Turkish Journal of Electrical Engineering and Computer Sciences




This paper presents the application of deep learning algorithms towards demand response management. Demand limit violation and voltage stability are the major problems associated with a secondary distribution system. These problems are solved using demand response models by day ahead scheduling loads at every 15 min interval through linear integer programming and based on short term forecasting of load (kW). A new architecture for short term load forecasting is presented namely gated recurrent unit in which statistical analysis is carried out to get the optimal architecture of the neural network model. Reliability indices such as loss of load probability (LOLP) is evaluated to handle uncertainties that may occur in forecasting due to overestimation and underestimation. Here a novel dynamic power flow is carried out to check limit violation of the voltage. Also, scheduling is performed for two types of loads namely deferrable with interruptible and deferrable with uninterruptable, when either of maximum demand or voltage limit violation occurs. Finally, the suggested model is validated on a modified 12 bus radial distribution system. The result analysis shows that the suggested gated recurrent unit minimizes the forecast error and demand response program schedules household appliances without a demand limit violation and ensures the prevention of voltage collapse.


Long short-term memory, load forecasting, gated recurrent unit, linear integer programming, voltage stability analysis, deep learning

First Page


Last Page