A discussion is presented on the nature of highly-centralized optimization of the closed-loop adaptive stability control technology and the influence of model uncertainties on its suitability.Industrial experiences in dissolving uncertainties in security defense practices are introduced.It is pointed out that for design of security defense framework time-space conditions applicable for highly-centralized optimization system should be taken into full consideration，the conceptual design of a large complex closed-loop system control framework should not be solely based on highly-centralized optimization，other methods of not relying on system modeling accuracy should be developed besides consistently improving modeling accuracy，time-space distributed decision-making should be used to share risks of wrong decisions due to uncertainties，the role of engineers in the decision-making process should be emphasized and on-line verification of the adaptability of emergency control decisions should be widely used，flexibility of control objectives should be considered to avoid seeking simultaneous satisfaction of multi optimal man-made objectives and risk based concept should be used to coordinate on-line and off-line pre-decision-making，and further research into decision support technology of manual emergency control should be conducted.