Edge Computing and Real‑time Load Regulation System in Smart Grid
TL;DR
This paper constructs a three-level collaborative architecture of edge, region, and cloud, designs an edge node deployment scheme based on load density orientation and a fusion communication mechanism, proposes an improved lightweight LSTM real-time load prediction model, and builds an actual distribution network sub-area experimental platform to complete system verification.
Abstract
The smart grid is confronted with real-time scheduling challenges brought about by the random fluctuations in the output of distributed power sources and the dynamic growth of diverse loads. The traditional centralized computing architecture has problems such as high communication latency and prominent computing capacity bottlenecks, making it difficult to meet the millisecond-level load optimization requirements of the power grid. Edge computing, as a distributed computing paradigm close to the terminal side, possesses advantages such as low latency, high reliability, and local processing capabilities, providing a new technical path to solve the bottleneck of real-time load scheduling in the smart grid. This paper constructs a three-level collaborative architecture of edge, region, and cloud, designs an edge node deployment scheme based on load density orientation and a fusion communication mechanism, proposes an improved lightweight LSTM real-time load prediction model, and builds an actual distribution network sub-area experimental platform to complete system verification. The results show that the average absolute percentage error of system prediction is reduced to 4.8%, the scheduling response time is only 12ms, and it still operates stably under extreme conditions, effectively enhancing the source-load coordination balance ability, operational safety, and economic efficiency of the smart grid.