DR Participants’ Actual Response Prediction Using Artificial Neural Networks

Abstract

Empowering the consumers will increase the complexity of local communities’ management. Enabling bidirectional communication and appliances to become smarter can be a huge step toward implementing demand response. However, a solution capable of providing the right knowledge and tools must be developed. The authors thereby propose a methodology to manage the active consumers on Demand Response (DR) events optimally, considering the context in which it is triggered. The distribution system operator detects a voltage violation and requests a load reduction to the aggregators. In this study, to test a performance rate designed by the authors to deal with response uncertainty, a comparison between requested and actual reduction is done. The proposed methodology was applied to three scenarios where the goal is predicting the response from the consumers using artificial neural networks, by changing the features used in the input.