A new strategy for power monitoring data collection based on data mining and its role in improving prediction accuracy

In this paper, the data collection and preprocessing process in the power monitoring process is first described, and the collected data are preprocessed using normalization processing technique and sliding sampling technique. After that, the Local Outlier Factor (LOF) and Isolated Forest (iForest) methods are used to monitor abnormal power values. Finally, the samples and labels obtained are inputted into the improved Transformer model for tuning, training, prediction, and evaluation of the model. The results show that the improved LOF algorithm is able to significantly recognize power anomaly data. For the application effect of the improved Transformer model, it is found that the MAPE of the model is improved by 65.2% and 61.13% over the other models, and the R2 is almost close to 1. In different datasets and validation experiments, the R2 of the model is 99.63% and 97.71%, respectively, and the model’s accuracy is still extremely high. It shows that the prediction of power monitoring data using the proposed power data monitoring hair method is effective and can be applied in practice.