The large variability in power consumption in electrical power systems (EPS) influences not only growth balance losses and technical losses, but also in some cases reduces energy security. Delayed restoration of power generation, combined with unpredictable weather events leading to the loss of generating power can lead to a situation in which to save the stability of the power system there must be introduced in the system a load power limit or even disconnection of end-user in a given area, which will significantly reduce the comfort of use of energy. This situation can be prevented through either the building of new intervention power units or the aggregated use of new energy technologies, such as distributed network resources (DER), which are part of an intelligent Smart Grid network. Such resources bring together virtual power plants (VPP) and demand side management (DSM). The article presents an alternative decentralized active demand response (DADR) system, that by acting on selected groups of loads reduces peak loads with minimized loss of comfort of energy in use for the end-user. The system operates without any communication. The effectiveness of the proposed solution has been confirmed, outlined in test results obtained by the authors from a developed analytical model, which also contains stochastic algorithms to decrease the negative impact of such DSM systems on the power system (power overshoot and oscillation).