Evaluation of Power Grid Social Risk Early Warning System Based on Deep Learning

Abstract

In the context of the continuous development of the power grid, the tasks of regulation, operation, and management are becoming increasingly complex, and the operation risks are also increasing dramatically. Sensor technology can deal with the impact of uncertain risk factors, such as extremely disastrous weather, equipment failure, and load fluctuation, on the power grid. Therefore, this article proposes a real-time risk analysis and early warning system for the power grid based on machine learning and combined with sensing technology—a stack self-coding (SSC) neural network prediction model—and introduces the functional composition of the system, clarifying the research content. The experiment compared the accuracy of power grid load forecasting between the SSC forecasting model and the fuzzy neural network (FNN) forecasting model and obtained the forecasting curves of a holiday, a workday, and a Sunday, as well as a comprehensive forecasting accuracy comparison. The experimental results showed that the SSC prediction model based on machine learning designed in this paper improved the prediction accuracy by 12.94% compared with the FNN model. The power grid risk can be assessed through load forecasting, and it is also of great significance for load dispatching and reducing generation costs.