Typical Industry Customers’ Demand Response Potential Evaluation Method Based on Integrated Empirical Mode Decomposition and Multi-Head Convolutional Self-Attention

Abstract

Accurate evaluation of typical industry customers’ demand response potential (DRP) is of great significance for promoting the electricity retail companies to achieve DR targets and supporting the balance regulation of the power system with a high penetration of renewable energy resources. Existing DRP evaluation methods ignore the differences in customers’ DR features and the correlation between DR features in different time periods. Moreover, the characterisation of DR willingness only considers the impact of electricity prices, which reduces the accuracy of DRP evaluation results. Given this background, a DRP evaluation method based on integrated empirical mode decomposition (IEMD) and the multi-head convolutional self-attention algorithm (MCSA) for typical industry customers is proposed in this paper. Firstly, an IEMD and DR willingness-based method for extracting DR features of industry customers is proposed. Then, an MCSA-based DRP evaluation method for typical industry customers, utilising the extracted DR features, is developed to realise accurate DRP evaluation by electricity retail companies. Finally, case studies on the industry customers in Zhejiang province, China, show that the proposed method can obtain higher accuracy in evaluating the typical industry customers’ DRP, thus providing technical support for the electricity retail companies to fully mobilise the flexible resources of the demand side.