CBGA: A deep learning method for power grid communication networks service activity prediction

Abstract

Power grid communication network service activity (PCNSA) refers to the active degree of access devices in the power grid communication networks that are actively providing services over a specific time period as a gauge of the operation status of power devices and load demand. Accurate monitoring of PCNSA helps optimize power dispatch, ensure supply and demand balance, and guide network planning and management. However, due to the complex nonlinear, multi-scale and multivariate characteristics of the PCNSA data, it is often difficult to capture its intrinsic patterns and dynamic changes through a simple model. To address this problem, this paper proposes a convolutional neural network(CNN), bi-directional gated recurrent unit(BiGRU), and attention mechanism(AM)-based service activity prediction method for power grid CBGA communication network, referred to as CBGA. CNN is used to extract the features of time-series data, BiGRU is used to capture the dependencies before and after the extracted data in both forward and backward directions and predict the service activity data of the next hour, and AM is used to capture the influence of feature states and enhance the extraction of important information. In this paper, experiments and comparisons are conducted on three real power grid communication network service datasets, and the experimental results demonstrate that the performance of this method is the best, with the smallest MAE and RMSE and the largest R2 score. In comparison to other models, CBGA model is better suited for the prediction of PCNSA and providing reliable data support for power communication network managers to make decisions .