The purpose of this work is to present a concept how to realize a decentral energy-efficient and selfsufficient energy supply system, consisting of a wind-photovoltaic (PV) or PV- micro-hydro power plants for a stable local energy supply. This comprises of subsystems for controlling the energy flows between power generators and consumers, sensor networks for keeping energy balance and to predict failures in particular subsystems (so called predictive maintenance), and optimization of energy demand tailored to remote areas in Central Asia. The energy production of this self-sufficient system shall be based on regenerative energy sources only such as wind, hydro and photovoltaic. The system will be supported by an appropriate energy storage system with, e.g., lithium ion or lead-gel or even lead-acid batteries. This study presents a higher-level management system based on IIoT networks and machine learning methods which control the individual energy generators (wind, hydro and photovoltaic) together with the connected consumers (load management) to avoid peak loads and to trigger suitable processes in case of energy surplus or deficiency. The dimensioning of each subsystem has to consider specific local energy demands and specific climate conditions with dedicated forecasting and machine learning methods.