Differential privacy distributed optimization algorithm against adversarial attacks for efficiency optimization of complex industrial processes

In order to improve the overall performance of large-scale industrial processes with complex constraints, vast amounts of production data are collected for distributed process optimization. Usually, this data contains a lot of enterprise sensitive information. Considering the conventional distributed algorithm has the issue of data privacy leakage, which weakens the safety of the entire manufacturing process and poses a serious threat to economic benefit, this paper proposes a privacy-preserving based distributed algorithm for a class of optimization problem of large-scale industrial processes. Additionally, the proposed method can be applied to the privacy of industrial process optimization problems between different enterprises, such as industrial value chain optimization problems. In specific, the differential privacy mechanism is adopted to protect the data privacy of the local node. Meanwhile, the mutual information technique is adopted to analyze the information loss in the communication data. Moreover, Lagrange primal-dual method is used to deal with the coupling inequality constraint. Subsequently, rigorous theoretical proof shows the convergence of the proposed algorithm. Then, the algorithm privacy metrics fully demonstrate the rationality and superiority of the mutual information technology used in this article for designing privacy parameters. Finally, experimental results of numerical cases and ethylene process optimization show the effectiveness of the proposed algorithm.