基于Attention-LSTM的电力负荷自动预测方法

Automatic prediction method of electric load based on Attention-LSTM

电力负荷预测方法异常数据筛选能力较差,导致预测精度较低,为此,基于Attention-LSTM研究一种新的电力负荷自动预测方法,构建ZigBee组网协议下的数据采集信息组网,采集电力负荷原始数据,建立自动预测模型,将电力负荷数据输入到数据处理模块中,构建模态序列,将各个子序列中的模态分量结果重组叠加,得到电力负荷功率的预测结果。实验结果表明,该方法能够筛选出绝大部分异常数据,异常数据筛选率在90%以上,预测精度在99%以上,预测时间低于15 s。

The power load forecasting method has poor ability to filter abnormal data,resulting in low forecasting accuracy.Therefore,a new automatic power load forecasting method based on Attention-LSTM is studied.The data acquisition information networking under ZigBee networking protocol is constructed,the original power load data is collected,the automatic forecasting model is established,the power load data is input into the data processing module,the modal sequence is constructed,the modal component results in each sub sequence are recombined and superimposed,and the power of power load is the prediction result.The experimental results show that this method can screen out most of the abnormal data,the screening rate of abnormal data is more than 90%,the prediction accuracy is more than 99%,and the prediction time is less than 15 s.