A Deep Neural Network-Based Adaptive Dispatch Optimization for Power Blockchain Systems

With the advent of big data and cloud computing, enterprises, organizations and individuals are increasingly interconnected, leading to a geometric increase in data resource sharing among institutions. However, this trend raises significant concerns regarding user data security and access control, particularly within power systems. This paper proposes leveraging deep neural network models integrated with blockchain technology to process and analyze security information in power big data. We introduce an enhanced approach by combining the Hopfield Neural Network (HNN) with the Simulated Annealing (SA) algorithm, addressing the limitations inherent in the traditional HNN model. Our proposed framework, SA-HNN, is designed to improve the adaptive capabilities of power systems through blockchain-based security strategies. In our experiments, we compared SA-HNN with other machine learning models, including XGBoost, LightGBM and Linear SVC, focusing on storage time and data integrity. The results indicate that SA-HNN outperforms these models in both metrics. Specifically, during a power system security defense test, SA-HNN achieved an algorithm recognition rate exceeding 95%. Furthermore, when evaluating the average transaction time consumption in power blockchain transactions, SA-HNN demonstrated superior performance, handling large volumes of transaction data efficiently with shorter processing times. In terms of user attribute revocation efficiency, SA-HNN exhibited greater file processing capacity and shorter time performance compared with other models. This research highlights the potential of integrating advanced neural networks with blockchain technology to enhance the security and efficiency of power systems. Future work will focus on further refining these models and exploring their applications in broader contexts.