From Traditional Design of Power Supply Systems to Automated Support of Information Decisions Based on a Systems Approach and Digital Twins

The aim of the study is to reduce the time required and ensure the quality of the systemic transformation of methods for intelligent support of managerial decision-making through the integration of physical and mathematical modelling and machine learning in a digital twin (DT) environment to overcome information fragmentation at the design stage of power supply systems for industrial enterprises. The paper addresses the problem of overcoming a priori uncertainty and the “information vacuum” effect when assessing the remaining service life of electrical equipment during the design stage under conditions where the use of historical failure data is limited. The research methods include systems analysis, physics-informed machine neural networks (PINN), ontological modelling (CIM), Bayesian inference for quantifying epistemic uncertainty, and surrogate multiphysics modelling. The novelty of the study lies in demonstrating the methodological inadequacy of deterministic normative calculations based on fixed coefficients for assessing nonlinear equipment degradation. A structural and functional model is proposed for converting routine heuristic operations into authorized software scenarios, forming a “digital pipeline.” The research results include a seven-step algorithm for eliminating the gap in design between projected reliability and actual service life. The work shows that the integration of a priori physical laws acts as a regularization agent in the small-data regime, reducing the prediction error of remaining useful life. The findings state that automating the parameterization of digital twins in designing power supply systems for industrial enterprises eliminates the “expert bottleneck” and ensures the scalability of intelligent decision support systems.