A Software Suite for Predictive Energy Control: Modular Architecture, Forecasting–Optimization Integration, and Reproducible Deployment Workflow

Predictive control studies often report forecasting or optimization algorithms in isolation, whereas deployment requires a software suite that coordinates telemetry ingestion, data-quality control, forecasting, optimization, monitoring, and model lifecycle management. This paper presents a modular software suite for predictive control of energy consumption systems and analyzes it as a reproducible data-to-decision workflow. The suite is organized around six interacting layers: the physical object, telemetry and preprocessing, the forecasting service, the optimization and supervisory decision service, monitoring, and lifecycle support. An implementation is demonstrated on an indoor ice rink refrigeration system, where a control-oriented state model and a nonlinear predictive optimizer are embedded as service components. The paper formalizes functional, operational, and reproducibility requirements, describes the architecture and orchestration logic, and reports runtime and traceability results for the full execution cycle. In the case study, the mean full cycle time is 0.0311 s and the maximum time is 0.1064 s for a 5 min supervisory interval, leaving a reserve above 2800x. The suite also records model versions, configuration manifests, runtime logs, and monitoring indicators that support retraining and release management.