关键词: 绝热压缩空气储能, 模型预测控制, 功率跟踪, 状态空间, 动态模型, 滚动优化, 并网功率调节

Abstract:

The adiabatic compressed air energy storage （A-CAES） systems face challenges including high response speed requirements， strong pressure-temperature coupling， and multiple operational safety constraints in grid-connected power scheduling and automatic generation control （AGC） frequency-regulation command tracking.To address these issues，a nonlinear model predictive control （NMPC）-based power tracking strategy considering multiple operational constraints was proposed. Dynamic models of the main components， including the compressor， turbine， air storage chamber， thermal storage system， and heat exchanger， was established according to their thermodynamic characteristics. The air storage chamber pressure， air storage chamber temperature， and thermal storage tank temperature were selected as state variables， while the air mass flow rate was adopted as a unified control input for charging and discharging. The continuous nonlinear equations were discretized to establish a state-space prediction model suitable for receding horizon optimization. With grid-connected power-tracking accuracy and control-input smoothness as the main objectives， the proposed controller considers constraints on air storage pressure， temperature， mass flow rate， and input variation rate. A terminal pressure cost was introduced to guide the storage state toward a reasonable operating region， and a pressure-constraint relaxation variable was added to improve the feasibility of the optimization problem under complex operating conditions. For numerical implementation， a symbolic modeling and automatic differentiation tool was used to formulate the nonlinear programming problem， and an interior-point solver was employed to perform the receding horizon optimization at each sampling instant. Only the first element of the optimal control sequence was applied to the system， after which the optimization problem was updated using the latest measured state， thereby forming a closed-loop process of prediction， optimization， execution， and feedback correction. With reference to the design parameters of a practical compressed air energy storage power station， a dynamic simulation model was developed in Matlab. Step power， ramp power， and AGC command tracking scenarios were designed to evaluate the proposed method and compare it with conventional proportional-integral-derivative （PID） control. The results show that the proposed NMPC strategy achieves higher tracking accuracy under all three scenarios， significantly reduces the root mean square error and maximum tracking deviation， suppresses transient error peaks during charging and discharging switching， and maintains stable operation within the allowable ranges of pressure， temperature， and actuator constraints. The average optimization time is well below the sampling interval， indicating satisfactory online computational performance. These results demonstrate that the controller can balance tracking performance， operational safety， and computational efficiency under rapidly changing power commands. The proposed method therefore provides a reliable control solution for A-CAES systems participating in grid frequency regulation and flexible grid-connected power regulation.

Key words: adiabatic compressed air energy storage, model predictive control, power tracking, state space, dynamic model, receding horizon optimization, grid-connected power regulation