Advanced wind-cooling coordination by tensor train based higher order dynamic mode decomposition

The global transition to a carbon-neutral energy system faces a significant challenge due to the rapid increase in building cooling requirements. Integrating wind energy, which is cost-competitive but highly intermittent, into building cooling processes presents a promising way to achieve sustainable development. However, because of the significant temporal and spatial mismatch between supply and demand, it still encounters challenges. To address this issue, a new framework is proposed that combines tensor train decomposition and higher order dynamic mode decomposition. This framework starts by compressing high-dimensional temporal and spatial dynamics into a low-rank tensor train format. Then, higher order dynamic mode decomposition is applied to identify the dominant temporal and spatial modes, and their dynamic behaviors. This enables the implementation of an effective model predictive control strategy for real-time coordination. Compared with existed methods such as canonical polyadic decomposition and long short-term memory networks, the proposed framework reduces the amount of wind curtailment by 7.2% and operating costs by 21.4%–25.2% while maintaining thermal comfort violation rates below 3.5%. Furthermore, the framework achieves 95.1% prediction accuracy with an online computation time of only 0.45 s per control step, fully meeting real-time requirements.