Abstract:

To address the issues of low energy regulation efficiency, slow response speed, and poor robustness faced by energy storage systems in the development of renewable energy and smart grids, an energy management strategy that integrates deep learning(DL) and model predictive control(MPC) algorithm is proposed. Firstly, an accurate mathematical model that considers temperature changes and aging effects is established. The optimal network structure is determined based on Kolmogorov’s theorem, and the weight coefficients are scientifically set using the analytic hierarchy process. Then, the dynamic adaptive adjustment of parameters is achieved by using the model predictive control algorithm. Finally, through comprehensive comparative experiments with 11 mainstream methods and robustness tests under 8 typical disturbances, the results show that the energy regulation efficiency of the proposed method reaches 99.6%, and the response time is only 28 ms, which is significantly superior to traditional strategy. It demonstrates excellent stability and anti-interference ability under various complex disturbance conditions.

Key words: grid energy storage system, deep learning algorithm, energy management strategy, model predictive control algorithm, robustness