Safety Performance Evaluation of Data Center Electromechanical Systems with Deep Learning
Guangzuo Yu · 2023
With various critical operations increasingly relying on data centers, ensuring the security and reliability of their electromechanical systems has become crucial. This article proposes a deep learning-based method for evaluating the security performance of data center electromechanical systems. By harnessing the power of artificial intelligence and data analytics, this method can assess potential risks and vulnerabilities more accurately and proactively. Firstly, the article introduces a Multi-Scale Convolutional Neural Network (MSCNN) for evaluating the security performance of data center electromechanical systems. It utilizes larger convolutional kernels to learn raw features and employs multi-scale modules to extract multi-scale features. This effectively enhances feature discriminability and robustness, thereby improving the network's performance. Secondly, this work enhances generalization capability by employing BN normalization and dropout techniques. Thirdly, various experiments are conducted on the proposed MSCNN model, and the experimental data validate the feasibility of applying this method to evaluate the security performance of data center electromechanical systems.