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

Introduction

The regional power grid plays various roles, such as allocating distributed power for users and allocating wind, solar, and other renewable energy for the regional power grid. The failure of the regional power grid would directly affect the reliable power supply of various users, so it is necessary to analyze the power grid’s social risk early warning system. There is already some research on power grid risk early warning systems. Tian et al. (2020) studied the early warning and suppression of subsequent commutation faults during the restoration process under a power grid fault. Gu et al. (2019) studied the lightning risk assessment and early warning of the UHVDC transmission channel. Y. Liu et al. (2021) developed the risk early warning technology for the overhead transmission line tripping process caused by wildfire. Zhang et al. (2017) conducted intelligent early warning on power system dynamic insecurity risks to achieve the best balance between accuracy and precision. Chen et al. (2019) designed a statistical risk assessment framework for distribution network resilience to address grid risks. Stone et al. (2021) combined the weather forecast regional climate model with the advanced building energy model to simulate the interior building temperature of more than 2.8 million residents in Atlanta, Detroit, Michigan, and Phoenix, Arizona, in heat wave and power failure conditions. The research results showed that 68% to 100% of the urban population was exposed to a high risk of heat exhaustion or heatstroke due to the simulated recent intensity and duration of the composite heat wave and power grid failure events. P. Liu et al. (2017) designed wind turbines’ dynamic tripping risk alarm mechanism in large-scale wind farms. The above research has analyzed the power grid risk early warning.

Many scholars have studied the application of deep learning power grids. He et al. (2017) designed an intelligent mechanism based on deep learning for real-time detection of false data injection attacks in smart grids. Jeyaraj and Nadar (2021) argued that the computer-aided demand-side energy management in the residential smart residential grid could adopt a new set-depth learning algorithm. Ibrahim et al. (2021) designed an intelligent grid power theft detection method based on deep learning to deal with grid risk. Alam et al. (2022) designed the best energy management scheme for photovoltaic and battery energy storage integrated home microgrid systems based on deep learning. Kishor et al. (2021) designed the intelligent grid operation on deep learning and data analysis to support renewable energy. The research of the above scholars has made fruitful progress in the in-depth study of grid applications. Compared with traditional load forecasting methods, the random forest algorithm, and the LSTM model, the load forecasting model based on the fuzzy neural network has higher forecasting accuracy in tasks related to time series data and can be applied to short-term load forecasting research of power systems.

At present, there is an urgent need to analyze and control the risks of regional power grids comprehensively. Currently, research on short-term load forecasting mainly focuses on deep learning. Still, the training cost of deep learning models is often high, especially when combined with other time series decomposition methods. The most intuitive manifestation is that the model requires a longer training time. The existing research has fully considered the impact of meteorological information and transmission line fault characteristics on power grid risks and has achieved some results. It is essential to formulate production operation plans and other risk checks to achieve planning optimization and safety control to eliminate and prevent known risks.

The innovations of this article include the following points:

• 1.

  This article proposes security analysis standards for transmission networks and successfully applies them to energy management systems in transmission networks.

• 2.

  This article studies the safety evaluation indicators of power system operation under the influence of sudden faults.

• 3.

  This article proposes a real-time risk analysis and warning system for power grids based on machine learning combined with sensing technology, which has higher accuracy than traditional methods.