效聚类历史数据,获取多指标特征数据;该模型可精准自适应预测低碳住宅建筑日用电负荷;应用该模型后,可提升负荷率与碳减排效益。

Considering the difference,periodicity,fluctuation and variability of load data,which are vulnerable to the impact of cyclical load,holiday effect and meteorological factors on load,a multi index adaptive forecasting model for daily power load of low-carbon residential buildings is designed.The fuzzy C-means clustering algorithm is used to process the historical data of periodic load,holiday effect and meteorological factors;By combining short and long term memory with gated loop unit,a multi-layer bidirectional recurrent neural network model is established;The grey wolf algorithm is used to optimize the network model parameters,and a mature multi index adaptive forecasting model of daily power load is established;Input the characteristic data of the three indicators in the mature model,and output the daily power load forecast results.The experiment shows that the model can effectively cluster historical data and obtain multi index feature data;The model can accurately and adaptively predict the daily power load of low carbon residential buildings;After applying the model,the load rate and carbon emission reduction benefits can be improved.