Analysis of Electrical Load Balancing by Simulation and Neural Network Forecast

The rising share of renewable energy poses new challenges to actors of electricity markets: wind and solar energy are not available without variation and interruption, so there is a rising need of high priced control energy. Smart grids are introduced to deal with this problem, by load balancing at the electricity consumers and producers. We analyze the capabilities of electrical load balancing and present initial results, starting with a short review of relevant literature. Second part is an analysis of load balancing potentials at consumer households. A software prototype is developed for simulating the reaction to dynamically changing electricity rates, by implementing two generic classes of smart devices: devices running once in a defined limited time slice, as as simplified model of real devices like dish washer, clothes washer, or laundry dryer; devices without time restriction and a given daily runtime, as a simplified model of water heater with large storage. Third part is an analysis of centrally controlled combined heat and power plants (CHPP) for load balancing in a virtual power plant composed of CHPPs, wind and solar energy plants. CHPP load is driven by heating requirements but we want to forecast the (uninfluenced) produced electricity. Our neural network forecast of CHPP load allows to alter the behavior of heat (and electricity) production. In times of low demand or high production by wind and solar energy, the CHPP can be switched off, provided that sufficient heat reserves have been accumulated before. Based on the neural network forecast, a software prototype simulates the effects of load balancing in virtual power plants by controlling the CHPPs.