Abstract: The gas-supercritical CO₂ combined cycle is clean and high-efficient, with a compact structure. As its components are of smaller volumes, the combined cycle has a stronger capability of rapid load change. As an accommodation source, it is a viable solution for large-scale accommodation of renewable energy by rapid load change. This combined cycle uses a supercritical CO₂ cycle and a transcritical CO₂ cycle to recover the exhaust heat from a gas turbine. To solve the issues of rapid prediction and optimization of off-design performance under rapid load change condition, this study proposes a solution procedure based on face-centered cubic design and back-propagation neural network. In terms of the particle swarm optimization algorithm, the optimal sliding pressure operation strategy is proposed. The results show the off-design performance prediction method is of a satisfied accuracy, and it can shorten the simulation time. Compared with the strategy of proportional mass flow rate operation, the optimal sliding pressure operation strategy has a higher efficiency and a wider operation range. It indicates that with this operation strategy the gas-supercritical CO₂ combined cycle can have a better off-design performance.