Data-driven real-time price-based demand response for industrial facilities energy management

Recent advances in smart grid technologies have highlighted demand response (DR) as an important tool to alleviate electricity demand–supply mismatches. In this paper, a real-time price (RTP)-based DR algorithm is proposed for industrial facilities, aiming to minimize the electricity cost while satisfying production requirements. In particular, due to future price uncertainties, a data-driven approach is adopted to forecast the future unknown prices for supporting global time horizon optimization, which is realized by long short-term memory recurrent neural network (LSTM RNN). With the aid of predicted prices, the industrial facility energy management is formulated as a mixed integer linear programming (MILP) problem, which is then solved by Gurobi over a rolling horizon basis. Finally, an entire practical steel powder manufacturing process is selected as a case study to verify the RTP-based DR scheme. Numerical simulation results show that the proposed scheme is able to effectively shift energy consumption from peak to off-peak periods and reduce the electricity cost of the facility, while satisfying all of the operating constraints. The performance of the presented data-driven RTP forecasting approach is compared to different prediction methods, and error sensitivity analyses are also conducted to evaluate the impact of the RTP uncertainties and the robustness of the proposed RTP-based DR algorithm. Moreover, the DR capability to RTPs is investigated. •Propose a data-driven real-time price-based industrial demand response scheme.•Long short-term memory network is adopted to overcome future price uncertainties.•The decision making process is solved over a rolling horizon basis.•Error sensitivity analyses are conducted to evaluate the robustness of the scheme.•The demand response capability to real-time prices is investigated.