Manufacturing SMEs face a structural challenge in energy management: despite the growing availability of enabling digital technologies, energy consumption monitoring, diagnostic analysis, and operational control remain separate functions, lacking systematic integration. This fragmentation prevents the construction of a coherent path from energy data to operational decisions, leaving invisible inefficiencies that directly affect economic margins. The literature highlights that this gap is not exclusively technological, but stems from a combination of organisational barriers, decision-making limitations, and the absence of replicable architectures oriented toward the constraints of SMEs.This paper proposes a modular framework for digital energy management in manufacturing SMEs, implemented in Python and structured around six interconnected functional blocks: multi-machine energy simulation with integration of the economic dimension, controlled generation of degradative scenarios with explicit ground truth, hybrid anomaly detection combining deterministic rules and unsupervised learning, CPS control logics for intelligent standby and dynamic power capping, comparative energy and economic evaluation, and an orchestration dashboard. The architecture is designed to be modular, parametrisable, and replicable across contexts with varying levels of digital maturity.Experimental validation is conducted in a controlled simulation environment on a four-machine heterogeneous plant over an annual horizon. The counterfactual comparison CPS ON vs. baseline reveals a reduction in energy consumption of 1.28% and in energy cost of 1.52%, with an improvement in energy intensity of 0.24 kWh/h working. The anomaly detection system achieves F1 = 1.000 on impulsive anomalies, F1 = 0.590 on moderate drift, and F1 = 0.621 under stress conditions, confirming the complementarity between physical rules and the unsupervised model as the nature of the degradative phenomenon varies.The framework demonstrates that an integrated, modular architecture can connect energy data, advanced diagnostics, and operational decisions in a structured and measurable way, offering a concrete response to the fragmentation identified in the literature. The main contribution lies not in the point-level performance of individual modules, but in demonstrating that this approach is achievable even in the most constrained production contexts, lowering the threshold of access to digital energy management for the entire manufacturing SME ecosystem.
A. Ferretto Parodi (Tue,) studied this question.