Dongxu Zhou, Jingming Zhao, Fanqin Zeng, Xiashan Feng, and Shuoyu Li
Topological structure of the substation area, topology generation, topology reconstruction, long short-term memory network (LSTM), lasso
The accurate topology structure of the power grid substation is crucial for the power system operating stably, but traditional topology identification methods often have low efficiency and accuracy. To improve the accuracy of topology generation and reconstruction in power grid substations, this study proposes a method for generating and reconstructing power grid substation topology based on long short-term memory networks and Lasso regression models. This method combines the time recognition advantage of long short-term memory with the sparsity feature of Lasso regression. It can handle the complex dynamic behaviour and internal relationships of power grid substations, while ensuring the readability and accuracy of the model. This study normalises and standardises real-time data from the power grid, and inputs it into an improved Lasso model for topology generation and reconstruction. The results demonstrated that the average absolute error estimated by the model was within 1.59%, with 80% of the errors distributed within the range of [0, 2%], and no significant outliers were observed in each time period. Compared to traditional artificial neural networks and support vector regression models, its predictive performance was more outstanding. The topology generation algorithm proposed in the study exhibits high accuracy and generation efficiency, enabling real-time monitoring of the power grid topology structure. This, in turn, effectively improves the stability and security of the power grid structure. The topology reconstruction algorithm has high reconstruction efficiency, and the reconstructed topology structure significantly reduces line losses and operating costs of the power grid, improving the economic efficiency of the power grid.
Important Links:
Go Back
